* [RFC PATCH 0/2] add an external testing library for unit tests
@ 2023-04-27 17:50 Calvin Wan
2023-04-27 17:50 ` [RFC PATCH 1/2] Add C TAP harness Calvin Wan
` (5 more replies)
0 siblings, 6 replies; 29+ messages in thread
From: Calvin Wan @ 2023-04-27 17:50 UTC (permalink / raw)
To: git; +Cc: Calvin Wan
In our current testing environment, we spend a significant amount of
effort crafting end-to-end tests for error conditions that could easily
be captured by unit tests (or we simply forgo some hard-to-setup and
rare error conditions). Unit tests additionally provide stability to the
codebase and can simplify debugging through isolation. Turning parts of
Git into libraries[1] gives us the ability to run unit tests on the
libraries and to write unit tests in C. Writing unit tests in pure C,
rather than with our current shell/test-tool helper setup, simplifies
test setup, simplifies passing data around (no shell-isms required), and
reduces testing runtime by not spawning a separate process for every
test invocation.
Unit testing in C requires a separate testing harness that we ideally
would like to be TAP-style and to come with a non-restrictive license.
Fortunately, there already exists a C TAP harness library[2] with an MIT
license (at least for the files included in this series).
This first patch introduces the C TAP harness and includes only the
necessary files. The second patch showcases a basic example of it. As an
RFC, I am wondering what the list thinks about using a second testing
library for unit testing? Are there any problems with this particular
external testing library and does it provide the necessary functionality
we would want for unit testing Git libraries? How should the folders be
structured and how should the new Makefile rules be written? Ultimately,
this will help us determine the setup of our unit tests in future
libification patches.
[1] https://lore.kernel.org/git/CAJoAoZ=Cig_kLocxKGax31sU7Xe4==BGzC__Bg2_pr7krNq6MA@mail.gmail.com/
[2] https://github.com/rra/c-tap-harness/
Calvin Wan (2):
Add C TAP harness
unit test: add basic example
Makefile | 11 +
t/TESTS | 1 +
t/runtests.c | 1789 ++++++++++++++++++++++++++++++++++++++++++++++++
t/tap/basic.c | 1029 ++++++++++++++++++++++++++++
t/tap/basic.h | 198 ++++++
t/tap/macros.h | 109 +++
t/unit-test.c | 14 +
7 files changed, 3151 insertions(+)
create mode 100644 t/TESTS
create mode 100644 t/runtests.c
create mode 100644 t/tap/basic.c
create mode 100644 t/tap/basic.h
create mode 100644 t/tap/macros.h
create mode 100644 t/unit-test.c
--
2.40.1.495.gc816e09b53d-goog
^ permalink raw reply [flat|nested] 29+ messages in thread
* [RFC PATCH 1/2] Add C TAP harness
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
@ 2023-04-27 17:50 ` Calvin Wan
2023-04-27 18:29 ` SZEDER Gábor
2023-04-27 20:15 ` Phillip Wood
2023-04-27 17:50 ` [RFC PATCH 2/2] unit test: add basic example Calvin Wan
` (4 subsequent siblings)
5 siblings, 2 replies; 29+ messages in thread
From: Calvin Wan @ 2023-04-27 17:50 UTC (permalink / raw)
To: git; +Cc: Calvin Wan
Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
There is also more complete documentation at
https://www.eyrie.org/~eagle/software/c-tap-harness/
Signed-off-by: Calvin Wan <calvinwan@google.com>
---
t/TESTS | 0
t/runtests.c | 1789 ++++++++++++++++++++++++++++++++++++++++++++++++
t/tap/basic.c | 1029 ++++++++++++++++++++++++++++
t/tap/basic.h | 198 ++++++
t/tap/macros.h | 109 +++
5 files changed, 3125 insertions(+)
create mode 100644 t/TESTS
create mode 100644 t/runtests.c
create mode 100644 t/tap/basic.c
create mode 100644 t/tap/basic.h
create mode 100644 t/tap/macros.h
diff --git a/t/TESTS b/t/TESTS
new file mode 100644
index 0000000000..e69de29bb2
diff --git a/t/runtests.c b/t/runtests.c
new file mode 100644
index 0000000000..4a55a801a6
--- /dev/null
+++ b/t/runtests.c
@@ -0,0 +1,1789 @@
+/*
+ * Run a set of tests, reporting results.
+ *
+ * Test suite driver that runs a set of tests implementing a subset of the
+ * Test Anything Protocol (TAP) and reports the results.
+ *
+ * Any bug reports, bug fixes, and improvements are very much welcome and
+ * should be sent to the e-mail address below. This program is part of C TAP
+ * Harness <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
+ *
+ * Copyright 2000-2001, 2004, 2006-2019, 2022 Russ Allbery <eagle@eyrie.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * SPDX-License-Identifier: MIT
+ */
+
+/*
+ * Usage:
+ *
+ * runtests [-hv] [-b <build-dir>] [-s <source-dir>] -l <test-list>
+ * runtests [-hv] [-b <build-dir>] [-s <source-dir>] <test> [<test> ...]
+ * runtests -o [-h] [-b <build-dir>] [-s <source-dir>] <test>
+ *
+ * In the first case, expects a list of executables located in the given file,
+ * one line per executable, possibly followed by a space-separated list of
+ * options. For each one, runs it as part of a test suite, reporting results.
+ * In the second case, use the same infrastructure, but run only the tests
+ * listed on the command line.
+ *
+ * Test output should start with a line containing the number of tests
+ * (numbered from 1 to this number), optionally preceded by "1..", although
+ * that line may be given anywhere in the output. Each additional line should
+ * be in the following format:
+ *
+ * ok <number>
+ * not ok <number>
+ * ok <number> # skip
+ * not ok <number> # todo
+ *
+ * where <number> is the number of the test. An optional comment is permitted
+ * after the number if preceded by whitespace. ok indicates success, not ok
+ * indicates failure. "# skip" and "# todo" are a special cases of a comment,
+ * and must start with exactly that formatting. They indicate the test was
+ * skipped for some reason (maybe because it doesn't apply to this platform)
+ * or is testing something known to currently fail. The text following either
+ * "# skip" or "# todo" and whitespace is the reason.
+ *
+ * As a special case, the first line of the output may be in the form:
+ *
+ * 1..0 # skip some reason
+ *
+ * which indicates that this entire test case should be skipped and gives a
+ * reason.
+ *
+ * Any other lines are ignored, although for compliance with the TAP protocol
+ * all lines other than the ones in the above format should be sent to
+ * standard error rather than standard output and start with #.
+ *
+ * This is a subset of TAP as documented in Test::Harness::TAP or
+ * TAP::Parser::Grammar, which comes with Perl.
+ *
+ * If the -o option is given, instead run a single test and display all of its
+ * output. This is intended for use with failing tests so that the person
+ * running the test suite can get more details about what failed.
+ *
+ * If built with the C preprocessor symbols C_TAP_SOURCE and C_TAP_BUILD
+ * defined, C TAP Harness will export those values in the environment so that
+ * tests can find the source and build directory and will look for tests under
+ * both directories. These paths can also be set with the -b and -s
+ * command-line options, which will override anything set at build time.
+ *
+ * If the -v option is given, or the C_TAP_VERBOSE environment variable is set,
+ * display the full output of each test as it runs rather than showing a
+ * summary of the results of each test.
+ */
+
+/* Required for fdopen(), getopt(), and putenv(). */
+#if defined(__STRICT_ANSI__) || defined(PEDANTIC)
+# ifndef _XOPEN_SOURCE
+# define _XOPEN_SOURCE 500
+# endif
+#endif
+
+#include <ctype.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <limits.h>
+#include <stdarg.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <strings.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <time.h>
+#include <unistd.h>
+
+/* sys/time.h must be included before sys/resource.h on some platforms. */
+#include <sys/resource.h>
+
+/* AIX 6.1 (and possibly later) doesn't have WCOREDUMP. */
+#ifndef WCOREDUMP
+# define WCOREDUMP(status) ((unsigned) (status) &0x80)
+#endif
+
+/*
+ * POSIX requires that these be defined in <unistd.h>, but they're not always
+ * available. If one of them has been defined, all the rest almost certainly
+ * have.
+ */
+#ifndef STDIN_FILENO
+# define STDIN_FILENO 0
+# define STDOUT_FILENO 1
+# define STDERR_FILENO 2
+#endif
+
+/*
+ * Used for iterating through arrays. Returns the number of elements in the
+ * array (useful for a < upper bound in a for loop).
+ */
+#define ARRAY_SIZE(array) (sizeof(array) / sizeof((array)[0]))
+
+/*
+ * The source and build versions of the tests directory. This is used to set
+ * the C_TAP_SOURCE and C_TAP_BUILD environment variables (and the SOURCE and
+ * BUILD environment variables set for backward compatibility) and find test
+ * programs, if set. Normally, this should be set as part of the build
+ * process to the test subdirectories of $(abs_top_srcdir) and
+ * $(abs_top_builddir) respectively.
+ */
+#ifndef C_TAP_SOURCE
+# define C_TAP_SOURCE NULL
+#endif
+#ifndef C_TAP_BUILD
+# define C_TAP_BUILD NULL
+#endif
+
+/* Test status codes. */
+enum test_status {
+ TEST_FAIL,
+ TEST_PASS,
+ TEST_SKIP,
+ TEST_INVALID
+};
+
+/* Really, just a boolean, but this is more self-documenting. */
+enum test_verbose {
+ CONCISE = 0,
+ VERBOSE = 1
+};
+
+/* Indicates the state of our plan. */
+enum plan_status {
+ PLAN_INIT, /* Nothing seen yet. */
+ PLAN_FIRST, /* Plan seen before any tests. */
+ PLAN_PENDING, /* Test seen and no plan yet. */
+ PLAN_FINAL /* Plan seen after some tests. */
+};
+
+/* Error exit statuses for test processes. */
+#define CHILDERR_DUP 100 /* Couldn't redirect stderr or stdout. */
+#define CHILDERR_EXEC 101 /* Couldn't exec child process. */
+#define CHILDERR_STDIN 102 /* Couldn't open stdin file. */
+#define CHILDERR_STDERR 103 /* Couldn't open stderr file. */
+
+/* Structure to hold data for a set of tests. */
+struct testset {
+ char *file; /* The file name of the test. */
+ char **command; /* The argv vector to run the command. */
+ enum plan_status plan; /* The status of our plan. */
+ unsigned long count; /* Expected count of tests. */
+ unsigned long current; /* The last seen test number. */
+ unsigned int length; /* The length of the last status message. */
+ unsigned long passed; /* Count of passing tests. */
+ unsigned long failed; /* Count of failing lists. */
+ unsigned long skipped; /* Count of skipped tests (passed). */
+ unsigned long allocated; /* The size of the results table. */
+ enum test_status *results; /* Table of results by test number. */
+ unsigned int aborted; /* Whether the set was aborted. */
+ unsigned int reported; /* Whether the results were reported. */
+ int status; /* The exit status of the test. */
+ unsigned int all_skipped; /* Whether all tests were skipped. */
+ char *reason; /* Why all tests were skipped. */
+};
+
+/* Structure to hold a linked list of test sets. */
+struct testlist {
+ struct testset *ts;
+ struct testlist *next;
+};
+
+/*
+ * Usage message. Should be used as a printf format with four arguments: the
+ * path to runtests, given three times, and the usage_description. This is
+ * split into variables to satisfy the pedantic ISO C90 limit on strings.
+ */
+static const char usage_message[] = "\
+Usage: %s [-hv] [-b <build-dir>] [-s <source-dir>] <test> ...\n\
+ %s [-hv] [-b <build-dir>] [-s <source-dir>] -l <test-list>\n\
+ %s -o [-h] [-b <build-dir>] [-s <source-dir>] <test>\n\
+\n\
+Options:\n\
+ -b <build-dir> Set the build directory to <build-dir>\n\
+%s";
+static const char usage_extra[] = "\
+ -l <list> Take the list of tests to run from <test-list>\n\
+ -o Run a single test rather than a list of tests\n\
+ -s <source-dir> Set the source directory to <source-dir>\n\
+ -v Show the full output of each test\n\
+\n\
+runtests normally runs each test listed on the command line. With the -l\n\
+option, it instead runs every test listed in a file. With the -o option,\n\
+it instead runs a single test and shows its complete output.\n";
+
+/*
+ * Header used for test output. %s is replaced by the file name of the list
+ * of tests.
+ */
+static const char banner[] = "\n\
+Running all tests listed in %s. If any tests fail, run the failing\n\
+test program with runtests -o to see more details.\n\n";
+
+/* Header for reports of failed tests. */
+static const char header[] = "\n\
+Failed Set Fail/Total (%) Skip Stat Failing Tests\n\
+-------------------------- -------------- ---- ---- ------------------------";
+
+/* Include the file name and line number in malloc failures. */
+#define xcalloc(n, type) \
+ ((type *) x_calloc((n), sizeof(type), __FILE__, __LINE__))
+#define xmalloc(size) ((char *) x_malloc((size), __FILE__, __LINE__))
+#define xstrdup(p) x_strdup((p), __FILE__, __LINE__)
+#define xstrndup(p, size) x_strndup((p), (size), __FILE__, __LINE__)
+#define xreallocarray(p, n, type) \
+ ((type *) x_reallocarray((p), (n), sizeof(type), __FILE__, __LINE__))
+
+/*
+ * __attribute__ is available in gcc 2.5 and later, but only with gcc 2.7
+ * could you use the __format__ form of the attributes, which is what we use
+ * (to avoid confusion with other macros).
+ */
+#ifndef __attribute__
+# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 7)
+# define __attribute__(spec) /* empty */
+# endif
+#endif
+
+/*
+ * We use __alloc_size__, but it was only available in fairly recent versions
+ * of GCC. Suppress warnings about the unknown attribute if GCC is too old.
+ * We know that we're GCC at this point, so we can use the GCC variadic macro
+ * extension, which will still work with versions of GCC too old to have C99
+ * variadic macro support.
+ */
+#if !defined(__attribute__) && !defined(__alloc_size__)
+# if defined(__GNUC__) && !defined(__clang__)
+# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3)
+# define __alloc_size__(spec, args...) /* empty */
+# endif
+# endif
+#endif
+
+/*
+ * Suppress the argument to __malloc__ in Clang (not supported in at least
+ * version 13) and GCC versions prior to 11.
+ */
+#if !defined(__attribute__) && !defined(__malloc__)
+# if defined(__clang__) || __GNUC__ < 11
+# define __malloc__(dalloc) __malloc__
+# endif
+#endif
+
+/*
+ * LLVM and Clang pretend to be GCC but don't support all of the __attribute__
+ * settings that GCC does. For them, suppress warnings about unknown
+ * attributes on declarations. This unfortunately will affect the entire
+ * compilation context, but there's no push and pop available.
+ */
+#if !defined(__attribute__) && (defined(__llvm__) || defined(__clang__))
+# pragma GCC diagnostic ignored "-Wattributes"
+#endif
+
+/* Declare internal functions that benefit from compiler attributes. */
+static void die(const char *, ...)
+ __attribute__((__nonnull__, __noreturn__, __format__(printf, 1, 2)));
+static void sysdie(const char *, ...)
+ __attribute__((__nonnull__, __noreturn__, __format__(printf, 1, 2)));
+static void *x_calloc(size_t, size_t, const char *, int)
+ __attribute__((__alloc_size__(1, 2), __malloc__(free), __nonnull__));
+static void *x_malloc(size_t, const char *, int)
+ __attribute__((__alloc_size__(1), __malloc__(free), __nonnull__));
+static void *x_reallocarray(void *, size_t, size_t, const char *, int)
+ __attribute__((__alloc_size__(2, 3), __malloc__(free), __nonnull__(4)));
+static char *x_strdup(const char *, const char *, int)
+ __attribute__((__malloc__(free), __nonnull__));
+static char *x_strndup(const char *, size_t, const char *, int)
+ __attribute__((__malloc__(free), __nonnull__));
+
+
+/*
+ * Report a fatal error and exit.
+ */
+static void
+die(const char *format, ...)
+{
+ va_list args;
+
+ fflush(stdout);
+ fprintf(stderr, "runtests: ");
+ va_start(args, format);
+ vfprintf(stderr, format, args);
+ va_end(args);
+ fprintf(stderr, "\n");
+ exit(1);
+}
+
+
+/*
+ * Report a fatal error, including the results of strerror, and exit.
+ */
+static void
+sysdie(const char *format, ...)
+{
+ int oerrno;
+ va_list args;
+
+ oerrno = errno;
+ fflush(stdout);
+ fprintf(stderr, "runtests: ");
+ va_start(args, format);
+ vfprintf(stderr, format, args);
+ va_end(args);
+ fprintf(stderr, ": %s\n", strerror(oerrno));
+ exit(1);
+}
+
+
+/*
+ * Allocate zeroed memory, reporting a fatal error and exiting on failure.
+ */
+static void *
+x_calloc(size_t n, size_t size, const char *file, int line)
+{
+ void *p;
+
+ n = (n > 0) ? n : 1;
+ size = (size > 0) ? size : 1;
+ p = calloc(n, size);
+ if (p == NULL)
+ sysdie("failed to calloc %lu bytes at %s line %d",
+ (unsigned long) size, file, line);
+ return p;
+}
+
+
+/*
+ * Allocate memory, reporting a fatal error and exiting on failure.
+ */
+static void *
+x_malloc(size_t size, const char *file, int line)
+{
+ void *p;
+
+ p = malloc(size);
+ if (p == NULL)
+ sysdie("failed to malloc %lu bytes at %s line %d",
+ (unsigned long) size, file, line);
+ return p;
+}
+
+
+/*
+ * Reallocate memory, reporting a fatal error and exiting on failure.
+ *
+ * We should technically use SIZE_MAX here for the overflow check, but
+ * SIZE_MAX is C99 and we're only assuming C89 + SUSv3, which does not
+ * guarantee that it exists. They do guarantee that UINT_MAX exists, and we
+ * can assume that UINT_MAX <= SIZE_MAX. And we should not be allocating
+ * anything anywhere near that large.
+ *
+ * (In theory, C89 and C99 permit size_t to be smaller than unsigned int, but
+ * I disbelieve in the existence of such systems and they will have to cope
+ * without overflow checks.)
+ */
+static void *
+x_reallocarray(void *p, size_t n, size_t size, const char *file, int line)
+{
+ n = (n > 0) ? n : 1;
+ size = (size > 0) ? size : 1;
+
+ if (n > 0 && UINT_MAX / n <= size)
+ sysdie("realloc too large at %s line %d", file, line);
+ p = realloc(p, n * size);
+ if (p == NULL)
+ sysdie("failed to realloc %lu bytes at %s line %d",
+ (unsigned long) (n * size), file, line);
+ return p;
+}
+
+
+/*
+ * Copy a string, reporting a fatal error and exiting on failure.
+ */
+static char *
+x_strdup(const char *s, const char *file, int line)
+{
+ char *p;
+ size_t len;
+
+ len = strlen(s) + 1;
+ p = (char *) malloc(len);
+ if (p == NULL)
+ sysdie("failed to strdup %lu bytes at %s line %d", (unsigned long) len,
+ file, line);
+ memcpy(p, s, len);
+ return p;
+}
+
+
+/*
+ * Copy the first n characters of a string, reporting a fatal error and
+ * existing on failure.
+ *
+ * Avoid using the system strndup function since it may not exist (on Mac OS
+ * X, for example), and there's no need to introduce another portability
+ * requirement.
+ */
+char *
+x_strndup(const char *s, size_t size, const char *file, int line)
+{
+ const char *p;
+ size_t len;
+ char *copy;
+
+ /* Don't assume that the source string is nul-terminated. */
+ for (p = s; (size_t) (p - s) < size && *p != '\0'; p++)
+ ;
+ len = (size_t) (p - s);
+ copy = (char *) malloc(len + 1);
+ if (copy == NULL)
+ sysdie("failed to strndup %lu bytes at %s line %d",
+ (unsigned long) len, file, line);
+ memcpy(copy, s, len);
+ copy[len] = '\0';
+ return copy;
+}
+
+
+/*
+ * Form a new string by concatenating multiple strings. The arguments must be
+ * terminated by (const char *) 0.
+ *
+ * This function only exists because we can't assume asprintf. We can't
+ * simulate asprintf with snprintf because we're only assuming SUSv3, which
+ * does not require that snprintf with a NULL buffer return the required
+ * length. When those constraints are relaxed, this should be ripped out and
+ * replaced with asprintf or a more trivial replacement with snprintf.
+ */
+static char *
+concat(const char *first, ...)
+{
+ va_list args;
+ char *result;
+ const char *string;
+ size_t offset;
+ size_t length = 0;
+
+ /*
+ * Find the total memory required. Ensure we don't overflow length. We
+ * aren't guaranteed to have SIZE_MAX, so use UINT_MAX as an acceptable
+ * substitute (see the x_nrealloc comments).
+ */
+ va_start(args, first);
+ for (string = first; string != NULL; string = va_arg(args, const char *)) {
+ if (length >= UINT_MAX - strlen(string)) {
+ errno = EINVAL;
+ sysdie("strings too long in concat");
+ }
+ length += strlen(string);
+ }
+ va_end(args);
+ length++;
+
+ /* Create the string. */
+ result = xmalloc(length);
+ va_start(args, first);
+ offset = 0;
+ for (string = first; string != NULL; string = va_arg(args, const char *)) {
+ memcpy(result + offset, string, strlen(string));
+ offset += strlen(string);
+ }
+ va_end(args);
+ result[offset] = '\0';
+ return result;
+}
+
+
+/*
+ * Given a struct timeval, return the number of seconds it represents as a
+ * double. Use difftime() to convert a time_t to a double.
+ */
+static double
+tv_seconds(const struct timeval *tv)
+{
+ return difftime(tv->tv_sec, 0) + (double) tv->tv_usec * 1e-6;
+}
+
+
+/*
+ * Given two struct timevals, return the difference in seconds.
+ */
+static double
+tv_diff(const struct timeval *tv1, const struct timeval *tv0)
+{
+ return tv_seconds(tv1) - tv_seconds(tv0);
+}
+
+
+/*
+ * Given two struct timevals, return the sum in seconds as a double.
+ */
+static double
+tv_sum(const struct timeval *tv1, const struct timeval *tv2)
+{
+ return tv_seconds(tv1) + tv_seconds(tv2);
+}
+
+
+/*
+ * Given a pointer to a string, skip any leading whitespace and return a
+ * pointer to the first non-whitespace character.
+ */
+static const char *
+skip_whitespace(const char *p)
+{
+ while (isspace((unsigned char) (*p)))
+ p++;
+ return p;
+}
+
+
+/*
+ * Given a pointer to a string, skip any non-whitespace characters and return
+ * a pointer to the first whitespace character, or to the end of the string.
+ */
+static const char *
+skip_non_whitespace(const char *p)
+{
+ while (*p != '\0' && !isspace((unsigned char) (*p)))
+ p++;
+ return p;
+}
+
+
+/*
+ * Start a program, connecting its stdout to a pipe on our end and its stderr
+ * to /dev/null, and storing the file descriptor to read from in the two
+ * argument. Returns the PID of the new process. Errors are fatal.
+ */
+static pid_t
+test_start(char *const *command, int *fd)
+{
+ int fds[2], infd, errfd;
+ pid_t child;
+
+ /* Create a pipe used to capture the output from the test program. */
+ if (pipe(fds) == -1) {
+ puts("ABORTED");
+ fflush(stdout);
+ sysdie("can't create pipe");
+ }
+
+ /* Fork a child process, massage the file descriptors, and exec. */
+ child = fork();
+ switch (child) {
+ case -1:
+ puts("ABORTED");
+ fflush(stdout);
+ sysdie("can't fork");
+
+ /* In the child. Set up our standard output. */
+ case 0:
+ close(fds[0]);
+ close(STDOUT_FILENO);
+ if (dup2(fds[1], STDOUT_FILENO) < 0)
+ _exit(CHILDERR_DUP);
+ close(fds[1]);
+
+ /* Point standard input at /dev/null. */
+ close(STDIN_FILENO);
+ infd = open("/dev/null", O_RDONLY);
+ if (infd < 0)
+ _exit(CHILDERR_STDIN);
+ if (infd != STDIN_FILENO) {
+ if (dup2(infd, STDIN_FILENO) < 0)
+ _exit(CHILDERR_DUP);
+ close(infd);
+ }
+
+ /* Point standard error at /dev/null. */
+ close(STDERR_FILENO);
+ errfd = open("/dev/null", O_WRONLY);
+ if (errfd < 0)
+ _exit(CHILDERR_STDERR);
+ if (errfd != STDERR_FILENO) {
+ if (dup2(errfd, STDERR_FILENO) < 0)
+ _exit(CHILDERR_DUP);
+ close(errfd);
+ }
+
+ /* Now, exec our process. */
+ if (execv(command[0], command) == -1)
+ _exit(CHILDERR_EXEC);
+ break;
+
+ /* In parent. Close the extra file descriptor. */
+ default:
+ close(fds[1]);
+ break;
+ }
+ *fd = fds[0];
+ return child;
+}
+
+
+/*
+ * Back up over the output saying what test we were executing.
+ */
+static void
+test_backspace(struct testset *ts)
+{
+ unsigned int i;
+
+ if (!isatty(STDOUT_FILENO))
+ return;
+ for (i = 0; i < ts->length; i++)
+ putchar('\b');
+ for (i = 0; i < ts->length; i++)
+ putchar(' ');
+ for (i = 0; i < ts->length; i++)
+ putchar('\b');
+ ts->length = 0;
+}
+
+
+/*
+ * Allocate or resize the array of test results to be large enough to contain
+ * the test number in.
+ */
+static void
+resize_results(struct testset *ts, unsigned long n)
+{
+ unsigned long i;
+ size_t s;
+
+ /* If there's already enough space, return quickly. */
+ if (n <= ts->allocated)
+ return;
+
+ /*
+ * If no space has been allocated, do the initial allocation. Otherwise,
+ * resize. Start with 32 test cases and then add 1024 with each resize to
+ * try to reduce the number of reallocations.
+ */
+ if (ts->allocated == 0) {
+ s = (n > 32) ? n : 32;
+ ts->results = xcalloc(s, enum test_status);
+ } else {
+ s = (n > ts->allocated + 1024) ? n : ts->allocated + 1024;
+ ts->results = xreallocarray(ts->results, s, enum test_status);
+ }
+
+ /* Set the results for the newly-allocated test array. */
+ for (i = ts->allocated; i < s; i++)
+ ts->results[i] = TEST_INVALID;
+ ts->allocated = s;
+}
+
+
+/*
+ * Report an invalid test number and set the appropriate flags. Pulled into a
+ * separate function since we do this in several places.
+ */
+static void
+invalid_test_number(struct testset *ts, long n, enum test_verbose verbose)
+{
+ if (!verbose)
+ test_backspace(ts);
+ printf("ABORTED (invalid test number %ld)\n", n);
+ ts->aborted = 1;
+ ts->reported = 1;
+}
+
+
+/*
+ * Read the plan line of test output, which should contain the range of test
+ * numbers. We may initialize the testset structure here if we haven't yet
+ * seen a test. Return true if initialization succeeded and the test should
+ * continue, false otherwise.
+ */
+static int
+test_plan(const char *line, struct testset *ts, enum test_verbose verbose)
+{
+ long n;
+
+ /*
+ * Accept a plan without the leading 1.. for compatibility with older
+ * versions of runtests. This will only be allowed if we've not yet seen
+ * a test result.
+ */
+ line = skip_whitespace(line);
+ if (strncmp(line, "1..", 3) == 0)
+ line += 3;
+
+ /*
+ * Get the count and check it for validity.
+ *
+ * If we have something of the form "1..0 # skip foo", the whole file was
+ * skipped; record that. If we do skip the whole file, zero out all of
+ * our statistics, since they're no longer relevant.
+ *
+ * strtol is called with a second argument to advance the line pointer
+ * past the count to make it simpler to detect the # skip case.
+ */
+ n = strtol(line, (char **) &line, 10);
+ if (n == 0) {
+ line = skip_whitespace(line);
+ if (*line == '#') {
+ line = skip_whitespace(line + 1);
+ if (strncasecmp(line, "skip", 4) == 0) {
+ line = skip_whitespace(line + 4);
+ if (*line != '\0') {
+ ts->reason = xstrdup(line);
+ ts->reason[strlen(ts->reason) - 1] = '\0';
+ }
+ ts->all_skipped = 1;
+ ts->aborted = 1;
+ ts->count = 0;
+ ts->passed = 0;
+ ts->skipped = 0;
+ ts->failed = 0;
+ return 0;
+ }
+ }
+ }
+ if (n <= 0) {
+ puts("ABORTED (invalid test count)");
+ ts->aborted = 1;
+ ts->reported = 1;
+ return 0;
+ }
+
+ /*
+ * If we are doing lazy planning, check the plan against the largest test
+ * number that we saw and fail now if we saw a check outside the plan
+ * range.
+ */
+ if (ts->plan == PLAN_PENDING && (unsigned long) n < ts->count) {
+ invalid_test_number(ts, (long) ts->count, verbose);
+ return 0;
+ }
+
+ /*
+ * Otherwise, allocated or resize the results if needed and update count,
+ * and then record that we've seen a plan.
+ */
+ resize_results(ts, (unsigned long) n);
+ ts->count = (unsigned long) n;
+ if (ts->plan == PLAN_INIT)
+ ts->plan = PLAN_FIRST;
+ else if (ts->plan == PLAN_PENDING)
+ ts->plan = PLAN_FINAL;
+ return 1;
+}
+
+
+/*
+ * Given a single line of output from a test, parse it and return the success
+ * status of that test. Anything printed to stdout not matching the form
+ * /^(not )?ok \d+/ is ignored. Sets ts->current to the test number that just
+ * reported status.
+ */
+static void
+test_checkline(const char *line, struct testset *ts, enum test_verbose verbose)
+{
+ enum test_status status = TEST_PASS;
+ const char *bail;
+ char *end;
+ long number;
+ unsigned long current;
+ int outlen;
+
+ /* Before anything, check for a test abort. */
+ bail = strstr(line, "Bail out!");
+ if (bail != NULL) {
+ bail = skip_whitespace(bail + strlen("Bail out!"));
+ if (*bail != '\0') {
+ size_t length;
+
+ length = strlen(bail);
+ if (bail[length - 1] == '\n')
+ length--;
+ if (!verbose)
+ test_backspace(ts);
+ printf("ABORTED (%.*s)\n", (int) length, bail);
+ ts->reported = 1;
+ }
+ ts->aborted = 1;
+ return;
+ }
+
+ /*
+ * If the given line isn't newline-terminated, it was too big for an
+ * fgets(), which means ignore it.
+ */
+ if (line[strlen(line) - 1] != '\n')
+ return;
+
+ /* If the line begins with a hash mark, ignore it. */
+ if (line[0] == '#')
+ return;
+
+ /* If we haven't yet seen a plan, look for one. */
+ if (ts->plan == PLAN_INIT && isdigit((unsigned char) (*line))) {
+ if (!test_plan(line, ts, verbose))
+ return;
+ } else if (strncmp(line, "1..", 3) == 0) {
+ if (ts->plan == PLAN_PENDING) {
+ if (!test_plan(line, ts, verbose))
+ return;
+ } else {
+ if (!verbose)
+ test_backspace(ts);
+ puts("ABORTED (multiple plans)");
+ ts->aborted = 1;
+ ts->reported = 1;
+ return;
+ }
+ }
+
+ /* Parse the line, ignoring something we can't parse. */
+ if (strncmp(line, "not ", 4) == 0) {
+ status = TEST_FAIL;
+ line += 4;
+ }
+ if (strncmp(line, "ok", 2) != 0)
+ return;
+ line = skip_whitespace(line + 2);
+ errno = 0;
+ number = strtol(line, &end, 10);
+ if (errno != 0 || end == line)
+ current = ts->current + 1;
+ else if (number <= 0) {
+ invalid_test_number(ts, number, verbose);
+ return;
+ } else
+ current = (unsigned long) number;
+ if (current > ts->count && ts->plan == PLAN_FIRST) {
+ invalid_test_number(ts, (long) current, verbose);
+ return;
+ }
+
+ /* We have a valid test result. Tweak the results array if needed. */
+ if (ts->plan == PLAN_INIT || ts->plan == PLAN_PENDING) {
+ ts->plan = PLAN_PENDING;
+ resize_results(ts, current);
+ if (current > ts->count)
+ ts->count = current;
+ }
+
+ /*
+ * Handle directives. We should probably do something more interesting
+ * with unexpected passes of todo tests.
+ */
+ while (isdigit((unsigned char) (*line)))
+ line++;
+ line = skip_whitespace(line);
+ if (*line == '#') {
+ line = skip_whitespace(line + 1);
+ if (strncasecmp(line, "skip", 4) == 0)
+ status = TEST_SKIP;
+ if (strncasecmp(line, "todo", 4) == 0)
+ status = (status == TEST_FAIL) ? TEST_SKIP : TEST_FAIL;
+ }
+
+ /* Make sure that the test number is in range and not a duplicate. */
+ if (ts->results[current - 1] != TEST_INVALID) {
+ if (!verbose)
+ test_backspace(ts);
+ printf("ABORTED (duplicate test number %lu)\n", current);
+ ts->aborted = 1;
+ ts->reported = 1;
+ return;
+ }
+
+ /* Good results. Increment our various counters. */
+ switch (status) {
+ case TEST_PASS:
+ ts->passed++;
+ break;
+ case TEST_FAIL:
+ ts->failed++;
+ break;
+ case TEST_SKIP:
+ ts->skipped++;
+ break;
+ case TEST_INVALID:
+ break;
+ }
+ ts->current = current;
+ ts->results[current - 1] = status;
+ if (!verbose && isatty(STDOUT_FILENO)) {
+ test_backspace(ts);
+ if (ts->plan == PLAN_PENDING)
+ outlen = printf("%lu/?", current);
+ else
+ outlen = printf("%lu/%lu", current, ts->count);
+ ts->length = (outlen >= 0) ? (unsigned int) outlen : 0;
+ fflush(stdout);
+ }
+}
+
+
+/*
+ * Print out a range of test numbers, returning the number of characters it
+ * took up. Takes the first number, the last number, the number of characters
+ * already printed on the line, and the limit of number of characters the line
+ * can hold. Add a comma and a space before the range if chars indicates that
+ * something has already been printed on the line, and print ... instead if
+ * chars plus the space needed would go over the limit (use a limit of 0 to
+ * disable this).
+ */
+static unsigned int
+test_print_range(unsigned long first, unsigned long last, unsigned long chars,
+ unsigned int limit)
+{
+ unsigned int needed = 0;
+ unsigned long n;
+
+ for (n = first; n > 0; n /= 10)
+ needed++;
+ if (last > first) {
+ for (n = last; n > 0; n /= 10)
+ needed++;
+ needed++;
+ }
+ if (chars > 0)
+ needed += 2;
+ if (limit > 0 && chars + needed > limit) {
+ needed = 0;
+ if (chars <= limit) {
+ if (chars > 0) {
+ printf(", ");
+ needed += 2;
+ }
+ printf("...");
+ needed += 3;
+ }
+ } else {
+ if (chars > 0)
+ printf(", ");
+ if (last > first)
+ printf("%lu-", first);
+ printf("%lu", last);
+ }
+ return needed;
+}
+
+
+/*
+ * Summarize a single test set. The second argument is 0 if the set exited
+ * cleanly, a positive integer representing the exit status if it exited
+ * with a non-zero status, and a negative integer representing the signal
+ * that terminated it if it was killed by a signal.
+ */
+static void
+test_summarize(struct testset *ts, int status)
+{
+ unsigned long i;
+ unsigned long missing = 0;
+ unsigned long failed = 0;
+ unsigned long first = 0;
+ unsigned long last = 0;
+
+ if (ts->aborted) {
+ fputs("ABORTED", stdout);
+ if (ts->count > 0)
+ printf(" (passed %lu/%lu)", ts->passed, ts->count - ts->skipped);
+ } else {
+ for (i = 0; i < ts->count; i++) {
+ if (ts->results[i] == TEST_INVALID) {
+ if (missing == 0)
+ fputs("MISSED ", stdout);
+ if (first && i == last)
+ last = i + 1;
+ else {
+ if (first)
+ test_print_range(first, last, missing - 1, 0);
+ missing++;
+ first = i + 1;
+ last = i + 1;
+ }
+ }
+ }
+ if (first)
+ test_print_range(first, last, missing - 1, 0);
+ first = 0;
+ last = 0;
+ for (i = 0; i < ts->count; i++) {
+ if (ts->results[i] == TEST_FAIL) {
+ if (missing && !failed)
+ fputs("; ", stdout);
+ if (failed == 0)
+ fputs("FAILED ", stdout);
+ if (first && i == last)
+ last = i + 1;
+ else {
+ if (first)
+ test_print_range(first, last, failed - 1, 0);
+ failed++;
+ first = i + 1;
+ last = i + 1;
+ }
+ }
+ }
+ if (first)
+ test_print_range(first, last, failed - 1, 0);
+ if (!missing && !failed) {
+ fputs(!status ? "ok" : "dubious", stdout);
+ if (ts->skipped > 0) {
+ if (ts->skipped == 1)
+ printf(" (skipped %lu test)", ts->skipped);
+ else
+ printf(" (skipped %lu tests)", ts->skipped);
+ }
+ }
+ }
+ if (status > 0)
+ printf(" (exit status %d)", status);
+ else if (status < 0)
+ printf(" (killed by signal %d%s)", -status,
+ WCOREDUMP(ts->status) ? ", core dumped" : "");
+ putchar('\n');
+}
+
+
+/*
+ * Given a test set, analyze the results, classify the exit status, handle a
+ * few special error messages, and then pass it along to test_summarize() for
+ * the regular output. Returns true if the test set ran successfully and all
+ * tests passed or were skipped, false otherwise.
+ */
+static int
+test_analyze(struct testset *ts)
+{
+ if (ts->reported)
+ return 0;
+ if (ts->all_skipped) {
+ if (ts->reason == NULL)
+ puts("skipped");
+ else
+ printf("skipped (%s)\n", ts->reason);
+ return 1;
+ } else if (WIFEXITED(ts->status) && WEXITSTATUS(ts->status) != 0) {
+ switch (WEXITSTATUS(ts->status)) {
+ case CHILDERR_DUP:
+ if (!ts->reported)
+ puts("ABORTED (can't dup file descriptors)");
+ break;
+ case CHILDERR_EXEC:
+ if (!ts->reported)
+ puts("ABORTED (execution failed -- not found?)");
+ break;
+ case CHILDERR_STDIN:
+ case CHILDERR_STDERR:
+ if (!ts->reported)
+ puts("ABORTED (can't open /dev/null)");
+ break;
+ default:
+ test_summarize(ts, WEXITSTATUS(ts->status));
+ break;
+ }
+ return 0;
+ } else if (WIFSIGNALED(ts->status)) {
+ test_summarize(ts, -WTERMSIG(ts->status));
+ return 0;
+ } else if (ts->plan != PLAN_FIRST && ts->plan != PLAN_FINAL) {
+ puts("ABORTED (no valid test plan)");
+ ts->aborted = 1;
+ return 0;
+ } else {
+ test_summarize(ts, 0);
+ return (ts->failed == 0);
+ }
+}
+
+
+/*
+ * Runs a single test set, accumulating and then reporting the results.
+ * Returns true if the test set was successfully run and all tests passed,
+ * false otherwise.
+ */
+static int
+test_run(struct testset *ts, enum test_verbose verbose)
+{
+ pid_t testpid, child;
+ int outfd, status;
+ unsigned long i;
+ FILE *output;
+ char buffer[BUFSIZ];
+
+ /* Run the test program. */
+ testpid = test_start(ts->command, &outfd);
+ output = fdopen(outfd, "r");
+ if (!output) {
+ puts("ABORTED");
+ fflush(stdout);
+ sysdie("fdopen failed");
+ }
+
+ /*
+ * Pass each line of output to test_checkline(), and print the line if
+ * verbosity is requested.
+ */
+ while (!ts->aborted && fgets(buffer, sizeof(buffer), output)) {
+ if (verbose)
+ printf("%s", buffer);
+ test_checkline(buffer, ts, verbose);
+ }
+ if (ferror(output) || ts->plan == PLAN_INIT)
+ ts->aborted = 1;
+ if (!verbose)
+ test_backspace(ts);
+
+ /*
+ * Consume the rest of the test output, close the output descriptor,
+ * retrieve the exit status, and pass that information to test_analyze()
+ * for eventual output.
+ */
+ while (fgets(buffer, sizeof(buffer), output))
+ if (verbose)
+ printf("%s", buffer);
+ fclose(output);
+ child = waitpid(testpid, &ts->status, 0);
+ if (child == (pid_t) -1) {
+ if (!ts->reported) {
+ puts("ABORTED");
+ fflush(stdout);
+ }
+ sysdie("waitpid for %u failed", (unsigned int) testpid);
+ }
+ if (ts->all_skipped)
+ ts->aborted = 0;
+ status = test_analyze(ts);
+
+ /* Convert missing tests to failed tests. */
+ for (i = 0; i < ts->count; i++) {
+ if (ts->results[i] == TEST_INVALID) {
+ ts->failed++;
+ ts->results[i] = TEST_FAIL;
+ status = 0;
+ }
+ }
+ return status;
+}
+
+
+/* Summarize a list of test failures. */
+static void
+test_fail_summary(const struct testlist *fails)
+{
+ struct testset *ts;
+ unsigned int chars;
+ unsigned long i, first, last, total;
+ double failed;
+
+ puts(header);
+
+ /* Failed Set Fail/Total (%) Skip Stat Failing (25)
+ -------------------------- -------------- ---- ---- -------------- */
+ for (; fails; fails = fails->next) {
+ ts = fails->ts;
+ total = ts->count - ts->skipped;
+ failed = (double) ts->failed;
+ printf("%-26.26s %4lu/%-4lu %3.0f%% %4lu ", ts->file, ts->failed,
+ total, total ? (failed * 100.0) / (double) total : 0,
+ ts->skipped);
+ if (WIFEXITED(ts->status))
+ printf("%4d ", WEXITSTATUS(ts->status));
+ else
+ printf(" -- ");
+ if (ts->aborted) {
+ puts("aborted");
+ continue;
+ }
+ chars = 0;
+ first = 0;
+ last = 0;
+ for (i = 0; i < ts->count; i++) {
+ if (ts->results[i] == TEST_FAIL) {
+ if (first != 0 && i == last)
+ last = i + 1;
+ else {
+ if (first != 0)
+ chars += test_print_range(first, last, chars, 19);
+ first = i + 1;
+ last = i + 1;
+ }
+ }
+ }
+ if (first != 0)
+ test_print_range(first, last, chars, 19);
+ putchar('\n');
+ }
+}
+
+
+/*
+ * Check whether a given file path is a valid test. Currently, this checks
+ * whether it is executable and is a regular file. Returns true or false.
+ */
+static int
+is_valid_test(const char *path)
+{
+ struct stat st;
+
+ if (access(path, X_OK) < 0)
+ return 0;
+ if (stat(path, &st) < 0)
+ return 0;
+ if (!S_ISREG(st.st_mode))
+ return 0;
+ return 1;
+}
+
+
+/*
+ * Given the name of a test, a pointer to the testset struct, and the source
+ * and build directories, find the test. We try first relative to the current
+ * directory, then in the build directory (if not NULL), then in the source
+ * directory. In each of those directories, we first try a "-t" extension and
+ * then a ".t" extension. When we find an executable program, we return the
+ * path to that program. If none of those paths are executable, just fill in
+ * the name of the test as is.
+ *
+ * The caller is responsible for freeing the path member of the testset
+ * struct.
+ */
+static char *
+find_test(const char *name, const char *source, const char *build)
+{
+ char *path = NULL;
+ const char *bases[3], *suffix, *base;
+ unsigned int i, j;
+ const char *suffixes[3] = {"-t", ".t", ""};
+
+ /* Possible base directories. */
+ bases[0] = ".";
+ bases[1] = build;
+ bases[2] = source;
+
+ /* Try each suffix with each base. */
+ for (i = 0; i < ARRAY_SIZE(suffixes); i++) {
+ suffix = suffixes[i];
+ for (j = 0; j < ARRAY_SIZE(bases); j++) {
+ base = bases[j];
+ if (base == NULL)
+ continue;
+ path = concat(base, "/", name, suffix, (const char *) 0);
+ if (is_valid_test(path))
+ return path;
+ free(path);
+ path = NULL;
+ }
+ }
+ if (path == NULL)
+ path = xstrdup(name);
+ return path;
+}
+
+
+/*
+ * Parse a single line of a test list and store the test name and command to
+ * execute it in the given testset struct.
+ *
+ * Normally, each line is just the name of the test, which is located in the
+ * test directory and turned into a command to run. However, each line may
+ * have whitespace-separated options, which change the command that's run.
+ * Current supported options are:
+ *
+ * valgrind
+ * Run the test under valgrind if C_TAP_VALGRIND is set. The contents
+ * of that environment variable are taken as the valgrind command (with
+ * options) to run. The command is parsed with a simple split on
+ * whitespace and no quoting is supported.
+ *
+ * libtool
+ * If running under valgrind, use libtool to invoke valgrind. This avoids
+ * running valgrind on the wrapper shell script generated by libtool. If
+ * set, C_TAP_LIBTOOL must be set to the full path to the libtool program
+ * to use to run valgrind and thus the test. Ignored if the test isn't
+ * being run under valgrind.
+ */
+static void
+parse_test_list_line(const char *line, struct testset *ts, const char *source,
+ const char *build)
+{
+ const char *p, *end, *option, *libtool;
+ const char *valgrind = NULL;
+ unsigned int use_libtool = 0;
+ unsigned int use_valgrind = 0;
+ size_t len, i;
+
+ /* Determine the name of the test. */
+ p = skip_non_whitespace(line);
+ ts->file = xstrndup(line, p - line);
+
+ /* Check if any test options are set. */
+ p = skip_whitespace(p);
+ while (*p != '\0') {
+ end = skip_non_whitespace(p);
+ if (strncmp(p, "libtool", end - p) == 0) {
+ use_libtool = 1;
+ } else if (strncmp(p, "valgrind", end - p) == 0) {
+ valgrind = getenv("C_TAP_VALGRIND");
+ use_valgrind = (valgrind != NULL);
+ } else {
+ option = xstrndup(p, end - p);
+ die("unknown test list option %s", option);
+ }
+ p = skip_whitespace(end);
+ }
+
+ /* Construct the argv to run the test. First, find the length. */
+ len = 1;
+ if (use_valgrind && valgrind != NULL) {
+ p = skip_whitespace(valgrind);
+ while (*p != '\0') {
+ len++;
+ p = skip_whitespace(skip_non_whitespace(p));
+ }
+ if (use_libtool)
+ len += 2;
+ }
+
+ /* Now, build the command. */
+ ts->command = xcalloc(len + 1, char *);
+ i = 0;
+ if (use_valgrind && valgrind != NULL) {
+ if (use_libtool) {
+ libtool = getenv("C_TAP_LIBTOOL");
+ if (libtool == NULL)
+ die("valgrind with libtool requested, but C_TAP_LIBTOOL is not"
+ " set");
+ ts->command[i++] = xstrdup(libtool);
+ ts->command[i++] = xstrdup("--mode=execute");
+ }
+ p = skip_whitespace(valgrind);
+ while (*p != '\0') {
+ end = skip_non_whitespace(p);
+ ts->command[i++] = xstrndup(p, end - p);
+ p = skip_whitespace(end);
+ }
+ }
+ if (i != len - 1)
+ die("internal error while constructing command line");
+ ts->command[i++] = find_test(ts->file, source, build);
+ ts->command[i] = NULL;
+}
+
+
+/*
+ * Read a list of tests from a file, returning the list of tests as a struct
+ * testlist, or NULL if there were no tests (such as a file containing only
+ * comments). Reports an error to standard error and exits if the list of
+ * tests cannot be read.
+ */
+static struct testlist *
+read_test_list(const char *filename, const char *source, const char *build)
+{
+ FILE *file;
+ unsigned int line;
+ size_t length;
+ char buffer[BUFSIZ];
+ const char *start;
+ struct testlist *listhead, *current;
+
+ /* Create the initial container list that will hold our results. */
+ listhead = xcalloc(1, struct testlist);
+ current = NULL;
+
+ /*
+ * Open our file of tests to run and read it line by line, creating a new
+ * struct testlist and struct testset for each line.
+ */
+ file = fopen(filename, "r");
+ if (file == NULL)
+ sysdie("can't open %s", filename);
+ line = 0;
+ while (fgets(buffer, sizeof(buffer), file)) {
+ line++;
+ length = strlen(buffer) - 1;
+ if (buffer[length] != '\n') {
+ fprintf(stderr, "%s:%u: line too long\n", filename, line);
+ exit(1);
+ }
+ buffer[length] = '\0';
+
+ /* Skip comments, leading spaces, and blank lines. */
+ start = skip_whitespace(buffer);
+ if (strlen(start) == 0)
+ continue;
+ if (start[0] == '#')
+ continue;
+
+ /* Allocate the new testset structure. */
+ if (current == NULL)
+ current = listhead;
+ else {
+ current->next = xcalloc(1, struct testlist);
+ current = current->next;
+ }
+ current->ts = xcalloc(1, struct testset);
+ current->ts->plan = PLAN_INIT;
+
+ /* Parse the line and store the results in the testset struct. */
+ parse_test_list_line(start, current->ts, source, build);
+ }
+ fclose(file);
+
+ /* If there were no tests, current is still NULL. */
+ if (current == NULL) {
+ free(listhead);
+ return NULL;
+ }
+
+ /* Return the results. */
+ return listhead;
+}
+
+
+/*
+ * Build a list of tests from command line arguments. Takes the argv and argc
+ * representing the command line arguments and returns a newly allocated test
+ * list, or NULL if there were no tests. The caller is responsible for
+ * freeing.
+ */
+static struct testlist *
+build_test_list(char *argv[], int argc, const char *source, const char *build)
+{
+ int i;
+ struct testlist *listhead, *current;
+
+ /* Create the initial container list that will hold our results. */
+ listhead = xcalloc(1, struct testlist);
+ current = NULL;
+
+ /* Walk the list of arguments and create test sets for them. */
+ for (i = 0; i < argc; i++) {
+ if (current == NULL)
+ current = listhead;
+ else {
+ current->next = xcalloc(1, struct testlist);
+ current = current->next;
+ }
+ current->ts = xcalloc(1, struct testset);
+ current->ts->plan = PLAN_INIT;
+ current->ts->file = xstrdup(argv[i]);
+ current->ts->command = xcalloc(2, char *);
+ current->ts->command[0] = find_test(current->ts->file, source, build);
+ current->ts->command[1] = NULL;
+ }
+
+ /* If there were no tests, current is still NULL. */
+ if (current == NULL) {
+ free(listhead);
+ return NULL;
+ }
+
+ /* Return the results. */
+ return listhead;
+}
+
+
+/* Free a struct testset. */
+static void
+free_testset(struct testset *ts)
+{
+ size_t i;
+
+ free(ts->file);
+ for (i = 0; ts->command[i] != NULL; i++)
+ free(ts->command[i]);
+ free(ts->command);
+ free(ts->results);
+ free(ts->reason);
+ free(ts);
+}
+
+
+/*
+ * Run a batch of tests. Takes two additional parameters: the root of the
+ * source directory and the root of the build directory. Test programs will
+ * be first searched for in the current directory, then the build directory,
+ * then the source directory. Returns true iff all tests passed, and always
+ * frees the test list that's passed in.
+ */
+static int
+test_batch(struct testlist *tests, enum test_verbose verbose)
+{
+ size_t length, i;
+ size_t longest = 0;
+ unsigned int count = 0;
+ struct testset *ts;
+ struct timeval start, end;
+ struct rusage stats;
+ struct testlist *failhead = NULL;
+ struct testlist *failtail = NULL;
+ struct testlist *current, *next;
+ int succeeded;
+ unsigned long total = 0;
+ unsigned long passed = 0;
+ unsigned long skipped = 0;
+ unsigned long failed = 0;
+ unsigned long aborted = 0;
+
+ /* Walk the list of tests to find the longest name. */
+ for (current = tests; current != NULL; current = current->next) {
+ length = strlen(current->ts->file);
+ if (length > longest)
+ longest = length;
+ }
+
+ /*
+ * Add two to longest and round up to the nearest tab stop. This is how
+ * wide the column for printing the current test name will be.
+ */
+ longest += 2;
+ if (longest % 8)
+ longest += 8 - (longest % 8);
+
+ /* Start the wall clock timer. */
+ gettimeofday(&start, NULL);
+
+ /* Now, plow through our tests again, running each one. */
+ for (current = tests; current != NULL; current = current->next) {
+ ts = current->ts;
+
+ /* Print out the name of the test file. */
+ fputs(ts->file, stdout);
+ if (verbose)
+ fputs("\n\n", stdout);
+ else
+ for (i = strlen(ts->file); i < longest; i++)
+ putchar('.');
+ if (isatty(STDOUT_FILENO))
+ fflush(stdout);
+
+ /* Run the test. */
+ succeeded = test_run(ts, verbose);
+ fflush(stdout);
+ if (verbose)
+ putchar('\n');
+
+ /* Record cumulative statistics. */
+ aborted += ts->aborted;
+ total += ts->count + ts->all_skipped;
+ passed += ts->passed;
+ skipped += ts->skipped + ts->all_skipped;
+ failed += ts->failed;
+ count++;
+
+ /* If the test fails, we shuffle it over to the fail list. */
+ if (!succeeded) {
+ if (failhead == NULL) {
+ failhead = xcalloc(1, struct testlist);
+ failtail = failhead;
+ } else {
+ failtail->next = xcalloc(1, struct testlist);
+ failtail = failtail->next;
+ }
+ failtail->ts = ts;
+ failtail->next = NULL;
+ }
+ }
+ total -= skipped;
+
+ /* Stop the timer and get our child resource statistics. */
+ gettimeofday(&end, NULL);
+ getrusage(RUSAGE_CHILDREN, &stats);
+
+ /* Summarize the failures and free the failure list. */
+ if (failhead != NULL) {
+ test_fail_summary(failhead);
+ while (failhead != NULL) {
+ next = failhead->next;
+ free(failhead);
+ failhead = next;
+ }
+ }
+
+ /* Free the memory used by the test lists. */
+ while (tests != NULL) {
+ next = tests->next;
+ free_testset(tests->ts);
+ free(tests);
+ tests = next;
+ }
+
+ /* Print out the final test summary. */
+ putchar('\n');
+ if (aborted != 0) {
+ if (aborted == 1)
+ printf("Aborted %lu test set", aborted);
+ else
+ printf("Aborted %lu test sets", aborted);
+ printf(", passed %lu/%lu tests", passed, total);
+ } else if (failed == 0)
+ fputs("All tests successful", stdout);
+ else
+ printf("Failed %lu/%lu tests, %.2f%% okay", failed, total,
+ (double) (total - failed) * 100.0 / (double) total);
+ if (skipped != 0) {
+ if (skipped == 1)
+ printf(", %lu test skipped", skipped);
+ else
+ printf(", %lu tests skipped", skipped);
+ }
+ puts(".");
+ printf("Files=%u, Tests=%lu", count, total);
+ printf(", %.2f seconds", tv_diff(&end, &start));
+ printf(" (%.2f usr + %.2f sys = %.2f CPU)\n", tv_seconds(&stats.ru_utime),
+ tv_seconds(&stats.ru_stime),
+ tv_sum(&stats.ru_utime, &stats.ru_stime));
+ return (failed == 0 && aborted == 0);
+}
+
+
+/*
+ * Run a single test case. This involves just running the test program after
+ * having done the environment setup and finding the test program.
+ */
+static void
+test_single(const char *program, const char *source, const char *build)
+{
+ char *path;
+
+ path = find_test(program, source, build);
+ if (execl(path, path, (char *) 0) == -1)
+ sysdie("cannot exec %s", path);
+}
+
+
+/*
+ * Main routine. Set the C_TAP_SOURCE, C_TAP_BUILD, SOURCE, and BUILD
+ * environment variables and then, given a file listing tests, run each test
+ * listed.
+ */
+int
+main(int argc, char *argv[])
+{
+ int option;
+ int status = 0;
+ int single = 0;
+ enum test_verbose verbose = CONCISE;
+ char *c_tap_source_env = NULL;
+ char *c_tap_build_env = NULL;
+ char *source_env = NULL;
+ char *build_env = NULL;
+ const char *program;
+ const char *shortlist;
+ const char *list = NULL;
+ const char *source = C_TAP_SOURCE;
+ const char *build = C_TAP_BUILD;
+ struct testlist *tests;
+
+ program = argv[0];
+ while ((option = getopt(argc, argv, "b:hl:os:v")) != EOF) {
+ switch (option) {
+ case 'b':
+ build = optarg;
+ break;
+ case 'h':
+ printf(usage_message, program, program, program, usage_extra);
+ exit(0);
+ case 'l':
+ list = optarg;
+ break;
+ case 'o':
+ single = 1;
+ break;
+ case 's':
+ source = optarg;
+ break;
+ case 'v':
+ verbose = VERBOSE;
+ break;
+ default:
+ exit(1);
+ }
+ }
+ argv += optind;
+ argc -= optind;
+ if ((list == NULL && argc < 1) || (list != NULL && argc > 0)) {
+ fprintf(stderr, usage_message, program, program, program, usage_extra);
+ exit(1);
+ }
+
+ /*
+ * If C_TAP_VERBOSE is set in the environment, that also turns on verbose
+ * mode.
+ */
+ if (getenv("C_TAP_VERBOSE") != NULL)
+ verbose = VERBOSE;
+
+ /*
+ * Set C_TAP_SOURCE and C_TAP_BUILD environment variables. Also set
+ * SOURCE and BUILD for backward compatibility, although we're trying to
+ * migrate to the ones with a C_TAP_* prefix.
+ */
+ if (source != NULL) {
+ c_tap_source_env = concat("C_TAP_SOURCE=", source, (const char *) 0);
+ if (putenv(c_tap_source_env) != 0)
+ sysdie("cannot set C_TAP_SOURCE in the environment");
+ source_env = concat("SOURCE=", source, (const char *) 0);
+ if (putenv(source_env) != 0)
+ sysdie("cannot set SOURCE in the environment");
+ }
+ if (build != NULL) {
+ c_tap_build_env = concat("C_TAP_BUILD=", build, (const char *) 0);
+ if (putenv(c_tap_build_env) != 0)
+ sysdie("cannot set C_TAP_BUILD in the environment");
+ build_env = concat("BUILD=", build, (const char *) 0);
+ if (putenv(build_env) != 0)
+ sysdie("cannot set BUILD in the environment");
+ }
+
+ /* Run the tests as instructed. */
+ if (single)
+ test_single(argv[0], source, build);
+ else if (list != NULL) {
+ shortlist = strrchr(list, '/');
+ if (shortlist == NULL)
+ shortlist = list;
+ else
+ shortlist++;
+ printf(banner, shortlist);
+ tests = read_test_list(list, source, build);
+ status = test_batch(tests, verbose) ? 0 : 1;
+ } else {
+ tests = build_test_list(argv, argc, source, build);
+ status = test_batch(tests, verbose) ? 0 : 1;
+ }
+
+ /* For valgrind cleanliness, free all our memory. */
+ if (source_env != NULL) {
+ putenv((char *) "C_TAP_SOURCE=");
+ putenv((char *) "SOURCE=");
+ free(c_tap_source_env);
+ free(source_env);
+ }
+ if (build_env != NULL) {
+ putenv((char *) "C_TAP_BUILD=");
+ putenv((char *) "BUILD=");
+ free(c_tap_build_env);
+ free(build_env);
+ }
+ exit(status);
+}
\ No newline at end of file
diff --git a/t/tap/basic.c b/t/tap/basic.c
new file mode 100644
index 0000000000..704282b9c1
--- /dev/null
+++ b/t/tap/basic.c
@@ -0,0 +1,1029 @@
+/*
+ * Some utility routines for writing tests.
+ *
+ * Here are a variety of utility routines for writing tests compatible with
+ * the TAP protocol. All routines of the form ok() or is*() take a test
+ * number and some number of appropriate arguments, check to be sure the
+ * results match the expected output using the arguments, and print out
+ * something appropriate for that test number. Other utility routines help in
+ * constructing more complex tests, skipping tests, reporting errors, setting
+ * up the TAP output format, or finding things in the test environment.
+ *
+ * This file is part of C TAP Harness. The current version plus supporting
+ * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
+ *
+ * Written by Russ Allbery <eagle@eyrie.org>
+ * Copyright 2009-2019, 2021 Russ Allbery <eagle@eyrie.org>
+ * Copyright 2001-2002, 2004-2008, 2011-2014
+ * The Board of Trustees of the Leland Stanford Junior University
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * SPDX-License-Identifier: MIT
+ */
+
+#include <errno.h>
+#include <limits.h>
+#include <stdarg.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#ifdef _WIN32
+# include <direct.h>
+#else
+# include <sys/stat.h>
+#endif
+#include <sys/types.h>
+#include <unistd.h>
+
+#include <tap/basic.h>
+
+/* Windows provides mkdir and rmdir under different names. */
+#ifdef _WIN32
+# define mkdir(p, m) _mkdir(p)
+# define rmdir(p) _rmdir(p)
+#endif
+
+/*
+ * The test count. Always contains the number that will be used for the next
+ * test status. This is exported to callers of the library.
+ */
+unsigned long testnum = 1;
+
+/*
+ * Status information stored so that we can give a test summary at the end of
+ * the test case. We store the planned final test and the count of failures.
+ * We can get the highest test count from testnum.
+ */
+static unsigned long _planned = 0;
+static unsigned long _failed = 0;
+
+/*
+ * Store the PID of the process that called plan() and only summarize
+ * results when that process exits, so as to not misreport results in forked
+ * processes.
+ */
+static pid_t _process = 0;
+
+/*
+ * If true, we're doing lazy planning and will print out the plan based on the
+ * last test number at the end of testing.
+ */
+static int _lazy = 0;
+
+/*
+ * If true, the test was aborted by calling bail(). Currently, this is only
+ * used to ensure that we pass a false value to any cleanup functions even if
+ * all tests to that point have passed.
+ */
+static int _aborted = 0;
+
+/*
+ * Registered cleanup functions. These are stored as a linked list and run in
+ * registered order by finish when the test program exits. Each function is
+ * passed a boolean value indicating whether all tests were successful.
+ */
+struct cleanup_func {
+ test_cleanup_func func;
+ test_cleanup_func_with_data func_with_data;
+ void *data;
+ struct cleanup_func *next;
+};
+static struct cleanup_func *cleanup_funcs = NULL;
+
+/*
+ * Registered diag files. Any output found in these files will be printed out
+ * as if it were passed to diag() before any other output we do. This allows
+ * background processes to log to a file and have that output interleaved with
+ * the test output.
+ */
+struct diag_file {
+ char *name;
+ FILE *file;
+ char *buffer;
+ size_t bufsize;
+ struct diag_file *next;
+};
+static struct diag_file *diag_files = NULL;
+
+/*
+ * Print a specified prefix and then the test description. Handles turning
+ * the argument list into a va_args structure suitable for passing to
+ * print_desc, which has to be done in a macro. Assumes that format is the
+ * argument immediately before the variadic arguments.
+ */
+#define PRINT_DESC(prefix, format) \
+ do { \
+ if (format != NULL) { \
+ va_list args; \
+ printf("%s", prefix); \
+ va_start(args, format); \
+ vprintf(format, args); \
+ va_end(args); \
+ } \
+ } while (0)
+
+
+/*
+ * Form a new string by concatenating multiple strings. The arguments must be
+ * terminated by (const char *) 0.
+ *
+ * This function only exists because we can't assume asprintf. We can't
+ * simulate asprintf with snprintf because we're only assuming SUSv3, which
+ * does not require that snprintf with a NULL buffer return the required
+ * length. When those constraints are relaxed, this should be ripped out and
+ * replaced with asprintf or a more trivial replacement with snprintf.
+ */
+static char *
+concat(const char *first, ...)
+{
+ va_list args;
+ char *result;
+ const char *string;
+ size_t offset;
+ size_t length = 0;
+
+ /*
+ * Find the total memory required. Ensure we don't overflow length. See
+ * the comment for breallocarray for why we're using UINT_MAX here.
+ */
+ va_start(args, first);
+ for (string = first; string != NULL; string = va_arg(args, const char *)) {
+ if (length >= UINT_MAX - strlen(string))
+ bail("strings too long in concat");
+ length += strlen(string);
+ }
+ va_end(args);
+ length++;
+
+ /* Create the string. */
+ result = bcalloc_type(length, char);
+ va_start(args, first);
+ offset = 0;
+ for (string = first; string != NULL; string = va_arg(args, const char *)) {
+ memcpy(result + offset, string, strlen(string));
+ offset += strlen(string);
+ }
+ va_end(args);
+ result[offset] = '\0';
+ return result;
+}
+
+
+/*
+ * Helper function for check_diag_files to handle a single line in a diag
+ * file.
+ *
+ * The general scheme here used is as follows: read one line of output. If we
+ * get NULL, check for an error. If there was one, bail out of the test
+ * program; otherwise, return, and the enclosing loop will check for EOF.
+ *
+ * If we get some data, see if it ends in a newline. If it doesn't end in a
+ * newline, we have one of two cases: our buffer isn't large enough, in which
+ * case we resize it and try again, or we have incomplete data in the file, in
+ * which case we rewind the file and will try again next time.
+ *
+ * Returns a boolean indicating whether the last line was incomplete.
+ */
+static int
+handle_diag_file_line(struct diag_file *file, fpos_t where)
+{
+ int size;
+ size_t length;
+
+ /* Read the next line from the file. */
+ size = file->bufsize > INT_MAX ? INT_MAX : (int) file->bufsize;
+ if (fgets(file->buffer, size, file->file) == NULL) {
+ if (ferror(file->file))
+ sysbail("cannot read from %s", file->name);
+ return 0;
+ }
+
+ /*
+ * See if the line ends in a newline. If not, see which error case we
+ * have.
+ */
+ length = strlen(file->buffer);
+ if (file->buffer[length - 1] != '\n') {
+ int incomplete = 0;
+
+ /* Check whether we ran out of buffer space and resize if so. */
+ if (length < file->bufsize - 1)
+ incomplete = 1;
+ else {
+ file->bufsize += BUFSIZ;
+ file->buffer =
+ breallocarray_type(file->buffer, file->bufsize, char);
+ }
+
+ /*
+ * On either incomplete lines or too small of a buffer, rewind
+ * and read the file again (on the next pass, if incomplete).
+ * It's simpler than trying to double-buffer the file.
+ */
+ if (fsetpos(file->file, &where) < 0)
+ sysbail("cannot set position in %s", file->name);
+ return incomplete;
+ }
+
+ /* We saw a complete line. Print it out. */
+ printf("# %s", file->buffer);
+ return 0;
+}
+
+
+/*
+ * Check all registered diag_files for any output. We only print out the
+ * output if we see a complete line; otherwise, we wait for the next newline.
+ */
+static void
+check_diag_files(void)
+{
+ struct diag_file *file;
+ fpos_t where;
+ int incomplete;
+
+ /*
+ * Walk through each file and read each line of output available.
+ */
+ for (file = diag_files; file != NULL; file = file->next) {
+ clearerr(file->file);
+
+ /* Store the current position in case we have to rewind. */
+ if (fgetpos(file->file, &where) < 0)
+ sysbail("cannot get position in %s", file->name);
+
+ /* Continue until we get EOF or an incomplete line of data. */
+ incomplete = 0;
+ while (!feof(file->file) && !incomplete) {
+ incomplete = handle_diag_file_line(file, where);
+ }
+ }
+}
+
+
+/*
+ * Our exit handler. Called on completion of the test to report a summary of
+ * results provided we're still in the original process. This also handles
+ * printing out the plan if we used plan_lazy(), although that's suppressed if
+ * we never ran a test (due to an early bail, for example), and running any
+ * registered cleanup functions.
+ */
+static void
+finish(void)
+{
+ int success, primary;
+ struct cleanup_func *current;
+ unsigned long highest = testnum - 1;
+ struct diag_file *file, *tmp;
+
+ /* Check for pending diag_file output. */
+ check_diag_files();
+
+ /* Free the diag_files. */
+ file = diag_files;
+ while (file != NULL) {
+ tmp = file;
+ file = file->next;
+ fclose(tmp->file);
+ free(tmp->name);
+ free(tmp->buffer);
+ free(tmp);
+ }
+ diag_files = NULL;
+
+ /*
+ * Determine whether all tests were successful, which is needed before
+ * calling cleanup functions since we pass that fact to the functions.
+ */
+ if (_planned == 0 && _lazy)
+ _planned = highest;
+ success = (!_aborted && _planned == highest && _failed == 0);
+
+ /*
+ * If there are any registered cleanup functions, we run those first. We
+ * always run them, even if we didn't run a test. Don't do anything
+ * except free the diag_files and call cleanup functions if we aren't the
+ * primary process (the process in which plan or plan_lazy was called),
+ * and tell the cleanup functions that fact.
+ */
+ primary = (_process == 0 || getpid() == _process);
+ while (cleanup_funcs != NULL) {
+ if (cleanup_funcs->func_with_data) {
+ void *data = cleanup_funcs->data;
+
+ cleanup_funcs->func_with_data(success, primary, data);
+ } else {
+ cleanup_funcs->func(success, primary);
+ }
+ current = cleanup_funcs;
+ cleanup_funcs = cleanup_funcs->next;
+ free(current);
+ }
+ if (!primary)
+ return;
+
+ /* Don't do anything further if we never planned a test. */
+ if (_planned == 0)
+ return;
+
+ /* If we're aborting due to bail, don't print summaries. */
+ if (_aborted)
+ return;
+
+ /* Print out the lazy plan if needed. */
+ fflush(stderr);
+ if (_lazy)
+ printf("1..%lu\n", _planned);
+
+ /* Print out a summary of the results. */
+ if (_planned > highest)
+ diag("Looks like you planned %lu test%s but only ran %lu", _planned,
+ (_planned > 1 ? "s" : ""), highest);
+ else if (_planned < highest)
+ diag("Looks like you planned %lu test%s but ran %lu extra", _planned,
+ (_planned > 1 ? "s" : ""), highest - _planned);
+ else if (_failed > 0)
+ diag("Looks like you failed %lu test%s of %lu", _failed,
+ (_failed > 1 ? "s" : ""), _planned);
+ else if (_planned != 1)
+ diag("All %lu tests successful or skipped", _planned);
+ else
+ diag("%lu test successful or skipped", _planned);
+}
+
+
+/*
+ * Initialize things. Turns on line buffering on stdout and then prints out
+ * the number of tests in the test suite. We intentionally don't check for
+ * pending diag_file output here, since it should really come after the plan.
+ */
+void
+plan(unsigned long count)
+{
+ if (setvbuf(stdout, NULL, _IOLBF, BUFSIZ) != 0)
+ sysdiag("cannot set stdout to line buffered");
+ fflush(stderr);
+ printf("1..%lu\n", count);
+ testnum = 1;
+ _planned = count;
+ _process = getpid();
+ if (atexit(finish) != 0) {
+ sysdiag("cannot register exit handler");
+ diag("cleanups will not be run");
+ }
+}
+
+
+/*
+ * Initialize things for lazy planning, where we'll automatically print out a
+ * plan at the end of the program. Turns on line buffering on stdout as well.
+ */
+void
+plan_lazy(void)
+{
+ if (setvbuf(stdout, NULL, _IOLBF, BUFSIZ) != 0)
+ sysdiag("cannot set stdout to line buffered");
+ testnum = 1;
+ _process = getpid();
+ _lazy = 1;
+ if (atexit(finish) != 0)
+ sysbail("cannot register exit handler to display plan");
+}
+
+
+/*
+ * Skip the entire test suite and exits. Should be called instead of plan(),
+ * not after it, since it prints out a special plan line. Ignore diag_file
+ * output here, since it's not clear if it's allowed before the plan.
+ */
+void
+skip_all(const char *format, ...)
+{
+ fflush(stderr);
+ printf("1..0 # skip");
+ PRINT_DESC(" ", format);
+ putchar('\n');
+ exit(0);
+}
+
+
+/*
+ * Takes a boolean success value and assumes the test passes if that value
+ * is true and fails if that value is false.
+ */
+int
+ok(int success, const char *format, ...)
+{
+ fflush(stderr);
+ check_diag_files();
+ printf("%sok %lu", success ? "" : "not ", testnum++);
+ if (!success)
+ _failed++;
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Same as ok(), but takes the format arguments as a va_list.
+ */
+int
+okv(int success, const char *format, va_list args)
+{
+ fflush(stderr);
+ check_diag_files();
+ printf("%sok %lu", success ? "" : "not ", testnum++);
+ if (!success)
+ _failed++;
+ if (format != NULL) {
+ printf(" - ");
+ vprintf(format, args);
+ }
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Skip a test.
+ */
+void
+skip(const char *reason, ...)
+{
+ fflush(stderr);
+ check_diag_files();
+ printf("ok %lu # skip", testnum++);
+ PRINT_DESC(" ", reason);
+ putchar('\n');
+}
+
+
+/*
+ * Report the same status on the next count tests.
+ */
+int
+ok_block(unsigned long count, int success, const char *format, ...)
+{
+ unsigned long i;
+
+ fflush(stderr);
+ check_diag_files();
+ for (i = 0; i < count; i++) {
+ printf("%sok %lu", success ? "" : "not ", testnum++);
+ if (!success)
+ _failed++;
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ }
+ return success;
+}
+
+
+/*
+ * Skip the next count tests.
+ */
+void
+skip_block(unsigned long count, const char *reason, ...)
+{
+ unsigned long i;
+
+ fflush(stderr);
+ check_diag_files();
+ for (i = 0; i < count; i++) {
+ printf("ok %lu # skip", testnum++);
+ PRINT_DESC(" ", reason);
+ putchar('\n');
+ }
+}
+
+
+/*
+ * Takes two boolean values and requires the truth value of both match.
+ */
+int
+is_bool(int left, int right, const char *format, ...)
+{
+ int success;
+
+ fflush(stderr);
+ check_diag_files();
+ success = (!!left == !!right);
+ if (success)
+ printf("ok %lu", testnum++);
+ else {
+ diag(" left: %s", !!left ? "true" : "false");
+ diag("right: %s", !!right ? "true" : "false");
+ printf("not ok %lu", testnum++);
+ _failed++;
+ }
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Takes two integer values and requires they match.
+ */
+int
+is_int(long left, long right, const char *format, ...)
+{
+ int success;
+
+ fflush(stderr);
+ check_diag_files();
+ success = (left == right);
+ if (success)
+ printf("ok %lu", testnum++);
+ else {
+ diag(" left: %ld", left);
+ diag("right: %ld", right);
+ printf("not ok %lu", testnum++);
+ _failed++;
+ }
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Takes two strings and requires they match (using strcmp). NULL arguments
+ * are permitted and handled correctly.
+ */
+int
+is_string(const char *left, const char *right, const char *format, ...)
+{
+ int success;
+
+ fflush(stderr);
+ check_diag_files();
+
+ /* Compare the strings, being careful of NULL. */
+ if (left == NULL)
+ success = (right == NULL);
+ else if (right == NULL)
+ success = 0;
+ else
+ success = (strcmp(left, right) == 0);
+
+ /* Report the results. */
+ if (success)
+ printf("ok %lu", testnum++);
+ else {
+ diag(" left: %s", left == NULL ? "(null)" : left);
+ diag("right: %s", right == NULL ? "(null)" : right);
+ printf("not ok %lu", testnum++);
+ _failed++;
+ }
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Takes two unsigned longs and requires they match. On failure, reports them
+ * in hex.
+ */
+int
+is_hex(unsigned long left, unsigned long right, const char *format, ...)
+{
+ int success;
+
+ fflush(stderr);
+ check_diag_files();
+ success = (left == right);
+ if (success)
+ printf("ok %lu", testnum++);
+ else {
+ diag(" left: %lx", (unsigned long) left);
+ diag("right: %lx", (unsigned long) right);
+ printf("not ok %lu", testnum++);
+ _failed++;
+ }
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Takes pointers to a regions of memory and requires that len bytes from each
+ * match. Otherwise reports any bytes which didn't match.
+ */
+int
+is_blob(const void *left, const void *right, size_t len, const char *format,
+ ...)
+{
+ int success;
+ size_t i;
+
+ fflush(stderr);
+ check_diag_files();
+ success = (memcmp(left, right, len) == 0);
+ if (success)
+ printf("ok %lu", testnum++);
+ else {
+ const unsigned char *left_c = (const unsigned char *) left;
+ const unsigned char *right_c = (const unsigned char *) right;
+
+ for (i = 0; i < len; i++) {
+ if (left_c[i] != right_c[i])
+ diag("offset %lu: left %02x, right %02x", (unsigned long) i,
+ left_c[i], right_c[i]);
+ }
+ printf("not ok %lu", testnum++);
+ _failed++;
+ }
+ PRINT_DESC(" - ", format);
+ putchar('\n');
+ return success;
+}
+
+
+/*
+ * Bail out with an error.
+ */
+void
+bail(const char *format, ...)
+{
+ va_list args;
+
+ _aborted = 1;
+ fflush(stderr);
+ check_diag_files();
+ fflush(stdout);
+ printf("Bail out! ");
+ va_start(args, format);
+ vprintf(format, args);
+ va_end(args);
+ printf("\n");
+ exit(255);
+}
+
+
+/*
+ * Bail out with an error, appending strerror(errno).
+ */
+void
+sysbail(const char *format, ...)
+{
+ va_list args;
+ int oerrno = errno;
+
+ _aborted = 1;
+ fflush(stderr);
+ check_diag_files();
+ fflush(stdout);
+ printf("Bail out! ");
+ va_start(args, format);
+ vprintf(format, args);
+ va_end(args);
+ printf(": %s\n", strerror(oerrno));
+ exit(255);
+}
+
+
+/*
+ * Report a diagnostic to stderr. Always returns 1 to allow embedding in
+ * compound statements.
+ */
+int
+diag(const char *format, ...)
+{
+ va_list args;
+
+ fflush(stderr);
+ check_diag_files();
+ fflush(stdout);
+ printf("# ");
+ va_start(args, format);
+ vprintf(format, args);
+ va_end(args);
+ printf("\n");
+ return 1;
+}
+
+
+/*
+ * Report a diagnostic to stderr, appending strerror(errno). Always returns 1
+ * to allow embedding in compound statements.
+ */
+int
+sysdiag(const char *format, ...)
+{
+ va_list args;
+ int oerrno = errno;
+
+ fflush(stderr);
+ check_diag_files();
+ fflush(stdout);
+ printf("# ");
+ va_start(args, format);
+ vprintf(format, args);
+ va_end(args);
+ printf(": %s\n", strerror(oerrno));
+ return 1;
+}
+
+
+/*
+ * Register a new file for diag_file processing.
+ */
+void
+diag_file_add(const char *name)
+{
+ struct diag_file *file, *prev;
+
+ file = bcalloc_type(1, struct diag_file);
+ file->name = bstrdup(name);
+ file->file = fopen(file->name, "r");
+ if (file->file == NULL)
+ sysbail("cannot open %s", name);
+ file->buffer = bcalloc_type(BUFSIZ, char);
+ file->bufsize = BUFSIZ;
+ if (diag_files == NULL)
+ diag_files = file;
+ else {
+ for (prev = diag_files; prev->next != NULL; prev = prev->next)
+ ;
+ prev->next = file;
+ }
+}
+
+
+/*
+ * Remove a file from diag_file processing. If the file is not found, do
+ * nothing, since there are some situations where it can be removed twice
+ * (such as if it's removed from a cleanup function, since cleanup functions
+ * are called after freeing all the diag_files).
+ */
+void
+diag_file_remove(const char *name)
+{
+ struct diag_file *file;
+ struct diag_file **prev = &diag_files;
+
+ for (file = diag_files; file != NULL; file = file->next) {
+ if (strcmp(file->name, name) == 0) {
+ *prev = file->next;
+ fclose(file->file);
+ free(file->name);
+ free(file->buffer);
+ free(file);
+ return;
+ }
+ prev = &file->next;
+ }
+}
+
+
+/*
+ * Allocate cleared memory, reporting a fatal error with bail on failure.
+ */
+void *
+bcalloc(size_t n, size_t size)
+{
+ void *p;
+
+ p = calloc(n, size);
+ if (p == NULL)
+ sysbail("failed to calloc %lu", (unsigned long) (n * size));
+ return p;
+}
+
+
+/*
+ * Allocate memory, reporting a fatal error with bail on failure.
+ */
+void *
+bmalloc(size_t size)
+{
+ void *p;
+
+ p = malloc(size);
+ if (p == NULL)
+ sysbail("failed to malloc %lu", (unsigned long) size);
+ return p;
+}
+
+
+/*
+ * Reallocate memory, reporting a fatal error with bail on failure.
+ */
+void *
+brealloc(void *p, size_t size)
+{
+ p = realloc(p, size);
+ if (p == NULL)
+ sysbail("failed to realloc %lu bytes", (unsigned long) size);
+ return p;
+}
+
+
+/*
+ * The same as brealloc, but determine the size by multiplying an element
+ * count by a size, similar to calloc. The multiplication is checked for
+ * integer overflow.
+ *
+ * We should technically use SIZE_MAX here for the overflow check, but
+ * SIZE_MAX is C99 and we're only assuming C89 + SUSv3, which does not
+ * guarantee that it exists. They do guarantee that UINT_MAX exists, and we
+ * can assume that UINT_MAX <= SIZE_MAX.
+ *
+ * (In theory, C89 and C99 permit size_t to be smaller than unsigned int, but
+ * I disbelieve in the existence of such systems and they will have to cope
+ * without overflow checks.)
+ */
+void *
+breallocarray(void *p, size_t n, size_t size)
+{
+ if (n > 0 && UINT_MAX / n <= size)
+ bail("reallocarray too large");
+ if (n == 0)
+ n = 1;
+ p = realloc(p, n * size);
+ if (p == NULL)
+ sysbail("failed to realloc %lu bytes", (unsigned long) (n * size));
+ return p;
+}
+
+
+/*
+ * Copy a string, reporting a fatal error with bail on failure.
+ */
+char *
+bstrdup(const char *s)
+{
+ char *p;
+ size_t len;
+
+ len = strlen(s) + 1;
+ p = (char *) malloc(len);
+ if (p == NULL)
+ sysbail("failed to strdup %lu bytes", (unsigned long) len);
+ memcpy(p, s, len);
+ return p;
+}
+
+
+/*
+ * Copy up to n characters of a string, reporting a fatal error with bail on
+ * failure. Don't use the system strndup function, since it may not exist and
+ * the TAP library doesn't assume any portability support.
+ */
+char *
+bstrndup(const char *s, size_t n)
+{
+ const char *p;
+ char *copy;
+ size_t length;
+
+ /* Don't assume that the source string is nul-terminated. */
+ for (p = s; (size_t) (p - s) < n && *p != '\0'; p++)
+ ;
+ length = (size_t) (p - s);
+ copy = (char *) malloc(length + 1);
+ if (copy == NULL)
+ sysbail("failed to strndup %lu bytes", (unsigned long) length);
+ memcpy(copy, s, length);
+ copy[length] = '\0';
+ return copy;
+}
+
+
+/*
+ * Locate a test file. Given the partial path to a file, look under
+ * C_TAP_BUILD and then C_TAP_SOURCE for the file and return the full path to
+ * the file. Returns NULL if the file doesn't exist. A non-NULL return
+ * should be freed with test_file_path_free().
+ */
+char *
+test_file_path(const char *file)
+{
+ char *base;
+ char *path = NULL;
+ const char *envs[] = {"C_TAP_BUILD", "C_TAP_SOURCE", NULL};
+ int i;
+
+ for (i = 0; envs[i] != NULL; i++) {
+ base = getenv(envs[i]);
+ if (base == NULL)
+ continue;
+ path = concat(base, "/", file, (const char *) 0);
+ if (access(path, R_OK) == 0)
+ break;
+ free(path);
+ path = NULL;
+ }
+ return path;
+}
+
+
+/*
+ * Free a path returned from test_file_path(). This function exists primarily
+ * for Windows, where memory must be freed from the same library domain that
+ * it was allocated from.
+ */
+void
+test_file_path_free(char *path)
+{
+ free(path);
+}
+
+
+/*
+ * Create a temporary directory, tmp, under C_TAP_BUILD if set and the current
+ * directory if it does not. Returns the path to the temporary directory in
+ * newly allocated memory, and calls bail on any failure. The return value
+ * should be freed with test_tmpdir_free.
+ *
+ * This function uses sprintf because it attempts to be independent of all
+ * other portability layers. The use immediately after a memory allocation
+ * should be safe without using snprintf or strlcpy/strlcat.
+ */
+char *
+test_tmpdir(void)
+{
+ const char *build;
+ char *path = NULL;
+
+ build = getenv("C_TAP_BUILD");
+ if (build == NULL)
+ build = ".";
+ path = concat(build, "/tmp", (const char *) 0);
+ if (access(path, X_OK) < 0)
+ if (mkdir(path, 0777) < 0)
+ sysbail("error creating temporary directory %s", path);
+ return path;
+}
+
+
+/*
+ * Free a path returned from test_tmpdir() and attempt to remove the
+ * directory. If we can't delete the directory, don't worry; something else
+ * that hasn't yet cleaned up may still be using it.
+ */
+void
+test_tmpdir_free(char *path)
+{
+ if (path != NULL)
+ rmdir(path);
+ free(path);
+}
+
+static void
+register_cleanup(test_cleanup_func func,
+ test_cleanup_func_with_data func_with_data, void *data)
+{
+ struct cleanup_func *cleanup, **last;
+
+ cleanup = bcalloc_type(1, struct cleanup_func);
+ cleanup->func = func;
+ cleanup->func_with_data = func_with_data;
+ cleanup->data = data;
+ cleanup->next = NULL;
+ last = &cleanup_funcs;
+ while (*last != NULL)
+ last = &(*last)->next;
+ *last = cleanup;
+}
+
+/*
+ * Register a cleanup function that is called when testing ends. All such
+ * registered functions will be run by finish.
+ */
+void
+test_cleanup_register(test_cleanup_func func)
+{
+ register_cleanup(func, NULL, NULL);
+}
+
+/*
+ * Same as above, but also allows an opaque pointer to be passed to the cleanup
+ * function.
+ */
+void
+test_cleanup_register_with_data(test_cleanup_func_with_data func, void *data)
+{
+ register_cleanup(NULL, func, data);
+}
diff --git a/t/tap/basic.h b/t/tap/basic.h
new file mode 100644
index 0000000000..afea8cb210
--- /dev/null
+++ b/t/tap/basic.h
@@ -0,0 +1,198 @@
+/*
+ * Basic utility routines for the TAP protocol.
+ *
+ * This file is part of C TAP Harness. The current version plus supporting
+ * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
+ *
+ * Written by Russ Allbery <eagle@eyrie.org>
+ * Copyright 2009-2019, 2022 Russ Allbery <eagle@eyrie.org>
+ * Copyright 2001-2002, 2004-2008, 2011-2012, 2014
+ * The Board of Trustees of the Leland Stanford Junior University
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * SPDX-License-Identifier: MIT
+ */
+
+#ifndef TAP_BASIC_H
+#define TAP_BASIC_H 1
+
+#include <stdarg.h> /* va_list */
+#include <stddef.h> /* size_t */
+#include <stdlib.h> /* free */
+#include <tap/macros.h>
+
+/*
+ * Used for iterating through arrays. ARRAY_SIZE returns the number of
+ * elements in the array (useful for a < upper bound in a for loop) and
+ * ARRAY_END returns a pointer to the element past the end (ISO C99 makes it
+ * legal to refer to such a pointer as long as it's never dereferenced).
+ */
+// #define ARRAY_SIZE(array) (sizeof(array) / sizeof((array)[0]))
+// #define ARRAY_END(array) (&(array)[ARRAY_SIZE(array)])
+
+BEGIN_DECLS
+
+/*
+ * The test count. Always contains the number that will be used for the next
+ * test status.
+ */
+extern unsigned long testnum;
+
+/* Print out the number of tests and set standard output to line buffered. */
+void plan(unsigned long count);
+
+/*
+ * Prepare for lazy planning, in which the plan will be printed automatically
+ * at the end of the test program.
+ */
+void plan_lazy(void);
+
+/* Skip the entire test suite. Call instead of plan. */
+void skip_all(const char *format, ...)
+ __attribute__((__noreturn__, __format__(printf, 1, 2)));
+
+/*
+ * Basic reporting functions. The okv() function is the same as ok() but
+ * takes the test description as a va_list to make it easier to reuse the
+ * reporting infrastructure when writing new tests. ok() and okv() return the
+ * value of the success argument.
+ */
+int ok(int success, const char *format, ...)
+ __attribute__((__format__(printf, 2, 3)));
+int okv(int success, const char *format, va_list args)
+ __attribute__((__format__(printf, 2, 0)));
+void skip(const char *reason, ...) __attribute__((__format__(printf, 1, 2)));
+
+/*
+ * Report the same status on, or skip, the next count tests. ok_block()
+ * returns the value of the success argument.
+ */
+int ok_block(unsigned long count, int success, const char *format, ...)
+ __attribute__((__format__(printf, 3, 4)));
+void skip_block(unsigned long count, const char *reason, ...)
+ __attribute__((__format__(printf, 2, 3)));
+
+/*
+ * Compare two values. Returns true if the test passes and false if it fails.
+ * is_bool takes an int since the bool type isn't fully portable yet, but
+ * interprets both arguments for their truth value, not for their numeric
+ * value.
+ */
+int is_bool(int, int, const char *format, ...)
+ __attribute__((__format__(printf, 3, 4)));
+int is_int(long, long, const char *format, ...)
+ __attribute__((__format__(printf, 3, 4)));
+int is_string(const char *, const char *, const char *format, ...)
+ __attribute__((__format__(printf, 3, 4)));
+int is_hex(unsigned long, unsigned long, const char *format, ...)
+ __attribute__((__format__(printf, 3, 4)));
+int is_blob(const void *, const void *, size_t, const char *format, ...)
+ __attribute__((__format__(printf, 4, 5)));
+
+/* Bail out with an error. sysbail appends strerror(errno). */
+void bail(const char *format, ...)
+ __attribute__((__noreturn__, __nonnull__, __format__(printf, 1, 2)));
+void sysbail(const char *format, ...)
+ __attribute__((__noreturn__, __nonnull__, __format__(printf, 1, 2)));
+
+/* Report a diagnostic to stderr prefixed with #. */
+int diag(const char *format, ...)
+ __attribute__((__nonnull__, __format__(printf, 1, 2)));
+int sysdiag(const char *format, ...)
+ __attribute__((__nonnull__, __format__(printf, 1, 2)));
+
+/*
+ * Register or unregister a file that contains supplementary diagnostics.
+ * Before any other output, all registered files will be read, line by line,
+ * and each line will be reported as a diagnostic as if it were passed to
+ * diag(). Nul characters are not supported in these files and will result in
+ * truncated output.
+ */
+void diag_file_add(const char *file) __attribute__((__nonnull__));
+void diag_file_remove(const char *file) __attribute__((__nonnull__));
+
+/* Allocate memory, reporting a fatal error with bail on failure. */
+void *bcalloc(size_t, size_t)
+ __attribute__((__alloc_size__(1, 2), __malloc__(free),
+ __warn_unused_result__));
+void *bmalloc(size_t) __attribute__((__alloc_size__(1), __malloc__(free),
+ __warn_unused_result__));
+void *breallocarray(void *, size_t, size_t)
+ __attribute__((__alloc_size__(2, 3), __malloc__(free),
+ __warn_unused_result__));
+void *brealloc(void *, size_t)
+ __attribute__((__alloc_size__(2), __malloc__(free),
+ __warn_unused_result__));
+char *bstrdup(const char *)
+ __attribute__((__malloc__(free), __nonnull__, __warn_unused_result__));
+char *bstrndup(const char *, size_t)
+ __attribute__((__malloc__(free), __nonnull__, __warn_unused_result__));
+
+/*
+ * Macros that cast the return value from b* memory functions, making them
+ * usable in C++ code and providing some additional type safety.
+ */
+#define bcalloc_type(n, type) ((type *) bcalloc((n), sizeof(type)))
+#define breallocarray_type(p, n, type) \
+ ((type *) breallocarray((p), (n), sizeof(type)))
+
+/*
+ * Find a test file under C_TAP_BUILD or C_TAP_SOURCE, returning the full
+ * path. The returned path should be freed with test_file_path_free().
+ */
+void test_file_path_free(char *path);
+char *test_file_path(const char *file)
+ __attribute__((__malloc__(test_file_path_free), __nonnull__,
+ __warn_unused_result__));
+
+/*
+ * Create a temporary directory relative to C_TAP_BUILD and return the path.
+ * The returned path should be freed with test_tmpdir_free().
+ */
+void test_tmpdir_free(char *path);
+char *test_tmpdir(void)
+ __attribute__((__malloc__(test_tmpdir_free), __warn_unused_result__));
+
+/*
+ * Register a cleanup function that is called when testing ends. All such
+ * registered functions will be run during atexit handling (and are therefore
+ * subject to all the same constraints and caveats as atexit functions).
+ *
+ * The function must return void and will be passed two arguments: an int that
+ * will be true if the test completed successfully and false otherwise, and an
+ * int that will be true if the cleanup function is run in the primary process
+ * (the one that called plan or plan_lazy) and false otherwise. If
+ * test_cleanup_register_with_data is used instead, a generic pointer can be
+ * provided and will be passed to the cleanup function as a third argument.
+ *
+ * test_cleanup_register_with_data is the better API and should have been the
+ * only API. test_cleanup_register was an API error preserved for backward
+ * cmpatibility.
+ */
+typedef void (*test_cleanup_func)(int, int);
+typedef void (*test_cleanup_func_with_data)(int, int, void *);
+
+void test_cleanup_register(test_cleanup_func) __attribute__((__nonnull__));
+void test_cleanup_register_with_data(test_cleanup_func_with_data, void *)
+ __attribute__((__nonnull__));
+
+END_DECLS
+
+#endif /* TAP_BASIC_H */
diff --git a/t/tap/macros.h b/t/tap/macros.h
new file mode 100644
index 0000000000..0eabcb5847
--- /dev/null
+++ b/t/tap/macros.h
@@ -0,0 +1,109 @@
+/*
+ * Helpful macros for TAP header files.
+ *
+ * This is not, strictly speaking, related to TAP, but any TAP add-on is
+ * probably going to need these macros, so define them in one place so that
+ * everyone can pull them in.
+ *
+ * This file is part of C TAP Harness. The current version plus supporting
+ * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
+ *
+ * Copyright 2008, 2012-2013, 2015, 2022 Russ Allbery <eagle@eyrie.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ *
+ * SPDX-License-Identifier: MIT
+ */
+
+#ifndef TAP_MACROS_H
+#define TAP_MACROS_H 1
+
+/*
+ * __attribute__ is available in gcc 2.5 and later, but only with gcc 2.7
+ * could you use the __format__ form of the attributes, which is what we use
+ * (to avoid confusion with other macros), and only with gcc 2.96 can you use
+ * the attribute __malloc__. 2.96 is very old, so don't bother trying to get
+ * the other attributes to work with GCC versions between 2.7 and 2.96.
+ */
+#ifndef __attribute__
+# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
+# define __attribute__(spec) /* empty */
+# endif
+#endif
+
+/*
+ * We use __alloc_size__, but it was only available in fairly recent versions
+ * of GCC. Suppress warnings about the unknown attribute if GCC is too old.
+ * We know that we're GCC at this point, so we can use the GCC variadic macro
+ * extension, which will still work with versions of GCC too old to have C99
+ * variadic macro support.
+ */
+#if !defined(__attribute__) && !defined(__alloc_size__)
+# if defined(__GNUC__) && !defined(__clang__)
+# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3)
+# define __alloc_size__(spec, args...) /* empty */
+# endif
+# endif
+#endif
+
+/* Suppress __warn_unused_result__ if gcc is too old. */
+#if !defined(__attribute__) && !defined(__warn_unused_result__)
+# if __GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4)
+# define __warn_unused_result__ /* empty */
+# endif
+#endif
+
+/*
+ * Suppress the argument to __malloc__ in Clang (not supported in at least
+ * version 13) and GCC versions prior to 11.
+ */
+#if !defined(__attribute__) && !defined(__malloc__)
+# if defined(__clang__) || __GNUC__ < 11
+# define __malloc__(dalloc) __malloc__
+# endif
+#endif
+
+/*
+ * LLVM and Clang pretend to be GCC but don't support all of the __attribute__
+ * settings that GCC does. For them, suppress warnings about unknown
+ * attributes on declarations. This unfortunately will affect the entire
+ * compilation context, but there's no push and pop available.
+ */
+#if !defined(__attribute__) && (defined(__llvm__) || defined(__clang__))
+# pragma GCC diagnostic ignored "-Wattributes"
+#endif
+
+/* Used for unused parameters to silence gcc warnings. */
+// #define UNUSED __attribute__((__unused__))
+
+/*
+ * BEGIN_DECLS is used at the beginning of declarations so that C++
+ * compilers don't mangle their names. END_DECLS is used at the end.
+ */
+#undef BEGIN_DECLS
+#undef END_DECLS
+#ifdef __cplusplus
+# define BEGIN_DECLS extern "C" {
+# define END_DECLS }
+#else
+# define BEGIN_DECLS /* empty */
+# define END_DECLS /* empty */
+#endif
+
+#endif /* TAP_MACROS_H */
--
2.40.1.495.gc816e09b53d-goog
^ permalink raw reply related [flat|nested] 29+ messages in thread
* [RFC PATCH 2/2] unit test: add basic example
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
2023-04-27 17:50 ` [RFC PATCH 1/2] Add C TAP harness Calvin Wan
@ 2023-04-27 17:50 ` Calvin Wan
2023-04-27 18:47 ` Junio C Hamano
2023-04-27 18:39 ` [RFC PATCH 0/2] add an external testing library for unit tests Junio C Hamano
` (3 subsequent siblings)
5 siblings, 1 reply; 29+ messages in thread
From: Calvin Wan @ 2023-04-27 17:50 UTC (permalink / raw)
To: git; +Cc: Calvin Wan
Although this commit doesn't showcase unit tests running against
anything in Git, locally, I have a smaller set of Git files compiling
into a library with unit tests running against it using this C TAP
harness.
However, you can run `make ctap` to get an idea of how the output looks
and use it as a baseline to play around with the other features of the
library.
Signed-off-by: Calvin Wan <calvinwan@google.com>
---
Makefile | 11 +++++++++++
t/TESTS | 1 +
t/unit-test.c | 14 ++++++++++++++
3 files changed, 26 insertions(+)
create mode 100644 t/unit-test.c
diff --git a/Makefile b/Makefile
index 60ab1a8b4f..b67b81e312 100644
--- a/Makefile
+++ b/Makefile
@@ -3831,3 +3831,14 @@ $(FUZZ_PROGRAMS): all
$(XDIFF_OBJS) $(EXTLIBS) git.o $@.o $(LIB_FUZZING_ENGINE) -o $@
fuzz-all: $(FUZZ_PROGRAMS)
+
+LIB_CTAP_SOURCES = t/tap/basic.c t/tap/basic.h t/tap/macros.h
+
+PHONY: ctap
+ctap:
+ cc -It -c $(LIB_CTAP_SOURCES)
+ ar -rc libtap.a basic.o
+ mkdir -p t/unit-test
+ cc -It -o t/unit-test/unit-test-t t/unit-test.c -L. -l:libtap.a
+ cc -o t/runtests t/runtests.c
+ cd t && ./runtests -l TESTS
diff --git a/t/TESTS b/t/TESTS
index e69de29bb2..4b1899a115 100644
--- a/t/TESTS
+++ b/t/TESTS
@@ -0,0 +1 @@
+unit-test/unit-test-t
diff --git a/t/unit-test.c b/t/unit-test.c
new file mode 100644
index 0000000000..6eceb50ee6
--- /dev/null
+++ b/t/unit-test.c
@@ -0,0 +1,14 @@
+#include <tap/basic.h>
+
+int unit_test() {
+ if (1 != 1)
+ return 0;
+ return 1;
+}
+
+int main(void)
+{
+ plan(1);
+ ok(unit_test(), "unit test runs successfully");
+ return 0;
+}
--
2.40.1.495.gc816e09b53d-goog
^ permalink raw reply related [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-27 17:50 ` [RFC PATCH 1/2] Add C TAP harness Calvin Wan
@ 2023-04-27 18:29 ` SZEDER Gábor
2023-04-27 18:38 ` Calvin Wan
2023-04-27 20:15 ` Phillip Wood
1 sibling, 1 reply; 29+ messages in thread
From: SZEDER Gábor @ 2023-04-27 18:29 UTC (permalink / raw)
To: Calvin Wan; +Cc: git
On Thu, Apr 27, 2023 at 05:50:06PM +0000, Calvin Wan wrote:
> Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
>
> There is also more complete documentation at
> https://www.eyrie.org/~eagle/software/c-tap-harness/
>
> Signed-off-by: Calvin Wan <calvinwan@google.com>
Why?
Please remember that the justification of a change belongs to the
commit message and that the contents of the cover letter will not be
included in the history.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-27 18:29 ` SZEDER Gábor
@ 2023-04-27 18:38 ` Calvin Wan
0 siblings, 0 replies; 29+ messages in thread
From: Calvin Wan @ 2023-04-27 18:38 UTC (permalink / raw)
To: SZEDER Gábor; +Cc: git
As an RFC, I didn't intend for this patch series to be taken as is.
The commits are intended as proof of concept to help readers visualize
what I am describing in the cover letter. If the general idea of
having a testing library for unit tests is acceptable, then I plan on
rerolling it as an actual series.
On Thu, Apr 27, 2023 at 11:29 AM SZEDER Gábor <szeder.dev@gmail.com> wrote:
>
> On Thu, Apr 27, 2023 at 05:50:06PM +0000, Calvin Wan wrote:
> > Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
> >
> > There is also more complete documentation at
> > https://www.eyrie.org/~eagle/software/c-tap-harness/
> >
> > Signed-off-by: Calvin Wan <calvinwan@google.com>
>
> Why?
>
> Please remember that the justification of a change belongs to the
> commit message and that the contents of the cover letter will not be
> included in the history.
>
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 0/2] add an external testing library for unit tests
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
2023-04-27 17:50 ` [RFC PATCH 1/2] Add C TAP harness Calvin Wan
2023-04-27 17:50 ` [RFC PATCH 2/2] unit test: add basic example Calvin Wan
@ 2023-04-27 18:39 ` Junio C Hamano
2023-04-27 18:46 ` Calvin Wan
2023-04-27 21:35 ` brian m. carlson
` (2 subsequent siblings)
5 siblings, 1 reply; 29+ messages in thread
From: Junio C Hamano @ 2023-04-27 18:39 UTC (permalink / raw)
To: Calvin Wan; +Cc: git
Calvin Wan <calvinwan@google.com> writes:
> ... Unit tests additionally provide stability to the
> codebase and can simplify debugging through isolation. Turning parts of
> Git into libraries[1] gives us the ability to run unit tests on the
> libraries and to write unit tests in C. Writing unit tests in pure C,
> rather than with our current shell/test-tool helper setup, simplifies
> test setup, simplifies passing data around (no shell-isms required), and
> reduces testing runtime by not spawning a separate process for every
> test invocation.
Good goal, except that unit tests are not panacea---our ultimate
obligation is to give a stable behaviour to the end users and
end-to-end testing is still needed. We would benefit from having
tests at both levels.
> Unit testing in C requires a separate testing harness that we ideally
> would like to be TAP-style and to come with a non-restrictive license.
>
> Fortunately, there already exists a C TAP harness library[2] with an MIT
> license (at least for the files included in this series).
Yup. Consistency with our existing test framework would make it easier
to adopt for all of us. Good goal.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 0/2] add an external testing library for unit tests
2023-04-27 18:39 ` [RFC PATCH 0/2] add an external testing library for unit tests Junio C Hamano
@ 2023-04-27 18:46 ` Calvin Wan
0 siblings, 0 replies; 29+ messages in thread
From: Calvin Wan @ 2023-04-27 18:46 UTC (permalink / raw)
To: Junio C Hamano; +Cc: git
On Thu, Apr 27, 2023 at 11:40 AM Junio C Hamano <gitster@pobox.com> wrote:
>
> Calvin Wan <calvinwan@google.com> writes:
>
> > ... Unit tests additionally provide stability to the
> > codebase and can simplify debugging through isolation. Turning parts of
> > Git into libraries[1] gives us the ability to run unit tests on the
> > libraries and to write unit tests in C. Writing unit tests in pure C,
> > rather than with our current shell/test-tool helper setup, simplifies
> > test setup, simplifies passing data around (no shell-isms required), and
> > reduces testing runtime by not spawning a separate process for every
> > test invocation.
>
> Good goal, except that unit tests are not panacea---our ultimate
> obligation is to give a stable behaviour to the end users and
> end-to-end testing is still needed. We would benefit from having
> tests at both levels.
Ah I seem to have misworded this section a bit. I didn't mean to say
that writing unit tests in C would replace our current shell/test-tool
setup. I meant to make the comparison between writing unit tests in C
vs writing unit tests with our current shell/test-tool setup. And I
agree that we would benefit from having
tests at both levels.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 2/2] unit test: add basic example
2023-04-27 17:50 ` [RFC PATCH 2/2] unit test: add basic example Calvin Wan
@ 2023-04-27 18:47 ` Junio C Hamano
2023-05-02 15:58 ` Felipe Contreras
0 siblings, 1 reply; 29+ messages in thread
From: Junio C Hamano @ 2023-04-27 18:47 UTC (permalink / raw)
To: Calvin Wan; +Cc: git
Calvin Wan <calvinwan@google.com> writes:
> Although this commit doesn't showcase unit tests running against
> anything in Git, locally, I have a smaller set of Git files compiling
> into a library with unit tests running against it using this C TAP
> harness.
>
> However, you can run `make ctap` to get an idea of how the output looks
> and use it as a baseline to play around with the other features of the
> library.
My aging eyes first read it as "make crap" ;-).
Should we ove most of it to some other Makefile that lives in t/tap
and keep it outside the primary Makefile? This is adding about half
a dozen lines to support a sample unit-test.c file to be tested, and
if we are going to have 200 of them would we add a thousand or more
lines?
As a technology demonstration, such a hello-world test may be fine,
but I would have liked to see a more realistic sample that links
with reasonably isolated part of Git codebase, perhaps testing say
functions from <quote.h>.
THanks.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-27 17:50 ` [RFC PATCH 1/2] Add C TAP harness Calvin Wan
2023-04-27 18:29 ` SZEDER Gábor
@ 2023-04-27 20:15 ` Phillip Wood
2023-04-28 16:31 ` Calvin Wan
` (2 more replies)
1 sibling, 3 replies; 29+ messages in thread
From: Phillip Wood @ 2023-04-27 20:15 UTC (permalink / raw)
To: Calvin Wan, git
Hi Calvin
On 27/04/2023 18:50, Calvin Wan wrote:
> Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
>
> There is also more complete documentation at
> https://www.eyrie.org/~eagle/software/c-tap-harness/
I'm afraid this reply is rather briefer than I'd like but I'm short of
time and about to go off-list for a couple of weeks. My ideal unit test
library would
- print the file and line number of failed assertions
- allow the test plan to be omitted by calling test_done() at the end
of the test file as we do in our main test suite.
- support the TODO directive
- allow named tests (this maybe more trouble that it is worth as I
think it inevitably leads to more boilerplate code calling the named
tests)
Unfortunately this library doesn't seem to offer any of those features.
It does support a lazy test plan but uses atexit() so will not detect if
the test program exits before all the tests have run. I think it would
be useful to add some unit tests to our test suite and maybe this
library could form the basis of that but I think printing the file and
line number of failed assertions is pretty essential.
Best Wishes
Phillip
> Signed-off-by: Calvin Wan <calvinwan@google.com>
> ---
> t/TESTS | 0
> t/runtests.c | 1789 ++++++++++++++++++++++++++++++++++++++++++++++++
> t/tap/basic.c | 1029 ++++++++++++++++++++++++++++
> t/tap/basic.h | 198 ++++++
> t/tap/macros.h | 109 +++
> 5 files changed, 3125 insertions(+)
> create mode 100644 t/TESTS
> create mode 100644 t/runtests.c
> create mode 100644 t/tap/basic.c
> create mode 100644 t/tap/basic.h
> create mode 100644 t/tap/macros.h
>
> diff --git a/t/TESTS b/t/TESTS
> new file mode 100644
> index 0000000000..e69de29bb2
> diff --git a/t/runtests.c b/t/runtests.c
> new file mode 100644
> index 0000000000..4a55a801a6
> --- /dev/null
> +++ b/t/runtests.c
> @@ -0,0 +1,1789 @@
> +/*
> + * Run a set of tests, reporting results.
> + *
> + * Test suite driver that runs a set of tests implementing a subset of the
> + * Test Anything Protocol (TAP) and reports the results.
> + *
> + * Any bug reports, bug fixes, and improvements are very much welcome and
> + * should be sent to the e-mail address below. This program is part of C TAP
> + * Harness <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> + *
> + * Copyright 2000-2001, 2004, 2006-2019, 2022 Russ Allbery <eagle@eyrie.org>
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> + * DEALINGS IN THE SOFTWARE.
> + *
> + * SPDX-License-Identifier: MIT
> + */
> +
> +/*
> + * Usage:
> + *
> + * runtests [-hv] [-b <build-dir>] [-s <source-dir>] -l <test-list>
> + * runtests [-hv] [-b <build-dir>] [-s <source-dir>] <test> [<test> ...]
> + * runtests -o [-h] [-b <build-dir>] [-s <source-dir>] <test>
> + *
> + * In the first case, expects a list of executables located in the given file,
> + * one line per executable, possibly followed by a space-separated list of
> + * options. For each one, runs it as part of a test suite, reporting results.
> + * In the second case, use the same infrastructure, but run only the tests
> + * listed on the command line.
> + *
> + * Test output should start with a line containing the number of tests
> + * (numbered from 1 to this number), optionally preceded by "1..", although
> + * that line may be given anywhere in the output. Each additional line should
> + * be in the following format:
> + *
> + * ok <number>
> + * not ok <number>
> + * ok <number> # skip
> + * not ok <number> # todo
> + *
> + * where <number> is the number of the test. An optional comment is permitted
> + * after the number if preceded by whitespace. ok indicates success, not ok
> + * indicates failure. "# skip" and "# todo" are a special cases of a comment,
> + * and must start with exactly that formatting. They indicate the test was
> + * skipped for some reason (maybe because it doesn't apply to this platform)
> + * or is testing something known to currently fail. The text following either
> + * "# skip" or "# todo" and whitespace is the reason.
> + *
> + * As a special case, the first line of the output may be in the form:
> + *
> + * 1..0 # skip some reason
> + *
> + * which indicates that this entire test case should be skipped and gives a
> + * reason.
> + *
> + * Any other lines are ignored, although for compliance with the TAP protocol
> + * all lines other than the ones in the above format should be sent to
> + * standard error rather than standard output and start with #.
> + *
> + * This is a subset of TAP as documented in Test::Harness::TAP or
> + * TAP::Parser::Grammar, which comes with Perl.
> + *
> + * If the -o option is given, instead run a single test and display all of its
> + * output. This is intended for use with failing tests so that the person
> + * running the test suite can get more details about what failed.
> + *
> + * If built with the C preprocessor symbols C_TAP_SOURCE and C_TAP_BUILD
> + * defined, C TAP Harness will export those values in the environment so that
> + * tests can find the source and build directory and will look for tests under
> + * both directories. These paths can also be set with the -b and -s
> + * command-line options, which will override anything set at build time.
> + *
> + * If the -v option is given, or the C_TAP_VERBOSE environment variable is set,
> + * display the full output of each test as it runs rather than showing a
> + * summary of the results of each test.
> + */
> +
> +/* Required for fdopen(), getopt(), and putenv(). */
> +#if defined(__STRICT_ANSI__) || defined(PEDANTIC)
> +# ifndef _XOPEN_SOURCE
> +# define _XOPEN_SOURCE 500
> +# endif
> +#endif
> +
> +#include <ctype.h>
> +#include <errno.h>
> +#include <fcntl.h>
> +#include <limits.h>
> +#include <stdarg.h>
> +#include <stddef.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <strings.h>
> +#include <sys/stat.h>
> +#include <sys/time.h>
> +#include <sys/types.h>
> +#include <sys/wait.h>
> +#include <time.h>
> +#include <unistd.h>
> +
> +/* sys/time.h must be included before sys/resource.h on some platforms. */
> +#include <sys/resource.h>
> +
> +/* AIX 6.1 (and possibly later) doesn't have WCOREDUMP. */
> +#ifndef WCOREDUMP
> +# define WCOREDUMP(status) ((unsigned) (status) &0x80)
> +#endif
> +
> +/*
> + * POSIX requires that these be defined in <unistd.h>, but they're not always
> + * available. If one of them has been defined, all the rest almost certainly
> + * have.
> + */
> +#ifndef STDIN_FILENO
> +# define STDIN_FILENO 0
> +# define STDOUT_FILENO 1
> +# define STDERR_FILENO 2
> +#endif
> +
> +/*
> + * Used for iterating through arrays. Returns the number of elements in the
> + * array (useful for a < upper bound in a for loop).
> + */
> +#define ARRAY_SIZE(array) (sizeof(array) / sizeof((array)[0]))
> +
> +/*
> + * The source and build versions of the tests directory. This is used to set
> + * the C_TAP_SOURCE and C_TAP_BUILD environment variables (and the SOURCE and
> + * BUILD environment variables set for backward compatibility) and find test
> + * programs, if set. Normally, this should be set as part of the build
> + * process to the test subdirectories of $(abs_top_srcdir) and
> + * $(abs_top_builddir) respectively.
> + */
> +#ifndef C_TAP_SOURCE
> +# define C_TAP_SOURCE NULL
> +#endif
> +#ifndef C_TAP_BUILD
> +# define C_TAP_BUILD NULL
> +#endif
> +
> +/* Test status codes. */
> +enum test_status {
> + TEST_FAIL,
> + TEST_PASS,
> + TEST_SKIP,
> + TEST_INVALID
> +};
> +
> +/* Really, just a boolean, but this is more self-documenting. */
> +enum test_verbose {
> + CONCISE = 0,
> + VERBOSE = 1
> +};
> +
> +/* Indicates the state of our plan. */
> +enum plan_status {
> + PLAN_INIT, /* Nothing seen yet. */
> + PLAN_FIRST, /* Plan seen before any tests. */
> + PLAN_PENDING, /* Test seen and no plan yet. */
> + PLAN_FINAL /* Plan seen after some tests. */
> +};
> +
> +/* Error exit statuses for test processes. */
> +#define CHILDERR_DUP 100 /* Couldn't redirect stderr or stdout. */
> +#define CHILDERR_EXEC 101 /* Couldn't exec child process. */
> +#define CHILDERR_STDIN 102 /* Couldn't open stdin file. */
> +#define CHILDERR_STDERR 103 /* Couldn't open stderr file. */
> +
> +/* Structure to hold data for a set of tests. */
> +struct testset {
> + char *file; /* The file name of the test. */
> + char **command; /* The argv vector to run the command. */
> + enum plan_status plan; /* The status of our plan. */
> + unsigned long count; /* Expected count of tests. */
> + unsigned long current; /* The last seen test number. */
> + unsigned int length; /* The length of the last status message. */
> + unsigned long passed; /* Count of passing tests. */
> + unsigned long failed; /* Count of failing lists. */
> + unsigned long skipped; /* Count of skipped tests (passed). */
> + unsigned long allocated; /* The size of the results table. */
> + enum test_status *results; /* Table of results by test number. */
> + unsigned int aborted; /* Whether the set was aborted. */
> + unsigned int reported; /* Whether the results were reported. */
> + int status; /* The exit status of the test. */
> + unsigned int all_skipped; /* Whether all tests were skipped. */
> + char *reason; /* Why all tests were skipped. */
> +};
> +
> +/* Structure to hold a linked list of test sets. */
> +struct testlist {
> + struct testset *ts;
> + struct testlist *next;
> +};
> +
> +/*
> + * Usage message. Should be used as a printf format with four arguments: the
> + * path to runtests, given three times, and the usage_description. This is
> + * split into variables to satisfy the pedantic ISO C90 limit on strings.
> + */
> +static const char usage_message[] = "\
> +Usage: %s [-hv] [-b <build-dir>] [-s <source-dir>] <test> ...\n\
> + %s [-hv] [-b <build-dir>] [-s <source-dir>] -l <test-list>\n\
> + %s -o [-h] [-b <build-dir>] [-s <source-dir>] <test>\n\
> +\n\
> +Options:\n\
> + -b <build-dir> Set the build directory to <build-dir>\n\
> +%s";
> +static const char usage_extra[] = "\
> + -l <list> Take the list of tests to run from <test-list>\n\
> + -o Run a single test rather than a list of tests\n\
> + -s <source-dir> Set the source directory to <source-dir>\n\
> + -v Show the full output of each test\n\
> +\n\
> +runtests normally runs each test listed on the command line. With the -l\n\
> +option, it instead runs every test listed in a file. With the -o option,\n\
> +it instead runs a single test and shows its complete output.\n";
> +
> +/*
> + * Header used for test output. %s is replaced by the file name of the list
> + * of tests.
> + */
> +static const char banner[] = "\n\
> +Running all tests listed in %s. If any tests fail, run the failing\n\
> +test program with runtests -o to see more details.\n\n";
> +
> +/* Header for reports of failed tests. */
> +static const char header[] = "\n\
> +Failed Set Fail/Total (%) Skip Stat Failing Tests\n\
> +-------------------------- -------------- ---- ---- ------------------------";
> +
> +/* Include the file name and line number in malloc failures. */
> +#define xcalloc(n, type) \
> + ((type *) x_calloc((n), sizeof(type), __FILE__, __LINE__))
> +#define xmalloc(size) ((char *) x_malloc((size), __FILE__, __LINE__))
> +#define xstrdup(p) x_strdup((p), __FILE__, __LINE__)
> +#define xstrndup(p, size) x_strndup((p), (size), __FILE__, __LINE__)
> +#define xreallocarray(p, n, type) \
> + ((type *) x_reallocarray((p), (n), sizeof(type), __FILE__, __LINE__))
> +
> +/*
> + * __attribute__ is available in gcc 2.5 and later, but only with gcc 2.7
> + * could you use the __format__ form of the attributes, which is what we use
> + * (to avoid confusion with other macros).
> + */
> +#ifndef __attribute__
> +# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 7)
> +# define __attribute__(spec) /* empty */
> +# endif
> +#endif
> +
> +/*
> + * We use __alloc_size__, but it was only available in fairly recent versions
> + * of GCC. Suppress warnings about the unknown attribute if GCC is too old.
> + * We know that we're GCC at this point, so we can use the GCC variadic macro
> + * extension, which will still work with versions of GCC too old to have C99
> + * variadic macro support.
> + */
> +#if !defined(__attribute__) && !defined(__alloc_size__)
> +# if defined(__GNUC__) && !defined(__clang__)
> +# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3)
> +# define __alloc_size__(spec, args...) /* empty */
> +# endif
> +# endif
> +#endif
> +
> +/*
> + * Suppress the argument to __malloc__ in Clang (not supported in at least
> + * version 13) and GCC versions prior to 11.
> + */
> +#if !defined(__attribute__) && !defined(__malloc__)
> +# if defined(__clang__) || __GNUC__ < 11
> +# define __malloc__(dalloc) __malloc__
> +# endif
> +#endif
> +
> +/*
> + * LLVM and Clang pretend to be GCC but don't support all of the __attribute__
> + * settings that GCC does. For them, suppress warnings about unknown
> + * attributes on declarations. This unfortunately will affect the entire
> + * compilation context, but there's no push and pop available.
> + */
> +#if !defined(__attribute__) && (defined(__llvm__) || defined(__clang__))
> +# pragma GCC diagnostic ignored "-Wattributes"
> +#endif
> +
> +/* Declare internal functions that benefit from compiler attributes. */
> +static void die(const char *, ...)
> + __attribute__((__nonnull__, __noreturn__, __format__(printf, 1, 2)));
> +static void sysdie(const char *, ...)
> + __attribute__((__nonnull__, __noreturn__, __format__(printf, 1, 2)));
> +static void *x_calloc(size_t, size_t, const char *, int)
> + __attribute__((__alloc_size__(1, 2), __malloc__(free), __nonnull__));
> +static void *x_malloc(size_t, const char *, int)
> + __attribute__((__alloc_size__(1), __malloc__(free), __nonnull__));
> +static void *x_reallocarray(void *, size_t, size_t, const char *, int)
> + __attribute__((__alloc_size__(2, 3), __malloc__(free), __nonnull__(4)));
> +static char *x_strdup(const char *, const char *, int)
> + __attribute__((__malloc__(free), __nonnull__));
> +static char *x_strndup(const char *, size_t, const char *, int)
> + __attribute__((__malloc__(free), __nonnull__));
> +
> +
> +/*
> + * Report a fatal error and exit.
> + */
> +static void
> +die(const char *format, ...)
> +{
> + va_list args;
> +
> + fflush(stdout);
> + fprintf(stderr, "runtests: ");
> + va_start(args, format);
> + vfprintf(stderr, format, args);
> + va_end(args);
> + fprintf(stderr, "\n");
> + exit(1);
> +}
> +
> +
> +/*
> + * Report a fatal error, including the results of strerror, and exit.
> + */
> +static void
> +sysdie(const char *format, ...)
> +{
> + int oerrno;
> + va_list args;
> +
> + oerrno = errno;
> + fflush(stdout);
> + fprintf(stderr, "runtests: ");
> + va_start(args, format);
> + vfprintf(stderr, format, args);
> + va_end(args);
> + fprintf(stderr, ": %s\n", strerror(oerrno));
> + exit(1);
> +}
> +
> +
> +/*
> + * Allocate zeroed memory, reporting a fatal error and exiting on failure.
> + */
> +static void *
> +x_calloc(size_t n, size_t size, const char *file, int line)
> +{
> + void *p;
> +
> + n = (n > 0) ? n : 1;
> + size = (size > 0) ? size : 1;
> + p = calloc(n, size);
> + if (p == NULL)
> + sysdie("failed to calloc %lu bytes at %s line %d",
> + (unsigned long) size, file, line);
> + return p;
> +}
> +
> +
> +/*
> + * Allocate memory, reporting a fatal error and exiting on failure.
> + */
> +static void *
> +x_malloc(size_t size, const char *file, int line)
> +{
> + void *p;
> +
> + p = malloc(size);
> + if (p == NULL)
> + sysdie("failed to malloc %lu bytes at %s line %d",
> + (unsigned long) size, file, line);
> + return p;
> +}
> +
> +
> +/*
> + * Reallocate memory, reporting a fatal error and exiting on failure.
> + *
> + * We should technically use SIZE_MAX here for the overflow check, but
> + * SIZE_MAX is C99 and we're only assuming C89 + SUSv3, which does not
> + * guarantee that it exists. They do guarantee that UINT_MAX exists, and we
> + * can assume that UINT_MAX <= SIZE_MAX. And we should not be allocating
> + * anything anywhere near that large.
> + *
> + * (In theory, C89 and C99 permit size_t to be smaller than unsigned int, but
> + * I disbelieve in the existence of such systems and they will have to cope
> + * without overflow checks.)
> + */
> +static void *
> +x_reallocarray(void *p, size_t n, size_t size, const char *file, int line)
> +{
> + n = (n > 0) ? n : 1;
> + size = (size > 0) ? size : 1;
> +
> + if (n > 0 && UINT_MAX / n <= size)
> + sysdie("realloc too large at %s line %d", file, line);
> + p = realloc(p, n * size);
> + if (p == NULL)
> + sysdie("failed to realloc %lu bytes at %s line %d",
> + (unsigned long) (n * size), file, line);
> + return p;
> +}
> +
> +
> +/*
> + * Copy a string, reporting a fatal error and exiting on failure.
> + */
> +static char *
> +x_strdup(const char *s, const char *file, int line)
> +{
> + char *p;
> + size_t len;
> +
> + len = strlen(s) + 1;
> + p = (char *) malloc(len);
> + if (p == NULL)
> + sysdie("failed to strdup %lu bytes at %s line %d", (unsigned long) len,
> + file, line);
> + memcpy(p, s, len);
> + return p;
> +}
> +
> +
> +/*
> + * Copy the first n characters of a string, reporting a fatal error and
> + * existing on failure.
> + *
> + * Avoid using the system strndup function since it may not exist (on Mac OS
> + * X, for example), and there's no need to introduce another portability
> + * requirement.
> + */
> +char *
> +x_strndup(const char *s, size_t size, const char *file, int line)
> +{
> + const char *p;
> + size_t len;
> + char *copy;
> +
> + /* Don't assume that the source string is nul-terminated. */
> + for (p = s; (size_t) (p - s) < size && *p != '\0'; p++)
> + ;
> + len = (size_t) (p - s);
> + copy = (char *) malloc(len + 1);
> + if (copy == NULL)
> + sysdie("failed to strndup %lu bytes at %s line %d",
> + (unsigned long) len, file, line);
> + memcpy(copy, s, len);
> + copy[len] = '\0';
> + return copy;
> +}
> +
> +
> +/*
> + * Form a new string by concatenating multiple strings. The arguments must be
> + * terminated by (const char *) 0.
> + *
> + * This function only exists because we can't assume asprintf. We can't
> + * simulate asprintf with snprintf because we're only assuming SUSv3, which
> + * does not require that snprintf with a NULL buffer return the required
> + * length. When those constraints are relaxed, this should be ripped out and
> + * replaced with asprintf or a more trivial replacement with snprintf.
> + */
> +static char *
> +concat(const char *first, ...)
> +{
> + va_list args;
> + char *result;
> + const char *string;
> + size_t offset;
> + size_t length = 0;
> +
> + /*
> + * Find the total memory required. Ensure we don't overflow length. We
> + * aren't guaranteed to have SIZE_MAX, so use UINT_MAX as an acceptable
> + * substitute (see the x_nrealloc comments).
> + */
> + va_start(args, first);
> + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> + if (length >= UINT_MAX - strlen(string)) {
> + errno = EINVAL;
> + sysdie("strings too long in concat");
> + }
> + length += strlen(string);
> + }
> + va_end(args);
> + length++;
> +
> + /* Create the string. */
> + result = xmalloc(length);
> + va_start(args, first);
> + offset = 0;
> + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> + memcpy(result + offset, string, strlen(string));
> + offset += strlen(string);
> + }
> + va_end(args);
> + result[offset] = '\0';
> + return result;
> +}
> +
> +
> +/*
> + * Given a struct timeval, return the number of seconds it represents as a
> + * double. Use difftime() to convert a time_t to a double.
> + */
> +static double
> +tv_seconds(const struct timeval *tv)
> +{
> + return difftime(tv->tv_sec, 0) + (double) tv->tv_usec * 1e-6;
> +}
> +
> +
> +/*
> + * Given two struct timevals, return the difference in seconds.
> + */
> +static double
> +tv_diff(const struct timeval *tv1, const struct timeval *tv0)
> +{
> + return tv_seconds(tv1) - tv_seconds(tv0);
> +}
> +
> +
> +/*
> + * Given two struct timevals, return the sum in seconds as a double.
> + */
> +static double
> +tv_sum(const struct timeval *tv1, const struct timeval *tv2)
> +{
> + return tv_seconds(tv1) + tv_seconds(tv2);
> +}
> +
> +
> +/*
> + * Given a pointer to a string, skip any leading whitespace and return a
> + * pointer to the first non-whitespace character.
> + */
> +static const char *
> +skip_whitespace(const char *p)
> +{
> + while (isspace((unsigned char) (*p)))
> + p++;
> + return p;
> +}
> +
> +
> +/*
> + * Given a pointer to a string, skip any non-whitespace characters and return
> + * a pointer to the first whitespace character, or to the end of the string.
> + */
> +static const char *
> +skip_non_whitespace(const char *p)
> +{
> + while (*p != '\0' && !isspace((unsigned char) (*p)))
> + p++;
> + return p;
> +}
> +
> +
> +/*
> + * Start a program, connecting its stdout to a pipe on our end and its stderr
> + * to /dev/null, and storing the file descriptor to read from in the two
> + * argument. Returns the PID of the new process. Errors are fatal.
> + */
> +static pid_t
> +test_start(char *const *command, int *fd)
> +{
> + int fds[2], infd, errfd;
> + pid_t child;
> +
> + /* Create a pipe used to capture the output from the test program. */
> + if (pipe(fds) == -1) {
> + puts("ABORTED");
> + fflush(stdout);
> + sysdie("can't create pipe");
> + }
> +
> + /* Fork a child process, massage the file descriptors, and exec. */
> + child = fork();
> + switch (child) {
> + case -1:
> + puts("ABORTED");
> + fflush(stdout);
> + sysdie("can't fork");
> +
> + /* In the child. Set up our standard output. */
> + case 0:
> + close(fds[0]);
> + close(STDOUT_FILENO);
> + if (dup2(fds[1], STDOUT_FILENO) < 0)
> + _exit(CHILDERR_DUP);
> + close(fds[1]);
> +
> + /* Point standard input at /dev/null. */
> + close(STDIN_FILENO);
> + infd = open("/dev/null", O_RDONLY);
> + if (infd < 0)
> + _exit(CHILDERR_STDIN);
> + if (infd != STDIN_FILENO) {
> + if (dup2(infd, STDIN_FILENO) < 0)
> + _exit(CHILDERR_DUP);
> + close(infd);
> + }
> +
> + /* Point standard error at /dev/null. */
> + close(STDERR_FILENO);
> + errfd = open("/dev/null", O_WRONLY);
> + if (errfd < 0)
> + _exit(CHILDERR_STDERR);
> + if (errfd != STDERR_FILENO) {
> + if (dup2(errfd, STDERR_FILENO) < 0)
> + _exit(CHILDERR_DUP);
> + close(errfd);
> + }
> +
> + /* Now, exec our process. */
> + if (execv(command[0], command) == -1)
> + _exit(CHILDERR_EXEC);
> + break;
> +
> + /* In parent. Close the extra file descriptor. */
> + default:
> + close(fds[1]);
> + break;
> + }
> + *fd = fds[0];
> + return child;
> +}
> +
> +
> +/*
> + * Back up over the output saying what test we were executing.
> + */
> +static void
> +test_backspace(struct testset *ts)
> +{
> + unsigned int i;
> +
> + if (!isatty(STDOUT_FILENO))
> + return;
> + for (i = 0; i < ts->length; i++)
> + putchar('\b');
> + for (i = 0; i < ts->length; i++)
> + putchar(' ');
> + for (i = 0; i < ts->length; i++)
> + putchar('\b');
> + ts->length = 0;
> +}
> +
> +
> +/*
> + * Allocate or resize the array of test results to be large enough to contain
> + * the test number in.
> + */
> +static void
> +resize_results(struct testset *ts, unsigned long n)
> +{
> + unsigned long i;
> + size_t s;
> +
> + /* If there's already enough space, return quickly. */
> + if (n <= ts->allocated)
> + return;
> +
> + /*
> + * If no space has been allocated, do the initial allocation. Otherwise,
> + * resize. Start with 32 test cases and then add 1024 with each resize to
> + * try to reduce the number of reallocations.
> + */
> + if (ts->allocated == 0) {
> + s = (n > 32) ? n : 32;
> + ts->results = xcalloc(s, enum test_status);
> + } else {
> + s = (n > ts->allocated + 1024) ? n : ts->allocated + 1024;
> + ts->results = xreallocarray(ts->results, s, enum test_status);
> + }
> +
> + /* Set the results for the newly-allocated test array. */
> + for (i = ts->allocated; i < s; i++)
> + ts->results[i] = TEST_INVALID;
> + ts->allocated = s;
> +}
> +
> +
> +/*
> + * Report an invalid test number and set the appropriate flags. Pulled into a
> + * separate function since we do this in several places.
> + */
> +static void
> +invalid_test_number(struct testset *ts, long n, enum test_verbose verbose)
> +{
> + if (!verbose)
> + test_backspace(ts);
> + printf("ABORTED (invalid test number %ld)\n", n);
> + ts->aborted = 1;
> + ts->reported = 1;
> +}
> +
> +
> +/*
> + * Read the plan line of test output, which should contain the range of test
> + * numbers. We may initialize the testset structure here if we haven't yet
> + * seen a test. Return true if initialization succeeded and the test should
> + * continue, false otherwise.
> + */
> +static int
> +test_plan(const char *line, struct testset *ts, enum test_verbose verbose)
> +{
> + long n;
> +
> + /*
> + * Accept a plan without the leading 1.. for compatibility with older
> + * versions of runtests. This will only be allowed if we've not yet seen
> + * a test result.
> + */
> + line = skip_whitespace(line);
> + if (strncmp(line, "1..", 3) == 0)
> + line += 3;
> +
> + /*
> + * Get the count and check it for validity.
> + *
> + * If we have something of the form "1..0 # skip foo", the whole file was
> + * skipped; record that. If we do skip the whole file, zero out all of
> + * our statistics, since they're no longer relevant.
> + *
> + * strtol is called with a second argument to advance the line pointer
> + * past the count to make it simpler to detect the # skip case.
> + */
> + n = strtol(line, (char **) &line, 10);
> + if (n == 0) {
> + line = skip_whitespace(line);
> + if (*line == '#') {
> + line = skip_whitespace(line + 1);
> + if (strncasecmp(line, "skip", 4) == 0) {
> + line = skip_whitespace(line + 4);
> + if (*line != '\0') {
> + ts->reason = xstrdup(line);
> + ts->reason[strlen(ts->reason) - 1] = '\0';
> + }
> + ts->all_skipped = 1;
> + ts->aborted = 1;
> + ts->count = 0;
> + ts->passed = 0;
> + ts->skipped = 0;
> + ts->failed = 0;
> + return 0;
> + }
> + }
> + }
> + if (n <= 0) {
> + puts("ABORTED (invalid test count)");
> + ts->aborted = 1;
> + ts->reported = 1;
> + return 0;
> + }
> +
> + /*
> + * If we are doing lazy planning, check the plan against the largest test
> + * number that we saw and fail now if we saw a check outside the plan
> + * range.
> + */
> + if (ts->plan == PLAN_PENDING && (unsigned long) n < ts->count) {
> + invalid_test_number(ts, (long) ts->count, verbose);
> + return 0;
> + }
> +
> + /*
> + * Otherwise, allocated or resize the results if needed and update count,
> + * and then record that we've seen a plan.
> + */
> + resize_results(ts, (unsigned long) n);
> + ts->count = (unsigned long) n;
> + if (ts->plan == PLAN_INIT)
> + ts->plan = PLAN_FIRST;
> + else if (ts->plan == PLAN_PENDING)
> + ts->plan = PLAN_FINAL;
> + return 1;
> +}
> +
> +
> +/*
> + * Given a single line of output from a test, parse it and return the success
> + * status of that test. Anything printed to stdout not matching the form
> + * /^(not )?ok \d+/ is ignored. Sets ts->current to the test number that just
> + * reported status.
> + */
> +static void
> +test_checkline(const char *line, struct testset *ts, enum test_verbose verbose)
> +{
> + enum test_status status = TEST_PASS;
> + const char *bail;
> + char *end;
> + long number;
> + unsigned long current;
> + int outlen;
> +
> + /* Before anything, check for a test abort. */
> + bail = strstr(line, "Bail out!");
> + if (bail != NULL) {
> + bail = skip_whitespace(bail + strlen("Bail out!"));
> + if (*bail != '\0') {
> + size_t length;
> +
> + length = strlen(bail);
> + if (bail[length - 1] == '\n')
> + length--;
> + if (!verbose)
> + test_backspace(ts);
> + printf("ABORTED (%.*s)\n", (int) length, bail);
> + ts->reported = 1;
> + }
> + ts->aborted = 1;
> + return;
> + }
> +
> + /*
> + * If the given line isn't newline-terminated, it was too big for an
> + * fgets(), which means ignore it.
> + */
> + if (line[strlen(line) - 1] != '\n')
> + return;
> +
> + /* If the line begins with a hash mark, ignore it. */
> + if (line[0] == '#')
> + return;
> +
> + /* If we haven't yet seen a plan, look for one. */
> + if (ts->plan == PLAN_INIT && isdigit((unsigned char) (*line))) {
> + if (!test_plan(line, ts, verbose))
> + return;
> + } else if (strncmp(line, "1..", 3) == 0) {
> + if (ts->plan == PLAN_PENDING) {
> + if (!test_plan(line, ts, verbose))
> + return;
> + } else {
> + if (!verbose)
> + test_backspace(ts);
> + puts("ABORTED (multiple plans)");
> + ts->aborted = 1;
> + ts->reported = 1;
> + return;
> + }
> + }
> +
> + /* Parse the line, ignoring something we can't parse. */
> + if (strncmp(line, "not ", 4) == 0) {
> + status = TEST_FAIL;
> + line += 4;
> + }
> + if (strncmp(line, "ok", 2) != 0)
> + return;
> + line = skip_whitespace(line + 2);
> + errno = 0;
> + number = strtol(line, &end, 10);
> + if (errno != 0 || end == line)
> + current = ts->current + 1;
> + else if (number <= 0) {
> + invalid_test_number(ts, number, verbose);
> + return;
> + } else
> + current = (unsigned long) number;
> + if (current > ts->count && ts->plan == PLAN_FIRST) {
> + invalid_test_number(ts, (long) current, verbose);
> + return;
> + }
> +
> + /* We have a valid test result. Tweak the results array if needed. */
> + if (ts->plan == PLAN_INIT || ts->plan == PLAN_PENDING) {
> + ts->plan = PLAN_PENDING;
> + resize_results(ts, current);
> + if (current > ts->count)
> + ts->count = current;
> + }
> +
> + /*
> + * Handle directives. We should probably do something more interesting
> + * with unexpected passes of todo tests.
> + */
> + while (isdigit((unsigned char) (*line)))
> + line++;
> + line = skip_whitespace(line);
> + if (*line == '#') {
> + line = skip_whitespace(line + 1);
> + if (strncasecmp(line, "skip", 4) == 0)
> + status = TEST_SKIP;
> + if (strncasecmp(line, "todo", 4) == 0)
> + status = (status == TEST_FAIL) ? TEST_SKIP : TEST_FAIL;
> + }
> +
> + /* Make sure that the test number is in range and not a duplicate. */
> + if (ts->results[current - 1] != TEST_INVALID) {
> + if (!verbose)
> + test_backspace(ts);
> + printf("ABORTED (duplicate test number %lu)\n", current);
> + ts->aborted = 1;
> + ts->reported = 1;
> + return;
> + }
> +
> + /* Good results. Increment our various counters. */
> + switch (status) {
> + case TEST_PASS:
> + ts->passed++;
> + break;
> + case TEST_FAIL:
> + ts->failed++;
> + break;
> + case TEST_SKIP:
> + ts->skipped++;
> + break;
> + case TEST_INVALID:
> + break;
> + }
> + ts->current = current;
> + ts->results[current - 1] = status;
> + if (!verbose && isatty(STDOUT_FILENO)) {
> + test_backspace(ts);
> + if (ts->plan == PLAN_PENDING)
> + outlen = printf("%lu/?", current);
> + else
> + outlen = printf("%lu/%lu", current, ts->count);
> + ts->length = (outlen >= 0) ? (unsigned int) outlen : 0;
> + fflush(stdout);
> + }
> +}
> +
> +
> +/*
> + * Print out a range of test numbers, returning the number of characters it
> + * took up. Takes the first number, the last number, the number of characters
> + * already printed on the line, and the limit of number of characters the line
> + * can hold. Add a comma and a space before the range if chars indicates that
> + * something has already been printed on the line, and print ... instead if
> + * chars plus the space needed would go over the limit (use a limit of 0 to
> + * disable this).
> + */
> +static unsigned int
> +test_print_range(unsigned long first, unsigned long last, unsigned long chars,
> + unsigned int limit)
> +{
> + unsigned int needed = 0;
> + unsigned long n;
> +
> + for (n = first; n > 0; n /= 10)
> + needed++;
> + if (last > first) {
> + for (n = last; n > 0; n /= 10)
> + needed++;
> + needed++;
> + }
> + if (chars > 0)
> + needed += 2;
> + if (limit > 0 && chars + needed > limit) {
> + needed = 0;
> + if (chars <= limit) {
> + if (chars > 0) {
> + printf(", ");
> + needed += 2;
> + }
> + printf("...");
> + needed += 3;
> + }
> + } else {
> + if (chars > 0)
> + printf(", ");
> + if (last > first)
> + printf("%lu-", first);
> + printf("%lu", last);
> + }
> + return needed;
> +}
> +
> +
> +/*
> + * Summarize a single test set. The second argument is 0 if the set exited
> + * cleanly, a positive integer representing the exit status if it exited
> + * with a non-zero status, and a negative integer representing the signal
> + * that terminated it if it was killed by a signal.
> + */
> +static void
> +test_summarize(struct testset *ts, int status)
> +{
> + unsigned long i;
> + unsigned long missing = 0;
> + unsigned long failed = 0;
> + unsigned long first = 0;
> + unsigned long last = 0;
> +
> + if (ts->aborted) {
> + fputs("ABORTED", stdout);
> + if (ts->count > 0)
> + printf(" (passed %lu/%lu)", ts->passed, ts->count - ts->skipped);
> + } else {
> + for (i = 0; i < ts->count; i++) {
> + if (ts->results[i] == TEST_INVALID) {
> + if (missing == 0)
> + fputs("MISSED ", stdout);
> + if (first && i == last)
> + last = i + 1;
> + else {
> + if (first)
> + test_print_range(first, last, missing - 1, 0);
> + missing++;
> + first = i + 1;
> + last = i + 1;
> + }
> + }
> + }
> + if (first)
> + test_print_range(first, last, missing - 1, 0);
> + first = 0;
> + last = 0;
> + for (i = 0; i < ts->count; i++) {
> + if (ts->results[i] == TEST_FAIL) {
> + if (missing && !failed)
> + fputs("; ", stdout);
> + if (failed == 0)
> + fputs("FAILED ", stdout);
> + if (first && i == last)
> + last = i + 1;
> + else {
> + if (first)
> + test_print_range(first, last, failed - 1, 0);
> + failed++;
> + first = i + 1;
> + last = i + 1;
> + }
> + }
> + }
> + if (first)
> + test_print_range(first, last, failed - 1, 0);
> + if (!missing && !failed) {
> + fputs(!status ? "ok" : "dubious", stdout);
> + if (ts->skipped > 0) {
> + if (ts->skipped == 1)
> + printf(" (skipped %lu test)", ts->skipped);
> + else
> + printf(" (skipped %lu tests)", ts->skipped);
> + }
> + }
> + }
> + if (status > 0)
> + printf(" (exit status %d)", status);
> + else if (status < 0)
> + printf(" (killed by signal %d%s)", -status,
> + WCOREDUMP(ts->status) ? ", core dumped" : "");
> + putchar('\n');
> +}
> +
> +
> +/*
> + * Given a test set, analyze the results, classify the exit status, handle a
> + * few special error messages, and then pass it along to test_summarize() for
> + * the regular output. Returns true if the test set ran successfully and all
> + * tests passed or were skipped, false otherwise.
> + */
> +static int
> +test_analyze(struct testset *ts)
> +{
> + if (ts->reported)
> + return 0;
> + if (ts->all_skipped) {
> + if (ts->reason == NULL)
> + puts("skipped");
> + else
> + printf("skipped (%s)\n", ts->reason);
> + return 1;
> + } else if (WIFEXITED(ts->status) && WEXITSTATUS(ts->status) != 0) {
> + switch (WEXITSTATUS(ts->status)) {
> + case CHILDERR_DUP:
> + if (!ts->reported)
> + puts("ABORTED (can't dup file descriptors)");
> + break;
> + case CHILDERR_EXEC:
> + if (!ts->reported)
> + puts("ABORTED (execution failed -- not found?)");
> + break;
> + case CHILDERR_STDIN:
> + case CHILDERR_STDERR:
> + if (!ts->reported)
> + puts("ABORTED (can't open /dev/null)");
> + break;
> + default:
> + test_summarize(ts, WEXITSTATUS(ts->status));
> + break;
> + }
> + return 0;
> + } else if (WIFSIGNALED(ts->status)) {
> + test_summarize(ts, -WTERMSIG(ts->status));
> + return 0;
> + } else if (ts->plan != PLAN_FIRST && ts->plan != PLAN_FINAL) {
> + puts("ABORTED (no valid test plan)");
> + ts->aborted = 1;
> + return 0;
> + } else {
> + test_summarize(ts, 0);
> + return (ts->failed == 0);
> + }
> +}
> +
> +
> +/*
> + * Runs a single test set, accumulating and then reporting the results.
> + * Returns true if the test set was successfully run and all tests passed,
> + * false otherwise.
> + */
> +static int
> +test_run(struct testset *ts, enum test_verbose verbose)
> +{
> + pid_t testpid, child;
> + int outfd, status;
> + unsigned long i;
> + FILE *output;
> + char buffer[BUFSIZ];
> +
> + /* Run the test program. */
> + testpid = test_start(ts->command, &outfd);
> + output = fdopen(outfd, "r");
> + if (!output) {
> + puts("ABORTED");
> + fflush(stdout);
> + sysdie("fdopen failed");
> + }
> +
> + /*
> + * Pass each line of output to test_checkline(), and print the line if
> + * verbosity is requested.
> + */
> + while (!ts->aborted && fgets(buffer, sizeof(buffer), output)) {
> + if (verbose)
> + printf("%s", buffer);
> + test_checkline(buffer, ts, verbose);
> + }
> + if (ferror(output) || ts->plan == PLAN_INIT)
> + ts->aborted = 1;
> + if (!verbose)
> + test_backspace(ts);
> +
> + /*
> + * Consume the rest of the test output, close the output descriptor,
> + * retrieve the exit status, and pass that information to test_analyze()
> + * for eventual output.
> + */
> + while (fgets(buffer, sizeof(buffer), output))
> + if (verbose)
> + printf("%s", buffer);
> + fclose(output);
> + child = waitpid(testpid, &ts->status, 0);
> + if (child == (pid_t) -1) {
> + if (!ts->reported) {
> + puts("ABORTED");
> + fflush(stdout);
> + }
> + sysdie("waitpid for %u failed", (unsigned int) testpid);
> + }
> + if (ts->all_skipped)
> + ts->aborted = 0;
> + status = test_analyze(ts);
> +
> + /* Convert missing tests to failed tests. */
> + for (i = 0; i < ts->count; i++) {
> + if (ts->results[i] == TEST_INVALID) {
> + ts->failed++;
> + ts->results[i] = TEST_FAIL;
> + status = 0;
> + }
> + }
> + return status;
> +}
> +
> +
> +/* Summarize a list of test failures. */
> +static void
> +test_fail_summary(const struct testlist *fails)
> +{
> + struct testset *ts;
> + unsigned int chars;
> + unsigned long i, first, last, total;
> + double failed;
> +
> + puts(header);
> +
> + /* Failed Set Fail/Total (%) Skip Stat Failing (25)
> + -------------------------- -------------- ---- ---- -------------- */
> + for (; fails; fails = fails->next) {
> + ts = fails->ts;
> + total = ts->count - ts->skipped;
> + failed = (double) ts->failed;
> + printf("%-26.26s %4lu/%-4lu %3.0f%% %4lu ", ts->file, ts->failed,
> + total, total ? (failed * 100.0) / (double) total : 0,
> + ts->skipped);
> + if (WIFEXITED(ts->status))
> + printf("%4d ", WEXITSTATUS(ts->status));
> + else
> + printf(" -- ");
> + if (ts->aborted) {
> + puts("aborted");
> + continue;
> + }
> + chars = 0;
> + first = 0;
> + last = 0;
> + for (i = 0; i < ts->count; i++) {
> + if (ts->results[i] == TEST_FAIL) {
> + if (first != 0 && i == last)
> + last = i + 1;
> + else {
> + if (first != 0)
> + chars += test_print_range(first, last, chars, 19);
> + first = i + 1;
> + last = i + 1;
> + }
> + }
> + }
> + if (first != 0)
> + test_print_range(first, last, chars, 19);
> + putchar('\n');
> + }
> +}
> +
> +
> +/*
> + * Check whether a given file path is a valid test. Currently, this checks
> + * whether it is executable and is a regular file. Returns true or false.
> + */
> +static int
> +is_valid_test(const char *path)
> +{
> + struct stat st;
> +
> + if (access(path, X_OK) < 0)
> + return 0;
> + if (stat(path, &st) < 0)
> + return 0;
> + if (!S_ISREG(st.st_mode))
> + return 0;
> + return 1;
> +}
> +
> +
> +/*
> + * Given the name of a test, a pointer to the testset struct, and the source
> + * and build directories, find the test. We try first relative to the current
> + * directory, then in the build directory (if not NULL), then in the source
> + * directory. In each of those directories, we first try a "-t" extension and
> + * then a ".t" extension. When we find an executable program, we return the
> + * path to that program. If none of those paths are executable, just fill in
> + * the name of the test as is.
> + *
> + * The caller is responsible for freeing the path member of the testset
> + * struct.
> + */
> +static char *
> +find_test(const char *name, const char *source, const char *build)
> +{
> + char *path = NULL;
> + const char *bases[3], *suffix, *base;
> + unsigned int i, j;
> + const char *suffixes[3] = {"-t", ".t", ""};
> +
> + /* Possible base directories. */
> + bases[0] = ".";
> + bases[1] = build;
> + bases[2] = source;
> +
> + /* Try each suffix with each base. */
> + for (i = 0; i < ARRAY_SIZE(suffixes); i++) {
> + suffix = suffixes[i];
> + for (j = 0; j < ARRAY_SIZE(bases); j++) {
> + base = bases[j];
> + if (base == NULL)
> + continue;
> + path = concat(base, "/", name, suffix, (const char *) 0);
> + if (is_valid_test(path))
> + return path;
> + free(path);
> + path = NULL;
> + }
> + }
> + if (path == NULL)
> + path = xstrdup(name);
> + return path;
> +}
> +
> +
> +/*
> + * Parse a single line of a test list and store the test name and command to
> + * execute it in the given testset struct.
> + *
> + * Normally, each line is just the name of the test, which is located in the
> + * test directory and turned into a command to run. However, each line may
> + * have whitespace-separated options, which change the command that's run.
> + * Current supported options are:
> + *
> + * valgrind
> + * Run the test under valgrind if C_TAP_VALGRIND is set. The contents
> + * of that environment variable are taken as the valgrind command (with
> + * options) to run. The command is parsed with a simple split on
> + * whitespace and no quoting is supported.
> + *
> + * libtool
> + * If running under valgrind, use libtool to invoke valgrind. This avoids
> + * running valgrind on the wrapper shell script generated by libtool. If
> + * set, C_TAP_LIBTOOL must be set to the full path to the libtool program
> + * to use to run valgrind and thus the test. Ignored if the test isn't
> + * being run under valgrind.
> + */
> +static void
> +parse_test_list_line(const char *line, struct testset *ts, const char *source,
> + const char *build)
> +{
> + const char *p, *end, *option, *libtool;
> + const char *valgrind = NULL;
> + unsigned int use_libtool = 0;
> + unsigned int use_valgrind = 0;
> + size_t len, i;
> +
> + /* Determine the name of the test. */
> + p = skip_non_whitespace(line);
> + ts->file = xstrndup(line, p - line);
> +
> + /* Check if any test options are set. */
> + p = skip_whitespace(p);
> + while (*p != '\0') {
> + end = skip_non_whitespace(p);
> + if (strncmp(p, "libtool", end - p) == 0) {
> + use_libtool = 1;
> + } else if (strncmp(p, "valgrind", end - p) == 0) {
> + valgrind = getenv("C_TAP_VALGRIND");
> + use_valgrind = (valgrind != NULL);
> + } else {
> + option = xstrndup(p, end - p);
> + die("unknown test list option %s", option);
> + }
> + p = skip_whitespace(end);
> + }
> +
> + /* Construct the argv to run the test. First, find the length. */
> + len = 1;
> + if (use_valgrind && valgrind != NULL) {
> + p = skip_whitespace(valgrind);
> + while (*p != '\0') {
> + len++;
> + p = skip_whitespace(skip_non_whitespace(p));
> + }
> + if (use_libtool)
> + len += 2;
> + }
> +
> + /* Now, build the command. */
> + ts->command = xcalloc(len + 1, char *);
> + i = 0;
> + if (use_valgrind && valgrind != NULL) {
> + if (use_libtool) {
> + libtool = getenv("C_TAP_LIBTOOL");
> + if (libtool == NULL)
> + die("valgrind with libtool requested, but C_TAP_LIBTOOL is not"
> + " set");
> + ts->command[i++] = xstrdup(libtool);
> + ts->command[i++] = xstrdup("--mode=execute");
> + }
> + p = skip_whitespace(valgrind);
> + while (*p != '\0') {
> + end = skip_non_whitespace(p);
> + ts->command[i++] = xstrndup(p, end - p);
> + p = skip_whitespace(end);
> + }
> + }
> + if (i != len - 1)
> + die("internal error while constructing command line");
> + ts->command[i++] = find_test(ts->file, source, build);
> + ts->command[i] = NULL;
> +}
> +
> +
> +/*
> + * Read a list of tests from a file, returning the list of tests as a struct
> + * testlist, or NULL if there were no tests (such as a file containing only
> + * comments). Reports an error to standard error and exits if the list of
> + * tests cannot be read.
> + */
> +static struct testlist *
> +read_test_list(const char *filename, const char *source, const char *build)
> +{
> + FILE *file;
> + unsigned int line;
> + size_t length;
> + char buffer[BUFSIZ];
> + const char *start;
> + struct testlist *listhead, *current;
> +
> + /* Create the initial container list that will hold our results. */
> + listhead = xcalloc(1, struct testlist);
> + current = NULL;
> +
> + /*
> + * Open our file of tests to run and read it line by line, creating a new
> + * struct testlist and struct testset for each line.
> + */
> + file = fopen(filename, "r");
> + if (file == NULL)
> + sysdie("can't open %s", filename);
> + line = 0;
> + while (fgets(buffer, sizeof(buffer), file)) {
> + line++;
> + length = strlen(buffer) - 1;
> + if (buffer[length] != '\n') {
> + fprintf(stderr, "%s:%u: line too long\n", filename, line);
> + exit(1);
> + }
> + buffer[length] = '\0';
> +
> + /* Skip comments, leading spaces, and blank lines. */
> + start = skip_whitespace(buffer);
> + if (strlen(start) == 0)
> + continue;
> + if (start[0] == '#')
> + continue;
> +
> + /* Allocate the new testset structure. */
> + if (current == NULL)
> + current = listhead;
> + else {
> + current->next = xcalloc(1, struct testlist);
> + current = current->next;
> + }
> + current->ts = xcalloc(1, struct testset);
> + current->ts->plan = PLAN_INIT;
> +
> + /* Parse the line and store the results in the testset struct. */
> + parse_test_list_line(start, current->ts, source, build);
> + }
> + fclose(file);
> +
> + /* If there were no tests, current is still NULL. */
> + if (current == NULL) {
> + free(listhead);
> + return NULL;
> + }
> +
> + /* Return the results. */
> + return listhead;
> +}
> +
> +
> +/*
> + * Build a list of tests from command line arguments. Takes the argv and argc
> + * representing the command line arguments and returns a newly allocated test
> + * list, or NULL if there were no tests. The caller is responsible for
> + * freeing.
> + */
> +static struct testlist *
> +build_test_list(char *argv[], int argc, const char *source, const char *build)
> +{
> + int i;
> + struct testlist *listhead, *current;
> +
> + /* Create the initial container list that will hold our results. */
> + listhead = xcalloc(1, struct testlist);
> + current = NULL;
> +
> + /* Walk the list of arguments and create test sets for them. */
> + for (i = 0; i < argc; i++) {
> + if (current == NULL)
> + current = listhead;
> + else {
> + current->next = xcalloc(1, struct testlist);
> + current = current->next;
> + }
> + current->ts = xcalloc(1, struct testset);
> + current->ts->plan = PLAN_INIT;
> + current->ts->file = xstrdup(argv[i]);
> + current->ts->command = xcalloc(2, char *);
> + current->ts->command[0] = find_test(current->ts->file, source, build);
> + current->ts->command[1] = NULL;
> + }
> +
> + /* If there were no tests, current is still NULL. */
> + if (current == NULL) {
> + free(listhead);
> + return NULL;
> + }
> +
> + /* Return the results. */
> + return listhead;
> +}
> +
> +
> +/* Free a struct testset. */
> +static void
> +free_testset(struct testset *ts)
> +{
> + size_t i;
> +
> + free(ts->file);
> + for (i = 0; ts->command[i] != NULL; i++)
> + free(ts->command[i]);
> + free(ts->command);
> + free(ts->results);
> + free(ts->reason);
> + free(ts);
> +}
> +
> +
> +/*
> + * Run a batch of tests. Takes two additional parameters: the root of the
> + * source directory and the root of the build directory. Test programs will
> + * be first searched for in the current directory, then the build directory,
> + * then the source directory. Returns true iff all tests passed, and always
> + * frees the test list that's passed in.
> + */
> +static int
> +test_batch(struct testlist *tests, enum test_verbose verbose)
> +{
> + size_t length, i;
> + size_t longest = 0;
> + unsigned int count = 0;
> + struct testset *ts;
> + struct timeval start, end;
> + struct rusage stats;
> + struct testlist *failhead = NULL;
> + struct testlist *failtail = NULL;
> + struct testlist *current, *next;
> + int succeeded;
> + unsigned long total = 0;
> + unsigned long passed = 0;
> + unsigned long skipped = 0;
> + unsigned long failed = 0;
> + unsigned long aborted = 0;
> +
> + /* Walk the list of tests to find the longest name. */
> + for (current = tests; current != NULL; current = current->next) {
> + length = strlen(current->ts->file);
> + if (length > longest)
> + longest = length;
> + }
> +
> + /*
> + * Add two to longest and round up to the nearest tab stop. This is how
> + * wide the column for printing the current test name will be.
> + */
> + longest += 2;
> + if (longest % 8)
> + longest += 8 - (longest % 8);
> +
> + /* Start the wall clock timer. */
> + gettimeofday(&start, NULL);
> +
> + /* Now, plow through our tests again, running each one. */
> + for (current = tests; current != NULL; current = current->next) {
> + ts = current->ts;
> +
> + /* Print out the name of the test file. */
> + fputs(ts->file, stdout);
> + if (verbose)
> + fputs("\n\n", stdout);
> + else
> + for (i = strlen(ts->file); i < longest; i++)
> + putchar('.');
> + if (isatty(STDOUT_FILENO))
> + fflush(stdout);
> +
> + /* Run the test. */
> + succeeded = test_run(ts, verbose);
> + fflush(stdout);
> + if (verbose)
> + putchar('\n');
> +
> + /* Record cumulative statistics. */
> + aborted += ts->aborted;
> + total += ts->count + ts->all_skipped;
> + passed += ts->passed;
> + skipped += ts->skipped + ts->all_skipped;
> + failed += ts->failed;
> + count++;
> +
> + /* If the test fails, we shuffle it over to the fail list. */
> + if (!succeeded) {
> + if (failhead == NULL) {
> + failhead = xcalloc(1, struct testlist);
> + failtail = failhead;
> + } else {
> + failtail->next = xcalloc(1, struct testlist);
> + failtail = failtail->next;
> + }
> + failtail->ts = ts;
> + failtail->next = NULL;
> + }
> + }
> + total -= skipped;
> +
> + /* Stop the timer and get our child resource statistics. */
> + gettimeofday(&end, NULL);
> + getrusage(RUSAGE_CHILDREN, &stats);
> +
> + /* Summarize the failures and free the failure list. */
> + if (failhead != NULL) {
> + test_fail_summary(failhead);
> + while (failhead != NULL) {
> + next = failhead->next;
> + free(failhead);
> + failhead = next;
> + }
> + }
> +
> + /* Free the memory used by the test lists. */
> + while (tests != NULL) {
> + next = tests->next;
> + free_testset(tests->ts);
> + free(tests);
> + tests = next;
> + }
> +
> + /* Print out the final test summary. */
> + putchar('\n');
> + if (aborted != 0) {
> + if (aborted == 1)
> + printf("Aborted %lu test set", aborted);
> + else
> + printf("Aborted %lu test sets", aborted);
> + printf(", passed %lu/%lu tests", passed, total);
> + } else if (failed == 0)
> + fputs("All tests successful", stdout);
> + else
> + printf("Failed %lu/%lu tests, %.2f%% okay", failed, total,
> + (double) (total - failed) * 100.0 / (double) total);
> + if (skipped != 0) {
> + if (skipped == 1)
> + printf(", %lu test skipped", skipped);
> + else
> + printf(", %lu tests skipped", skipped);
> + }
> + puts(".");
> + printf("Files=%u, Tests=%lu", count, total);
> + printf(", %.2f seconds", tv_diff(&end, &start));
> + printf(" (%.2f usr + %.2f sys = %.2f CPU)\n", tv_seconds(&stats.ru_utime),
> + tv_seconds(&stats.ru_stime),
> + tv_sum(&stats.ru_utime, &stats.ru_stime));
> + return (failed == 0 && aborted == 0);
> +}
> +
> +
> +/*
> + * Run a single test case. This involves just running the test program after
> + * having done the environment setup and finding the test program.
> + */
> +static void
> +test_single(const char *program, const char *source, const char *build)
> +{
> + char *path;
> +
> + path = find_test(program, source, build);
> + if (execl(path, path, (char *) 0) == -1)
> + sysdie("cannot exec %s", path);
> +}
> +
> +
> +/*
> + * Main routine. Set the C_TAP_SOURCE, C_TAP_BUILD, SOURCE, and BUILD
> + * environment variables and then, given a file listing tests, run each test
> + * listed.
> + */
> +int
> +main(int argc, char *argv[])
> +{
> + int option;
> + int status = 0;
> + int single = 0;
> + enum test_verbose verbose = CONCISE;
> + char *c_tap_source_env = NULL;
> + char *c_tap_build_env = NULL;
> + char *source_env = NULL;
> + char *build_env = NULL;
> + const char *program;
> + const char *shortlist;
> + const char *list = NULL;
> + const char *source = C_TAP_SOURCE;
> + const char *build = C_TAP_BUILD;
> + struct testlist *tests;
> +
> + program = argv[0];
> + while ((option = getopt(argc, argv, "b:hl:os:v")) != EOF) {
> + switch (option) {
> + case 'b':
> + build = optarg;
> + break;
> + case 'h':
> + printf(usage_message, program, program, program, usage_extra);
> + exit(0);
> + case 'l':
> + list = optarg;
> + break;
> + case 'o':
> + single = 1;
> + break;
> + case 's':
> + source = optarg;
> + break;
> + case 'v':
> + verbose = VERBOSE;
> + break;
> + default:
> + exit(1);
> + }
> + }
> + argv += optind;
> + argc -= optind;
> + if ((list == NULL && argc < 1) || (list != NULL && argc > 0)) {
> + fprintf(stderr, usage_message, program, program, program, usage_extra);
> + exit(1);
> + }
> +
> + /*
> + * If C_TAP_VERBOSE is set in the environment, that also turns on verbose
> + * mode.
> + */
> + if (getenv("C_TAP_VERBOSE") != NULL)
> + verbose = VERBOSE;
> +
> + /*
> + * Set C_TAP_SOURCE and C_TAP_BUILD environment variables. Also set
> + * SOURCE and BUILD for backward compatibility, although we're trying to
> + * migrate to the ones with a C_TAP_* prefix.
> + */
> + if (source != NULL) {
> + c_tap_source_env = concat("C_TAP_SOURCE=", source, (const char *) 0);
> + if (putenv(c_tap_source_env) != 0)
> + sysdie("cannot set C_TAP_SOURCE in the environment");
> + source_env = concat("SOURCE=", source, (const char *) 0);
> + if (putenv(source_env) != 0)
> + sysdie("cannot set SOURCE in the environment");
> + }
> + if (build != NULL) {
> + c_tap_build_env = concat("C_TAP_BUILD=", build, (const char *) 0);
> + if (putenv(c_tap_build_env) != 0)
> + sysdie("cannot set C_TAP_BUILD in the environment");
> + build_env = concat("BUILD=", build, (const char *) 0);
> + if (putenv(build_env) != 0)
> + sysdie("cannot set BUILD in the environment");
> + }
> +
> + /* Run the tests as instructed. */
> + if (single)
> + test_single(argv[0], source, build);
> + else if (list != NULL) {
> + shortlist = strrchr(list, '/');
> + if (shortlist == NULL)
> + shortlist = list;
> + else
> + shortlist++;
> + printf(banner, shortlist);
> + tests = read_test_list(list, source, build);
> + status = test_batch(tests, verbose) ? 0 : 1;
> + } else {
> + tests = build_test_list(argv, argc, source, build);
> + status = test_batch(tests, verbose) ? 0 : 1;
> + }
> +
> + /* For valgrind cleanliness, free all our memory. */
> + if (source_env != NULL) {
> + putenv((char *) "C_TAP_SOURCE=");
> + putenv((char *) "SOURCE=");
> + free(c_tap_source_env);
> + free(source_env);
> + }
> + if (build_env != NULL) {
> + putenv((char *) "C_TAP_BUILD=");
> + putenv((char *) "BUILD=");
> + free(c_tap_build_env);
> + free(build_env);
> + }
> + exit(status);
> +}
> \ No newline at end of file
> diff --git a/t/tap/basic.c b/t/tap/basic.c
> new file mode 100644
> index 0000000000..704282b9c1
> --- /dev/null
> +++ b/t/tap/basic.c
> @@ -0,0 +1,1029 @@
> +/*
> + * Some utility routines for writing tests.
> + *
> + * Here are a variety of utility routines for writing tests compatible with
> + * the TAP protocol. All routines of the form ok() or is*() take a test
> + * number and some number of appropriate arguments, check to be sure the
> + * results match the expected output using the arguments, and print out
> + * something appropriate for that test number. Other utility routines help in
> + * constructing more complex tests, skipping tests, reporting errors, setting
> + * up the TAP output format, or finding things in the test environment.
> + *
> + * This file is part of C TAP Harness. The current version plus supporting
> + * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> + *
> + * Written by Russ Allbery <eagle@eyrie.org>
> + * Copyright 2009-2019, 2021 Russ Allbery <eagle@eyrie.org>
> + * Copyright 2001-2002, 2004-2008, 2011-2014
> + * The Board of Trustees of the Leland Stanford Junior University
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> + * DEALINGS IN THE SOFTWARE.
> + *
> + * SPDX-License-Identifier: MIT
> + */
> +
> +#include <errno.h>
> +#include <limits.h>
> +#include <stdarg.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#ifdef _WIN32
> +# include <direct.h>
> +#else
> +# include <sys/stat.h>
> +#endif
> +#include <sys/types.h>
> +#include <unistd.h>
> +
> +#include <tap/basic.h>
> +
> +/* Windows provides mkdir and rmdir under different names. */
> +#ifdef _WIN32
> +# define mkdir(p, m) _mkdir(p)
> +# define rmdir(p) _rmdir(p)
> +#endif
> +
> +/*
> + * The test count. Always contains the number that will be used for the next
> + * test status. This is exported to callers of the library.
> + */
> +unsigned long testnum = 1;
> +
> +/*
> + * Status information stored so that we can give a test summary at the end of
> + * the test case. We store the planned final test and the count of failures.
> + * We can get the highest test count from testnum.
> + */
> +static unsigned long _planned = 0;
> +static unsigned long _failed = 0;
> +
> +/*
> + * Store the PID of the process that called plan() and only summarize
> + * results when that process exits, so as to not misreport results in forked
> + * processes.
> + */
> +static pid_t _process = 0;
> +
> +/*
> + * If true, we're doing lazy planning and will print out the plan based on the
> + * last test number at the end of testing.
> + */
> +static int _lazy = 0;
> +
> +/*
> + * If true, the test was aborted by calling bail(). Currently, this is only
> + * used to ensure that we pass a false value to any cleanup functions even if
> + * all tests to that point have passed.
> + */
> +static int _aborted = 0;
> +
> +/*
> + * Registered cleanup functions. These are stored as a linked list and run in
> + * registered order by finish when the test program exits. Each function is
> + * passed a boolean value indicating whether all tests were successful.
> + */
> +struct cleanup_func {
> + test_cleanup_func func;
> + test_cleanup_func_with_data func_with_data;
> + void *data;
> + struct cleanup_func *next;
> +};
> +static struct cleanup_func *cleanup_funcs = NULL;
> +
> +/*
> + * Registered diag files. Any output found in these files will be printed out
> + * as if it were passed to diag() before any other output we do. This allows
> + * background processes to log to a file and have that output interleaved with
> + * the test output.
> + */
> +struct diag_file {
> + char *name;
> + FILE *file;
> + char *buffer;
> + size_t bufsize;
> + struct diag_file *next;
> +};
> +static struct diag_file *diag_files = NULL;
> +
> +/*
> + * Print a specified prefix and then the test description. Handles turning
> + * the argument list into a va_args structure suitable for passing to
> + * print_desc, which has to be done in a macro. Assumes that format is the
> + * argument immediately before the variadic arguments.
> + */
> +#define PRINT_DESC(prefix, format) \
> + do { \
> + if (format != NULL) { \
> + va_list args; \
> + printf("%s", prefix); \
> + va_start(args, format); \
> + vprintf(format, args); \
> + va_end(args); \
> + } \
> + } while (0)
> +
> +
> +/*
> + * Form a new string by concatenating multiple strings. The arguments must be
> + * terminated by (const char *) 0.
> + *
> + * This function only exists because we can't assume asprintf. We can't
> + * simulate asprintf with snprintf because we're only assuming SUSv3, which
> + * does not require that snprintf with a NULL buffer return the required
> + * length. When those constraints are relaxed, this should be ripped out and
> + * replaced with asprintf or a more trivial replacement with snprintf.
> + */
> +static char *
> +concat(const char *first, ...)
> +{
> + va_list args;
> + char *result;
> + const char *string;
> + size_t offset;
> + size_t length = 0;
> +
> + /*
> + * Find the total memory required. Ensure we don't overflow length. See
> + * the comment for breallocarray for why we're using UINT_MAX here.
> + */
> + va_start(args, first);
> + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> + if (length >= UINT_MAX - strlen(string))
> + bail("strings too long in concat");
> + length += strlen(string);
> + }
> + va_end(args);
> + length++;
> +
> + /* Create the string. */
> + result = bcalloc_type(length, char);
> + va_start(args, first);
> + offset = 0;
> + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> + memcpy(result + offset, string, strlen(string));
> + offset += strlen(string);
> + }
> + va_end(args);
> + result[offset] = '\0';
> + return result;
> +}
> +
> +
> +/*
> + * Helper function for check_diag_files to handle a single line in a diag
> + * file.
> + *
> + * The general scheme here used is as follows: read one line of output. If we
> + * get NULL, check for an error. If there was one, bail out of the test
> + * program; otherwise, return, and the enclosing loop will check for EOF.
> + *
> + * If we get some data, see if it ends in a newline. If it doesn't end in a
> + * newline, we have one of two cases: our buffer isn't large enough, in which
> + * case we resize it and try again, or we have incomplete data in the file, in
> + * which case we rewind the file and will try again next time.
> + *
> + * Returns a boolean indicating whether the last line was incomplete.
> + */
> +static int
> +handle_diag_file_line(struct diag_file *file, fpos_t where)
> +{
> + int size;
> + size_t length;
> +
> + /* Read the next line from the file. */
> + size = file->bufsize > INT_MAX ? INT_MAX : (int) file->bufsize;
> + if (fgets(file->buffer, size, file->file) == NULL) {
> + if (ferror(file->file))
> + sysbail("cannot read from %s", file->name);
> + return 0;
> + }
> +
> + /*
> + * See if the line ends in a newline. If not, see which error case we
> + * have.
> + */
> + length = strlen(file->buffer);
> + if (file->buffer[length - 1] != '\n') {
> + int incomplete = 0;
> +
> + /* Check whether we ran out of buffer space and resize if so. */
> + if (length < file->bufsize - 1)
> + incomplete = 1;
> + else {
> + file->bufsize += BUFSIZ;
> + file->buffer =
> + breallocarray_type(file->buffer, file->bufsize, char);
> + }
> +
> + /*
> + * On either incomplete lines or too small of a buffer, rewind
> + * and read the file again (on the next pass, if incomplete).
> + * It's simpler than trying to double-buffer the file.
> + */
> + if (fsetpos(file->file, &where) < 0)
> + sysbail("cannot set position in %s", file->name);
> + return incomplete;
> + }
> +
> + /* We saw a complete line. Print it out. */
> + printf("# %s", file->buffer);
> + return 0;
> +}
> +
> +
> +/*
> + * Check all registered diag_files for any output. We only print out the
> + * output if we see a complete line; otherwise, we wait for the next newline.
> + */
> +static void
> +check_diag_files(void)
> +{
> + struct diag_file *file;
> + fpos_t where;
> + int incomplete;
> +
> + /*
> + * Walk through each file and read each line of output available.
> + */
> + for (file = diag_files; file != NULL; file = file->next) {
> + clearerr(file->file);
> +
> + /* Store the current position in case we have to rewind. */
> + if (fgetpos(file->file, &where) < 0)
> + sysbail("cannot get position in %s", file->name);
> +
> + /* Continue until we get EOF or an incomplete line of data. */
> + incomplete = 0;
> + while (!feof(file->file) && !incomplete) {
> + incomplete = handle_diag_file_line(file, where);
> + }
> + }
> +}
> +
> +
> +/*
> + * Our exit handler. Called on completion of the test to report a summary of
> + * results provided we're still in the original process. This also handles
> + * printing out the plan if we used plan_lazy(), although that's suppressed if
> + * we never ran a test (due to an early bail, for example), and running any
> + * registered cleanup functions.
> + */
> +static void
> +finish(void)
> +{
> + int success, primary;
> + struct cleanup_func *current;
> + unsigned long highest = testnum - 1;
> + struct diag_file *file, *tmp;
> +
> + /* Check for pending diag_file output. */
> + check_diag_files();
> +
> + /* Free the diag_files. */
> + file = diag_files;
> + while (file != NULL) {
> + tmp = file;
> + file = file->next;
> + fclose(tmp->file);
> + free(tmp->name);
> + free(tmp->buffer);
> + free(tmp);
> + }
> + diag_files = NULL;
> +
> + /*
> + * Determine whether all tests were successful, which is needed before
> + * calling cleanup functions since we pass that fact to the functions.
> + */
> + if (_planned == 0 && _lazy)
> + _planned = highest;
> + success = (!_aborted && _planned == highest && _failed == 0);
> +
> + /*
> + * If there are any registered cleanup functions, we run those first. We
> + * always run them, even if we didn't run a test. Don't do anything
> + * except free the diag_files and call cleanup functions if we aren't the
> + * primary process (the process in which plan or plan_lazy was called),
> + * and tell the cleanup functions that fact.
> + */
> + primary = (_process == 0 || getpid() == _process);
> + while (cleanup_funcs != NULL) {
> + if (cleanup_funcs->func_with_data) {
> + void *data = cleanup_funcs->data;
> +
> + cleanup_funcs->func_with_data(success, primary, data);
> + } else {
> + cleanup_funcs->func(success, primary);
> + }
> + current = cleanup_funcs;
> + cleanup_funcs = cleanup_funcs->next;
> + free(current);
> + }
> + if (!primary)
> + return;
> +
> + /* Don't do anything further if we never planned a test. */
> + if (_planned == 0)
> + return;
> +
> + /* If we're aborting due to bail, don't print summaries. */
> + if (_aborted)
> + return;
> +
> + /* Print out the lazy plan if needed. */
> + fflush(stderr);
> + if (_lazy)
> + printf("1..%lu\n", _planned);
> +
> + /* Print out a summary of the results. */
> + if (_planned > highest)
> + diag("Looks like you planned %lu test%s but only ran %lu", _planned,
> + (_planned > 1 ? "s" : ""), highest);
> + else if (_planned < highest)
> + diag("Looks like you planned %lu test%s but ran %lu extra", _planned,
> + (_planned > 1 ? "s" : ""), highest - _planned);
> + else if (_failed > 0)
> + diag("Looks like you failed %lu test%s of %lu", _failed,
> + (_failed > 1 ? "s" : ""), _planned);
> + else if (_planned != 1)
> + diag("All %lu tests successful or skipped", _planned);
> + else
> + diag("%lu test successful or skipped", _planned);
> +}
> +
> +
> +/*
> + * Initialize things. Turns on line buffering on stdout and then prints out
> + * the number of tests in the test suite. We intentionally don't check for
> + * pending diag_file output here, since it should really come after the plan.
> + */
> +void
> +plan(unsigned long count)
> +{
> + if (setvbuf(stdout, NULL, _IOLBF, BUFSIZ) != 0)
> + sysdiag("cannot set stdout to line buffered");
> + fflush(stderr);
> + printf("1..%lu\n", count);
> + testnum = 1;
> + _planned = count;
> + _process = getpid();
> + if (atexit(finish) != 0) {
> + sysdiag("cannot register exit handler");
> + diag("cleanups will not be run");
> + }
> +}
> +
> +
> +/*
> + * Initialize things for lazy planning, where we'll automatically print out a
> + * plan at the end of the program. Turns on line buffering on stdout as well.
> + */
> +void
> +plan_lazy(void)
> +{
> + if (setvbuf(stdout, NULL, _IOLBF, BUFSIZ) != 0)
> + sysdiag("cannot set stdout to line buffered");
> + testnum = 1;
> + _process = getpid();
> + _lazy = 1;
> + if (atexit(finish) != 0)
> + sysbail("cannot register exit handler to display plan");
> +}
> +
> +
> +/*
> + * Skip the entire test suite and exits. Should be called instead of plan(),
> + * not after it, since it prints out a special plan line. Ignore diag_file
> + * output here, since it's not clear if it's allowed before the plan.
> + */
> +void
> +skip_all(const char *format, ...)
> +{
> + fflush(stderr);
> + printf("1..0 # skip");
> + PRINT_DESC(" ", format);
> + putchar('\n');
> + exit(0);
> +}
> +
> +
> +/*
> + * Takes a boolean success value and assumes the test passes if that value
> + * is true and fails if that value is false.
> + */
> +int
> +ok(int success, const char *format, ...)
> +{
> + fflush(stderr);
> + check_diag_files();
> + printf("%sok %lu", success ? "" : "not ", testnum++);
> + if (!success)
> + _failed++;
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Same as ok(), but takes the format arguments as a va_list.
> + */
> +int
> +okv(int success, const char *format, va_list args)
> +{
> + fflush(stderr);
> + check_diag_files();
> + printf("%sok %lu", success ? "" : "not ", testnum++);
> + if (!success)
> + _failed++;
> + if (format != NULL) {
> + printf(" - ");
> + vprintf(format, args);
> + }
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Skip a test.
> + */
> +void
> +skip(const char *reason, ...)
> +{
> + fflush(stderr);
> + check_diag_files();
> + printf("ok %lu # skip", testnum++);
> + PRINT_DESC(" ", reason);
> + putchar('\n');
> +}
> +
> +
> +/*
> + * Report the same status on the next count tests.
> + */
> +int
> +ok_block(unsigned long count, int success, const char *format, ...)
> +{
> + unsigned long i;
> +
> + fflush(stderr);
> + check_diag_files();
> + for (i = 0; i < count; i++) {
> + printf("%sok %lu", success ? "" : "not ", testnum++);
> + if (!success)
> + _failed++;
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + }
> + return success;
> +}
> +
> +
> +/*
> + * Skip the next count tests.
> + */
> +void
> +skip_block(unsigned long count, const char *reason, ...)
> +{
> + unsigned long i;
> +
> + fflush(stderr);
> + check_diag_files();
> + for (i = 0; i < count; i++) {
> + printf("ok %lu # skip", testnum++);
> + PRINT_DESC(" ", reason);
> + putchar('\n');
> + }
> +}
> +
> +
> +/*
> + * Takes two boolean values and requires the truth value of both match.
> + */
> +int
> +is_bool(int left, int right, const char *format, ...)
> +{
> + int success;
> +
> + fflush(stderr);
> + check_diag_files();
> + success = (!!left == !!right);
> + if (success)
> + printf("ok %lu", testnum++);
> + else {
> + diag(" left: %s", !!left ? "true" : "false");
> + diag("right: %s", !!right ? "true" : "false");
> + printf("not ok %lu", testnum++);
> + _failed++;
> + }
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Takes two integer values and requires they match.
> + */
> +int
> +is_int(long left, long right, const char *format, ...)
> +{
> + int success;
> +
> + fflush(stderr);
> + check_diag_files();
> + success = (left == right);
> + if (success)
> + printf("ok %lu", testnum++);
> + else {
> + diag(" left: %ld", left);
> + diag("right: %ld", right);
> + printf("not ok %lu", testnum++);
> + _failed++;
> + }
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Takes two strings and requires they match (using strcmp). NULL arguments
> + * are permitted and handled correctly.
> + */
> +int
> +is_string(const char *left, const char *right, const char *format, ...)
> +{
> + int success;
> +
> + fflush(stderr);
> + check_diag_files();
> +
> + /* Compare the strings, being careful of NULL. */
> + if (left == NULL)
> + success = (right == NULL);
> + else if (right == NULL)
> + success = 0;
> + else
> + success = (strcmp(left, right) == 0);
> +
> + /* Report the results. */
> + if (success)
> + printf("ok %lu", testnum++);
> + else {
> + diag(" left: %s", left == NULL ? "(null)" : left);
> + diag("right: %s", right == NULL ? "(null)" : right);
> + printf("not ok %lu", testnum++);
> + _failed++;
> + }
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Takes two unsigned longs and requires they match. On failure, reports them
> + * in hex.
> + */
> +int
> +is_hex(unsigned long left, unsigned long right, const char *format, ...)
> +{
> + int success;
> +
> + fflush(stderr);
> + check_diag_files();
> + success = (left == right);
> + if (success)
> + printf("ok %lu", testnum++);
> + else {
> + diag(" left: %lx", (unsigned long) left);
> + diag("right: %lx", (unsigned long) right);
> + printf("not ok %lu", testnum++);
> + _failed++;
> + }
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Takes pointers to a regions of memory and requires that len bytes from each
> + * match. Otherwise reports any bytes which didn't match.
> + */
> +int
> +is_blob(const void *left, const void *right, size_t len, const char *format,
> + ...)
> +{
> + int success;
> + size_t i;
> +
> + fflush(stderr);
> + check_diag_files();
> + success = (memcmp(left, right, len) == 0);
> + if (success)
> + printf("ok %lu", testnum++);
> + else {
> + const unsigned char *left_c = (const unsigned char *) left;
> + const unsigned char *right_c = (const unsigned char *) right;
> +
> + for (i = 0; i < len; i++) {
> + if (left_c[i] != right_c[i])
> + diag("offset %lu: left %02x, right %02x", (unsigned long) i,
> + left_c[i], right_c[i]);
> + }
> + printf("not ok %lu", testnum++);
> + _failed++;
> + }
> + PRINT_DESC(" - ", format);
> + putchar('\n');
> + return success;
> +}
> +
> +
> +/*
> + * Bail out with an error.
> + */
> +void
> +bail(const char *format, ...)
> +{
> + va_list args;
> +
> + _aborted = 1;
> + fflush(stderr);
> + check_diag_files();
> + fflush(stdout);
> + printf("Bail out! ");
> + va_start(args, format);
> + vprintf(format, args);
> + va_end(args);
> + printf("\n");
> + exit(255);
> +}
> +
> +
> +/*
> + * Bail out with an error, appending strerror(errno).
> + */
> +void
> +sysbail(const char *format, ...)
> +{
> + va_list args;
> + int oerrno = errno;
> +
> + _aborted = 1;
> + fflush(stderr);
> + check_diag_files();
> + fflush(stdout);
> + printf("Bail out! ");
> + va_start(args, format);
> + vprintf(format, args);
> + va_end(args);
> + printf(": %s\n", strerror(oerrno));
> + exit(255);
> +}
> +
> +
> +/*
> + * Report a diagnostic to stderr. Always returns 1 to allow embedding in
> + * compound statements.
> + */
> +int
> +diag(const char *format, ...)
> +{
> + va_list args;
> +
> + fflush(stderr);
> + check_diag_files();
> + fflush(stdout);
> + printf("# ");
> + va_start(args, format);
> + vprintf(format, args);
> + va_end(args);
> + printf("\n");
> + return 1;
> +}
> +
> +
> +/*
> + * Report a diagnostic to stderr, appending strerror(errno). Always returns 1
> + * to allow embedding in compound statements.
> + */
> +int
> +sysdiag(const char *format, ...)
> +{
> + va_list args;
> + int oerrno = errno;
> +
> + fflush(stderr);
> + check_diag_files();
> + fflush(stdout);
> + printf("# ");
> + va_start(args, format);
> + vprintf(format, args);
> + va_end(args);
> + printf(": %s\n", strerror(oerrno));
> + return 1;
> +}
> +
> +
> +/*
> + * Register a new file for diag_file processing.
> + */
> +void
> +diag_file_add(const char *name)
> +{
> + struct diag_file *file, *prev;
> +
> + file = bcalloc_type(1, struct diag_file);
> + file->name = bstrdup(name);
> + file->file = fopen(file->name, "r");
> + if (file->file == NULL)
> + sysbail("cannot open %s", name);
> + file->buffer = bcalloc_type(BUFSIZ, char);
> + file->bufsize = BUFSIZ;
> + if (diag_files == NULL)
> + diag_files = file;
> + else {
> + for (prev = diag_files; prev->next != NULL; prev = prev->next)
> + ;
> + prev->next = file;
> + }
> +}
> +
> +
> +/*
> + * Remove a file from diag_file processing. If the file is not found, do
> + * nothing, since there are some situations where it can be removed twice
> + * (such as if it's removed from a cleanup function, since cleanup functions
> + * are called after freeing all the diag_files).
> + */
> +void
> +diag_file_remove(const char *name)
> +{
> + struct diag_file *file;
> + struct diag_file **prev = &diag_files;
> +
> + for (file = diag_files; file != NULL; file = file->next) {
> + if (strcmp(file->name, name) == 0) {
> + *prev = file->next;
> + fclose(file->file);
> + free(file->name);
> + free(file->buffer);
> + free(file);
> + return;
> + }
> + prev = &file->next;
> + }
> +}
> +
> +
> +/*
> + * Allocate cleared memory, reporting a fatal error with bail on failure.
> + */
> +void *
> +bcalloc(size_t n, size_t size)
> +{
> + void *p;
> +
> + p = calloc(n, size);
> + if (p == NULL)
> + sysbail("failed to calloc %lu", (unsigned long) (n * size));
> + return p;
> +}
> +
> +
> +/*
> + * Allocate memory, reporting a fatal error with bail on failure.
> + */
> +void *
> +bmalloc(size_t size)
> +{
> + void *p;
> +
> + p = malloc(size);
> + if (p == NULL)
> + sysbail("failed to malloc %lu", (unsigned long) size);
> + return p;
> +}
> +
> +
> +/*
> + * Reallocate memory, reporting a fatal error with bail on failure.
> + */
> +void *
> +brealloc(void *p, size_t size)
> +{
> + p = realloc(p, size);
> + if (p == NULL)
> + sysbail("failed to realloc %lu bytes", (unsigned long) size);
> + return p;
> +}
> +
> +
> +/*
> + * The same as brealloc, but determine the size by multiplying an element
> + * count by a size, similar to calloc. The multiplication is checked for
> + * integer overflow.
> + *
> + * We should technically use SIZE_MAX here for the overflow check, but
> + * SIZE_MAX is C99 and we're only assuming C89 + SUSv3, which does not
> + * guarantee that it exists. They do guarantee that UINT_MAX exists, and we
> + * can assume that UINT_MAX <= SIZE_MAX.
> + *
> + * (In theory, C89 and C99 permit size_t to be smaller than unsigned int, but
> + * I disbelieve in the existence of such systems and they will have to cope
> + * without overflow checks.)
> + */
> +void *
> +breallocarray(void *p, size_t n, size_t size)
> +{
> + if (n > 0 && UINT_MAX / n <= size)
> + bail("reallocarray too large");
> + if (n == 0)
> + n = 1;
> + p = realloc(p, n * size);
> + if (p == NULL)
> + sysbail("failed to realloc %lu bytes", (unsigned long) (n * size));
> + return p;
> +}
> +
> +
> +/*
> + * Copy a string, reporting a fatal error with bail on failure.
> + */
> +char *
> +bstrdup(const char *s)
> +{
> + char *p;
> + size_t len;
> +
> + len = strlen(s) + 1;
> + p = (char *) malloc(len);
> + if (p == NULL)
> + sysbail("failed to strdup %lu bytes", (unsigned long) len);
> + memcpy(p, s, len);
> + return p;
> +}
> +
> +
> +/*
> + * Copy up to n characters of a string, reporting a fatal error with bail on
> + * failure. Don't use the system strndup function, since it may not exist and
> + * the TAP library doesn't assume any portability support.
> + */
> +char *
> +bstrndup(const char *s, size_t n)
> +{
> + const char *p;
> + char *copy;
> + size_t length;
> +
> + /* Don't assume that the source string is nul-terminated. */
> + for (p = s; (size_t) (p - s) < n && *p != '\0'; p++)
> + ;
> + length = (size_t) (p - s);
> + copy = (char *) malloc(length + 1);
> + if (copy == NULL)
> + sysbail("failed to strndup %lu bytes", (unsigned long) length);
> + memcpy(copy, s, length);
> + copy[length] = '\0';
> + return copy;
> +}
> +
> +
> +/*
> + * Locate a test file. Given the partial path to a file, look under
> + * C_TAP_BUILD and then C_TAP_SOURCE for the file and return the full path to
> + * the file. Returns NULL if the file doesn't exist. A non-NULL return
> + * should be freed with test_file_path_free().
> + */
> +char *
> +test_file_path(const char *file)
> +{
> + char *base;
> + char *path = NULL;
> + const char *envs[] = {"C_TAP_BUILD", "C_TAP_SOURCE", NULL};
> + int i;
> +
> + for (i = 0; envs[i] != NULL; i++) {
> + base = getenv(envs[i]);
> + if (base == NULL)
> + continue;
> + path = concat(base, "/", file, (const char *) 0);
> + if (access(path, R_OK) == 0)
> + break;
> + free(path);
> + path = NULL;
> + }
> + return path;
> +}
> +
> +
> +/*
> + * Free a path returned from test_file_path(). This function exists primarily
> + * for Windows, where memory must be freed from the same library domain that
> + * it was allocated from.
> + */
> +void
> +test_file_path_free(char *path)
> +{
> + free(path);
> +}
> +
> +
> +/*
> + * Create a temporary directory, tmp, under C_TAP_BUILD if set and the current
> + * directory if it does not. Returns the path to the temporary directory in
> + * newly allocated memory, and calls bail on any failure. The return value
> + * should be freed with test_tmpdir_free.
> + *
> + * This function uses sprintf because it attempts to be independent of all
> + * other portability layers. The use immediately after a memory allocation
> + * should be safe without using snprintf or strlcpy/strlcat.
> + */
> +char *
> +test_tmpdir(void)
> +{
> + const char *build;
> + char *path = NULL;
> +
> + build = getenv("C_TAP_BUILD");
> + if (build == NULL)
> + build = ".";
> + path = concat(build, "/tmp", (const char *) 0);
> + if (access(path, X_OK) < 0)
> + if (mkdir(path, 0777) < 0)
> + sysbail("error creating temporary directory %s", path);
> + return path;
> +}
> +
> +
> +/*
> + * Free a path returned from test_tmpdir() and attempt to remove the
> + * directory. If we can't delete the directory, don't worry; something else
> + * that hasn't yet cleaned up may still be using it.
> + */
> +void
> +test_tmpdir_free(char *path)
> +{
> + if (path != NULL)
> + rmdir(path);
> + free(path);
> +}
> +
> +static void
> +register_cleanup(test_cleanup_func func,
> + test_cleanup_func_with_data func_with_data, void *data)
> +{
> + struct cleanup_func *cleanup, **last;
> +
> + cleanup = bcalloc_type(1, struct cleanup_func);
> + cleanup->func = func;
> + cleanup->func_with_data = func_with_data;
> + cleanup->data = data;
> + cleanup->next = NULL;
> + last = &cleanup_funcs;
> + while (*last != NULL)
> + last = &(*last)->next;
> + *last = cleanup;
> +}
> +
> +/*
> + * Register a cleanup function that is called when testing ends. All such
> + * registered functions will be run by finish.
> + */
> +void
> +test_cleanup_register(test_cleanup_func func)
> +{
> + register_cleanup(func, NULL, NULL);
> +}
> +
> +/*
> + * Same as above, but also allows an opaque pointer to be passed to the cleanup
> + * function.
> + */
> +void
> +test_cleanup_register_with_data(test_cleanup_func_with_data func, void *data)
> +{
> + register_cleanup(NULL, func, data);
> +}
> diff --git a/t/tap/basic.h b/t/tap/basic.h
> new file mode 100644
> index 0000000000..afea8cb210
> --- /dev/null
> +++ b/t/tap/basic.h
> @@ -0,0 +1,198 @@
> +/*
> + * Basic utility routines for the TAP protocol.
> + *
> + * This file is part of C TAP Harness. The current version plus supporting
> + * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> + *
> + * Written by Russ Allbery <eagle@eyrie.org>
> + * Copyright 2009-2019, 2022 Russ Allbery <eagle@eyrie.org>
> + * Copyright 2001-2002, 2004-2008, 2011-2012, 2014
> + * The Board of Trustees of the Leland Stanford Junior University
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> + * DEALINGS IN THE SOFTWARE.
> + *
> + * SPDX-License-Identifier: MIT
> + */
> +
> +#ifndef TAP_BASIC_H
> +#define TAP_BASIC_H 1
> +
> +#include <stdarg.h> /* va_list */
> +#include <stddef.h> /* size_t */
> +#include <stdlib.h> /* free */
> +#include <tap/macros.h>
> +
> +/*
> + * Used for iterating through arrays. ARRAY_SIZE returns the number of
> + * elements in the array (useful for a < upper bound in a for loop) and
> + * ARRAY_END returns a pointer to the element past the end (ISO C99 makes it
> + * legal to refer to such a pointer as long as it's never dereferenced).
> + */
> +// #define ARRAY_SIZE(array) (sizeof(array) / sizeof((array)[0]))
> +// #define ARRAY_END(array) (&(array)[ARRAY_SIZE(array)])
> +
> +BEGIN_DECLS
> +
> +/*
> + * The test count. Always contains the number that will be used for the next
> + * test status.
> + */
> +extern unsigned long testnum;
> +
> +/* Print out the number of tests and set standard output to line buffered. */
> +void plan(unsigned long count);
> +
> +/*
> + * Prepare for lazy planning, in which the plan will be printed automatically
> + * at the end of the test program.
> + */
> +void plan_lazy(void);
> +
> +/* Skip the entire test suite. Call instead of plan. */
> +void skip_all(const char *format, ...)
> + __attribute__((__noreturn__, __format__(printf, 1, 2)));
> +
> +/*
> + * Basic reporting functions. The okv() function is the same as ok() but
> + * takes the test description as a va_list to make it easier to reuse the
> + * reporting infrastructure when writing new tests. ok() and okv() return the
> + * value of the success argument.
> + */
> +int ok(int success, const char *format, ...)
> + __attribute__((__format__(printf, 2, 3)));
> +int okv(int success, const char *format, va_list args)
> + __attribute__((__format__(printf, 2, 0)));
> +void skip(const char *reason, ...) __attribute__((__format__(printf, 1, 2)));
> +
> +/*
> + * Report the same status on, or skip, the next count tests. ok_block()
> + * returns the value of the success argument.
> + */
> +int ok_block(unsigned long count, int success, const char *format, ...)
> + __attribute__((__format__(printf, 3, 4)));
> +void skip_block(unsigned long count, const char *reason, ...)
> + __attribute__((__format__(printf, 2, 3)));
> +
> +/*
> + * Compare two values. Returns true if the test passes and false if it fails.
> + * is_bool takes an int since the bool type isn't fully portable yet, but
> + * interprets both arguments for their truth value, not for their numeric
> + * value.
> + */
> +int is_bool(int, int, const char *format, ...)
> + __attribute__((__format__(printf, 3, 4)));
> +int is_int(long, long, const char *format, ...)
> + __attribute__((__format__(printf, 3, 4)));
> +int is_string(const char *, const char *, const char *format, ...)
> + __attribute__((__format__(printf, 3, 4)));
> +int is_hex(unsigned long, unsigned long, const char *format, ...)
> + __attribute__((__format__(printf, 3, 4)));
> +int is_blob(const void *, const void *, size_t, const char *format, ...)
> + __attribute__((__format__(printf, 4, 5)));
> +
> +/* Bail out with an error. sysbail appends strerror(errno). */
> +void bail(const char *format, ...)
> + __attribute__((__noreturn__, __nonnull__, __format__(printf, 1, 2)));
> +void sysbail(const char *format, ...)
> + __attribute__((__noreturn__, __nonnull__, __format__(printf, 1, 2)));
> +
> +/* Report a diagnostic to stderr prefixed with #. */
> +int diag(const char *format, ...)
> + __attribute__((__nonnull__, __format__(printf, 1, 2)));
> +int sysdiag(const char *format, ...)
> + __attribute__((__nonnull__, __format__(printf, 1, 2)));
> +
> +/*
> + * Register or unregister a file that contains supplementary diagnostics.
> + * Before any other output, all registered files will be read, line by line,
> + * and each line will be reported as a diagnostic as if it were passed to
> + * diag(). Nul characters are not supported in these files and will result in
> + * truncated output.
> + */
> +void diag_file_add(const char *file) __attribute__((__nonnull__));
> +void diag_file_remove(const char *file) __attribute__((__nonnull__));
> +
> +/* Allocate memory, reporting a fatal error with bail on failure. */
> +void *bcalloc(size_t, size_t)
> + __attribute__((__alloc_size__(1, 2), __malloc__(free),
> + __warn_unused_result__));
> +void *bmalloc(size_t) __attribute__((__alloc_size__(1), __malloc__(free),
> + __warn_unused_result__));
> +void *breallocarray(void *, size_t, size_t)
> + __attribute__((__alloc_size__(2, 3), __malloc__(free),
> + __warn_unused_result__));
> +void *brealloc(void *, size_t)
> + __attribute__((__alloc_size__(2), __malloc__(free),
> + __warn_unused_result__));
> +char *bstrdup(const char *)
> + __attribute__((__malloc__(free), __nonnull__, __warn_unused_result__));
> +char *bstrndup(const char *, size_t)
> + __attribute__((__malloc__(free), __nonnull__, __warn_unused_result__));
> +
> +/*
> + * Macros that cast the return value from b* memory functions, making them
> + * usable in C++ code and providing some additional type safety.
> + */
> +#define bcalloc_type(n, type) ((type *) bcalloc((n), sizeof(type)))
> +#define breallocarray_type(p, n, type) \
> + ((type *) breallocarray((p), (n), sizeof(type)))
> +
> +/*
> + * Find a test file under C_TAP_BUILD or C_TAP_SOURCE, returning the full
> + * path. The returned path should be freed with test_file_path_free().
> + */
> +void test_file_path_free(char *path);
> +char *test_file_path(const char *file)
> + __attribute__((__malloc__(test_file_path_free), __nonnull__,
> + __warn_unused_result__));
> +
> +/*
> + * Create a temporary directory relative to C_TAP_BUILD and return the path.
> + * The returned path should be freed with test_tmpdir_free().
> + */
> +void test_tmpdir_free(char *path);
> +char *test_tmpdir(void)
> + __attribute__((__malloc__(test_tmpdir_free), __warn_unused_result__));
> +
> +/*
> + * Register a cleanup function that is called when testing ends. All such
> + * registered functions will be run during atexit handling (and are therefore
> + * subject to all the same constraints and caveats as atexit functions).
> + *
> + * The function must return void and will be passed two arguments: an int that
> + * will be true if the test completed successfully and false otherwise, and an
> + * int that will be true if the cleanup function is run in the primary process
> + * (the one that called plan or plan_lazy) and false otherwise. If
> + * test_cleanup_register_with_data is used instead, a generic pointer can be
> + * provided and will be passed to the cleanup function as a third argument.
> + *
> + * test_cleanup_register_with_data is the better API and should have been the
> + * only API. test_cleanup_register was an API error preserved for backward
> + * cmpatibility.
> + */
> +typedef void (*test_cleanup_func)(int, int);
> +typedef void (*test_cleanup_func_with_data)(int, int, void *);
> +
> +void test_cleanup_register(test_cleanup_func) __attribute__((__nonnull__));
> +void test_cleanup_register_with_data(test_cleanup_func_with_data, void *)
> + __attribute__((__nonnull__));
> +
> +END_DECLS
> +
> +#endif /* TAP_BASIC_H */
> diff --git a/t/tap/macros.h b/t/tap/macros.h
> new file mode 100644
> index 0000000000..0eabcb5847
> --- /dev/null
> +++ b/t/tap/macros.h
> @@ -0,0 +1,109 @@
> +/*
> + * Helpful macros for TAP header files.
> + *
> + * This is not, strictly speaking, related to TAP, but any TAP add-on is
> + * probably going to need these macros, so define them in one place so that
> + * everyone can pull them in.
> + *
> + * This file is part of C TAP Harness. The current version plus supporting
> + * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> + *
> + * Copyright 2008, 2012-2013, 2015, 2022 Russ Allbery <eagle@eyrie.org>
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> + * DEALINGS IN THE SOFTWARE.
> + *
> + * SPDX-License-Identifier: MIT
> + */
> +
> +#ifndef TAP_MACROS_H
> +#define TAP_MACROS_H 1
> +
> +/*
> + * __attribute__ is available in gcc 2.5 and later, but only with gcc 2.7
> + * could you use the __format__ form of the attributes, which is what we use
> + * (to avoid confusion with other macros), and only with gcc 2.96 can you use
> + * the attribute __malloc__. 2.96 is very old, so don't bother trying to get
> + * the other attributes to work with GCC versions between 2.7 and 2.96.
> + */
> +#ifndef __attribute__
> +# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
> +# define __attribute__(spec) /* empty */
> +# endif
> +#endif
> +
> +/*
> + * We use __alloc_size__, but it was only available in fairly recent versions
> + * of GCC. Suppress warnings about the unknown attribute if GCC is too old.
> + * We know that we're GCC at this point, so we can use the GCC variadic macro
> + * extension, which will still work with versions of GCC too old to have C99
> + * variadic macro support.
> + */
> +#if !defined(__attribute__) && !defined(__alloc_size__)
> +# if defined(__GNUC__) && !defined(__clang__)
> +# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3)
> +# define __alloc_size__(spec, args...) /* empty */
> +# endif
> +# endif
> +#endif
> +
> +/* Suppress __warn_unused_result__ if gcc is too old. */
> +#if !defined(__attribute__) && !defined(__warn_unused_result__)
> +# if __GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4)
> +# define __warn_unused_result__ /* empty */
> +# endif
> +#endif
> +
> +/*
> + * Suppress the argument to __malloc__ in Clang (not supported in at least
> + * version 13) and GCC versions prior to 11.
> + */
> +#if !defined(__attribute__) && !defined(__malloc__)
> +# if defined(__clang__) || __GNUC__ < 11
> +# define __malloc__(dalloc) __malloc__
> +# endif
> +#endif
> +
> +/*
> + * LLVM and Clang pretend to be GCC but don't support all of the __attribute__
> + * settings that GCC does. For them, suppress warnings about unknown
> + * attributes on declarations. This unfortunately will affect the entire
> + * compilation context, but there's no push and pop available.
> + */
> +#if !defined(__attribute__) && (defined(__llvm__) || defined(__clang__))
> +# pragma GCC diagnostic ignored "-Wattributes"
> +#endif
> +
> +/* Used for unused parameters to silence gcc warnings. */
> +// #define UNUSED __attribute__((__unused__))
> +
> +/*
> + * BEGIN_DECLS is used at the beginning of declarations so that C++
> + * compilers don't mangle their names. END_DECLS is used at the end.
> + */
> +#undef BEGIN_DECLS
> +#undef END_DECLS
> +#ifdef __cplusplus
> +# define BEGIN_DECLS extern "C" {
> +# define END_DECLS }
> +#else
> +# define BEGIN_DECLS /* empty */
> +# define END_DECLS /* empty */
> +#endif
> +
> +#endif /* TAP_MACROS_H */
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 0/2] add an external testing library for unit tests
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
` (2 preceding siblings ...)
2023-04-27 18:39 ` [RFC PATCH 0/2] add an external testing library for unit tests Junio C Hamano
@ 2023-04-27 21:35 ` brian m. carlson
2023-05-02 4:18 ` Felipe Contreras
2023-05-02 13:52 ` Ævar Arnfjörð Bjarmason
5 siblings, 0 replies; 29+ messages in thread
From: brian m. carlson @ 2023-04-27 21:35 UTC (permalink / raw)
To: Calvin Wan; +Cc: git
[-- Attachment #1: Type: text/plain, Size: 2065 bytes --]
On 2023-04-27 at 17:50:05, Calvin Wan wrote:
> In our current testing environment, we spend a significant amount of
> effort crafting end-to-end tests for error conditions that could easily
> be captured by unit tests (or we simply forgo some hard-to-setup and
> rare error conditions). Unit tests additionally provide stability to the
> codebase and can simplify debugging through isolation. Turning parts of
> Git into libraries[1] gives us the ability to run unit tests on the
> libraries and to write unit tests in C. Writing unit tests in pure C,
> rather than with our current shell/test-tool helper setup, simplifies
> test setup, simplifies passing data around (no shell-isms required), and
> reduces testing runtime by not spawning a separate process for every
> test invocation.
I think this is a good idea. Like Junio said downthread, we need to
have integration and end-to-end tests, and I think those will continue
to compose the majority of our tests. However, having simple and easy
to use unit tests would be really valuable for testing things like our
hash and HMAC implementations, as well as a variety of other library
functions, including our strbuf code.
At work, I recently ported a project from C with no unit tests to Rust
with unit tests (and in both cases, our existing integration and
functional testsuite) and found that writing unit tests let us have
substantially more confidence in the correct functioning of our code.
I think it's great that we're using existing TAP functionality as well.
If you're looking for some proof-of-concept projects to illustrate why
this is useful in v1, might I suggest some of the subsystems above?
Assuming it lands, I plan on sending some tests for the percent-encoding
in the strbuf code and some more aggressive testing of our block SHA-256
and HMAC implementations if nobody gets to it before me. (Mostly
because this is stuff I wrote or touched and would like to have more
confidence in.)
--
brian m. carlson (he/him or they/them)
Toronto, Ontario, CA
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 263 bytes --]
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-27 20:15 ` Phillip Wood
@ 2023-04-28 16:31 ` Calvin Wan
2023-05-02 15:46 ` Felipe Contreras
2023-05-10 15:46 ` Phillip Wood
2023-05-02 15:54 ` Felipe Contreras
2023-05-02 16:34 ` Ævar Arnfjörð Bjarmason
2 siblings, 2 replies; 29+ messages in thread
From: Calvin Wan @ 2023-04-28 16:31 UTC (permalink / raw)
To: phillip.wood; +Cc: git
On Thu, Apr 27, 2023 at 1:15 PM Phillip Wood <phillip.wood123@gmail.com> wrote:
>
> Hi Calvin
>
> On 27/04/2023 18:50, Calvin Wan wrote:
> > Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
> >
> > There is also more complete documentation at
> > https://www.eyrie.org/~eagle/software/c-tap-harness/
>
> I'm afraid this reply is rather briefer than I'd like but I'm short of
> time and about to go off-list for a couple of weeks. My ideal unit test
> library would
>
> - print the file and line number of failed assertions
This library does print the file and line number of failed assertions.
You can add a failing test in a new test file to check that this is
the case.
> - allow the test plan to be omitted by calling test_done() at the end
> of the test file as we do in our main test suite.
The same functionality can be achieved by adding some logic to adjust
the number of planned tests and return early if necessary.
> - support the TODO directive
This would be a nice feature to add to the library, however, I don't
think it's necessary in a first pass.
> - allow named tests (this maybe more trouble that it is worth as I
> think it inevitably leads to more boilerplate code calling the named
> tests)
I'm not quite sure what you're referring to with "named tests". Could
you clarify (possibly with an example)?
>
> Unfortunately this library doesn't seem to offer any of those features.
> It does support a lazy test plan but uses atexit() so will not detect if
> the test program exits before all the tests have run. I think it would
> be useful to add some unit tests to our test suite and maybe this
> library could form the basis of that but I think printing the file and
> line number of failed assertions is pretty essential.
Since I've addressed your concerns regarding printing file and line
numbers for failed assertions, do you think this test library is
sufficient as a base to iterate off of? I personally believe this
library has the necessary features to serve as a basis to write good
unit tests, and we can make adjustments to the library as we see fit
in the future.
Thanks,
Calvin
>
> Best Wishes
>
> Phillip
>
> > Signed-off-by: Calvin Wan <calvinwan@google.com>
> > ---
> > t/TESTS | 0
> > t/runtests.c | 1789 ++++++++++++++++++++++++++++++++++++++++++++++++
> > t/tap/basic.c | 1029 ++++++++++++++++++++++++++++
> > t/tap/basic.h | 198 ++++++
> > t/tap/macros.h | 109 +++
> > 5 files changed, 3125 insertions(+)
> > create mode 100644 t/TESTS
> > create mode 100644 t/runtests.c
> > create mode 100644 t/tap/basic.c
> > create mode 100644 t/tap/basic.h
> > create mode 100644 t/tap/macros.h
> >
> > diff --git a/t/TESTS b/t/TESTS
> > new file mode 100644
> > index 0000000000..e69de29bb2
> > diff --git a/t/runtests.c b/t/runtests.c
> > new file mode 100644
> > index 0000000000..4a55a801a6
> > --- /dev/null
> > +++ b/t/runtests.c
> > @@ -0,0 +1,1789 @@
> > +/*
> > + * Run a set of tests, reporting results.
> > + *
> > + * Test suite driver that runs a set of tests implementing a subset of the
> > + * Test Anything Protocol (TAP) and reports the results.
> > + *
> > + * Any bug reports, bug fixes, and improvements are very much welcome and
> > + * should be sent to the e-mail address below. This program is part of C TAP
> > + * Harness <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> > + *
> > + * Copyright 2000-2001, 2004, 2006-2019, 2022 Russ Allbery <eagle@eyrie.org>
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the "Software"),
> > + * to deal in the Software without restriction, including without limitation
> > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice shall be included in
> > + * all copies or substantial portions of the Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> > + * DEALINGS IN THE SOFTWARE.
> > + *
> > + * SPDX-License-Identifier: MIT
> > + */
> > +
> > +/*
> > + * Usage:
> > + *
> > + * runtests [-hv] [-b <build-dir>] [-s <source-dir>] -l <test-list>
> > + * runtests [-hv] [-b <build-dir>] [-s <source-dir>] <test> [<test> ...]
> > + * runtests -o [-h] [-b <build-dir>] [-s <source-dir>] <test>
> > + *
> > + * In the first case, expects a list of executables located in the given file,
> > + * one line per executable, possibly followed by a space-separated list of
> > + * options. For each one, runs it as part of a test suite, reporting results.
> > + * In the second case, use the same infrastructure, but run only the tests
> > + * listed on the command line.
> > + *
> > + * Test output should start with a line containing the number of tests
> > + * (numbered from 1 to this number), optionally preceded by "1..", although
> > + * that line may be given anywhere in the output. Each additional line should
> > + * be in the following format:
> > + *
> > + * ok <number>
> > + * not ok <number>
> > + * ok <number> # skip
> > + * not ok <number> # todo
> > + *
> > + * where <number> is the number of the test. An optional comment is permitted
> > + * after the number if preceded by whitespace. ok indicates success, not ok
> > + * indicates failure. "# skip" and "# todo" are a special cases of a comment,
> > + * and must start with exactly that formatting. They indicate the test was
> > + * skipped for some reason (maybe because it doesn't apply to this platform)
> > + * or is testing something known to currently fail. The text following either
> > + * "# skip" or "# todo" and whitespace is the reason.
> > + *
> > + * As a special case, the first line of the output may be in the form:
> > + *
> > + * 1..0 # skip some reason
> > + *
> > + * which indicates that this entire test case should be skipped and gives a
> > + * reason.
> > + *
> > + * Any other lines are ignored, although for compliance with the TAP protocol
> > + * all lines other than the ones in the above format should be sent to
> > + * standard error rather than standard output and start with #.
> > + *
> > + * This is a subset of TAP as documented in Test::Harness::TAP or
> > + * TAP::Parser::Grammar, which comes with Perl.
> > + *
> > + * If the -o option is given, instead run a single test and display all of its
> > + * output. This is intended for use with failing tests so that the person
> > + * running the test suite can get more details about what failed.
> > + *
> > + * If built with the C preprocessor symbols C_TAP_SOURCE and C_TAP_BUILD
> > + * defined, C TAP Harness will export those values in the environment so that
> > + * tests can find the source and build directory and will look for tests under
> > + * both directories. These paths can also be set with the -b and -s
> > + * command-line options, which will override anything set at build time.
> > + *
> > + * If the -v option is given, or the C_TAP_VERBOSE environment variable is set,
> > + * display the full output of each test as it runs rather than showing a
> > + * summary of the results of each test.
> > + */
> > +
> > +/* Required for fdopen(), getopt(), and putenv(). */
> > +#if defined(__STRICT_ANSI__) || defined(PEDANTIC)
> > +# ifndef _XOPEN_SOURCE
> > +# define _XOPEN_SOURCE 500
> > +# endif
> > +#endif
> > +
> > +#include <ctype.h>
> > +#include <errno.h>
> > +#include <fcntl.h>
> > +#include <limits.h>
> > +#include <stdarg.h>
> > +#include <stddef.h>
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <string.h>
> > +#include <strings.h>
> > +#include <sys/stat.h>
> > +#include <sys/time.h>
> > +#include <sys/types.h>
> > +#include <sys/wait.h>
> > +#include <time.h>
> > +#include <unistd.h>
> > +
> > +/* sys/time.h must be included before sys/resource.h on some platforms. */
> > +#include <sys/resource.h>
> > +
> > +/* AIX 6.1 (and possibly later) doesn't have WCOREDUMP. */
> > +#ifndef WCOREDUMP
> > +# define WCOREDUMP(status) ((unsigned) (status) &0x80)
> > +#endif
> > +
> > +/*
> > + * POSIX requires that these be defined in <unistd.h>, but they're not always
> > + * available. If one of them has been defined, all the rest almost certainly
> > + * have.
> > + */
> > +#ifndef STDIN_FILENO
> > +# define STDIN_FILENO 0
> > +# define STDOUT_FILENO 1
> > +# define STDERR_FILENO 2
> > +#endif
> > +
> > +/*
> > + * Used for iterating through arrays. Returns the number of elements in the
> > + * array (useful for a < upper bound in a for loop).
> > + */
> > +#define ARRAY_SIZE(array) (sizeof(array) / sizeof((array)[0]))
> > +
> > +/*
> > + * The source and build versions of the tests directory. This is used to set
> > + * the C_TAP_SOURCE and C_TAP_BUILD environment variables (and the SOURCE and
> > + * BUILD environment variables set for backward compatibility) and find test
> > + * programs, if set. Normally, this should be set as part of the build
> > + * process to the test subdirectories of $(abs_top_srcdir) and
> > + * $(abs_top_builddir) respectively.
> > + */
> > +#ifndef C_TAP_SOURCE
> > +# define C_TAP_SOURCE NULL
> > +#endif
> > +#ifndef C_TAP_BUILD
> > +# define C_TAP_BUILD NULL
> > +#endif
> > +
> > +/* Test status codes. */
> > +enum test_status {
> > + TEST_FAIL,
> > + TEST_PASS,
> > + TEST_SKIP,
> > + TEST_INVALID
> > +};
> > +
> > +/* Really, just a boolean, but this is more self-documenting. */
> > +enum test_verbose {
> > + CONCISE = 0,
> > + VERBOSE = 1
> > +};
> > +
> > +/* Indicates the state of our plan. */
> > +enum plan_status {
> > + PLAN_INIT, /* Nothing seen yet. */
> > + PLAN_FIRST, /* Plan seen before any tests. */
> > + PLAN_PENDING, /* Test seen and no plan yet. */
> > + PLAN_FINAL /* Plan seen after some tests. */
> > +};
> > +
> > +/* Error exit statuses for test processes. */
> > +#define CHILDERR_DUP 100 /* Couldn't redirect stderr or stdout. */
> > +#define CHILDERR_EXEC 101 /* Couldn't exec child process. */
> > +#define CHILDERR_STDIN 102 /* Couldn't open stdin file. */
> > +#define CHILDERR_STDERR 103 /* Couldn't open stderr file. */
> > +
> > +/* Structure to hold data for a set of tests. */
> > +struct testset {
> > + char *file; /* The file name of the test. */
> > + char **command; /* The argv vector to run the command. */
> > + enum plan_status plan; /* The status of our plan. */
> > + unsigned long count; /* Expected count of tests. */
> > + unsigned long current; /* The last seen test number. */
> > + unsigned int length; /* The length of the last status message. */
> > + unsigned long passed; /* Count of passing tests. */
> > + unsigned long failed; /* Count of failing lists. */
> > + unsigned long skipped; /* Count of skipped tests (passed). */
> > + unsigned long allocated; /* The size of the results table. */
> > + enum test_status *results; /* Table of results by test number. */
> > + unsigned int aborted; /* Whether the set was aborted. */
> > + unsigned int reported; /* Whether the results were reported. */
> > + int status; /* The exit status of the test. */
> > + unsigned int all_skipped; /* Whether all tests were skipped. */
> > + char *reason; /* Why all tests were skipped. */
> > +};
> > +
> > +/* Structure to hold a linked list of test sets. */
> > +struct testlist {
> > + struct testset *ts;
> > + struct testlist *next;
> > +};
> > +
> > +/*
> > + * Usage message. Should be used as a printf format with four arguments: the
> > + * path to runtests, given three times, and the usage_description. This is
> > + * split into variables to satisfy the pedantic ISO C90 limit on strings.
> > + */
> > +static const char usage_message[] = "\
> > +Usage: %s [-hv] [-b <build-dir>] [-s <source-dir>] <test> ...\n\
> > + %s [-hv] [-b <build-dir>] [-s <source-dir>] -l <test-list>\n\
> > + %s -o [-h] [-b <build-dir>] [-s <source-dir>] <test>\n\
> > +\n\
> > +Options:\n\
> > + -b <build-dir> Set the build directory to <build-dir>\n\
> > +%s";
> > +static const char usage_extra[] = "\
> > + -l <list> Take the list of tests to run from <test-list>\n\
> > + -o Run a single test rather than a list of tests\n\
> > + -s <source-dir> Set the source directory to <source-dir>\n\
> > + -v Show the full output of each test\n\
> > +\n\
> > +runtests normally runs each test listed on the command line. With the -l\n\
> > +option, it instead runs every test listed in a file. With the -o option,\n\
> > +it instead runs a single test and shows its complete output.\n";
> > +
> > +/*
> > + * Header used for test output. %s is replaced by the file name of the list
> > + * of tests.
> > + */
> > +static const char banner[] = "\n\
> > +Running all tests listed in %s. If any tests fail, run the failing\n\
> > +test program with runtests -o to see more details.\n\n";
> > +
> > +/* Header for reports of failed tests. */
> > +static const char header[] = "\n\
> > +Failed Set Fail/Total (%) Skip Stat Failing Tests\n\
> > +-------------------------- -------------- ---- ---- ------------------------";
> > +
> > +/* Include the file name and line number in malloc failures. */
> > +#define xcalloc(n, type) \
> > + ((type *) x_calloc((n), sizeof(type), __FILE__, __LINE__))
> > +#define xmalloc(size) ((char *) x_malloc((size), __FILE__, __LINE__))
> > +#define xstrdup(p) x_strdup((p), __FILE__, __LINE__)
> > +#define xstrndup(p, size) x_strndup((p), (size), __FILE__, __LINE__)
> > +#define xreallocarray(p, n, type) \
> > + ((type *) x_reallocarray((p), (n), sizeof(type), __FILE__, __LINE__))
> > +
> > +/*
> > + * __attribute__ is available in gcc 2.5 and later, but only with gcc 2.7
> > + * could you use the __format__ form of the attributes, which is what we use
> > + * (to avoid confusion with other macros).
> > + */
> > +#ifndef __attribute__
> > +# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 7)
> > +# define __attribute__(spec) /* empty */
> > +# endif
> > +#endif
> > +
> > +/*
> > + * We use __alloc_size__, but it was only available in fairly recent versions
> > + * of GCC. Suppress warnings about the unknown attribute if GCC is too old.
> > + * We know that we're GCC at this point, so we can use the GCC variadic macro
> > + * extension, which will still work with versions of GCC too old to have C99
> > + * variadic macro support.
> > + */
> > +#if !defined(__attribute__) && !defined(__alloc_size__)
> > +# if defined(__GNUC__) && !defined(__clang__)
> > +# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3)
> > +# define __alloc_size__(spec, args...) /* empty */
> > +# endif
> > +# endif
> > +#endif
> > +
> > +/*
> > + * Suppress the argument to __malloc__ in Clang (not supported in at least
> > + * version 13) and GCC versions prior to 11.
> > + */
> > +#if !defined(__attribute__) && !defined(__malloc__)
> > +# if defined(__clang__) || __GNUC__ < 11
> > +# define __malloc__(dalloc) __malloc__
> > +# endif
> > +#endif
> > +
> > +/*
> > + * LLVM and Clang pretend to be GCC but don't support all of the __attribute__
> > + * settings that GCC does. For them, suppress warnings about unknown
> > + * attributes on declarations. This unfortunately will affect the entire
> > + * compilation context, but there's no push and pop available.
> > + */
> > +#if !defined(__attribute__) && (defined(__llvm__) || defined(__clang__))
> > +# pragma GCC diagnostic ignored "-Wattributes"
> > +#endif
> > +
> > +/* Declare internal functions that benefit from compiler attributes. */
> > +static void die(const char *, ...)
> > + __attribute__((__nonnull__, __noreturn__, __format__(printf, 1, 2)));
> > +static void sysdie(const char *, ...)
> > + __attribute__((__nonnull__, __noreturn__, __format__(printf, 1, 2)));
> > +static void *x_calloc(size_t, size_t, const char *, int)
> > + __attribute__((__alloc_size__(1, 2), __malloc__(free), __nonnull__));
> > +static void *x_malloc(size_t, const char *, int)
> > + __attribute__((__alloc_size__(1), __malloc__(free), __nonnull__));
> > +static void *x_reallocarray(void *, size_t, size_t, const char *, int)
> > + __attribute__((__alloc_size__(2, 3), __malloc__(free), __nonnull__(4)));
> > +static char *x_strdup(const char *, const char *, int)
> > + __attribute__((__malloc__(free), __nonnull__));
> > +static char *x_strndup(const char *, size_t, const char *, int)
> > + __attribute__((__malloc__(free), __nonnull__));
> > +
> > +
> > +/*
> > + * Report a fatal error and exit.
> > + */
> > +static void
> > +die(const char *format, ...)
> > +{
> > + va_list args;
> > +
> > + fflush(stdout);
> > + fprintf(stderr, "runtests: ");
> > + va_start(args, format);
> > + vfprintf(stderr, format, args);
> > + va_end(args);
> > + fprintf(stderr, "\n");
> > + exit(1);
> > +}
> > +
> > +
> > +/*
> > + * Report a fatal error, including the results of strerror, and exit.
> > + */
> > +static void
> > +sysdie(const char *format, ...)
> > +{
> > + int oerrno;
> > + va_list args;
> > +
> > + oerrno = errno;
> > + fflush(stdout);
> > + fprintf(stderr, "runtests: ");
> > + va_start(args, format);
> > + vfprintf(stderr, format, args);
> > + va_end(args);
> > + fprintf(stderr, ": %s\n", strerror(oerrno));
> > + exit(1);
> > +}
> > +
> > +
> > +/*
> > + * Allocate zeroed memory, reporting a fatal error and exiting on failure.
> > + */
> > +static void *
> > +x_calloc(size_t n, size_t size, const char *file, int line)
> > +{
> > + void *p;
> > +
> > + n = (n > 0) ? n : 1;
> > + size = (size > 0) ? size : 1;
> > + p = calloc(n, size);
> > + if (p == NULL)
> > + sysdie("failed to calloc %lu bytes at %s line %d",
> > + (unsigned long) size, file, line);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Allocate memory, reporting a fatal error and exiting on failure.
> > + */
> > +static void *
> > +x_malloc(size_t size, const char *file, int line)
> > +{
> > + void *p;
> > +
> > + p = malloc(size);
> > + if (p == NULL)
> > + sysdie("failed to malloc %lu bytes at %s line %d",
> > + (unsigned long) size, file, line);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Reallocate memory, reporting a fatal error and exiting on failure.
> > + *
> > + * We should technically use SIZE_MAX here for the overflow check, but
> > + * SIZE_MAX is C99 and we're only assuming C89 + SUSv3, which does not
> > + * guarantee that it exists. They do guarantee that UINT_MAX exists, and we
> > + * can assume that UINT_MAX <= SIZE_MAX. And we should not be allocating
> > + * anything anywhere near that large.
> > + *
> > + * (In theory, C89 and C99 permit size_t to be smaller than unsigned int, but
> > + * I disbelieve in the existence of such systems and they will have to cope
> > + * without overflow checks.)
> > + */
> > +static void *
> > +x_reallocarray(void *p, size_t n, size_t size, const char *file, int line)
> > +{
> > + n = (n > 0) ? n : 1;
> > + size = (size > 0) ? size : 1;
> > +
> > + if (n > 0 && UINT_MAX / n <= size)
> > + sysdie("realloc too large at %s line %d", file, line);
> > + p = realloc(p, n * size);
> > + if (p == NULL)
> > + sysdie("failed to realloc %lu bytes at %s line %d",
> > + (unsigned long) (n * size), file, line);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Copy a string, reporting a fatal error and exiting on failure.
> > + */
> > +static char *
> > +x_strdup(const char *s, const char *file, int line)
> > +{
> > + char *p;
> > + size_t len;
> > +
> > + len = strlen(s) + 1;
> > + p = (char *) malloc(len);
> > + if (p == NULL)
> > + sysdie("failed to strdup %lu bytes at %s line %d", (unsigned long) len,
> > + file, line);
> > + memcpy(p, s, len);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Copy the first n characters of a string, reporting a fatal error and
> > + * existing on failure.
> > + *
> > + * Avoid using the system strndup function since it may not exist (on Mac OS
> > + * X, for example), and there's no need to introduce another portability
> > + * requirement.
> > + */
> > +char *
> > +x_strndup(const char *s, size_t size, const char *file, int line)
> > +{
> > + const char *p;
> > + size_t len;
> > + char *copy;
> > +
> > + /* Don't assume that the source string is nul-terminated. */
> > + for (p = s; (size_t) (p - s) < size && *p != '\0'; p++)
> > + ;
> > + len = (size_t) (p - s);
> > + copy = (char *) malloc(len + 1);
> > + if (copy == NULL)
> > + sysdie("failed to strndup %lu bytes at %s line %d",
> > + (unsigned long) len, file, line);
> > + memcpy(copy, s, len);
> > + copy[len] = '\0';
> > + return copy;
> > +}
> > +
> > +
> > +/*
> > + * Form a new string by concatenating multiple strings. The arguments must be
> > + * terminated by (const char *) 0.
> > + *
> > + * This function only exists because we can't assume asprintf. We can't
> > + * simulate asprintf with snprintf because we're only assuming SUSv3, which
> > + * does not require that snprintf with a NULL buffer return the required
> > + * length. When those constraints are relaxed, this should be ripped out and
> > + * replaced with asprintf or a more trivial replacement with snprintf.
> > + */
> > +static char *
> > +concat(const char *first, ...)
> > +{
> > + va_list args;
> > + char *result;
> > + const char *string;
> > + size_t offset;
> > + size_t length = 0;
> > +
> > + /*
> > + * Find the total memory required. Ensure we don't overflow length. We
> > + * aren't guaranteed to have SIZE_MAX, so use UINT_MAX as an acceptable
> > + * substitute (see the x_nrealloc comments).
> > + */
> > + va_start(args, first);
> > + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> > + if (length >= UINT_MAX - strlen(string)) {
> > + errno = EINVAL;
> > + sysdie("strings too long in concat");
> > + }
> > + length += strlen(string);
> > + }
> > + va_end(args);
> > + length++;
> > +
> > + /* Create the string. */
> > + result = xmalloc(length);
> > + va_start(args, first);
> > + offset = 0;
> > + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> > + memcpy(result + offset, string, strlen(string));
> > + offset += strlen(string);
> > + }
> > + va_end(args);
> > + result[offset] = '\0';
> > + return result;
> > +}
> > +
> > +
> > +/*
> > + * Given a struct timeval, return the number of seconds it represents as a
> > + * double. Use difftime() to convert a time_t to a double.
> > + */
> > +static double
> > +tv_seconds(const struct timeval *tv)
> > +{
> > + return difftime(tv->tv_sec, 0) + (double) tv->tv_usec * 1e-6;
> > +}
> > +
> > +
> > +/*
> > + * Given two struct timevals, return the difference in seconds.
> > + */
> > +static double
> > +tv_diff(const struct timeval *tv1, const struct timeval *tv0)
> > +{
> > + return tv_seconds(tv1) - tv_seconds(tv0);
> > +}
> > +
> > +
> > +/*
> > + * Given two struct timevals, return the sum in seconds as a double.
> > + */
> > +static double
> > +tv_sum(const struct timeval *tv1, const struct timeval *tv2)
> > +{
> > + return tv_seconds(tv1) + tv_seconds(tv2);
> > +}
> > +
> > +
> > +/*
> > + * Given a pointer to a string, skip any leading whitespace and return a
> > + * pointer to the first non-whitespace character.
> > + */
> > +static const char *
> > +skip_whitespace(const char *p)
> > +{
> > + while (isspace((unsigned char) (*p)))
> > + p++;
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Given a pointer to a string, skip any non-whitespace characters and return
> > + * a pointer to the first whitespace character, or to the end of the string.
> > + */
> > +static const char *
> > +skip_non_whitespace(const char *p)
> > +{
> > + while (*p != '\0' && !isspace((unsigned char) (*p)))
> > + p++;
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Start a program, connecting its stdout to a pipe on our end and its stderr
> > + * to /dev/null, and storing the file descriptor to read from in the two
> > + * argument. Returns the PID of the new process. Errors are fatal.
> > + */
> > +static pid_t
> > +test_start(char *const *command, int *fd)
> > +{
> > + int fds[2], infd, errfd;
> > + pid_t child;
> > +
> > + /* Create a pipe used to capture the output from the test program. */
> > + if (pipe(fds) == -1) {
> > + puts("ABORTED");
> > + fflush(stdout);
> > + sysdie("can't create pipe");
> > + }
> > +
> > + /* Fork a child process, massage the file descriptors, and exec. */
> > + child = fork();
> > + switch (child) {
> > + case -1:
> > + puts("ABORTED");
> > + fflush(stdout);
> > + sysdie("can't fork");
> > +
> > + /* In the child. Set up our standard output. */
> > + case 0:
> > + close(fds[0]);
> > + close(STDOUT_FILENO);
> > + if (dup2(fds[1], STDOUT_FILENO) < 0)
> > + _exit(CHILDERR_DUP);
> > + close(fds[1]);
> > +
> > + /* Point standard input at /dev/null. */
> > + close(STDIN_FILENO);
> > + infd = open("/dev/null", O_RDONLY);
> > + if (infd < 0)
> > + _exit(CHILDERR_STDIN);
> > + if (infd != STDIN_FILENO) {
> > + if (dup2(infd, STDIN_FILENO) < 0)
> > + _exit(CHILDERR_DUP);
> > + close(infd);
> > + }
> > +
> > + /* Point standard error at /dev/null. */
> > + close(STDERR_FILENO);
> > + errfd = open("/dev/null", O_WRONLY);
> > + if (errfd < 0)
> > + _exit(CHILDERR_STDERR);
> > + if (errfd != STDERR_FILENO) {
> > + if (dup2(errfd, STDERR_FILENO) < 0)
> > + _exit(CHILDERR_DUP);
> > + close(errfd);
> > + }
> > +
> > + /* Now, exec our process. */
> > + if (execv(command[0], command) == -1)
> > + _exit(CHILDERR_EXEC);
> > + break;
> > +
> > + /* In parent. Close the extra file descriptor. */
> > + default:
> > + close(fds[1]);
> > + break;
> > + }
> > + *fd = fds[0];
> > + return child;
> > +}
> > +
> > +
> > +/*
> > + * Back up over the output saying what test we were executing.
> > + */
> > +static void
> > +test_backspace(struct testset *ts)
> > +{
> > + unsigned int i;
> > +
> > + if (!isatty(STDOUT_FILENO))
> > + return;
> > + for (i = 0; i < ts->length; i++)
> > + putchar('\b');
> > + for (i = 0; i < ts->length; i++)
> > + putchar(' ');
> > + for (i = 0; i < ts->length; i++)
> > + putchar('\b');
> > + ts->length = 0;
> > +}
> > +
> > +
> > +/*
> > + * Allocate or resize the array of test results to be large enough to contain
> > + * the test number in.
> > + */
> > +static void
> > +resize_results(struct testset *ts, unsigned long n)
> > +{
> > + unsigned long i;
> > + size_t s;
> > +
> > + /* If there's already enough space, return quickly. */
> > + if (n <= ts->allocated)
> > + return;
> > +
> > + /*
> > + * If no space has been allocated, do the initial allocation. Otherwise,
> > + * resize. Start with 32 test cases and then add 1024 with each resize to
> > + * try to reduce the number of reallocations.
> > + */
> > + if (ts->allocated == 0) {
> > + s = (n > 32) ? n : 32;
> > + ts->results = xcalloc(s, enum test_status);
> > + } else {
> > + s = (n > ts->allocated + 1024) ? n : ts->allocated + 1024;
> > + ts->results = xreallocarray(ts->results, s, enum test_status);
> > + }
> > +
> > + /* Set the results for the newly-allocated test array. */
> > + for (i = ts->allocated; i < s; i++)
> > + ts->results[i] = TEST_INVALID;
> > + ts->allocated = s;
> > +}
> > +
> > +
> > +/*
> > + * Report an invalid test number and set the appropriate flags. Pulled into a
> > + * separate function since we do this in several places.
> > + */
> > +static void
> > +invalid_test_number(struct testset *ts, long n, enum test_verbose verbose)
> > +{
> > + if (!verbose)
> > + test_backspace(ts);
> > + printf("ABORTED (invalid test number %ld)\n", n);
> > + ts->aborted = 1;
> > + ts->reported = 1;
> > +}
> > +
> > +
> > +/*
> > + * Read the plan line of test output, which should contain the range of test
> > + * numbers. We may initialize the testset structure here if we haven't yet
> > + * seen a test. Return true if initialization succeeded and the test should
> > + * continue, false otherwise.
> > + */
> > +static int
> > +test_plan(const char *line, struct testset *ts, enum test_verbose verbose)
> > +{
> > + long n;
> > +
> > + /*
> > + * Accept a plan without the leading 1.. for compatibility with older
> > + * versions of runtests. This will only be allowed if we've not yet seen
> > + * a test result.
> > + */
> > + line = skip_whitespace(line);
> > + if (strncmp(line, "1..", 3) == 0)
> > + line += 3;
> > +
> > + /*
> > + * Get the count and check it for validity.
> > + *
> > + * If we have something of the form "1..0 # skip foo", the whole file was
> > + * skipped; record that. If we do skip the whole file, zero out all of
> > + * our statistics, since they're no longer relevant.
> > + *
> > + * strtol is called with a second argument to advance the line pointer
> > + * past the count to make it simpler to detect the # skip case.
> > + */
> > + n = strtol(line, (char **) &line, 10);
> > + if (n == 0) {
> > + line = skip_whitespace(line);
> > + if (*line == '#') {
> > + line = skip_whitespace(line + 1);
> > + if (strncasecmp(line, "skip", 4) == 0) {
> > + line = skip_whitespace(line + 4);
> > + if (*line != '\0') {
> > + ts->reason = xstrdup(line);
> > + ts->reason[strlen(ts->reason) - 1] = '\0';
> > + }
> > + ts->all_skipped = 1;
> > + ts->aborted = 1;
> > + ts->count = 0;
> > + ts->passed = 0;
> > + ts->skipped = 0;
> > + ts->failed = 0;
> > + return 0;
> > + }
> > + }
> > + }
> > + if (n <= 0) {
> > + puts("ABORTED (invalid test count)");
> > + ts->aborted = 1;
> > + ts->reported = 1;
> > + return 0;
> > + }
> > +
> > + /*
> > + * If we are doing lazy planning, check the plan against the largest test
> > + * number that we saw and fail now if we saw a check outside the plan
> > + * range.
> > + */
> > + if (ts->plan == PLAN_PENDING && (unsigned long) n < ts->count) {
> > + invalid_test_number(ts, (long) ts->count, verbose);
> > + return 0;
> > + }
> > +
> > + /*
> > + * Otherwise, allocated or resize the results if needed and update count,
> > + * and then record that we've seen a plan.
> > + */
> > + resize_results(ts, (unsigned long) n);
> > + ts->count = (unsigned long) n;
> > + if (ts->plan == PLAN_INIT)
> > + ts->plan = PLAN_FIRST;
> > + else if (ts->plan == PLAN_PENDING)
> > + ts->plan = PLAN_FINAL;
> > + return 1;
> > +}
> > +
> > +
> > +/*
> > + * Given a single line of output from a test, parse it and return the success
> > + * status of that test. Anything printed to stdout not matching the form
> > + * /^(not )?ok \d+/ is ignored. Sets ts->current to the test number that just
> > + * reported status.
> > + */
> > +static void
> > +test_checkline(const char *line, struct testset *ts, enum test_verbose verbose)
> > +{
> > + enum test_status status = TEST_PASS;
> > + const char *bail;
> > + char *end;
> > + long number;
> > + unsigned long current;
> > + int outlen;
> > +
> > + /* Before anything, check for a test abort. */
> > + bail = strstr(line, "Bail out!");
> > + if (bail != NULL) {
> > + bail = skip_whitespace(bail + strlen("Bail out!"));
> > + if (*bail != '\0') {
> > + size_t length;
> > +
> > + length = strlen(bail);
> > + if (bail[length - 1] == '\n')
> > + length--;
> > + if (!verbose)
> > + test_backspace(ts);
> > + printf("ABORTED (%.*s)\n", (int) length, bail);
> > + ts->reported = 1;
> > + }
> > + ts->aborted = 1;
> > + return;
> > + }
> > +
> > + /*
> > + * If the given line isn't newline-terminated, it was too big for an
> > + * fgets(), which means ignore it.
> > + */
> > + if (line[strlen(line) - 1] != '\n')
> > + return;
> > +
> > + /* If the line begins with a hash mark, ignore it. */
> > + if (line[0] == '#')
> > + return;
> > +
> > + /* If we haven't yet seen a plan, look for one. */
> > + if (ts->plan == PLAN_INIT && isdigit((unsigned char) (*line))) {
> > + if (!test_plan(line, ts, verbose))
> > + return;
> > + } else if (strncmp(line, "1..", 3) == 0) {
> > + if (ts->plan == PLAN_PENDING) {
> > + if (!test_plan(line, ts, verbose))
> > + return;
> > + } else {
> > + if (!verbose)
> > + test_backspace(ts);
> > + puts("ABORTED (multiple plans)");
> > + ts->aborted = 1;
> > + ts->reported = 1;
> > + return;
> > + }
> > + }
> > +
> > + /* Parse the line, ignoring something we can't parse. */
> > + if (strncmp(line, "not ", 4) == 0) {
> > + status = TEST_FAIL;
> > + line += 4;
> > + }
> > + if (strncmp(line, "ok", 2) != 0)
> > + return;
> > + line = skip_whitespace(line + 2);
> > + errno = 0;
> > + number = strtol(line, &end, 10);
> > + if (errno != 0 || end == line)
> > + current = ts->current + 1;
> > + else if (number <= 0) {
> > + invalid_test_number(ts, number, verbose);
> > + return;
> > + } else
> > + current = (unsigned long) number;
> > + if (current > ts->count && ts->plan == PLAN_FIRST) {
> > + invalid_test_number(ts, (long) current, verbose);
> > + return;
> > + }
> > +
> > + /* We have a valid test result. Tweak the results array if needed. */
> > + if (ts->plan == PLAN_INIT || ts->plan == PLAN_PENDING) {
> > + ts->plan = PLAN_PENDING;
> > + resize_results(ts, current);
> > + if (current > ts->count)
> > + ts->count = current;
> > + }
> > +
> > + /*
> > + * Handle directives. We should probably do something more interesting
> > + * with unexpected passes of todo tests.
> > + */
> > + while (isdigit((unsigned char) (*line)))
> > + line++;
> > + line = skip_whitespace(line);
> > + if (*line == '#') {
> > + line = skip_whitespace(line + 1);
> > + if (strncasecmp(line, "skip", 4) == 0)
> > + status = TEST_SKIP;
> > + if (strncasecmp(line, "todo", 4) == 0)
> > + status = (status == TEST_FAIL) ? TEST_SKIP : TEST_FAIL;
> > + }
> > +
> > + /* Make sure that the test number is in range and not a duplicate. */
> > + if (ts->results[current - 1] != TEST_INVALID) {
> > + if (!verbose)
> > + test_backspace(ts);
> > + printf("ABORTED (duplicate test number %lu)\n", current);
> > + ts->aborted = 1;
> > + ts->reported = 1;
> > + return;
> > + }
> > +
> > + /* Good results. Increment our various counters. */
> > + switch (status) {
> > + case TEST_PASS:
> > + ts->passed++;
> > + break;
> > + case TEST_FAIL:
> > + ts->failed++;
> > + break;
> > + case TEST_SKIP:
> > + ts->skipped++;
> > + break;
> > + case TEST_INVALID:
> > + break;
> > + }
> > + ts->current = current;
> > + ts->results[current - 1] = status;
> > + if (!verbose && isatty(STDOUT_FILENO)) {
> > + test_backspace(ts);
> > + if (ts->plan == PLAN_PENDING)
> > + outlen = printf("%lu/?", current);
> > + else
> > + outlen = printf("%lu/%lu", current, ts->count);
> > + ts->length = (outlen >= 0) ? (unsigned int) outlen : 0;
> > + fflush(stdout);
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Print out a range of test numbers, returning the number of characters it
> > + * took up. Takes the first number, the last number, the number of characters
> > + * already printed on the line, and the limit of number of characters the line
> > + * can hold. Add a comma and a space before the range if chars indicates that
> > + * something has already been printed on the line, and print ... instead if
> > + * chars plus the space needed would go over the limit (use a limit of 0 to
> > + * disable this).
> > + */
> > +static unsigned int
> > +test_print_range(unsigned long first, unsigned long last, unsigned long chars,
> > + unsigned int limit)
> > +{
> > + unsigned int needed = 0;
> > + unsigned long n;
> > +
> > + for (n = first; n > 0; n /= 10)
> > + needed++;
> > + if (last > first) {
> > + for (n = last; n > 0; n /= 10)
> > + needed++;
> > + needed++;
> > + }
> > + if (chars > 0)
> > + needed += 2;
> > + if (limit > 0 && chars + needed > limit) {
> > + needed = 0;
> > + if (chars <= limit) {
> > + if (chars > 0) {
> > + printf(", ");
> > + needed += 2;
> > + }
> > + printf("...");
> > + needed += 3;
> > + }
> > + } else {
> > + if (chars > 0)
> > + printf(", ");
> > + if (last > first)
> > + printf("%lu-", first);
> > + printf("%lu", last);
> > + }
> > + return needed;
> > +}
> > +
> > +
> > +/*
> > + * Summarize a single test set. The second argument is 0 if the set exited
> > + * cleanly, a positive integer representing the exit status if it exited
> > + * with a non-zero status, and a negative integer representing the signal
> > + * that terminated it if it was killed by a signal.
> > + */
> > +static void
> > +test_summarize(struct testset *ts, int status)
> > +{
> > + unsigned long i;
> > + unsigned long missing = 0;
> > + unsigned long failed = 0;
> > + unsigned long first = 0;
> > + unsigned long last = 0;
> > +
> > + if (ts->aborted) {
> > + fputs("ABORTED", stdout);
> > + if (ts->count > 0)
> > + printf(" (passed %lu/%lu)", ts->passed, ts->count - ts->skipped);
> > + } else {
> > + for (i = 0; i < ts->count; i++) {
> > + if (ts->results[i] == TEST_INVALID) {
> > + if (missing == 0)
> > + fputs("MISSED ", stdout);
> > + if (first && i == last)
> > + last = i + 1;
> > + else {
> > + if (first)
> > + test_print_range(first, last, missing - 1, 0);
> > + missing++;
> > + first = i + 1;
> > + last = i + 1;
> > + }
> > + }
> > + }
> > + if (first)
> > + test_print_range(first, last, missing - 1, 0);
> > + first = 0;
> > + last = 0;
> > + for (i = 0; i < ts->count; i++) {
> > + if (ts->results[i] == TEST_FAIL) {
> > + if (missing && !failed)
> > + fputs("; ", stdout);
> > + if (failed == 0)
> > + fputs("FAILED ", stdout);
> > + if (first && i == last)
> > + last = i + 1;
> > + else {
> > + if (first)
> > + test_print_range(first, last, failed - 1, 0);
> > + failed++;
> > + first = i + 1;
> > + last = i + 1;
> > + }
> > + }
> > + }
> > + if (first)
> > + test_print_range(first, last, failed - 1, 0);
> > + if (!missing && !failed) {
> > + fputs(!status ? "ok" : "dubious", stdout);
> > + if (ts->skipped > 0) {
> > + if (ts->skipped == 1)
> > + printf(" (skipped %lu test)", ts->skipped);
> > + else
> > + printf(" (skipped %lu tests)", ts->skipped);
> > + }
> > + }
> > + }
> > + if (status > 0)
> > + printf(" (exit status %d)", status);
> > + else if (status < 0)
> > + printf(" (killed by signal %d%s)", -status,
> > + WCOREDUMP(ts->status) ? ", core dumped" : "");
> > + putchar('\n');
> > +}
> > +
> > +
> > +/*
> > + * Given a test set, analyze the results, classify the exit status, handle a
> > + * few special error messages, and then pass it along to test_summarize() for
> > + * the regular output. Returns true if the test set ran successfully and all
> > + * tests passed or were skipped, false otherwise.
> > + */
> > +static int
> > +test_analyze(struct testset *ts)
> > +{
> > + if (ts->reported)
> > + return 0;
> > + if (ts->all_skipped) {
> > + if (ts->reason == NULL)
> > + puts("skipped");
> > + else
> > + printf("skipped (%s)\n", ts->reason);
> > + return 1;
> > + } else if (WIFEXITED(ts->status) && WEXITSTATUS(ts->status) != 0) {
> > + switch (WEXITSTATUS(ts->status)) {
> > + case CHILDERR_DUP:
> > + if (!ts->reported)
> > + puts("ABORTED (can't dup file descriptors)");
> > + break;
> > + case CHILDERR_EXEC:
> > + if (!ts->reported)
> > + puts("ABORTED (execution failed -- not found?)");
> > + break;
> > + case CHILDERR_STDIN:
> > + case CHILDERR_STDERR:
> > + if (!ts->reported)
> > + puts("ABORTED (can't open /dev/null)");
> > + break;
> > + default:
> > + test_summarize(ts, WEXITSTATUS(ts->status));
> > + break;
> > + }
> > + return 0;
> > + } else if (WIFSIGNALED(ts->status)) {
> > + test_summarize(ts, -WTERMSIG(ts->status));
> > + return 0;
> > + } else if (ts->plan != PLAN_FIRST && ts->plan != PLAN_FINAL) {
> > + puts("ABORTED (no valid test plan)");
> > + ts->aborted = 1;
> > + return 0;
> > + } else {
> > + test_summarize(ts, 0);
> > + return (ts->failed == 0);
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Runs a single test set, accumulating and then reporting the results.
> > + * Returns true if the test set was successfully run and all tests passed,
> > + * false otherwise.
> > + */
> > +static int
> > +test_run(struct testset *ts, enum test_verbose verbose)
> > +{
> > + pid_t testpid, child;
> > + int outfd, status;
> > + unsigned long i;
> > + FILE *output;
> > + char buffer[BUFSIZ];
> > +
> > + /* Run the test program. */
> > + testpid = test_start(ts->command, &outfd);
> > + output = fdopen(outfd, "r");
> > + if (!output) {
> > + puts("ABORTED");
> > + fflush(stdout);
> > + sysdie("fdopen failed");
> > + }
> > +
> > + /*
> > + * Pass each line of output to test_checkline(), and print the line if
> > + * verbosity is requested.
> > + */
> > + while (!ts->aborted && fgets(buffer, sizeof(buffer), output)) {
> > + if (verbose)
> > + printf("%s", buffer);
> > + test_checkline(buffer, ts, verbose);
> > + }
> > + if (ferror(output) || ts->plan == PLAN_INIT)
> > + ts->aborted = 1;
> > + if (!verbose)
> > + test_backspace(ts);
> > +
> > + /*
> > + * Consume the rest of the test output, close the output descriptor,
> > + * retrieve the exit status, and pass that information to test_analyze()
> > + * for eventual output.
> > + */
> > + while (fgets(buffer, sizeof(buffer), output))
> > + if (verbose)
> > + printf("%s", buffer);
> > + fclose(output);
> > + child = waitpid(testpid, &ts->status, 0);
> > + if (child == (pid_t) -1) {
> > + if (!ts->reported) {
> > + puts("ABORTED");
> > + fflush(stdout);
> > + }
> > + sysdie("waitpid for %u failed", (unsigned int) testpid);
> > + }
> > + if (ts->all_skipped)
> > + ts->aborted = 0;
> > + status = test_analyze(ts);
> > +
> > + /* Convert missing tests to failed tests. */
> > + for (i = 0; i < ts->count; i++) {
> > + if (ts->results[i] == TEST_INVALID) {
> > + ts->failed++;
> > + ts->results[i] = TEST_FAIL;
> > + status = 0;
> > + }
> > + }
> > + return status;
> > +}
> > +
> > +
> > +/* Summarize a list of test failures. */
> > +static void
> > +test_fail_summary(const struct testlist *fails)
> > +{
> > + struct testset *ts;
> > + unsigned int chars;
> > + unsigned long i, first, last, total;
> > + double failed;
> > +
> > + puts(header);
> > +
> > + /* Failed Set Fail/Total (%) Skip Stat Failing (25)
> > + -------------------------- -------------- ---- ---- -------------- */
> > + for (; fails; fails = fails->next) {
> > + ts = fails->ts;
> > + total = ts->count - ts->skipped;
> > + failed = (double) ts->failed;
> > + printf("%-26.26s %4lu/%-4lu %3.0f%% %4lu ", ts->file, ts->failed,
> > + total, total ? (failed * 100.0) / (double) total : 0,
> > + ts->skipped);
> > + if (WIFEXITED(ts->status))
> > + printf("%4d ", WEXITSTATUS(ts->status));
> > + else
> > + printf(" -- ");
> > + if (ts->aborted) {
> > + puts("aborted");
> > + continue;
> > + }
> > + chars = 0;
> > + first = 0;
> > + last = 0;
> > + for (i = 0; i < ts->count; i++) {
> > + if (ts->results[i] == TEST_FAIL) {
> > + if (first != 0 && i == last)
> > + last = i + 1;
> > + else {
> > + if (first != 0)
> > + chars += test_print_range(first, last, chars, 19);
> > + first = i + 1;
> > + last = i + 1;
> > + }
> > + }
> > + }
> > + if (first != 0)
> > + test_print_range(first, last, chars, 19);
> > + putchar('\n');
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Check whether a given file path is a valid test. Currently, this checks
> > + * whether it is executable and is a regular file. Returns true or false.
> > + */
> > +static int
> > +is_valid_test(const char *path)
> > +{
> > + struct stat st;
> > +
> > + if (access(path, X_OK) < 0)
> > + return 0;
> > + if (stat(path, &st) < 0)
> > + return 0;
> > + if (!S_ISREG(st.st_mode))
> > + return 0;
> > + return 1;
> > +}
> > +
> > +
> > +/*
> > + * Given the name of a test, a pointer to the testset struct, and the source
> > + * and build directories, find the test. We try first relative to the current
> > + * directory, then in the build directory (if not NULL), then in the source
> > + * directory. In each of those directories, we first try a "-t" extension and
> > + * then a ".t" extension. When we find an executable program, we return the
> > + * path to that program. If none of those paths are executable, just fill in
> > + * the name of the test as is.
> > + *
> > + * The caller is responsible for freeing the path member of the testset
> > + * struct.
> > + */
> > +static char *
> > +find_test(const char *name, const char *source, const char *build)
> > +{
> > + char *path = NULL;
> > + const char *bases[3], *suffix, *base;
> > + unsigned int i, j;
> > + const char *suffixes[3] = {"-t", ".t", ""};
> > +
> > + /* Possible base directories. */
> > + bases[0] = ".";
> > + bases[1] = build;
> > + bases[2] = source;
> > +
> > + /* Try each suffix with each base. */
> > + for (i = 0; i < ARRAY_SIZE(suffixes); i++) {
> > + suffix = suffixes[i];
> > + for (j = 0; j < ARRAY_SIZE(bases); j++) {
> > + base = bases[j];
> > + if (base == NULL)
> > + continue;
> > + path = concat(base, "/", name, suffix, (const char *) 0);
> > + if (is_valid_test(path))
> > + return path;
> > + free(path);
> > + path = NULL;
> > + }
> > + }
> > + if (path == NULL)
> > + path = xstrdup(name);
> > + return path;
> > +}
> > +
> > +
> > +/*
> > + * Parse a single line of a test list and store the test name and command to
> > + * execute it in the given testset struct.
> > + *
> > + * Normally, each line is just the name of the test, which is located in the
> > + * test directory and turned into a command to run. However, each line may
> > + * have whitespace-separated options, which change the command that's run.
> > + * Current supported options are:
> > + *
> > + * valgrind
> > + * Run the test under valgrind if C_TAP_VALGRIND is set. The contents
> > + * of that environment variable are taken as the valgrind command (with
> > + * options) to run. The command is parsed with a simple split on
> > + * whitespace and no quoting is supported.
> > + *
> > + * libtool
> > + * If running under valgrind, use libtool to invoke valgrind. This avoids
> > + * running valgrind on the wrapper shell script generated by libtool. If
> > + * set, C_TAP_LIBTOOL must be set to the full path to the libtool program
> > + * to use to run valgrind and thus the test. Ignored if the test isn't
> > + * being run under valgrind.
> > + */
> > +static void
> > +parse_test_list_line(const char *line, struct testset *ts, const char *source,
> > + const char *build)
> > +{
> > + const char *p, *end, *option, *libtool;
> > + const char *valgrind = NULL;
> > + unsigned int use_libtool = 0;
> > + unsigned int use_valgrind = 0;
> > + size_t len, i;
> > +
> > + /* Determine the name of the test. */
> > + p = skip_non_whitespace(line);
> > + ts->file = xstrndup(line, p - line);
> > +
> > + /* Check if any test options are set. */
> > + p = skip_whitespace(p);
> > + while (*p != '\0') {
> > + end = skip_non_whitespace(p);
> > + if (strncmp(p, "libtool", end - p) == 0) {
> > + use_libtool = 1;
> > + } else if (strncmp(p, "valgrind", end - p) == 0) {
> > + valgrind = getenv("C_TAP_VALGRIND");
> > + use_valgrind = (valgrind != NULL);
> > + } else {
> > + option = xstrndup(p, end - p);
> > + die("unknown test list option %s", option);
> > + }
> > + p = skip_whitespace(end);
> > + }
> > +
> > + /* Construct the argv to run the test. First, find the length. */
> > + len = 1;
> > + if (use_valgrind && valgrind != NULL) {
> > + p = skip_whitespace(valgrind);
> > + while (*p != '\0') {
> > + len++;
> > + p = skip_whitespace(skip_non_whitespace(p));
> > + }
> > + if (use_libtool)
> > + len += 2;
> > + }
> > +
> > + /* Now, build the command. */
> > + ts->command = xcalloc(len + 1, char *);
> > + i = 0;
> > + if (use_valgrind && valgrind != NULL) {
> > + if (use_libtool) {
> > + libtool = getenv("C_TAP_LIBTOOL");
> > + if (libtool == NULL)
> > + die("valgrind with libtool requested, but C_TAP_LIBTOOL is not"
> > + " set");
> > + ts->command[i++] = xstrdup(libtool);
> > + ts->command[i++] = xstrdup("--mode=execute");
> > + }
> > + p = skip_whitespace(valgrind);
> > + while (*p != '\0') {
> > + end = skip_non_whitespace(p);
> > + ts->command[i++] = xstrndup(p, end - p);
> > + p = skip_whitespace(end);
> > + }
> > + }
> > + if (i != len - 1)
> > + die("internal error while constructing command line");
> > + ts->command[i++] = find_test(ts->file, source, build);
> > + ts->command[i] = NULL;
> > +}
> > +
> > +
> > +/*
> > + * Read a list of tests from a file, returning the list of tests as a struct
> > + * testlist, or NULL if there were no tests (such as a file containing only
> > + * comments). Reports an error to standard error and exits if the list of
> > + * tests cannot be read.
> > + */
> > +static struct testlist *
> > +read_test_list(const char *filename, const char *source, const char *build)
> > +{
> > + FILE *file;
> > + unsigned int line;
> > + size_t length;
> > + char buffer[BUFSIZ];
> > + const char *start;
> > + struct testlist *listhead, *current;
> > +
> > + /* Create the initial container list that will hold our results. */
> > + listhead = xcalloc(1, struct testlist);
> > + current = NULL;
> > +
> > + /*
> > + * Open our file of tests to run and read it line by line, creating a new
> > + * struct testlist and struct testset for each line.
> > + */
> > + file = fopen(filename, "r");
> > + if (file == NULL)
> > + sysdie("can't open %s", filename);
> > + line = 0;
> > + while (fgets(buffer, sizeof(buffer), file)) {
> > + line++;
> > + length = strlen(buffer) - 1;
> > + if (buffer[length] != '\n') {
> > + fprintf(stderr, "%s:%u: line too long\n", filename, line);
> > + exit(1);
> > + }
> > + buffer[length] = '\0';
> > +
> > + /* Skip comments, leading spaces, and blank lines. */
> > + start = skip_whitespace(buffer);
> > + if (strlen(start) == 0)
> > + continue;
> > + if (start[0] == '#')
> > + continue;
> > +
> > + /* Allocate the new testset structure. */
> > + if (current == NULL)
> > + current = listhead;
> > + else {
> > + current->next = xcalloc(1, struct testlist);
> > + current = current->next;
> > + }
> > + current->ts = xcalloc(1, struct testset);
> > + current->ts->plan = PLAN_INIT;
> > +
> > + /* Parse the line and store the results in the testset struct. */
> > + parse_test_list_line(start, current->ts, source, build);
> > + }
> > + fclose(file);
> > +
> > + /* If there were no tests, current is still NULL. */
> > + if (current == NULL) {
> > + free(listhead);
> > + return NULL;
> > + }
> > +
> > + /* Return the results. */
> > + return listhead;
> > +}
> > +
> > +
> > +/*
> > + * Build a list of tests from command line arguments. Takes the argv and argc
> > + * representing the command line arguments and returns a newly allocated test
> > + * list, or NULL if there were no tests. The caller is responsible for
> > + * freeing.
> > + */
> > +static struct testlist *
> > +build_test_list(char *argv[], int argc, const char *source, const char *build)
> > +{
> > + int i;
> > + struct testlist *listhead, *current;
> > +
> > + /* Create the initial container list that will hold our results. */
> > + listhead = xcalloc(1, struct testlist);
> > + current = NULL;
> > +
> > + /* Walk the list of arguments and create test sets for them. */
> > + for (i = 0; i < argc; i++) {
> > + if (current == NULL)
> > + current = listhead;
> > + else {
> > + current->next = xcalloc(1, struct testlist);
> > + current = current->next;
> > + }
> > + current->ts = xcalloc(1, struct testset);
> > + current->ts->plan = PLAN_INIT;
> > + current->ts->file = xstrdup(argv[i]);
> > + current->ts->command = xcalloc(2, char *);
> > + current->ts->command[0] = find_test(current->ts->file, source, build);
> > + current->ts->command[1] = NULL;
> > + }
> > +
> > + /* If there were no tests, current is still NULL. */
> > + if (current == NULL) {
> > + free(listhead);
> > + return NULL;
> > + }
> > +
> > + /* Return the results. */
> > + return listhead;
> > +}
> > +
> > +
> > +/* Free a struct testset. */
> > +static void
> > +free_testset(struct testset *ts)
> > +{
> > + size_t i;
> > +
> > + free(ts->file);
> > + for (i = 0; ts->command[i] != NULL; i++)
> > + free(ts->command[i]);
> > + free(ts->command);
> > + free(ts->results);
> > + free(ts->reason);
> > + free(ts);
> > +}
> > +
> > +
> > +/*
> > + * Run a batch of tests. Takes two additional parameters: the root of the
> > + * source directory and the root of the build directory. Test programs will
> > + * be first searched for in the current directory, then the build directory,
> > + * then the source directory. Returns true iff all tests passed, and always
> > + * frees the test list that's passed in.
> > + */
> > +static int
> > +test_batch(struct testlist *tests, enum test_verbose verbose)
> > +{
> > + size_t length, i;
> > + size_t longest = 0;
> > + unsigned int count = 0;
> > + struct testset *ts;
> > + struct timeval start, end;
> > + struct rusage stats;
> > + struct testlist *failhead = NULL;
> > + struct testlist *failtail = NULL;
> > + struct testlist *current, *next;
> > + int succeeded;
> > + unsigned long total = 0;
> > + unsigned long passed = 0;
> > + unsigned long skipped = 0;
> > + unsigned long failed = 0;
> > + unsigned long aborted = 0;
> > +
> > + /* Walk the list of tests to find the longest name. */
> > + for (current = tests; current != NULL; current = current->next) {
> > + length = strlen(current->ts->file);
> > + if (length > longest)
> > + longest = length;
> > + }
> > +
> > + /*
> > + * Add two to longest and round up to the nearest tab stop. This is how
> > + * wide the column for printing the current test name will be.
> > + */
> > + longest += 2;
> > + if (longest % 8)
> > + longest += 8 - (longest % 8);
> > +
> > + /* Start the wall clock timer. */
> > + gettimeofday(&start, NULL);
> > +
> > + /* Now, plow through our tests again, running each one. */
> > + for (current = tests; current != NULL; current = current->next) {
> > + ts = current->ts;
> > +
> > + /* Print out the name of the test file. */
> > + fputs(ts->file, stdout);
> > + if (verbose)
> > + fputs("\n\n", stdout);
> > + else
> > + for (i = strlen(ts->file); i < longest; i++)
> > + putchar('.');
> > + if (isatty(STDOUT_FILENO))
> > + fflush(stdout);
> > +
> > + /* Run the test. */
> > + succeeded = test_run(ts, verbose);
> > + fflush(stdout);
> > + if (verbose)
> > + putchar('\n');
> > +
> > + /* Record cumulative statistics. */
> > + aborted += ts->aborted;
> > + total += ts->count + ts->all_skipped;
> > + passed += ts->passed;
> > + skipped += ts->skipped + ts->all_skipped;
> > + failed += ts->failed;
> > + count++;
> > +
> > + /* If the test fails, we shuffle it over to the fail list. */
> > + if (!succeeded) {
> > + if (failhead == NULL) {
> > + failhead = xcalloc(1, struct testlist);
> > + failtail = failhead;
> > + } else {
> > + failtail->next = xcalloc(1, struct testlist);
> > + failtail = failtail->next;
> > + }
> > + failtail->ts = ts;
> > + failtail->next = NULL;
> > + }
> > + }
> > + total -= skipped;
> > +
> > + /* Stop the timer and get our child resource statistics. */
> > + gettimeofday(&end, NULL);
> > + getrusage(RUSAGE_CHILDREN, &stats);
> > +
> > + /* Summarize the failures and free the failure list. */
> > + if (failhead != NULL) {
> > + test_fail_summary(failhead);
> > + while (failhead != NULL) {
> > + next = failhead->next;
> > + free(failhead);
> > + failhead = next;
> > + }
> > + }
> > +
> > + /* Free the memory used by the test lists. */
> > + while (tests != NULL) {
> > + next = tests->next;
> > + free_testset(tests->ts);
> > + free(tests);
> > + tests = next;
> > + }
> > +
> > + /* Print out the final test summary. */
> > + putchar('\n');
> > + if (aborted != 0) {
> > + if (aborted == 1)
> > + printf("Aborted %lu test set", aborted);
> > + else
> > + printf("Aborted %lu test sets", aborted);
> > + printf(", passed %lu/%lu tests", passed, total);
> > + } else if (failed == 0)
> > + fputs("All tests successful", stdout);
> > + else
> > + printf("Failed %lu/%lu tests, %.2f%% okay", failed, total,
> > + (double) (total - failed) * 100.0 / (double) total);
> > + if (skipped != 0) {
> > + if (skipped == 1)
> > + printf(", %lu test skipped", skipped);
> > + else
> > + printf(", %lu tests skipped", skipped);
> > + }
> > + puts(".");
> > + printf("Files=%u, Tests=%lu", count, total);
> > + printf(", %.2f seconds", tv_diff(&end, &start));
> > + printf(" (%.2f usr + %.2f sys = %.2f CPU)\n", tv_seconds(&stats.ru_utime),
> > + tv_seconds(&stats.ru_stime),
> > + tv_sum(&stats.ru_utime, &stats.ru_stime));
> > + return (failed == 0 && aborted == 0);
> > +}
> > +
> > +
> > +/*
> > + * Run a single test case. This involves just running the test program after
> > + * having done the environment setup and finding the test program.
> > + */
> > +static void
> > +test_single(const char *program, const char *source, const char *build)
> > +{
> > + char *path;
> > +
> > + path = find_test(program, source, build);
> > + if (execl(path, path, (char *) 0) == -1)
> > + sysdie("cannot exec %s", path);
> > +}
> > +
> > +
> > +/*
> > + * Main routine. Set the C_TAP_SOURCE, C_TAP_BUILD, SOURCE, and BUILD
> > + * environment variables and then, given a file listing tests, run each test
> > + * listed.
> > + */
> > +int
> > +main(int argc, char *argv[])
> > +{
> > + int option;
> > + int status = 0;
> > + int single = 0;
> > + enum test_verbose verbose = CONCISE;
> > + char *c_tap_source_env = NULL;
> > + char *c_tap_build_env = NULL;
> > + char *source_env = NULL;
> > + char *build_env = NULL;
> > + const char *program;
> > + const char *shortlist;
> > + const char *list = NULL;
> > + const char *source = C_TAP_SOURCE;
> > + const char *build = C_TAP_BUILD;
> > + struct testlist *tests;
> > +
> > + program = argv[0];
> > + while ((option = getopt(argc, argv, "b:hl:os:v")) != EOF) {
> > + switch (option) {
> > + case 'b':
> > + build = optarg;
> > + break;
> > + case 'h':
> > + printf(usage_message, program, program, program, usage_extra);
> > + exit(0);
> > + case 'l':
> > + list = optarg;
> > + break;
> > + case 'o':
> > + single = 1;
> > + break;
> > + case 's':
> > + source = optarg;
> > + break;
> > + case 'v':
> > + verbose = VERBOSE;
> > + break;
> > + default:
> > + exit(1);
> > + }
> > + }
> > + argv += optind;
> > + argc -= optind;
> > + if ((list == NULL && argc < 1) || (list != NULL && argc > 0)) {
> > + fprintf(stderr, usage_message, program, program, program, usage_extra);
> > + exit(1);
> > + }
> > +
> > + /*
> > + * If C_TAP_VERBOSE is set in the environment, that also turns on verbose
> > + * mode.
> > + */
> > + if (getenv("C_TAP_VERBOSE") != NULL)
> > + verbose = VERBOSE;
> > +
> > + /*
> > + * Set C_TAP_SOURCE and C_TAP_BUILD environment variables. Also set
> > + * SOURCE and BUILD for backward compatibility, although we're trying to
> > + * migrate to the ones with a C_TAP_* prefix.
> > + */
> > + if (source != NULL) {
> > + c_tap_source_env = concat("C_TAP_SOURCE=", source, (const char *) 0);
> > + if (putenv(c_tap_source_env) != 0)
> > + sysdie("cannot set C_TAP_SOURCE in the environment");
> > + source_env = concat("SOURCE=", source, (const char *) 0);
> > + if (putenv(source_env) != 0)
> > + sysdie("cannot set SOURCE in the environment");
> > + }
> > + if (build != NULL) {
> > + c_tap_build_env = concat("C_TAP_BUILD=", build, (const char *) 0);
> > + if (putenv(c_tap_build_env) != 0)
> > + sysdie("cannot set C_TAP_BUILD in the environment");
> > + build_env = concat("BUILD=", build, (const char *) 0);
> > + if (putenv(build_env) != 0)
> > + sysdie("cannot set BUILD in the environment");
> > + }
> > +
> > + /* Run the tests as instructed. */
> > + if (single)
> > + test_single(argv[0], source, build);
> > + else if (list != NULL) {
> > + shortlist = strrchr(list, '/');
> > + if (shortlist == NULL)
> > + shortlist = list;
> > + else
> > + shortlist++;
> > + printf(banner, shortlist);
> > + tests = read_test_list(list, source, build);
> > + status = test_batch(tests, verbose) ? 0 : 1;
> > + } else {
> > + tests = build_test_list(argv, argc, source, build);
> > + status = test_batch(tests, verbose) ? 0 : 1;
> > + }
> > +
> > + /* For valgrind cleanliness, free all our memory. */
> > + if (source_env != NULL) {
> > + putenv((char *) "C_TAP_SOURCE=");
> > + putenv((char *) "SOURCE=");
> > + free(c_tap_source_env);
> > + free(source_env);
> > + }
> > + if (build_env != NULL) {
> > + putenv((char *) "C_TAP_BUILD=");
> > + putenv((char *) "BUILD=");
> > + free(c_tap_build_env);
> > + free(build_env);
> > + }
> > + exit(status);
> > +}
> > \ No newline at end of file
> > diff --git a/t/tap/basic.c b/t/tap/basic.c
> > new file mode 100644
> > index 0000000000..704282b9c1
> > --- /dev/null
> > +++ b/t/tap/basic.c
> > @@ -0,0 +1,1029 @@
> > +/*
> > + * Some utility routines for writing tests.
> > + *
> > + * Here are a variety of utility routines for writing tests compatible with
> > + * the TAP protocol. All routines of the form ok() or is*() take a test
> > + * number and some number of appropriate arguments, check to be sure the
> > + * results match the expected output using the arguments, and print out
> > + * something appropriate for that test number. Other utility routines help in
> > + * constructing more complex tests, skipping tests, reporting errors, setting
> > + * up the TAP output format, or finding things in the test environment.
> > + *
> > + * This file is part of C TAP Harness. The current version plus supporting
> > + * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> > + *
> > + * Written by Russ Allbery <eagle@eyrie.org>
> > + * Copyright 2009-2019, 2021 Russ Allbery <eagle@eyrie.org>
> > + * Copyright 2001-2002, 2004-2008, 2011-2014
> > + * The Board of Trustees of the Leland Stanford Junior University
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the "Software"),
> > + * to deal in the Software without restriction, including without limitation
> > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice shall be included in
> > + * all copies or substantial portions of the Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> > + * DEALINGS IN THE SOFTWARE.
> > + *
> > + * SPDX-License-Identifier: MIT
> > + */
> > +
> > +#include <errno.h>
> > +#include <limits.h>
> > +#include <stdarg.h>
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +#include <string.h>
> > +#ifdef _WIN32
> > +# include <direct.h>
> > +#else
> > +# include <sys/stat.h>
> > +#endif
> > +#include <sys/types.h>
> > +#include <unistd.h>
> > +
> > +#include <tap/basic.h>
> > +
> > +/* Windows provides mkdir and rmdir under different names. */
> > +#ifdef _WIN32
> > +# define mkdir(p, m) _mkdir(p)
> > +# define rmdir(p) _rmdir(p)
> > +#endif
> > +
> > +/*
> > + * The test count. Always contains the number that will be used for the next
> > + * test status. This is exported to callers of the library.
> > + */
> > +unsigned long testnum = 1;
> > +
> > +/*
> > + * Status information stored so that we can give a test summary at the end of
> > + * the test case. We store the planned final test and the count of failures.
> > + * We can get the highest test count from testnum.
> > + */
> > +static unsigned long _planned = 0;
> > +static unsigned long _failed = 0;
> > +
> > +/*
> > + * Store the PID of the process that called plan() and only summarize
> > + * results when that process exits, so as to not misreport results in forked
> > + * processes.
> > + */
> > +static pid_t _process = 0;
> > +
> > +/*
> > + * If true, we're doing lazy planning and will print out the plan based on the
> > + * last test number at the end of testing.
> > + */
> > +static int _lazy = 0;
> > +
> > +/*
> > + * If true, the test was aborted by calling bail(). Currently, this is only
> > + * used to ensure that we pass a false value to any cleanup functions even if
> > + * all tests to that point have passed.
> > + */
> > +static int _aborted = 0;
> > +
> > +/*
> > + * Registered cleanup functions. These are stored as a linked list and run in
> > + * registered order by finish when the test program exits. Each function is
> > + * passed a boolean value indicating whether all tests were successful.
> > + */
> > +struct cleanup_func {
> > + test_cleanup_func func;
> > + test_cleanup_func_with_data func_with_data;
> > + void *data;
> > + struct cleanup_func *next;
> > +};
> > +static struct cleanup_func *cleanup_funcs = NULL;
> > +
> > +/*
> > + * Registered diag files. Any output found in these files will be printed out
> > + * as if it were passed to diag() before any other output we do. This allows
> > + * background processes to log to a file and have that output interleaved with
> > + * the test output.
> > + */
> > +struct diag_file {
> > + char *name;
> > + FILE *file;
> > + char *buffer;
> > + size_t bufsize;
> > + struct diag_file *next;
> > +};
> > +static struct diag_file *diag_files = NULL;
> > +
> > +/*
> > + * Print a specified prefix and then the test description. Handles turning
> > + * the argument list into a va_args structure suitable for passing to
> > + * print_desc, which has to be done in a macro. Assumes that format is the
> > + * argument immediately before the variadic arguments.
> > + */
> > +#define PRINT_DESC(prefix, format) \
> > + do { \
> > + if (format != NULL) { \
> > + va_list args; \
> > + printf("%s", prefix); \
> > + va_start(args, format); \
> > + vprintf(format, args); \
> > + va_end(args); \
> > + } \
> > + } while (0)
> > +
> > +
> > +/*
> > + * Form a new string by concatenating multiple strings. The arguments must be
> > + * terminated by (const char *) 0.
> > + *
> > + * This function only exists because we can't assume asprintf. We can't
> > + * simulate asprintf with snprintf because we're only assuming SUSv3, which
> > + * does not require that snprintf with a NULL buffer return the required
> > + * length. When those constraints are relaxed, this should be ripped out and
> > + * replaced with asprintf or a more trivial replacement with snprintf.
> > + */
> > +static char *
> > +concat(const char *first, ...)
> > +{
> > + va_list args;
> > + char *result;
> > + const char *string;
> > + size_t offset;
> > + size_t length = 0;
> > +
> > + /*
> > + * Find the total memory required. Ensure we don't overflow length. See
> > + * the comment for breallocarray for why we're using UINT_MAX here.
> > + */
> > + va_start(args, first);
> > + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> > + if (length >= UINT_MAX - strlen(string))
> > + bail("strings too long in concat");
> > + length += strlen(string);
> > + }
> > + va_end(args);
> > + length++;
> > +
> > + /* Create the string. */
> > + result = bcalloc_type(length, char);
> > + va_start(args, first);
> > + offset = 0;
> > + for (string = first; string != NULL; string = va_arg(args, const char *)) {
> > + memcpy(result + offset, string, strlen(string));
> > + offset += strlen(string);
> > + }
> > + va_end(args);
> > + result[offset] = '\0';
> > + return result;
> > +}
> > +
> > +
> > +/*
> > + * Helper function for check_diag_files to handle a single line in a diag
> > + * file.
> > + *
> > + * The general scheme here used is as follows: read one line of output. If we
> > + * get NULL, check for an error. If there was one, bail out of the test
> > + * program; otherwise, return, and the enclosing loop will check for EOF.
> > + *
> > + * If we get some data, see if it ends in a newline. If it doesn't end in a
> > + * newline, we have one of two cases: our buffer isn't large enough, in which
> > + * case we resize it and try again, or we have incomplete data in the file, in
> > + * which case we rewind the file and will try again next time.
> > + *
> > + * Returns a boolean indicating whether the last line was incomplete.
> > + */
> > +static int
> > +handle_diag_file_line(struct diag_file *file, fpos_t where)
> > +{
> > + int size;
> > + size_t length;
> > +
> > + /* Read the next line from the file. */
> > + size = file->bufsize > INT_MAX ? INT_MAX : (int) file->bufsize;
> > + if (fgets(file->buffer, size, file->file) == NULL) {
> > + if (ferror(file->file))
> > + sysbail("cannot read from %s", file->name);
> > + return 0;
> > + }
> > +
> > + /*
> > + * See if the line ends in a newline. If not, see which error case we
> > + * have.
> > + */
> > + length = strlen(file->buffer);
> > + if (file->buffer[length - 1] != '\n') {
> > + int incomplete = 0;
> > +
> > + /* Check whether we ran out of buffer space and resize if so. */
> > + if (length < file->bufsize - 1)
> > + incomplete = 1;
> > + else {
> > + file->bufsize += BUFSIZ;
> > + file->buffer =
> > + breallocarray_type(file->buffer, file->bufsize, char);
> > + }
> > +
> > + /*
> > + * On either incomplete lines or too small of a buffer, rewind
> > + * and read the file again (on the next pass, if incomplete).
> > + * It's simpler than trying to double-buffer the file.
> > + */
> > + if (fsetpos(file->file, &where) < 0)
> > + sysbail("cannot set position in %s", file->name);
> > + return incomplete;
> > + }
> > +
> > + /* We saw a complete line. Print it out. */
> > + printf("# %s", file->buffer);
> > + return 0;
> > +}
> > +
> > +
> > +/*
> > + * Check all registered diag_files for any output. We only print out the
> > + * output if we see a complete line; otherwise, we wait for the next newline.
> > + */
> > +static void
> > +check_diag_files(void)
> > +{
> > + struct diag_file *file;
> > + fpos_t where;
> > + int incomplete;
> > +
> > + /*
> > + * Walk through each file and read each line of output available.
> > + */
> > + for (file = diag_files; file != NULL; file = file->next) {
> > + clearerr(file->file);
> > +
> > + /* Store the current position in case we have to rewind. */
> > + if (fgetpos(file->file, &where) < 0)
> > + sysbail("cannot get position in %s", file->name);
> > +
> > + /* Continue until we get EOF or an incomplete line of data. */
> > + incomplete = 0;
> > + while (!feof(file->file) && !incomplete) {
> > + incomplete = handle_diag_file_line(file, where);
> > + }
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Our exit handler. Called on completion of the test to report a summary of
> > + * results provided we're still in the original process. This also handles
> > + * printing out the plan if we used plan_lazy(), although that's suppressed if
> > + * we never ran a test (due to an early bail, for example), and running any
> > + * registered cleanup functions.
> > + */
> > +static void
> > +finish(void)
> > +{
> > + int success, primary;
> > + struct cleanup_func *current;
> > + unsigned long highest = testnum - 1;
> > + struct diag_file *file, *tmp;
> > +
> > + /* Check for pending diag_file output. */
> > + check_diag_files();
> > +
> > + /* Free the diag_files. */
> > + file = diag_files;
> > + while (file != NULL) {
> > + tmp = file;
> > + file = file->next;
> > + fclose(tmp->file);
> > + free(tmp->name);
> > + free(tmp->buffer);
> > + free(tmp);
> > + }
> > + diag_files = NULL;
> > +
> > + /*
> > + * Determine whether all tests were successful, which is needed before
> > + * calling cleanup functions since we pass that fact to the functions.
> > + */
> > + if (_planned == 0 && _lazy)
> > + _planned = highest;
> > + success = (!_aborted && _planned == highest && _failed == 0);
> > +
> > + /*
> > + * If there are any registered cleanup functions, we run those first. We
> > + * always run them, even if we didn't run a test. Don't do anything
> > + * except free the diag_files and call cleanup functions if we aren't the
> > + * primary process (the process in which plan or plan_lazy was called),
> > + * and tell the cleanup functions that fact.
> > + */
> > + primary = (_process == 0 || getpid() == _process);
> > + while (cleanup_funcs != NULL) {
> > + if (cleanup_funcs->func_with_data) {
> > + void *data = cleanup_funcs->data;
> > +
> > + cleanup_funcs->func_with_data(success, primary, data);
> > + } else {
> > + cleanup_funcs->func(success, primary);
> > + }
> > + current = cleanup_funcs;
> > + cleanup_funcs = cleanup_funcs->next;
> > + free(current);
> > + }
> > + if (!primary)
> > + return;
> > +
> > + /* Don't do anything further if we never planned a test. */
> > + if (_planned == 0)
> > + return;
> > +
> > + /* If we're aborting due to bail, don't print summaries. */
> > + if (_aborted)
> > + return;
> > +
> > + /* Print out the lazy plan if needed. */
> > + fflush(stderr);
> > + if (_lazy)
> > + printf("1..%lu\n", _planned);
> > +
> > + /* Print out a summary of the results. */
> > + if (_planned > highest)
> > + diag("Looks like you planned %lu test%s but only ran %lu", _planned,
> > + (_planned > 1 ? "s" : ""), highest);
> > + else if (_planned < highest)
> > + diag("Looks like you planned %lu test%s but ran %lu extra", _planned,
> > + (_planned > 1 ? "s" : ""), highest - _planned);
> > + else if (_failed > 0)
> > + diag("Looks like you failed %lu test%s of %lu", _failed,
> > + (_failed > 1 ? "s" : ""), _planned);
> > + else if (_planned != 1)
> > + diag("All %lu tests successful or skipped", _planned);
> > + else
> > + diag("%lu test successful or skipped", _planned);
> > +}
> > +
> > +
> > +/*
> > + * Initialize things. Turns on line buffering on stdout and then prints out
> > + * the number of tests in the test suite. We intentionally don't check for
> > + * pending diag_file output here, since it should really come after the plan.
> > + */
> > +void
> > +plan(unsigned long count)
> > +{
> > + if (setvbuf(stdout, NULL, _IOLBF, BUFSIZ) != 0)
> > + sysdiag("cannot set stdout to line buffered");
> > + fflush(stderr);
> > + printf("1..%lu\n", count);
> > + testnum = 1;
> > + _planned = count;
> > + _process = getpid();
> > + if (atexit(finish) != 0) {
> > + sysdiag("cannot register exit handler");
> > + diag("cleanups will not be run");
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Initialize things for lazy planning, where we'll automatically print out a
> > + * plan at the end of the program. Turns on line buffering on stdout as well.
> > + */
> > +void
> > +plan_lazy(void)
> > +{
> > + if (setvbuf(stdout, NULL, _IOLBF, BUFSIZ) != 0)
> > + sysdiag("cannot set stdout to line buffered");
> > + testnum = 1;
> > + _process = getpid();
> > + _lazy = 1;
> > + if (atexit(finish) != 0)
> > + sysbail("cannot register exit handler to display plan");
> > +}
> > +
> > +
> > +/*
> > + * Skip the entire test suite and exits. Should be called instead of plan(),
> > + * not after it, since it prints out a special plan line. Ignore diag_file
> > + * output here, since it's not clear if it's allowed before the plan.
> > + */
> > +void
> > +skip_all(const char *format, ...)
> > +{
> > + fflush(stderr);
> > + printf("1..0 # skip");
> > + PRINT_DESC(" ", format);
> > + putchar('\n');
> > + exit(0);
> > +}
> > +
> > +
> > +/*
> > + * Takes a boolean success value and assumes the test passes if that value
> > + * is true and fails if that value is false.
> > + */
> > +int
> > +ok(int success, const char *format, ...)
> > +{
> > + fflush(stderr);
> > + check_diag_files();
> > + printf("%sok %lu", success ? "" : "not ", testnum++);
> > + if (!success)
> > + _failed++;
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Same as ok(), but takes the format arguments as a va_list.
> > + */
> > +int
> > +okv(int success, const char *format, va_list args)
> > +{
> > + fflush(stderr);
> > + check_diag_files();
> > + printf("%sok %lu", success ? "" : "not ", testnum++);
> > + if (!success)
> > + _failed++;
> > + if (format != NULL) {
> > + printf(" - ");
> > + vprintf(format, args);
> > + }
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Skip a test.
> > + */
> > +void
> > +skip(const char *reason, ...)
> > +{
> > + fflush(stderr);
> > + check_diag_files();
> > + printf("ok %lu # skip", testnum++);
> > + PRINT_DESC(" ", reason);
> > + putchar('\n');
> > +}
> > +
> > +
> > +/*
> > + * Report the same status on the next count tests.
> > + */
> > +int
> > +ok_block(unsigned long count, int success, const char *format, ...)
> > +{
> > + unsigned long i;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + for (i = 0; i < count; i++) {
> > + printf("%sok %lu", success ? "" : "not ", testnum++);
> > + if (!success)
> > + _failed++;
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + }
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Skip the next count tests.
> > + */
> > +void
> > +skip_block(unsigned long count, const char *reason, ...)
> > +{
> > + unsigned long i;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + for (i = 0; i < count; i++) {
> > + printf("ok %lu # skip", testnum++);
> > + PRINT_DESC(" ", reason);
> > + putchar('\n');
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Takes two boolean values and requires the truth value of both match.
> > + */
> > +int
> > +is_bool(int left, int right, const char *format, ...)
> > +{
> > + int success;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + success = (!!left == !!right);
> > + if (success)
> > + printf("ok %lu", testnum++);
> > + else {
> > + diag(" left: %s", !!left ? "true" : "false");
> > + diag("right: %s", !!right ? "true" : "false");
> > + printf("not ok %lu", testnum++);
> > + _failed++;
> > + }
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Takes two integer values and requires they match.
> > + */
> > +int
> > +is_int(long left, long right, const char *format, ...)
> > +{
> > + int success;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + success = (left == right);
> > + if (success)
> > + printf("ok %lu", testnum++);
> > + else {
> > + diag(" left: %ld", left);
> > + diag("right: %ld", right);
> > + printf("not ok %lu", testnum++);
> > + _failed++;
> > + }
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Takes two strings and requires they match (using strcmp). NULL arguments
> > + * are permitted and handled correctly.
> > + */
> > +int
> > +is_string(const char *left, const char *right, const char *format, ...)
> > +{
> > + int success;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > +
> > + /* Compare the strings, being careful of NULL. */
> > + if (left == NULL)
> > + success = (right == NULL);
> > + else if (right == NULL)
> > + success = 0;
> > + else
> > + success = (strcmp(left, right) == 0);
> > +
> > + /* Report the results. */
> > + if (success)
> > + printf("ok %lu", testnum++);
> > + else {
> > + diag(" left: %s", left == NULL ? "(null)" : left);
> > + diag("right: %s", right == NULL ? "(null)" : right);
> > + printf("not ok %lu", testnum++);
> > + _failed++;
> > + }
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Takes two unsigned longs and requires they match. On failure, reports them
> > + * in hex.
> > + */
> > +int
> > +is_hex(unsigned long left, unsigned long right, const char *format, ...)
> > +{
> > + int success;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + success = (left == right);
> > + if (success)
> > + printf("ok %lu", testnum++);
> > + else {
> > + diag(" left: %lx", (unsigned long) left);
> > + diag("right: %lx", (unsigned long) right);
> > + printf("not ok %lu", testnum++);
> > + _failed++;
> > + }
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Takes pointers to a regions of memory and requires that len bytes from each
> > + * match. Otherwise reports any bytes which didn't match.
> > + */
> > +int
> > +is_blob(const void *left, const void *right, size_t len, const char *format,
> > + ...)
> > +{
> > + int success;
> > + size_t i;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + success = (memcmp(left, right, len) == 0);
> > + if (success)
> > + printf("ok %lu", testnum++);
> > + else {
> > + const unsigned char *left_c = (const unsigned char *) left;
> > + const unsigned char *right_c = (const unsigned char *) right;
> > +
> > + for (i = 0; i < len; i++) {
> > + if (left_c[i] != right_c[i])
> > + diag("offset %lu: left %02x, right %02x", (unsigned long) i,
> > + left_c[i], right_c[i]);
> > + }
> > + printf("not ok %lu", testnum++);
> > + _failed++;
> > + }
> > + PRINT_DESC(" - ", format);
> > + putchar('\n');
> > + return success;
> > +}
> > +
> > +
> > +/*
> > + * Bail out with an error.
> > + */
> > +void
> > +bail(const char *format, ...)
> > +{
> > + va_list args;
> > +
> > + _aborted = 1;
> > + fflush(stderr);
> > + check_diag_files();
> > + fflush(stdout);
> > + printf("Bail out! ");
> > + va_start(args, format);
> > + vprintf(format, args);
> > + va_end(args);
> > + printf("\n");
> > + exit(255);
> > +}
> > +
> > +
> > +/*
> > + * Bail out with an error, appending strerror(errno).
> > + */
> > +void
> > +sysbail(const char *format, ...)
> > +{
> > + va_list args;
> > + int oerrno = errno;
> > +
> > + _aborted = 1;
> > + fflush(stderr);
> > + check_diag_files();
> > + fflush(stdout);
> > + printf("Bail out! ");
> > + va_start(args, format);
> > + vprintf(format, args);
> > + va_end(args);
> > + printf(": %s\n", strerror(oerrno));
> > + exit(255);
> > +}
> > +
> > +
> > +/*
> > + * Report a diagnostic to stderr. Always returns 1 to allow embedding in
> > + * compound statements.
> > + */
> > +int
> > +diag(const char *format, ...)
> > +{
> > + va_list args;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + fflush(stdout);
> > + printf("# ");
> > + va_start(args, format);
> > + vprintf(format, args);
> > + va_end(args);
> > + printf("\n");
> > + return 1;
> > +}
> > +
> > +
> > +/*
> > + * Report a diagnostic to stderr, appending strerror(errno). Always returns 1
> > + * to allow embedding in compound statements.
> > + */
> > +int
> > +sysdiag(const char *format, ...)
> > +{
> > + va_list args;
> > + int oerrno = errno;
> > +
> > + fflush(stderr);
> > + check_diag_files();
> > + fflush(stdout);
> > + printf("# ");
> > + va_start(args, format);
> > + vprintf(format, args);
> > + va_end(args);
> > + printf(": %s\n", strerror(oerrno));
> > + return 1;
> > +}
> > +
> > +
> > +/*
> > + * Register a new file for diag_file processing.
> > + */
> > +void
> > +diag_file_add(const char *name)
> > +{
> > + struct diag_file *file, *prev;
> > +
> > + file = bcalloc_type(1, struct diag_file);
> > + file->name = bstrdup(name);
> > + file->file = fopen(file->name, "r");
> > + if (file->file == NULL)
> > + sysbail("cannot open %s", name);
> > + file->buffer = bcalloc_type(BUFSIZ, char);
> > + file->bufsize = BUFSIZ;
> > + if (diag_files == NULL)
> > + diag_files = file;
> > + else {
> > + for (prev = diag_files; prev->next != NULL; prev = prev->next)
> > + ;
> > + prev->next = file;
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Remove a file from diag_file processing. If the file is not found, do
> > + * nothing, since there are some situations where it can be removed twice
> > + * (such as if it's removed from a cleanup function, since cleanup functions
> > + * are called after freeing all the diag_files).
> > + */
> > +void
> > +diag_file_remove(const char *name)
> > +{
> > + struct diag_file *file;
> > + struct diag_file **prev = &diag_files;
> > +
> > + for (file = diag_files; file != NULL; file = file->next) {
> > + if (strcmp(file->name, name) == 0) {
> > + *prev = file->next;
> > + fclose(file->file);
> > + free(file->name);
> > + free(file->buffer);
> > + free(file);
> > + return;
> > + }
> > + prev = &file->next;
> > + }
> > +}
> > +
> > +
> > +/*
> > + * Allocate cleared memory, reporting a fatal error with bail on failure.
> > + */
> > +void *
> > +bcalloc(size_t n, size_t size)
> > +{
> > + void *p;
> > +
> > + p = calloc(n, size);
> > + if (p == NULL)
> > + sysbail("failed to calloc %lu", (unsigned long) (n * size));
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Allocate memory, reporting a fatal error with bail on failure.
> > + */
> > +void *
> > +bmalloc(size_t size)
> > +{
> > + void *p;
> > +
> > + p = malloc(size);
> > + if (p == NULL)
> > + sysbail("failed to malloc %lu", (unsigned long) size);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Reallocate memory, reporting a fatal error with bail on failure.
> > + */
> > +void *
> > +brealloc(void *p, size_t size)
> > +{
> > + p = realloc(p, size);
> > + if (p == NULL)
> > + sysbail("failed to realloc %lu bytes", (unsigned long) size);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * The same as brealloc, but determine the size by multiplying an element
> > + * count by a size, similar to calloc. The multiplication is checked for
> > + * integer overflow.
> > + *
> > + * We should technically use SIZE_MAX here for the overflow check, but
> > + * SIZE_MAX is C99 and we're only assuming C89 + SUSv3, which does not
> > + * guarantee that it exists. They do guarantee that UINT_MAX exists, and we
> > + * can assume that UINT_MAX <= SIZE_MAX.
> > + *
> > + * (In theory, C89 and C99 permit size_t to be smaller than unsigned int, but
> > + * I disbelieve in the existence of such systems and they will have to cope
> > + * without overflow checks.)
> > + */
> > +void *
> > +breallocarray(void *p, size_t n, size_t size)
> > +{
> > + if (n > 0 && UINT_MAX / n <= size)
> > + bail("reallocarray too large");
> > + if (n == 0)
> > + n = 1;
> > + p = realloc(p, n * size);
> > + if (p == NULL)
> > + sysbail("failed to realloc %lu bytes", (unsigned long) (n * size));
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Copy a string, reporting a fatal error with bail on failure.
> > + */
> > +char *
> > +bstrdup(const char *s)
> > +{
> > + char *p;
> > + size_t len;
> > +
> > + len = strlen(s) + 1;
> > + p = (char *) malloc(len);
> > + if (p == NULL)
> > + sysbail("failed to strdup %lu bytes", (unsigned long) len);
> > + memcpy(p, s, len);
> > + return p;
> > +}
> > +
> > +
> > +/*
> > + * Copy up to n characters of a string, reporting a fatal error with bail on
> > + * failure. Don't use the system strndup function, since it may not exist and
> > + * the TAP library doesn't assume any portability support.
> > + */
> > +char *
> > +bstrndup(const char *s, size_t n)
> > +{
> > + const char *p;
> > + char *copy;
> > + size_t length;
> > +
> > + /* Don't assume that the source string is nul-terminated. */
> > + for (p = s; (size_t) (p - s) < n && *p != '\0'; p++)
> > + ;
> > + length = (size_t) (p - s);
> > + copy = (char *) malloc(length + 1);
> > + if (copy == NULL)
> > + sysbail("failed to strndup %lu bytes", (unsigned long) length);
> > + memcpy(copy, s, length);
> > + copy[length] = '\0';
> > + return copy;
> > +}
> > +
> > +
> > +/*
> > + * Locate a test file. Given the partial path to a file, look under
> > + * C_TAP_BUILD and then C_TAP_SOURCE for the file and return the full path to
> > + * the file. Returns NULL if the file doesn't exist. A non-NULL return
> > + * should be freed with test_file_path_free().
> > + */
> > +char *
> > +test_file_path(const char *file)
> > +{
> > + char *base;
> > + char *path = NULL;
> > + const char *envs[] = {"C_TAP_BUILD", "C_TAP_SOURCE", NULL};
> > + int i;
> > +
> > + for (i = 0; envs[i] != NULL; i++) {
> > + base = getenv(envs[i]);
> > + if (base == NULL)
> > + continue;
> > + path = concat(base, "/", file, (const char *) 0);
> > + if (access(path, R_OK) == 0)
> > + break;
> > + free(path);
> > + path = NULL;
> > + }
> > + return path;
> > +}
> > +
> > +
> > +/*
> > + * Free a path returned from test_file_path(). This function exists primarily
> > + * for Windows, where memory must be freed from the same library domain that
> > + * it was allocated from.
> > + */
> > +void
> > +test_file_path_free(char *path)
> > +{
> > + free(path);
> > +}
> > +
> > +
> > +/*
> > + * Create a temporary directory, tmp, under C_TAP_BUILD if set and the current
> > + * directory if it does not. Returns the path to the temporary directory in
> > + * newly allocated memory, and calls bail on any failure. The return value
> > + * should be freed with test_tmpdir_free.
> > + *
> > + * This function uses sprintf because it attempts to be independent of all
> > + * other portability layers. The use immediately after a memory allocation
> > + * should be safe without using snprintf or strlcpy/strlcat.
> > + */
> > +char *
> > +test_tmpdir(void)
> > +{
> > + const char *build;
> > + char *path = NULL;
> > +
> > + build = getenv("C_TAP_BUILD");
> > + if (build == NULL)
> > + build = ".";
> > + path = concat(build, "/tmp", (const char *) 0);
> > + if (access(path, X_OK) < 0)
> > + if (mkdir(path, 0777) < 0)
> > + sysbail("error creating temporary directory %s", path);
> > + return path;
> > +}
> > +
> > +
> > +/*
> > + * Free a path returned from test_tmpdir() and attempt to remove the
> > + * directory. If we can't delete the directory, don't worry; something else
> > + * that hasn't yet cleaned up may still be using it.
> > + */
> > +void
> > +test_tmpdir_free(char *path)
> > +{
> > + if (path != NULL)
> > + rmdir(path);
> > + free(path);
> > +}
> > +
> > +static void
> > +register_cleanup(test_cleanup_func func,
> > + test_cleanup_func_with_data func_with_data, void *data)
> > +{
> > + struct cleanup_func *cleanup, **last;
> > +
> > + cleanup = bcalloc_type(1, struct cleanup_func);
> > + cleanup->func = func;
> > + cleanup->func_with_data = func_with_data;
> > + cleanup->data = data;
> > + cleanup->next = NULL;
> > + last = &cleanup_funcs;
> > + while (*last != NULL)
> > + last = &(*last)->next;
> > + *last = cleanup;
> > +}
> > +
> > +/*
> > + * Register a cleanup function that is called when testing ends. All such
> > + * registered functions will be run by finish.
> > + */
> > +void
> > +test_cleanup_register(test_cleanup_func func)
> > +{
> > + register_cleanup(func, NULL, NULL);
> > +}
> > +
> > +/*
> > + * Same as above, but also allows an opaque pointer to be passed to the cleanup
> > + * function.
> > + */
> > +void
> > +test_cleanup_register_with_data(test_cleanup_func_with_data func, void *data)
> > +{
> > + register_cleanup(NULL, func, data);
> > +}
> > diff --git a/t/tap/basic.h b/t/tap/basic.h
> > new file mode 100644
> > index 0000000000..afea8cb210
> > --- /dev/null
> > +++ b/t/tap/basic.h
> > @@ -0,0 +1,198 @@
> > +/*
> > + * Basic utility routines for the TAP protocol.
> > + *
> > + * This file is part of C TAP Harness. The current version plus supporting
> > + * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> > + *
> > + * Written by Russ Allbery <eagle@eyrie.org>
> > + * Copyright 2009-2019, 2022 Russ Allbery <eagle@eyrie.org>
> > + * Copyright 2001-2002, 2004-2008, 2011-2012, 2014
> > + * The Board of Trustees of the Leland Stanford Junior University
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the "Software"),
> > + * to deal in the Software without restriction, including without limitation
> > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice shall be included in
> > + * all copies or substantial portions of the Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> > + * DEALINGS IN THE SOFTWARE.
> > + *
> > + * SPDX-License-Identifier: MIT
> > + */
> > +
> > +#ifndef TAP_BASIC_H
> > +#define TAP_BASIC_H 1
> > +
> > +#include <stdarg.h> /* va_list */
> > +#include <stddef.h> /* size_t */
> > +#include <stdlib.h> /* free */
> > +#include <tap/macros.h>
> > +
> > +/*
> > + * Used for iterating through arrays. ARRAY_SIZE returns the number of
> > + * elements in the array (useful for a < upper bound in a for loop) and
> > + * ARRAY_END returns a pointer to the element past the end (ISO C99 makes it
> > + * legal to refer to such a pointer as long as it's never dereferenced).
> > + */
> > +// #define ARRAY_SIZE(array) (sizeof(array) / sizeof((array)[0]))
> > +// #define ARRAY_END(array) (&(array)[ARRAY_SIZE(array)])
> > +
> > +BEGIN_DECLS
> > +
> > +/*
> > + * The test count. Always contains the number that will be used for the next
> > + * test status.
> > + */
> > +extern unsigned long testnum;
> > +
> > +/* Print out the number of tests and set standard output to line buffered. */
> > +void plan(unsigned long count);
> > +
> > +/*
> > + * Prepare for lazy planning, in which the plan will be printed automatically
> > + * at the end of the test program.
> > + */
> > +void plan_lazy(void);
> > +
> > +/* Skip the entire test suite. Call instead of plan. */
> > +void skip_all(const char *format, ...)
> > + __attribute__((__noreturn__, __format__(printf, 1, 2)));
> > +
> > +/*
> > + * Basic reporting functions. The okv() function is the same as ok() but
> > + * takes the test description as a va_list to make it easier to reuse the
> > + * reporting infrastructure when writing new tests. ok() and okv() return the
> > + * value of the success argument.
> > + */
> > +int ok(int success, const char *format, ...)
> > + __attribute__((__format__(printf, 2, 3)));
> > +int okv(int success, const char *format, va_list args)
> > + __attribute__((__format__(printf, 2, 0)));
> > +void skip(const char *reason, ...) __attribute__((__format__(printf, 1, 2)));
> > +
> > +/*
> > + * Report the same status on, or skip, the next count tests. ok_block()
> > + * returns the value of the success argument.
> > + */
> > +int ok_block(unsigned long count, int success, const char *format, ...)
> > + __attribute__((__format__(printf, 3, 4)));
> > +void skip_block(unsigned long count, const char *reason, ...)
> > + __attribute__((__format__(printf, 2, 3)));
> > +
> > +/*
> > + * Compare two values. Returns true if the test passes and false if it fails.
> > + * is_bool takes an int since the bool type isn't fully portable yet, but
> > + * interprets both arguments for their truth value, not for their numeric
> > + * value.
> > + */
> > +int is_bool(int, int, const char *format, ...)
> > + __attribute__((__format__(printf, 3, 4)));
> > +int is_int(long, long, const char *format, ...)
> > + __attribute__((__format__(printf, 3, 4)));
> > +int is_string(const char *, const char *, const char *format, ...)
> > + __attribute__((__format__(printf, 3, 4)));
> > +int is_hex(unsigned long, unsigned long, const char *format, ...)
> > + __attribute__((__format__(printf, 3, 4)));
> > +int is_blob(const void *, const void *, size_t, const char *format, ...)
> > + __attribute__((__format__(printf, 4, 5)));
> > +
> > +/* Bail out with an error. sysbail appends strerror(errno). */
> > +void bail(const char *format, ...)
> > + __attribute__((__noreturn__, __nonnull__, __format__(printf, 1, 2)));
> > +void sysbail(const char *format, ...)
> > + __attribute__((__noreturn__, __nonnull__, __format__(printf, 1, 2)));
> > +
> > +/* Report a diagnostic to stderr prefixed with #. */
> > +int diag(const char *format, ...)
> > + __attribute__((__nonnull__, __format__(printf, 1, 2)));
> > +int sysdiag(const char *format, ...)
> > + __attribute__((__nonnull__, __format__(printf, 1, 2)));
> > +
> > +/*
> > + * Register or unregister a file that contains supplementary diagnostics.
> > + * Before any other output, all registered files will be read, line by line,
> > + * and each line will be reported as a diagnostic as if it were passed to
> > + * diag(). Nul characters are not supported in these files and will result in
> > + * truncated output.
> > + */
> > +void diag_file_add(const char *file) __attribute__((__nonnull__));
> > +void diag_file_remove(const char *file) __attribute__((__nonnull__));
> > +
> > +/* Allocate memory, reporting a fatal error with bail on failure. */
> > +void *bcalloc(size_t, size_t)
> > + __attribute__((__alloc_size__(1, 2), __malloc__(free),
> > + __warn_unused_result__));
> > +void *bmalloc(size_t) __attribute__((__alloc_size__(1), __malloc__(free),
> > + __warn_unused_result__));
> > +void *breallocarray(void *, size_t, size_t)
> > + __attribute__((__alloc_size__(2, 3), __malloc__(free),
> > + __warn_unused_result__));
> > +void *brealloc(void *, size_t)
> > + __attribute__((__alloc_size__(2), __malloc__(free),
> > + __warn_unused_result__));
> > +char *bstrdup(const char *)
> > + __attribute__((__malloc__(free), __nonnull__, __warn_unused_result__));
> > +char *bstrndup(const char *, size_t)
> > + __attribute__((__malloc__(free), __nonnull__, __warn_unused_result__));
> > +
> > +/*
> > + * Macros that cast the return value from b* memory functions, making them
> > + * usable in C++ code and providing some additional type safety.
> > + */
> > +#define bcalloc_type(n, type) ((type *) bcalloc((n), sizeof(type)))
> > +#define breallocarray_type(p, n, type) \
> > + ((type *) breallocarray((p), (n), sizeof(type)))
> > +
> > +/*
> > + * Find a test file under C_TAP_BUILD or C_TAP_SOURCE, returning the full
> > + * path. The returned path should be freed with test_file_path_free().
> > + */
> > +void test_file_path_free(char *path);
> > +char *test_file_path(const char *file)
> > + __attribute__((__malloc__(test_file_path_free), __nonnull__,
> > + __warn_unused_result__));
> > +
> > +/*
> > + * Create a temporary directory relative to C_TAP_BUILD and return the path.
> > + * The returned path should be freed with test_tmpdir_free().
> > + */
> > +void test_tmpdir_free(char *path);
> > +char *test_tmpdir(void)
> > + __attribute__((__malloc__(test_tmpdir_free), __warn_unused_result__));
> > +
> > +/*
> > + * Register a cleanup function that is called when testing ends. All such
> > + * registered functions will be run during atexit handling (and are therefore
> > + * subject to all the same constraints and caveats as atexit functions).
> > + *
> > + * The function must return void and will be passed two arguments: an int that
> > + * will be true if the test completed successfully and false otherwise, and an
> > + * int that will be true if the cleanup function is run in the primary process
> > + * (the one that called plan or plan_lazy) and false otherwise. If
> > + * test_cleanup_register_with_data is used instead, a generic pointer can be
> > + * provided and will be passed to the cleanup function as a third argument.
> > + *
> > + * test_cleanup_register_with_data is the better API and should have been the
> > + * only API. test_cleanup_register was an API error preserved for backward
> > + * cmpatibility.
> > + */
> > +typedef void (*test_cleanup_func)(int, int);
> > +typedef void (*test_cleanup_func_with_data)(int, int, void *);
> > +
> > +void test_cleanup_register(test_cleanup_func) __attribute__((__nonnull__));
> > +void test_cleanup_register_with_data(test_cleanup_func_with_data, void *)
> > + __attribute__((__nonnull__));
> > +
> > +END_DECLS
> > +
> > +#endif /* TAP_BASIC_H */
> > diff --git a/t/tap/macros.h b/t/tap/macros.h
> > new file mode 100644
> > index 0000000000..0eabcb5847
> > --- /dev/null
> > +++ b/t/tap/macros.h
> > @@ -0,0 +1,109 @@
> > +/*
> > + * Helpful macros for TAP header files.
> > + *
> > + * This is not, strictly speaking, related to TAP, but any TAP add-on is
> > + * probably going to need these macros, so define them in one place so that
> > + * everyone can pull them in.
> > + *
> > + * This file is part of C TAP Harness. The current version plus supporting
> > + * documentation is at <https://www.eyrie.org/~eagle/software/c-tap-harness/>.
> > + *
> > + * Copyright 2008, 2012-2013, 2015, 2022 Russ Allbery <eagle@eyrie.org>
> > + *
> > + * Permission is hereby granted, free of charge, to any person obtaining a
> > + * copy of this software and associated documentation files (the "Software"),
> > + * to deal in the Software without restriction, including without limitation
> > + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> > + * and/or sell copies of the Software, and to permit persons to whom the
> > + * Software is furnished to do so, subject to the following conditions:
> > + *
> > + * The above copyright notice and this permission notice shall be included in
> > + * all copies or substantial portions of the Software.
> > + *
> > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
> > + * DEALINGS IN THE SOFTWARE.
> > + *
> > + * SPDX-License-Identifier: MIT
> > + */
> > +
> > +#ifndef TAP_MACROS_H
> > +#define TAP_MACROS_H 1
> > +
> > +/*
> > + * __attribute__ is available in gcc 2.5 and later, but only with gcc 2.7
> > + * could you use the __format__ form of the attributes, which is what we use
> > + * (to avoid confusion with other macros), and only with gcc 2.96 can you use
> > + * the attribute __malloc__. 2.96 is very old, so don't bother trying to get
> > + * the other attributes to work with GCC versions between 2.7 and 2.96.
> > + */
> > +#ifndef __attribute__
> > +# if __GNUC__ < 2 || (__GNUC__ == 2 && __GNUC_MINOR__ < 96)
> > +# define __attribute__(spec) /* empty */
> > +# endif
> > +#endif
> > +
> > +/*
> > + * We use __alloc_size__, but it was only available in fairly recent versions
> > + * of GCC. Suppress warnings about the unknown attribute if GCC is too old.
> > + * We know that we're GCC at this point, so we can use the GCC variadic macro
> > + * extension, which will still work with versions of GCC too old to have C99
> > + * variadic macro support.
> > + */
> > +#if !defined(__attribute__) && !defined(__alloc_size__)
> > +# if defined(__GNUC__) && !defined(__clang__)
> > +# if __GNUC__ < 4 || (__GNUC__ == 4 && __GNUC_MINOR__ < 3)
> > +# define __alloc_size__(spec, args...) /* empty */
> > +# endif
> > +# endif
> > +#endif
> > +
> > +/* Suppress __warn_unused_result__ if gcc is too old. */
> > +#if !defined(__attribute__) && !defined(__warn_unused_result__)
> > +# if __GNUC__ < 3 || (__GNUC__ == 3 && __GNUC_MINOR__ < 4)
> > +# define __warn_unused_result__ /* empty */
> > +# endif
> > +#endif
> > +
> > +/*
> > + * Suppress the argument to __malloc__ in Clang (not supported in at least
> > + * version 13) and GCC versions prior to 11.
> > + */
> > +#if !defined(__attribute__) && !defined(__malloc__)
> > +# if defined(__clang__) || __GNUC__ < 11
> > +# define __malloc__(dalloc) __malloc__
> > +# endif
> > +#endif
> > +
> > +/*
> > + * LLVM and Clang pretend to be GCC but don't support all of the __attribute__
> > + * settings that GCC does. For them, suppress warnings about unknown
> > + * attributes on declarations. This unfortunately will affect the entire
> > + * compilation context, but there's no push and pop available.
> > + */
> > +#if !defined(__attribute__) && (defined(__llvm__) || defined(__clang__))
> > +# pragma GCC diagnostic ignored "-Wattributes"
> > +#endif
> > +
> > +/* Used for unused parameters to silence gcc warnings. */
> > +// #define UNUSED __attribute__((__unused__))
> > +
> > +/*
> > + * BEGIN_DECLS is used at the beginning of declarations so that C++
> > + * compilers don't mangle their names. END_DECLS is used at the end.
> > + */
> > +#undef BEGIN_DECLS
> > +#undef END_DECLS
> > +#ifdef __cplusplus
> > +# define BEGIN_DECLS extern "C" {
> > +# define END_DECLS }
> > +#else
> > +# define BEGIN_DECLS /* empty */
> > +# define END_DECLS /* empty */
> > +#endif
> > +
> > +#endif /* TAP_MACROS_H */
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 0/2] add an external testing library for unit tests
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
` (3 preceding siblings ...)
2023-04-27 21:35 ` brian m. carlson
@ 2023-05-02 4:18 ` Felipe Contreras
2023-05-02 13:52 ` Ævar Arnfjörð Bjarmason
5 siblings, 0 replies; 29+ messages in thread
From: Felipe Contreras @ 2023-05-02 4:18 UTC (permalink / raw)
To: Calvin Wan, git; +Cc: Calvin Wan
Calvin Wan wrote:
> In our current testing environment, we spend a significant amount of
> effort crafting end-to-end tests for error conditions that could easily
> be captured by unit tests (or we simply forgo some hard-to-setup and
> rare error conditions). Unit tests additionally provide stability to the
> codebase and can simplify debugging through isolation. Turning parts of
> Git into libraries[1] gives us the ability to run unit tests on the
> libraries and to write unit tests in C. Writing unit tests in pure C,
> rather than with our current shell/test-tool helper setup, simplifies
> test setup, simplifies passing data around (no shell-isms required), and
> reduces testing runtime by not spawning a separate process for every
> test invocation.
I agree unit tests would be very helpful, but they don't need to be written in C.
I sent a RFC patch series [1] attempting to write unit tests, but using Ruby, the
testing framework is less than 100 lines of code, and uses Ruby bindings to use
the C functions.
I think writing everything in C adds a ton of unnecessary complexity for no
gain. In fact some things are simply not possible, like dealing with crashes
without forks.
Modern languages exist for a reason: C isn't the best tool in every situation,
and I believe this is one of them.
Cheers.
[1] https://lore.kernel.org/git/20230502041113.103385-1-felipe.contreras@gmail.com/T/#t
--
Felipe Contreras
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 0/2] add an external testing library for unit tests
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
` (4 preceding siblings ...)
2023-05-02 4:18 ` Felipe Contreras
@ 2023-05-02 13:52 ` Ævar Arnfjörð Bjarmason
2023-05-02 15:28 ` Felipe Contreras
5 siblings, 1 reply; 29+ messages in thread
From: Ævar Arnfjörð Bjarmason @ 2023-05-02 13:52 UTC (permalink / raw)
To: Calvin Wan; +Cc: git
On Thu, Apr 27 2023, Calvin Wan wrote:
> In our current testing environment, we spend a significant amount of
> effort crafting end-to-end tests for error conditions that could easily
> be captured by unit tests (or we simply forgo some hard-to-setup and
> rare error conditions). Unit tests additionally provide stability to the
> codebase and can simplify debugging through isolation. Turning parts of
> Git into libraries[1] gives us the ability to run unit tests on the
> libraries and to write unit tests in C. Writing unit tests in pure C,
> rather than with our current shell/test-tool helper setup, simplifies
> test setup, simplifies passing data around (no shell-isms required), and
> reduces testing runtime by not spawning a separate process for every
> test invocation.
>
> Unit testing in C requires a separate testing harness that we ideally
> would like to be TAP-style and to come with a non-restrictive license.
> Fortunately, there already exists a C TAP harness library[2] with an MIT
> license (at least for the files included in this series).
>
> This first patch introduces the C TAP harness and includes only the
> necessary files. The second patch showcases a basic example of it. As an
> RFC, I am wondering what the list thinks about using a second testing
> library for unit testing? Are there any problems with this particular
> external testing library and does it provide the necessary functionality
> we would want for unit testing Git libraries? How should the folders be
> structured and how should the new Makefile rules be written? Ultimately,
> this will help us determine the setup of our unit tests in future
> libification patches.
>
> [1] https://lore.kernel.org/git/CAJoAoZ=Cig_kLocxKGax31sU7Xe4==BGzC__Bg2_pr7krNq6MA@mail.gmail.com/
> [2] https://github.com/rra/c-tap-harness/
I have some out-of-tree patches I've been meaning to submit that massage
some of our TAP output, and I'd really prefer if we don't end up with
two TAP emitters in-tree if we can help it.
We can support such a thing, but nothing about your goals or your
explanation here provides the "why".
Or rather, I'm not really buying the "passing data around" or "recuding
[...] runtime [overhead]". I think you're describing how *some* of our
*.sh to *.c interop goes, but we have test-lib.sh driving C code without
those issues.
We already have pure-C libraries that we add a really shim to unit test,
the most extreme example of this is t0032-reftable-unittest.sh, whose
body is simply (excluding comments):
#!/bin/sh
test_description='reftable unittests'
TEST_PASSES_SANITIZE_LEAK=true
. ./test-lib.sh
test_expect_success 'unittests' '
TMPDIR=$(pwd) && export TMPDIR &&
test-tool reftable
'
test_done
Now, that goes into reftable/test_framework.h which basically implements
its own mini-test framework, so that's at least a *partial* argument for
what you're suggesting here, but note that it doesn't emit TAP, it just
returns an exit code, the EXPECT() etc. is purely internal. I.e. "what
should we return?".
Probably a more git-y example is t0071-sort.sh, whose body is similar
(skipping most of the boilerplate):
test_expect_success 'DEFINE_LIST_SORT_DEBUG' '
test-tool mergesort test
'
We then have similar library tests as e.g. t0063-string-list.sh,
t0061-run-command.sh, t3070-wildmatch.sh etc.
None of those are perfect, but I think the current arrangement is rather
ideal. We can write most or all of the test in C, but we just do so by
calling a function that returns an exit code.
It does mean we need to spawn a shell for test-lib.sh, and call a
test-tool at least once, but the overhead of that is trivial. It's not
trivial in some cases where we call the helper in a loop, but that's
much more easily addressed than hoisting all of TAP generation into the
C space.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 0/2] add an external testing library for unit tests
2023-05-02 13:52 ` Ævar Arnfjörð Bjarmason
@ 2023-05-02 15:28 ` Felipe Contreras
0 siblings, 0 replies; 29+ messages in thread
From: Felipe Contreras @ 2023-05-02 15:28 UTC (permalink / raw)
To: Ævar Arnfjörð Bjarmason, Calvin Wan; +Cc: git
Ævar Arnfjörð Bjarmason wrote:
> We already have pure-C libraries that we add a really shim to unit test,
> the most extreme example of this is t0032-reftable-unittest.sh, whose
> body is simply (excluding comments):
>
> #!/bin/sh
> test_description='reftable unittests'
>
> TEST_PASSES_SANITIZE_LEAK=true
> . ./test-lib.sh
>
> test_expect_success 'unittests' '
> TMPDIR=$(pwd) && export TMPDIR &&
> test-tool reftable
> '
>
> test_done
Yeah, but an output of 'ok 1 - unittests' is not very useful, neither is an
output of 'not ok 1 - unittests'.
This completely misses the point of a TAP interface, which is to parse the
status of individual test cases, or even individual assertions.
If all we are doing is check the exit code of some program, then we don't need
TAP.
> Now, that goes into reftable/test_framework.h which basically implements
> its own mini-test framework, so that's at least a *partial* argument for
> what you're suggesting here, but note that it doesn't emit TAP, it just
> returns an exit code, the EXPECT() etc. is purely internal. I.e. "what
> should we return?".
You are just describing the status quo.
I think this is a naturaistic fallacy: confusing what is versus what ought to
be.
Can we test C code with our current testing framework? Yes.
Should we? That's the actual question.
> None of those are perfect, but I think the current arrangement is rather
> ideal.
I think misuing TAP is far from ideal.
In my view an ideal framework would:
1. Be able to test C code
2. Report individual test cases success/failure
3. Report relevant context in the case of failure (actual vs. expected)
4. Don't create forks on every individual test case or assertion
5. Properly handle crashes
Doing basically the following is not an ideal framework:
echo '1..1'
test-tool reftable > /dev/null 2>&1 && echo 'ok 1' || { echo 'not ok 1'; false; }
Yes, it works. But "it works" has never been a good enough reason to stay on
the status quo.
> We can write most or all of the test in C, but we just do so by
> calling a function that returns an exit code.
Yes, we *can*, but should we?
---
If we can test C code in a way that individual test case failures are reported
(as is the intention with TAP), why would we reject that in favor of the status
quo which basically is just reporting the exit code of the whole test file?
--
Felipe Contreras
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-28 16:31 ` Calvin Wan
@ 2023-05-02 15:46 ` Felipe Contreras
2023-05-10 15:46 ` Phillip Wood
1 sibling, 0 replies; 29+ messages in thread
From: Felipe Contreras @ 2023-05-02 15:46 UTC (permalink / raw)
To: Calvin Wan, phillip.wood; +Cc: git
Calvin Wan wrote:
> On Thu, Apr 27, 2023 at 1:15 PM Phillip Wood <phillip.wood123@gmail.com> wrote:
> >
> > Hi Calvin
> >
> > On 27/04/2023 18:50, Calvin Wan wrote:
> > > Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
> > >
> > > There is also more complete documentation at
> > > https://www.eyrie.org/~eagle/software/c-tap-harness/
> >
> > I'm afraid this reply is rather briefer than I'd like but I'm short of
> > time and about to go off-list for a couple of weeks. My ideal unit test
> > library would
> > - allow named tests (this maybe more trouble that it is worth as I
> > think it inevitably leads to more boilerplate code calling the named
> > tests)
>
> I'm not quite sure what you're referring to with "named tests". Could
> you clarify (possibly with an example)?
Many test frameworks have a way of specifying a name for a test case, for
example in JavaScript's QUnit:
test('basic test case', t => {
t.is('actual', 'expected');
});
In this case "basic test case" is the name of the test case.
In git's testing framework:
test_expect_success 'basic test case' '
echo actual > actual &&
echo expected > expected &&
test_cmp actual expected
'
When running this test case the TAP output would be:
ok 1 - basic test case
Other testing frameworks report on a per-assertion basis, for example Perl's
Test::More:
is('actual', 'expected', 'assertion check');
These are more like named assertions rather than named test cases.
The example of what you proposed shows:
ok(unit_test(), "unit test runs successfully");
Which is similar to Test::More, so it seems it runs on a concept of assertions
rather than test cases.
That's what I presume Phillip meant.
Cheers.
--
Felipe Contreras
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-27 20:15 ` Phillip Wood
2023-04-28 16:31 ` Calvin Wan
@ 2023-05-02 15:54 ` Felipe Contreras
2023-05-02 16:39 ` Ævar Arnfjörð Bjarmason
2023-05-02 16:34 ` Ævar Arnfjörð Bjarmason
2 siblings, 1 reply; 29+ messages in thread
From: Felipe Contreras @ 2023-05-02 15:54 UTC (permalink / raw)
To: Phillip Wood, Calvin Wan, git
Phillip Wood wrote:
> Unfortunately this library doesn't seem to offer any of those features.
> It does support a lazy test plan but uses atexit() so will not detect if
> the test program exits before all the tests have run.
I think there's a fundamental misunderstanding of how we use TAP.
If a program generates this output:
1..3
ok 1 - test 1
ok 2 - test 2
That's clearly not complete. It shouldn't be the job a test script to check for
those cases.
If you run the programm through a TAP harness such as prove, you get:
foo.t .. Failed 1/3 subtests
Test Summary Report
-------------------
foo.t (Wstat: 0 Tests: 2 Failed: 0)
Parse errors: Bad plan. You planned 3 tests but ran 2.
Files=1, Tests=2, 0 wallclock secs ( 0.01 usr + 0.00 sys = 0.01 CPU)
Result: FAIL
Why do we bother generaing a TAP output if we are not going to take advantage
of it?
--
Felipe Contreras
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 2/2] unit test: add basic example
2023-04-27 18:47 ` Junio C Hamano
@ 2023-05-02 15:58 ` Felipe Contreras
0 siblings, 0 replies; 29+ messages in thread
From: Felipe Contreras @ 2023-05-02 15:58 UTC (permalink / raw)
To: Junio C Hamano, Calvin Wan; +Cc: git
Junio C Hamano wrote:
> As a technology demonstration, such a hello-world test may be fine,
> but I would have liked to see a more realistic sample that links
> with reasonably isolated part of Git codebase, perhaps testing say
> functions from <quote.h>.
I don't know if that's the case for Calvin's proposal, but at least when I
tried to do that for my proposal, one major issue is that there's no shared
library to link to, making it difficult to play along with other shared
libraries.
--
Felipe Contreras
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-27 20:15 ` Phillip Wood
2023-04-28 16:31 ` Calvin Wan
2023-05-02 15:54 ` Felipe Contreras
@ 2023-05-02 16:34 ` Ævar Arnfjörð Bjarmason
2023-05-10 8:18 ` Phillip Wood
2 siblings, 1 reply; 29+ messages in thread
From: Ævar Arnfjörð Bjarmason @ 2023-05-02 16:34 UTC (permalink / raw)
To: phillip.wood; +Cc: Calvin Wan, git
On Thu, Apr 27 2023, Phillip Wood wrote:
> Hi Calvin
>
> On 27/04/2023 18:50, Calvin Wan wrote:
>> Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
>> There is also more complete documentation at
>> https://www.eyrie.org/~eagle/software/c-tap-harness/
>
> I'm afraid this reply is rather briefer than I'd like but I'm short of
> time and about to go off-list for a couple of weeks. My ideal unit
> test library would
>
> - print the file and line number of failed assertions
> - allow the test plan to be omitted by calling test_done() at the end
> of the test file as we do in our main test suite.
> - support the TODO directive
> - allow named tests (this maybe more trouble that it is worth as I
> think it inevitably leads to more boilerplate code calling the named
> tests)
>
> Unfortunately this library doesn't seem to offer any of those
> features. It does support a lazy test plan but uses atexit() so will
> not detect if the test program exits before all the tests have run. I
> think it would be useful to add some unit tests to our test suite and
> maybe this library could form the basis of that but I think printing
> the file and line number of failed assertions is pretty essential.
Other things aside, I prefer our explicit "test_done", but I don't see
why you think an atexit() isn't enough to catch incomplete tests.
For a C program you'd just do something like this (somewhat pseudocode,
I didn't check if it compiled etc):
static int done; /* read by atexit() handler */
void on_atexit(void)
{
if (!done)
BUG();
print_plan_line();
}
int main(void)
{
int ret;
setup_atexit(a_handler);
ret = do_tests();
done = 1;
return ret;
}
If I'm understanding you correctly you're concerned that if some user
code within do_test() calls exit() we won't return from "do_test()", but
we *would* call print_plan_line().
That's a valid concern, we want to distinguish such "early return" from
cases where we run to completion, that's why we use "test_done" in the
shell code.
But in the C case I think just using something like the "done" variable
pattern above should cover that, without the need for an explicit
"test_done".
But maybe I'm missing something.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-05-02 15:54 ` Felipe Contreras
@ 2023-05-02 16:39 ` Ævar Arnfjörð Bjarmason
2023-05-02 18:11 ` Felipe Contreras
0 siblings, 1 reply; 29+ messages in thread
From: Ævar Arnfjörð Bjarmason @ 2023-05-02 16:39 UTC (permalink / raw)
To: Felipe Contreras; +Cc: Phillip Wood, Calvin Wan, git
On Tue, May 02 2023, Felipe Contreras wrote:
> Phillip Wood wrote:
>
>> Unfortunately this library doesn't seem to offer any of those features.
>> It does support a lazy test plan but uses atexit() so will not detect if
>> the test program exits before all the tests have run.
>
> I think there's a fundamental misunderstanding of how we use TAP.
>
> If a program generates this output:
>
> 1..3
> ok 1 - test 1
> ok 2 - test 2
>
> That's clearly not complete. It shouldn't be the job a test script to check for
> those cases.
>
> If you run the programm through a TAP harness such as prove, you get:
>
> foo.t .. Failed 1/3 subtests
>
> Test Summary Report
> -------------------
> foo.t (Wstat: 0 Tests: 2 Failed: 0)
> Parse errors: Bad plan. You planned 3 tests but ran 2.
> Files=1, Tests=2, 0 wallclock secs ( 0.01 usr + 0.00 sys = 0.01 CPU)
> Result: FAIL
>
> Why do we bother generaing a TAP output if we are not going to take advantage
> of it?
(As the person who added the TAP output to git.git)
Yeah, we could do the "plan ahead", but it would mean that tests would
need to pre-declare the number of tests they have.
In the Perl world that's the usual pattern, but as it involves having a:
plan tests => 123;
At the top of the file that's guaranteed to give you merge conflicts if
two topics add/remove tests in different parts of the file.
It also doesn't work well in cases where the number of tests is hard to
determine in advance, i.e. when they're determined programatically.
I don't think there's any practical downside to using the "test_done"
pattern to print a plan at the end as far as missing tests go.
There *is* a minor practical downside to it though, which is that we'll
get output like "25/?" or whatever, but not "25/100", as we don't know
yet that we've got a total of 100 tests.
But I think that's a minor drawback, and really only matters if you're
eyeballing the prove(1) output of a very slow test as it scrolls by.
I think on balance the "plan at the end" approach we're using now is
much better, and would also be better in a future (or hypothetical)
pure-C test framework.
Well, there are ways to avoid the painful conflicts, e.g. by mandating
that all tests are driven by callbacks in an array or something, so then
you won't get merge conflicts on the "plan" line, as it'll just be "the
number of tests is the number of items in this array".
But such a thing is painful to mandate, and has its own drawbacks,
i.e. not being able to do a "test" at anything less than a function
callback level of granularity.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-05-02 16:39 ` Ævar Arnfjörð Bjarmason
@ 2023-05-02 18:11 ` Felipe Contreras
0 siblings, 0 replies; 29+ messages in thread
From: Felipe Contreras @ 2023-05-02 18:11 UTC (permalink / raw)
To: Ævar Arnfjörð Bjarmason, Felipe Contreras
Cc: Phillip Wood, Calvin Wan, git
Ævar Arnfjörð Bjarmason wrote:
> On Tue, May 02 2023, Felipe Contreras wrote:
>
> > Phillip Wood wrote:
> >
> >> Unfortunately this library doesn't seem to offer any of those features.
> >> It does support a lazy test plan but uses atexit() so will not detect if
> >> the test program exits before all the tests have run.
> >
> > I think there's a fundamental misunderstanding of how we use TAP.
> >
> > If a program generates this output:
> >
> > 1..3
> > ok 1 - test 1
> > ok 2 - test 2
> >
> > That's clearly not complete. It shouldn't be the job a test script to check for
> > those cases.
> >
> > If you run the programm through a TAP harness such as prove, you get:
> >
> > foo.t .. Failed 1/3 subtests
> >
> > Test Summary Report
> > -------------------
> > foo.t (Wstat: 0 Tests: 2 Failed: 0)
> > Parse errors: Bad plan. You planned 3 tests but ran 2.
> > Files=1, Tests=2, 0 wallclock secs ( 0.01 usr + 0.00 sys = 0.01 CPU)
> > Result: FAIL
> >
> > Why do we bother generaing a TAP output if we are not going to take advantage
> > of it?
>
> (As the person who added the TAP output to git.git)
>
> Yeah, we could do the "plan ahead", but it would mean that tests would
> need to pre-declare the number of tests they have.
I'm not advocating for planning ahead.
In the case of the shell testing framework it makes sense to not plan
ahead and just print the test plan at the end with test_done(), as
that's very convenient.
But $subject is a new proposal for C unit tests with a test plan
specified ahead, therefore if the program exits before all the tests
have run, that should be obvious from the output. I'm not saying that is
good or desirable, merely that it is the case.
I personaly don't even think C is the way to write unit tests.
> In the Perl world that's the usual pattern, but as it involves having a:
>
> plan tests => 123;
>
> At the top of the file that's guaranteed to give you merge conflicts if
> two topics add/remove tests in different parts of the file.
>
> It also doesn't work well in cases where the number of tests is hard to
> determine in advance, i.e. when they're determined programatically.
>
> I don't think there's any practical downside to using the "test_done"
> pattern to print a plan at the end as far as missing tests go.
>
> There *is* a minor practical downside to it though, which is that we'll
> get output like "25/?" or whatever, but not "25/100", as we don't know
> yet that we've got a total of 100 tests.
>
> But I think that's a minor drawback, and really only matters if you're
> eyeballing the prove(1) output of a very slow test as it scrolls by.
>
> I think on balance the "plan at the end" approach we're using now is
> much better, and would also be better in a future (or hypothetical)
> pure-C test framework.
>
> Well, there are ways to avoid the painful conflicts, e.g. by mandating
> that all tests are driven by callbacks in an array or something, so then
> you won't get merge conflicts on the "plan" line, as it'll just be "the
> number of tests is the number of items in this array".
>
> But such a thing is painful to mandate, and has its own drawbacks,
> i.e. not being able to do a "test" at anything less than a function
> callback level of granularity.
If you are using an old-school language that doesn't have support for
anonymous functions, then yes, that is pretty much your only nice
maintenable option.
But the list of languages that don't have support for that is rather
small, it's pretty much limited to shell, C, Fortran and COBOL. Even
Perl has support for anonymous functions.
In all modern languages you can write test cases using anonymous
functions such as (JavaScript's ava [1]):
test('test case', t => {
t.is('actual', 'expected');
});
This can be done in Perl as well (see Test::Functional [2]).
So at the end of the script it is known how many test cases there are
*before* running any of them.
This is precisely the approach I followed in my new testing framework,
which I sent as a proposal to the ml [3].
require 'testox'
test('one') { ok(true) }
test('two') { ok(true) }
test('three') { ok(true) }
test('four') { ok(true) }
puts '# num tests: %s' % TestOx.tests.length
No test case is run until the script ends in at_exit:
# num tests: 4
1..4
ok 1 - one
ok 2 - two
ok 3 - three
ok 4 - four
There's no need to specify the number of test cases.
I don't see why we insist in using ancient languages like C and shell
for things that neither C nor shell are good at. Other languages exist
in 2023. Other languages exist for a reason.
Cheers.
[1] https://github.com/avajs/ava
[2] https://metacpan.org/pod/Test::Functional
[3] https://lore.kernel.org/git/20230502041113.103385-1-felipe.contreras@gmail.com/T/#t
--
Felipe Contreras
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-05-02 16:34 ` Ævar Arnfjörð Bjarmason
@ 2023-05-10 8:18 ` Phillip Wood
0 siblings, 0 replies; 29+ messages in thread
From: Phillip Wood @ 2023-05-10 8:18 UTC (permalink / raw)
To: Ævar Arnfjörð Bjarmason, phillip.wood; +Cc: Calvin Wan, git
Hi Ævar
On 02/05/2023 17:34, Ævar Arnfjörð Bjarmason wrote:
>
> On Thu, Apr 27 2023, Phillip Wood wrote:
>
>> Hi Calvin
>>
>> On 27/04/2023 18:50, Calvin Wan wrote:
>>> Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
>>> There is also more complete documentation at
>>> https://www.eyrie.org/~eagle/software/c-tap-harness/
>>
>> I'm afraid this reply is rather briefer than I'd like but I'm short of
>> time and about to go off-list for a couple of weeks. My ideal unit
>> test library would
>>
>> - print the file and line number of failed assertions
>> - allow the test plan to be omitted by calling test_done() at the end
>> of the test file as we do in our main test suite.
>> - support the TODO directive
>> - allow named tests (this maybe more trouble that it is worth as I
>> think it inevitably leads to more boilerplate code calling the named
>> tests)
>>
>> Unfortunately this library doesn't seem to offer any of those
>> features. It does support a lazy test plan but uses atexit() so will
>> not detect if the test program exits before all the tests have run. I
>> think it would be useful to add some unit tests to our test suite and
>> maybe this library could form the basis of that but I think printing
>> the file and line number of failed assertions is pretty essential.
>
> Other things aside, I prefer our explicit "test_done", but I don't see
> why you think an atexit() isn't enough to catch incomplete tests.
>
> For a C program you'd just do something like this (somewhat pseudocode,
> I didn't check if it compiled etc):
>
> static int done; /* read by atexit() handler */
>
> void on_atexit(void)
> {
> if (!done)
> BUG();
> print_plan_line();
> }
>
> int main(void)
> {
> int ret;
>
> setup_atexit(a_handler);
> ret = do_tests();
> done = 1;
>
> return ret;
> }
>
> If I'm understanding you correctly you're concerned that if some user
> code within do_test() calls exit() we won't return from "do_test()", but
> we *would* call print_plan_line().
Exactly
> That's a valid concern, we want to distinguish such "early return" from
> cases where we run to completion, that's why we use "test_done" in the
> shell code.
>
> But in the C case I think just using something like the "done" variable
> pattern above should cover that, without the need for an explicit
> "test_done".
We could do that. My complaint is that the code being proposed does not
and so prints a valid plan if any code being tested calls exit()
Best Wishes
Phillip
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-04-28 16:31 ` Calvin Wan
2023-05-02 15:46 ` Felipe Contreras
@ 2023-05-10 15:46 ` Phillip Wood
2023-05-11 23:16 ` Glen Choo
2023-06-21 15:57 ` Linus Arver
1 sibling, 2 replies; 29+ messages in thread
From: Phillip Wood @ 2023-05-10 15:46 UTC (permalink / raw)
To: Calvin Wan, phillip.wood; +Cc: git, Ævar Arnfjörð Bjarmason
Hi Calvin
On 28/04/2023 17:31, Calvin Wan wrote:
> On Thu, Apr 27, 2023 at 1:15 PM Phillip Wood <phillip.wood123@gmail.com> wrote:
>>
>> Hi Calvin
>>
>> On 27/04/2023 18:50, Calvin Wan wrote:
>>> Introduces the C TAP harness from https://github.com/rra/c-tap-harness/
>>>
>>> There is also more complete documentation at
>>> https://www.eyrie.org/~eagle/software/c-tap-harness/
>>
>> I'm afraid this reply is rather briefer than I'd like but I'm short of
>> time and about to go off-list for a couple of weeks. My ideal unit test
>> library would
>>
>> - print the file and line number of failed assertions
>
> This library does print the file and line number of failed assertions.
> You can add a failing test in a new test file to check that this is
> the case.
I should have been clear that I was talking about is_int()
etc. provided by this library. They do not print the location of the
failing call as they are functions and therefore cannot know the
source location of the call unless they are wrapped in a macro which
passes the source location to them. The is_int() functions also print
an ok line for each check meaning that unless the test author comes up
with a unique description for each comparison it is hard to tell which
line a failing test comes from.
>> - allow the test plan to be omitted by calling test_done() at the end
>> of the test file as we do in our main test suite.
>
> The same functionality can be achieved by adding some logic to adjust
> the number of planned tests and return early if necessary.
I'm afraid I don't understand your comment. My concern is that having
an explicit plan is a pain to maintain. Æver has pointed out elsewhere
in this thread that requiring an explicit plan is a good source of
merge conflicts.
>> - support the TODO directive
>
> This would be a nice feature to add to the library, however, I don't
> think it's necessary in a first pass.
It may not be necessary now, but if we think we want to add it in the
future we should look to see how easy that would be.
>> - allow named tests (this maybe more trouble that it is worth as I
>> think it inevitably leads to more boilerplate code calling the named
>> tests)
>
> I'm not quite sure what you're referring to with "named tests". Could
> you clarify (possibly with an example)?
I'd like to be able to have several checks within a test and get one
line of output if it passes. I'd also like the test framework to
manage the pass/fail state based on the checks in the test. For
example
static void t_static_init(void)
{
struct strbuf buf = STRBUF_INIT;
check_uint(buf.len, ==, 0);
check_uint(buf.alloc, ==, 0);
if (check(buf.buf == strbuf_slopbuf))
return; /* avoid de-referencing buf.buf */
check_char(buf.buf[0], ==, '\0');
}
Would produce
ok 1 - static initialization works
Using is_int() etc from this library would result in one line of output
for each check and require a description for each check which would
quickly become tedious.
The patch at the end of this email implements a test framework that
supports tests like the one above. The pass/fail state of tests is
managed by the framework based on the checks in each test
test. Failing checks produce informative messages to aid debugging,
for example
# check "!strcmp(x, "hello")" failed at my-test.c:102
# left: "bye"
# right: "hello"
So one can quickly find the failing line in the test file and see what
the arguments to the test were.
>> Unfortunately this library doesn't seem to offer any of those features.
>> It does support a lazy test plan but uses atexit() so will not detect if
>> the test program exits before all the tests have run. I think it would
>> be useful to add some unit tests to our test suite and maybe this
>> library could form the basis of that but I think printing the file and
>> line number of failed assertions is pretty essential.
>
> Since I've addressed your concerns regarding printing file and line
> numbers for failed assertions, do you think this test library is
> sufficient as a base to iterate off of?
It would be possible, but I'd want to see some macros for checking
invariants that print the source location as well as their
arguments. I'd be interested to hear what you think of the alternative
approach in the patch below. The patch can be fetched with
git fetch https://github.com/phillipwood/git unit-tests
Best Wishes
Phillip
---- >8 ----
From: Phillip Wood <phillip.wood@dunelm.org.uk>
[RFC PATCH] unit-tests: proof of concept test framework
This patch contains a proof of concept implementation for writing unit
tests with TAP output. Each test is a function that contains one or more
checks. The test is run with the TEST() macro and if any of the checks
fail then the test will fail. A complete program that tests STRBUF_INIT
would look like
#include "test-lib.h"
#include "strbuf.h"
static void t_static_init(void)
{
struct strbuf buf = STRBUF_INIT;
check_uint(buf.len, ==, 0);
check_uint(buf.alloc, ==, 0);
if (check(buf.buf == strbuf_slopbuf))
return; /* avoid SIGSEV */
check_char(buf.buf[0], ==, '\0');
}
int main(void)
{
TEST(t_static_init(), "static initialization works);
return test_done();
}
The output of this program would be
ok 1 - static initialization works
1..1
If any of the checks in a test fail then they print a diagnostic message
to aid debugging and the test will be reported as failing. For example a
failing integer check would look like
# check "x >= 3" failed at my-test.c:102
# left: 2
# right: 3
not ok 1 - x is greater than or equal to three
There are a number of check functions implemented so far. check() checks
a boolean condition, check_int(), check_uint() and check_char() take two
values to compare and a comparison operator. check_str() will check if
two strings are equal. Custom checks are simple to implement as shown in
the comments above test_assert() in test-lib.h.
Tests can be skipped with test_skip() which can be supplied with a
reason for skipping which it will print. Tests can print diagnostic
messages with test_msg(). Checks that are known to fail can be wrapped
in TEST_TODO().
There are a couple of example test programs included in this
patch. t-basic.c implements some self-tests and demonstrates the
diagnostic output for failing test. The output of this program is
checked by t0080-unit-test-output.sh. t-strbuf.c shows some example
unit tests for strbuf.c
The unit tests can be built with "make unit-tests" (this works but the
Makefile changes need some further work). Once they have been built they
can be run manually (e.g t/unit-tests/t-strbuf) or with prove.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
---
Makefile | 23 ++-
t/t0080-unit-test-output.sh | 58 +++++++
t/unit-tests/.gitignore | 2 +
t/unit-tests/t-basic.c | 87 ++++++++++
t/unit-tests/t-strbuf.c | 75 ++++++++
t/unit-tests/test-lib.c | 329 ++++++++++++++++++++++++++++++++++++
t/unit-tests/test-lib.h | 143 ++++++++++++++++
7 files changed, 715 insertions(+), 2 deletions(-)
create mode 100755 t/t0080-unit-test-output.sh
create mode 100644 t/unit-tests/.gitignore
create mode 100644 t/unit-tests/t-basic.c
create mode 100644 t/unit-tests/t-strbuf.c
create mode 100644 t/unit-tests/test-lib.c
create mode 100644 t/unit-tests/test-lib.h
diff --git a/Makefile b/Makefile
index e440728c24..f0ca5bae4b 100644
--- a/Makefile
+++ b/Makefile
@@ -681,6 +681,7 @@ SCRIPT_LIB =
TEST_BUILTINS_OBJS =
TEST_OBJS =
TEST_PROGRAMS_NEED_X =
+UNIT_TEST_PROGRAMS =
THIRD_PARTY_SOURCES =
# Having this variable in your environment would break pipelines because
@@ -2665,13 +2666,20 @@ SCALAR_OBJS += scalar.o
.PHONY: scalar-objs
scalar-objs: $(SCALAR_OBJS)
+UNIT_TEST_PROGRAMS += t-basic
+UNIT_TEST_PROGRAMS += t-strbuf
+UNIT_TEST_PROGS = $(patsubst %,t/unit-tests/%$X,$(UNIT_TEST_PROGRAMS))
+UNIT_TEST_OBJS = $(patsubst %,t/unit-tests/%.o,$(UNIT_TEST_PROGRAMS))
+UNIT_TEST_OBJS += t/unit-tests/test-lib.o
+
OBJECTS += $(GIT_OBJS)
OBJECTS += $(SCALAR_OBJS)
OBJECTS += $(PROGRAM_OBJS)
OBJECTS += $(TEST_OBJS)
OBJECTS += $(XDIFF_OBJS)
OBJECTS += $(FUZZ_OBJS)
OBJECTS += $(REFTABLE_OBJS) $(REFTABLE_TEST_OBJS)
+OBJECTS += $(UNIT_TEST_OBJS)
ifndef NO_CURL
OBJECTS += http.o http-walker.o remote-curl.o
@@ -3167,7 +3175,7 @@ endif
test_bindir_programs := $(patsubst %,bin-wrappers/%,$(BINDIR_PROGRAMS_NEED_X) $(BINDIR_PROGRAMS_NO_X) $(TEST_PROGRAMS_NEED_X))
-all:: $(TEST_PROGRAMS) $(test_bindir_programs)
+all:: $(TEST_PROGRAMS) $(test_bindir_programs) $(UNIT_TEST_PROGS)
bin-wrappers/%: wrap-for-bin.sh
$(call mkdir_p_parent_template)
@@ -3592,7 +3600,7 @@ endif
artifacts-tar:: $(ALL_COMMANDS_TO_INSTALL) $(SCRIPT_LIB) $(OTHER_PROGRAMS) \
GIT-BUILD-OPTIONS $(TEST_PROGRAMS) $(test_bindir_programs) \
- $(MOFILES)
+ $(UNIT_TEST_PROGS) $(MOFILES)
$(QUIET_SUBDIR0)templates $(QUIET_SUBDIR1) \
SHELL_PATH='$(SHELL_PATH_SQ)' PERL_PATH='$(PERL_PATH_SQ)'
test -n "$(ARTIFACTS_DIRECTORY)"
@@ -3831,3 +3839,14 @@ $(FUZZ_PROGRAMS): all
$(XDIFF_OBJS) $(EXTLIBS) git.o $@.o $(LIB_FUZZING_ENGINE) -o $@
fuzz-all: $(FUZZ_PROGRAMS)
+
+t/unit-tests/t-basic$X: t/unit-tests/t-basic.o t/unit-tests/test-lib.o $(GITLIBS) GIT-LDFLAGS
+ $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
+ $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
+
+t/unit-tests/t-strbuf$X: t/unit-tests/t-strbuf.o t/unit-tests/test-lib.o $(GITLIBS) GIT-LDFLAGS
+ $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
+ $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
+
+.PHONY: unit-tests
+unit-tests: $(UNIT_TEST_PROGS)
diff --git a/t/t0080-unit-test-output.sh b/t/t0080-unit-test-output.sh
new file mode 100755
index 0000000000..c60e402260
--- /dev/null
+++ b/t/t0080-unit-test-output.sh
@@ -0,0 +1,58 @@
+#!/bin/sh
+
+test_description='Test the output of the unit test framework'
+
+. ./test-lib.sh
+
+test_expect_success 'TAP output from unit tests' '
+ cat >expect <<-EOF &&
+ ok 1 - passing test
+ ok 2 - passing test and assertion return 0
+ # check "1 == 2" failed at t/unit-tests/t-basic.c:68
+ # left: 1
+ # right: 2
+ not ok 3 - failing test
+ ok 4 - failing test and assertion return -1
+ not ok 5 - passing TEST_TODO() # TODO
+ ok 6 - passing TEST_TODO() returns 0
+ # todo check ${SQ}check(x)${SQ} succeeded at t/unit-tests/t-basic.c:17
+ not ok 7 - failing TEST_TODO()
+ ok 8 - failing TEST_TODO() returns -1
+ # check "0" failed at t/unit-tests/t-basic.c:22
+ # skipping test - missing prerequisite
+ # skipping check ${SQ}1${SQ} at t/unit-tests/t-basic.c:24
+ ok 9 - test_skip() # SKIP
+ ok 10 - skipped test returns 0
+ # skipping test - missing prerequisite
+ ok 11 - test_skip() inside TEST_TODO() # SKIP
+ ok 12 - test_skip() inside TEST_TODO() returns 0
+ # check "0" failed at t/unit-tests/t-basic.c:40
+ not ok 13 - TEST_TODO() after failing check
+ ok 14 - TEST_TODO() after failing check returns -1
+ # check "0" failed at t/unit-tests/t-basic.c:48
+ not ok 15 - failing check after TEST_TODO()
+ ok 16 - failing check after TEST_TODO() returns -1
+ # check "!strcmp("\thello\\\\", "there\"\n")" failed at t/unit-tests/t-basic.c:53
+ # left: "\011hello\\\\"
+ # right: "there\"\012"
+ # check "!strcmp("NULL", NULL)" failed at t/unit-tests/t-basic.c:54
+ # left: "NULL"
+ # right: NULL
+ # check "${SQ}a${SQ} == ${SQ}\n${SQ}" failed at t/unit-tests/t-basic.c:55
+ # left: ${SQ}a${SQ}
+ # right: ${SQ}\012${SQ}
+ # check "${SQ}\\\\${SQ} == ${SQ}\\${SQ}${SQ}" failed at t/unit-tests/t-basic.c:56
+ # left: ${SQ}\\\\${SQ}
+ # right: ${SQ}\\${SQ}${SQ}
+ not ok 17 - messages from failing string and char comparison
+ # BUG: test has no checks at t/unit-tests/t-basic.c:83
+ not ok 18 - test with no checks
+ ok 19 - test with no checks returns -1
+ 1..19
+ EOF
+
+ ! "$GIT_BUILD_DIR"/t/unit-tests/t-basic >actual &&
+ test_cmp expect actual
+'
+
+test_done
diff --git a/t/unit-tests/.gitignore b/t/unit-tests/.gitignore
new file mode 100644
index 0000000000..e292d58348
--- /dev/null
+++ b/t/unit-tests/.gitignore
@@ -0,0 +1,2 @@
+/t-basic
+/t-strbuf
diff --git a/t/unit-tests/t-basic.c b/t/unit-tests/t-basic.c
new file mode 100644
index 0000000000..ab0b7682c4
--- /dev/null
+++ b/t/unit-tests/t-basic.c
@@ -0,0 +1,87 @@
+#include "test-lib.h"
+
+/* Used to store the return value of check_int(). */
+static int check_res;
+
+/* Used to store the return value of TEST(). */
+static int test_res;
+
+static void t_res(int expect)
+{
+ check_int(check_res, ==, expect);
+ check_int(test_res, ==, expect);
+}
+
+static void t_todo(int x)
+{
+ check_res = TEST_TODO(check(x));
+}
+
+static void t_skip(void)
+{
+ check(0);
+ test_skip("missing prerequisite");
+ check(1);
+}
+
+static int do_skip(void)
+{
+ test_skip("missing prerequisite");
+ return 0;
+}
+
+static void t_skip_todo(void)
+{
+ check_res = TEST_TODO(do_skip());
+}
+
+static void t_todo_after_fail(void)
+{
+ check(0);
+ TEST_TODO(check(0));
+}
+
+static void t_fail_after_todo(void)
+{
+ check(1);
+ TEST_TODO(check(0));
+ check(0);
+}
+
+static void t_messages(void)
+{
+ check_str("\thello\\", "there\"\n");
+ check_str("NULL", NULL);
+ check_char('a', ==, '\n');
+ check_char('\\', ==, '\'');
+}
+
+static void t_empty(void)
+{
+ ; /* empty */
+}
+
+int cmd_main(int argc, const char **argv)
+{
+ test_res = TEST(check_res = check_int(1, ==, 1), "passing test");
+ TEST(t_res(0), "passing test and assertion return 0");
+ test_res = TEST(check_res = check_int(1, ==, 2), "failing test");
+ TEST(t_res(-1), "failing test and assertion return -1");
+ test_res = TEST(t_todo(0), "passing TEST_TODO()");
+ TEST(t_res(0), "passing TEST_TODO() returns 0");
+ test_res = TEST(t_todo(1), "failing TEST_TODO()");
+ TEST(t_res(-1), "failing TEST_TODO() returns -1");
+ test_res = TEST(t_skip(), "test_skip()");
+ TEST(check_int(test_res, ==, 0), "skipped test returns 0");
+ test_res = TEST(t_skip_todo(), "test_skip() inside TEST_TODO()");
+ TEST(t_res(0), "test_skip() inside TEST_TODO() returns 0");
+ test_res = TEST(t_todo_after_fail(), "TEST_TODO() after failing check");
+ TEST(check_int(test_res, ==, -1), "TEST_TODO() after failing check returns -1");
+ test_res = TEST(t_fail_after_todo(), "failing check after TEST_TODO()");
+ TEST(check_int(test_res, ==, -1), "failing check after TEST_TODO() returns -1");
+ TEST(t_messages(), "messages from failing string and char comparison");
+ test_res = TEST(t_empty(), "test with no checks");
+ TEST(check_int(test_res, ==, -1), "test with no checks returns -1");
+
+ return test_done();
+}
diff --git a/t/unit-tests/t-strbuf.c b/t/unit-tests/t-strbuf.c
new file mode 100644
index 0000000000..561611e242
--- /dev/null
+++ b/t/unit-tests/t-strbuf.c
@@ -0,0 +1,75 @@
+#include "test-lib.h"
+#include "strbuf.h"
+
+/* wrapper that supplies tests with an initialized strbuf */
+static void setup(void (*f)(struct strbuf*, void*), void *data)
+{
+ struct strbuf buf = STRBUF_INIT;
+
+ f(&buf, data);
+ strbuf_release(&buf);
+ check_uint(buf.len, ==, 0);
+ check_uint(buf.alloc, ==, 0);
+ check(buf.buf == strbuf_slopbuf);
+ check_char(buf.buf[0], ==, '\0');
+}
+
+static void t_static_init(void)
+{
+ struct strbuf buf = STRBUF_INIT;
+
+ check_uint(buf.len, ==, 0);
+ check_uint(buf.alloc, ==, 0);
+ if (check(buf.buf == strbuf_slopbuf))
+ return; /* avoid de-referencing buf.buf */
+ check_char(buf.buf[0], ==, '\0');
+}
+
+static void t_dynamic_init(void)
+{
+ struct strbuf buf;
+
+ strbuf_init(&buf, 1024);
+ check_uint(buf.len, ==, 0);
+ check_uint(buf.alloc, >=, 1024);
+ check_char(buf.buf[0], ==, '\0');
+ strbuf_release(&buf);
+}
+
+static void t_addch(struct strbuf *buf, void *data)
+{
+ const char *p_ch = data;
+ const char ch = *p_ch;
+
+ strbuf_addch(buf, ch);
+ if (check_uint(buf->len, ==, 1) ||
+ check_uint(buf->alloc, >, 1))
+ return; /* avoid de-referencing buf->buf */
+ check_char(buf->buf[0], ==, ch);
+ check_char(buf->buf[1], ==, '\0');
+}
+
+static void t_addstr(struct strbuf *buf, void *data)
+{
+ const char *text = data;
+ size_t len = strlen(text);
+
+ strbuf_addstr(buf, text);
+ if (check_uint(buf->len, ==, len) ||
+ check_uint(buf->alloc, >, len) ||
+ check_char(buf->buf[len], ==, '\0'))
+ return;
+ check_str(buf->buf, text);
+}
+
+int cmd_main(int argc, const char **argv)
+{
+ if (TEST(t_static_init(), "static initialization works"))
+ test_skip_all("STRBUF_INIT is broken");
+ TEST(t_dynamic_init(), "dynamic initialization works");
+ TEST(setup(t_addch, "a"), "strbuf_addch adds char");
+ TEST(setup(t_addch, ""), "strbuf_addch adds NUL char");
+ TEST(setup(t_addstr, "hello there"), "strbuf_addstr adds string");
+
+ return test_done();
+}
diff --git a/t/unit-tests/test-lib.c b/t/unit-tests/test-lib.c
new file mode 100644
index 0000000000..70030d587f
--- /dev/null
+++ b/t/unit-tests/test-lib.c
@@ -0,0 +1,329 @@
+#include "test-lib.h"
+
+enum result {
+ RESULT_NONE,
+ RESULT_FAILURE,
+ RESULT_SKIP,
+ RESULT_SUCCESS,
+ RESULT_TODO
+};
+
+static struct {
+ enum result result;
+ int count;
+ unsigned failed :1;
+ unsigned lazy_plan :1;
+ unsigned running :1;
+ unsigned skip_all :1;
+ unsigned todo :1;
+} ctx = {
+ .lazy_plan = 1,
+ .result = RESULT_NONE,
+};
+
+static void msg_with_prefix(const char *prefix, const char *format, va_list ap)
+{
+ fflush(stderr);
+ if (prefix)
+ fprintf(stdout, "%s", prefix);
+ vprintf(format, ap); /* TODO: handle newlines */
+ putc('\n', stdout);
+ fflush(stdout);
+}
+
+void test_msg(const char *format, ...)
+{
+ va_list ap;
+
+ va_start(ap, format);
+ msg_with_prefix("# ", format, ap);
+ va_end(ap);
+}
+
+void test_plan(int count)
+{
+ assert(!ctx.running);
+
+ fflush(stderr);
+ printf("1..%d\n", count);
+ fflush(stdout);
+ ctx.lazy_plan = 0;
+}
+
+int test_done(void)
+{
+ assert(!ctx.running);
+
+ if (ctx.lazy_plan)
+ test_plan(ctx.count);
+
+ return ctx.failed;
+}
+
+void test_skip(const char *format, ...)
+{
+ va_list ap;
+
+ assert(ctx.running);
+
+ ctx.result = RESULT_SKIP;
+ va_start(ap, format);
+ if (format)
+ msg_with_prefix("# skipping test - ", format, ap);
+ va_end(ap);
+}
+
+void test_skip_all(const char *format, ...)
+{
+ va_list ap;
+ const char *prefix;
+
+ if (!ctx.count && ctx.lazy_plan) {
+ /* We have not printed a test plan yet */
+ prefix = "1..0 # SKIP ";
+ ctx.lazy_plan = 0;
+ } else {
+ /* We have already printed a test plan */
+ prefix = "Bail out! # ";
+ ctx.failed = 1;
+ }
+ ctx.skip_all = 1;
+ ctx.result = RESULT_SKIP;
+ va_start(ap, format);
+ msg_with_prefix(prefix, format, ap);
+ va_end(ap);
+}
+
+int test__run_begin(void)
+{
+ assert(!ctx.running);
+
+ ctx.count++;
+ ctx.result = RESULT_NONE;
+ ctx.running = 1;
+
+ return ctx.skip_all;
+}
+
+static void print_description(const char *format, va_list ap)
+{
+ if (format) {
+ fputs(" - ", stdout);
+ vprintf(format, ap);
+ }
+}
+
+int test__run_end(int was_run UNUSED, const char *location, const char *format, ...)
+{
+ va_list ap;
+
+ assert(ctx.running);
+ assert(!ctx.todo);
+
+ fflush(stderr);
+ va_start(ap, format);
+ if (!ctx.skip_all) {
+ switch (ctx.result) {
+ case RESULT_SUCCESS:
+ printf("ok %d", ctx.count);
+ print_description(format, ap);
+ break;
+
+ case RESULT_FAILURE:
+ printf("not ok %d", ctx.count);
+ print_description(format, ap);
+ break;
+
+ case RESULT_TODO:
+ printf("not ok %d", ctx.count);
+ print_description(format, ap);
+ printf(" # TODO");
+ break;
+
+ case RESULT_SKIP:
+ printf("ok %d", ctx.count);
+ print_description(format, ap);
+ printf(" # SKIP");
+ break;
+
+ case RESULT_NONE:
+ test_msg("BUG: test has no checks at %s", location);
+ printf("not ok %d", ctx.count);
+ print_description(format, ap);
+ ctx.result = RESULT_FAILURE;
+ break;
+ }
+ }
+ va_end(ap);
+ ctx.running = 0;
+ if (ctx.skip_all)
+ return 0;
+ putc('\n', stdout);
+ fflush(stdout);
+ ctx.failed |= ctx.result == RESULT_FAILURE;
+
+ return -(ctx.result == RESULT_FAILURE);
+}
+
+static void test_fail(void)
+{
+ assert(ctx.result != RESULT_SKIP);
+
+ ctx.result = RESULT_FAILURE;
+}
+
+static void test_pass(void)
+{
+ assert(ctx.result != RESULT_SKIP);
+
+ if (ctx.result == RESULT_NONE)
+ ctx.result = RESULT_SUCCESS;
+}
+
+static void test_todo(void)
+{
+ assert(ctx.result != RESULT_SKIP);
+
+ if (ctx.result != RESULT_FAILURE)
+ ctx.result = RESULT_TODO;
+}
+
+int test_assert(const char *location, const char *check, int ok)
+{
+ assert(ctx.running);
+
+ if (ctx.result == RESULT_SKIP) {
+ test_msg("skipping check '%s' at %s", check, location);
+ return 0;
+ } else if (!ctx.todo) {
+ if (ok) {
+ test_pass();
+ } else {
+ test_msg("check \"%s\" failed at %s", check, location);
+ test_fail();
+ }
+ }
+
+ return -!ok;
+}
+
+void test__todo_begin(void)
+{
+ assert(ctx.running);
+ assert(!ctx.todo);
+
+ ctx.todo = 1;
+}
+
+int test__todo_end(const char *location, const char *check, int res)
+{
+ assert(ctx.running);
+ assert(ctx.todo);
+
+ ctx.todo = 0;
+ if (ctx.result == RESULT_SKIP)
+ return 0;
+ if (!res) {
+ test_msg("todo check '%s' succeeded at %s", check, location);
+ test_fail();
+ } else {
+ test_todo();
+ }
+
+ return -!res;
+}
+
+int check_bool_loc(const char *loc, const char *check, int ok)
+{
+ return test_assert(loc, check, ok);
+}
+
+union test__tmp test__tmp[2];
+
+int check_int_loc(const char *loc, const char *check, int ok,
+ intmax_t a, intmax_t b)
+{
+ int ret = test_assert(loc, check, ok);
+
+ if (ret) {
+ test_msg(" left: %"PRIdMAX, a);
+ test_msg(" right: %"PRIdMAX, b);
+ }
+
+ return ret;
+}
+
+int check_uint_loc(const char *loc, const char *check, int ok,
+ uintmax_t a, uintmax_t b)
+{
+ int ret = test_assert(loc, check, ok);
+
+ if (ret) {
+ test_msg(" left: %"PRIuMAX, a);
+ test_msg(" right: %"PRIuMAX, b);
+ }
+
+ return ret;
+}
+
+static void print_one_char(char ch, char quote)
+{
+ if ((unsigned char)ch < 0x20u || ch == 0x7f) {
+ /* TODO: improve handling of \a, \b, \f ... */
+ printf("\\%03o", (unsigned char)ch);
+ } else {
+ if (ch == '\\' || ch == quote)
+ putc('\\', stdout);
+ putc(ch, stdout);
+ }
+}
+
+static void print_char(const char *prefix, char ch)
+{
+ printf("# %s: '", prefix);
+ print_one_char(ch, '\'');
+ fputs("'\n", stdout);
+}
+
+int check_char_loc(const char *loc, const char *check, int ok, char a, char b)
+{
+ int ret = test_assert(loc, check, ok);
+
+ if (ret) {
+ fflush(stderr);
+ print_char(" left", a);
+ print_char(" right", b);
+ fflush(stdout);
+ }
+
+ return ret;
+}
+
+static void print_str(const char *prefix, const char *str)
+{
+ printf("# %s: ", prefix);
+ if (!str) {
+ fputs("NULL\n", stdout);
+ } else {
+ putc('"', stdout);
+ while (*str)
+ print_one_char(*str++, '"');
+ fputs("\"\n", stdout);
+ }
+}
+
+int check_str_loc(const char *loc, const char *check,
+ const char *a, const char *b)
+{
+ int ok = (!a && !b) || (a && b && !strcmp(a, b));
+ int ret = test_assert(loc, check, ok);
+
+ if (ret) {
+ fflush(stderr);
+ print_str(" left", a);
+ print_str(" right", b);
+ fflush(stdout);
+ }
+
+ return ret;
+}
diff --git a/t/unit-tests/test-lib.h b/t/unit-tests/test-lib.h
new file mode 100644
index 0000000000..720c97c6f8
--- /dev/null
+++ b/t/unit-tests/test-lib.h
@@ -0,0 +1,143 @@
+#ifndef TEST_LIB_H
+#define TEST_LIB_H
+
+#include "git-compat-util.h"
+
+/*
+ * Run a test function, returns 0 if the test succeeds, -1 if it
+ * fails. If test_skip_all() has been called then the test will not be
+ * run. The description for each test should be unique. For example:
+ *
+ * TEST(test_something(arg1, arg2), "something %d %d", arg1, arg2)
+ */
+#define TEST(t, ...) \
+ test__run_end(test__run_begin() ? 0 : (t, 1), \
+ TEST_LOCATION(), __VA_ARGS__)
+
+/*
+ * Print a test plan, should be called before any tests. If the number
+ * of tests is not known in advance test_done() will automatically
+ * print a plan at the end of the test program.
+ */
+void test_plan(int count);
+
+/*
+ * test_done() must be called at the end of main(). It will print the
+ * plan if plan() was not called at the beginning of the test program
+ * and returns the exit code for the test program.
+ */
+int test_done(void);
+
+/* Skip the current test. */
+__attribute__((format (printf, 1, 2)))
+void test_skip(const char *format, ...);
+
+/* Skip all remaining tests. */
+__attribute__((format (printf, 1, 2)))
+void test_skip_all(const char *format, ...);
+
+/* Print a diagnostic message to stdout. */
+__attribute__((format (printf, 1, 2)))
+void test_msg(const char *format, ...);
+
+/*
+ * Test checks are built around test_assert(). checks return 0 on
+ * success, -1 on failure. If any check fails then the test will
+ * fail. To create a custom check define a function that wraps
+ * test_assert() and a macro to wrap that function. For example:
+ *
+ * static int check_oid_loc(const char *loc, const char *check,
+ * struct object_id *a, struct object_id *b)
+ * {
+ * int res = test_assert(loc, check, oideq(a, b));
+ *
+ * if (res) {
+ * test_msg(" left: %s", oid_to_hex(a);
+ * test_msg(" right: %s", oid_to_hex(a);
+ *
+ * }
+ * return res;
+ * }
+ *
+ * #define check_oid(a, b) \
+ * check_oid_loc(TEST_LOCATION(), "oideq("#a", "#b")", a, b)
+ */
+int test_assert(const char *location, const char *check, int ok);
+
+/* Helper macro to pass the location to checks */
+#define TEST_LOCATION() TEST__MAKE_LOCATION(__LINE__)
+
+/* Check a boolean condition. */
+#define check(x) \
+ check_bool_loc(TEST_LOCATION(), #x, x)
+int check_bool_loc(const char *loc, const char *check, int ok);
+
+/*
+ * Compare two integers. Prints a message with the two values if the
+ * comparison fails. NB this is not thread safe.
+ */
+#define check_int(a, op, b) \
+ (test__tmp[0].i = (a), test__tmp[1].i = (b), \
+ check_int_loc(TEST_LOCATION(), #a" "#op" "#b, \
+ test__tmp[0].i op test__tmp[1].i, a, b))
+int check_int_loc(const char *loc, const char *check, int ok,
+ intmax_t a, intmax_t b);
+
+/*
+ * Compare two unsigned integers. Prints a message with the two values
+ * if the comparison fails. NB this is not thread safe.
+ */
+#define check_uint(a, op, b) \
+ (test__tmp[0].u = (a), test__tmp[1].u = (b), \
+ check_uint_loc(TEST_LOCATION(), #a" "#op" "#b, \
+ test__tmp[0].u op test__tmp[1].u, a, b))
+int check_uint_loc(const char *loc, const char *check, int ok,
+ uintmax_t a, uintmax_t b);
+
+/*
+ * Compare two chars. Prints a message with the two values if the
+ * comparison fails. NB this is not thread safe.
+ */
+#define check_char(a, op, b) \
+ (test__tmp[0].c = (a), test__tmp[1].c = (b), \
+ check_char_loc(TEST_LOCATION(), #a" "#op" "#b, \
+ test__tmp[0].c op test__tmp[1].c, a, b))
+int check_char_loc(const char *loc, const char *check, int ok,
+ char a, char b);
+
+/* Check whether two strings are equal. */
+#define check_str(a, b) \
+ check_str_loc(TEST_LOCATION(), "!strcmp("#a", "#b")", a, b)
+int check_str_loc(const char *loc, const char *check,
+ const char *a, const char *b);
+
+/*
+ * Wrap a check that is known to fail. If the check succeeds then the
+ * test will fail. Returns 0 if the check fails, -1 if it
+ * succeeds. For example:
+ *
+ * TEST_TODO(check(0));
+ */
+#define TEST_TODO(check) \
+ (test__todo_begin(), test__todo_end(TEST_LOCATION(), #check, check))
+
+/* Private helpers */
+
+#define TEST__STR(x) #x
+#define TEST__MAKE_LOCATION(line) __FILE__ ":" TEST__STR(line)
+
+union test__tmp {
+ intmax_t i;
+ uintmax_t u;
+ char c;
+};
+
+extern union test__tmp test__tmp[2];
+
+int test__run_begin(void);
+__attribute__((format (printf, 3, 4)))
+int test__run_end(int, const char *, const char *, ...);
+void test__todo_begin(void);
+int test__todo_end(const char *, const char *, int);
+
+#endif /* TEST_LIB_H */
--
2.40.0.685.gf2f8f6435f
^ permalink raw reply related [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-05-10 15:46 ` Phillip Wood
@ 2023-05-11 23:16 ` Glen Choo
2023-05-18 10:04 ` Phillip Wood
2023-06-21 15:57 ` Linus Arver
1 sibling, 1 reply; 29+ messages in thread
From: Glen Choo @ 2023-05-11 23:16 UTC (permalink / raw)
To: Phillip Wood, Calvin Wan, phillip.wood
Cc: git, Ævar Arnfjörð Bjarmason
Phillip Wood <phillip.wood123@gmail.com> writes:
> It would be possible, but I'd want to see some macros for checking
> invariants that print the source location as well as their
> arguments. I'd be interested to hear what you think of the alternative
> approach in the patch below.
I personally (not representing the Google folks as a whole) find the
option of implementing our own C TAP suite quite appealing. As your
patch shows, we can get pretty far even with a rudimentary initial
version, and I don't see why we couldn't refine it as we go along.
I'm concerned about the overhead of relying on a third party library
that implements something moderately complex and isn't widely used,
It's much easier to patch and read code that we maintain ourselves, e.g.
if we need to fix bugs/add features/understand its behavior. These risks
are quite real, especially when it's not clear what kind of support we
can rely on from C TAP Harness's side.
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-05-11 23:16 ` Glen Choo
@ 2023-05-18 10:04 ` Phillip Wood
0 siblings, 0 replies; 29+ messages in thread
From: Phillip Wood @ 2023-05-18 10:04 UTC (permalink / raw)
To: Glen Choo, Calvin Wan, phillip.wood
Cc: git, Ævar Arnfjörð Bjarmason, Josh Steadmon
On 12/05/2023 00:16, Glen Choo wrote:
> Phillip Wood <phillip.wood123@gmail.com> writes:
>
>> It would be possible, but I'd want to see some macros for checking
>> invariants that print the source location as well as their
>> arguments. I'd be interested to hear what you think of the alternative
>> approach in the patch below.
>
> I personally (not representing the Google folks as a whole) find the
> option of implementing our own C TAP suite quite appealing. As your
> patch shows, we can get pretty far even with a rudimentary initial
> version, and I don't see why we couldn't refine it as we go along.
>
> I'm concerned about the overhead of relying on a third party library
> that implements something moderately complex and isn't widely used,
> It's much easier to patch and read code that we maintain ourselves, e.g.
> if we need to fix bugs/add features/understand its behavior. These risks
> are quite real, especially when it's not clear what kind of support we
> can rely on from C TAP Harness's side.
Yes I think for something like this the advantages of having our own
implementation that implements exactly what we want and runs on all the
platforms we support outweigh the effort of writing the implementation.
As there is already talk of having to customize c-tap-harness it is not
clear there is even much effort saved going that route.
Best Wishes
Phillip
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-05-10 15:46 ` Phillip Wood
2023-05-11 23:16 ` Glen Choo
@ 2023-06-21 15:57 ` Linus Arver
2023-06-26 13:15 ` Phillip Wood
1 sibling, 1 reply; 29+ messages in thread
From: Linus Arver @ 2023-06-21 15:57 UTC (permalink / raw)
To: Phillip Wood, Calvin Wan, phillip.wood
Cc: git, Ævar Arnfjörð Bjarmason
Hello,
Phillip Wood <phillip.wood123@gmail.com> writes:
> I'd be interested to hear what you think of the alternative
> approach in the patch below. The patch can be fetched with
>
> git fetch https://github.com/phillipwood/git unit-tests
I tried out your unit testing framework to see how it works in practice.
It has most of what I'd expect in a test framework, e.g. showing file
and line number in a test failure. Here are some changes I'd like to
see:
- Make the 'TEST' macro accept the test description first. Or, keep the
'TEST' macro but also name a new macro 'IT' that accepts the
description first, to encourage usage that reads in a
behavior-driven-development (BDD) style, like 'IT("should accept foo",
t_bar(...))'. I find some test descriptions easier to write this way.
- The 'check_int(result, ==, expected)' usage would be better without
the commas, like 'check_int(result == expected)'. Is this possible?
- It would be nice to report the test name for failing tests, so that
this output can be parsed to check for the failing test name.
There are other smaller changes I'd like (such as optional
colorization), but I think these bits can be sorted out much later.
My patch can be found at
git fetch https://github.com/listx/git trailer-unit-tests
and also inline at the bottom of this email message.
Build and test with
make unit-tests && t/unit-tests/t-trailer
Sample output:
$ make -j47 unit-tests && t/unit-tests/t-trailer
CC t/unit-tests/t-trailer.o
LINK t/unit-tests/t-trailer
ok 1 - empty trailer_item and arg_item
ok 2 - identical
ok 3 - case should not matter
ok 4 - arg_item is longer than trailer_item
ok 5 - trailer_item is longer than arg_item
ok 6 - no similarity
ok 7 - accept WHERE_AFTER
ok 8 - accept WHERE_END
ok 9 - reject WHERE_END
ok 10 - token with trailing punctuation (colon)
ok 11 - token without trailing punctuation
ok 12 - token with spaces with trailing punctuation (colon)
ok 13 - token with spaces without trailing punctuation
ok 14 - token with leading non-separator punctuation
ok 15 - token with leading non-separator punctuation and space
ok 16 - one-letter token
ok 17 - token with punctuation in-between
ok 18 - token with multiple leading punctuation chars
# check "result == expected" failed at t/unit-tests/t-trailer.c:25
# left: 1
# right: 0
not ok 19 - empty trailer_item
# check "result == expected" failed at t/unit-tests/t-trailer.c:25
# left: 1
# right: 0
not ok 20 - empty arg_item
1..20
Note that some of the tests I expect to fail are passing. These may be
bugs but I am not sure because I don't fully understand the
interpret-trailers machinery yet.
Linus
---- >8 ----
From e5907b328d35eb37cfd1feb3a2431cb672beb4c8 Mon Sep 17 00:00:00 2001
From: Linus Arver <linusa@google.com>
Date: Thu, 15 Jun 2023 18:25:47 -0700
Subject: [PATCH] add some unit tests for interpret-trailers
Signed-off-by: Linus Arver <linusa@google.com>
---
Makefile | 5 +++
t/unit-tests/t-trailer.c | 77 ++++++++++++++++++++++++++++++++++++++++
trailer.c | 33 ++---------------
trailer.h | 32 +++++++++++++++++
4 files changed, 117 insertions(+), 30 deletions(-)
create mode 100644 t/unit-tests/t-trailer.c
diff --git a/Makefile b/Makefile
index f0ca5bae4b..9d68518168 100644
--- a/Makefile
+++ b/Makefile
@@ -2668,6 +2668,7 @@ scalar-objs: $(SCALAR_OBJS)
UNIT_TEST_PROGRAMS += t-basic
UNIT_TEST_PROGRAMS += t-strbuf
+UNIT_TEST_PROGRAMS += t-trailer
UNIT_TEST_PROGS = $(patsubst %,t/unit-tests/%$X,$(UNIT_TEST_PROGRAMS))
UNIT_TEST_OBJS = $(patsubst %,t/unit-tests/%.o,$(UNIT_TEST_PROGRAMS))
UNIT_TEST_OBJS += t/unit-tests/test-lib.o
@@ -3848,5 +3849,9 @@ t/unit-tests/t-strbuf$X: t/unit-tests/t-strbuf.o t/unit-tests/test-lib.o $(GITLI
$(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
$(filter %.o,$^) $(filter %.a,$^) $(LIBS)
+t/unit-tests/t-trailer$X: t/unit-tests/t-trailer.o t/unit-tests/test-lib.o builtin/interpret-trailers.o $(GITLIBS) GIT-LDFLAGS
+ $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
+ $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
+
.PHONY: unit-tests
unit-tests: $(UNIT_TEST_PROGS)
diff --git a/t/unit-tests/t-trailer.c b/t/unit-tests/t-trailer.c
new file mode 100644
index 0000000000..150b5606fa
--- /dev/null
+++ b/t/unit-tests/t-trailer.c
@@ -0,0 +1,77 @@
+#include "test-lib.h"
+#include "trailer.h"
+
+static void t_token_len_without_separator(const char *token, size_t expected)
+{
+ size_t result;
+ result = token_len_without_separator(token, strlen(token));
+ check_uint(result, ==, expected);
+}
+
+static void t_after_or_end(enum trailer_where where, int expected)
+{
+ size_t result;
+ result = after_or_end(where);
+ check_int(result, ==, expected);
+}
+
+
+static void t_same_token(char *a, char *b, int expected)
+{
+ struct trailer_item trailer_item = { .token = a };
+ struct arg_item arg_item = { .token = b };
+ size_t result;
+ result = same_token(&trailer_item, &arg_item);
+ check_int(result, ==, expected);
+}
+
+int cmd_main(int argc, const char **argv)
+{
+ TEST(t_same_token("", "", 1),
+ "empty trailer_item and arg_item");
+ TEST(t_same_token("foo", "foo", 1),
+ "identical");
+ TEST(t_same_token("foo", "FOO", 1),
+ "case should not matter");
+ TEST(t_same_token("foo", "foobar", 1),
+ "arg_item is longer than trailer_item");
+ TEST(t_same_token("foobar", "foo", 1),
+ "trailer_item is longer than arg_item");
+ TEST(t_same_token("foo", "bar", 0),
+ "no similarity");
+
+ TEST(t_after_or_end(WHERE_AFTER, 1), "accept WHERE_AFTER");
+ TEST(t_after_or_end(WHERE_END, 1), "accept WHERE_END");
+ TEST(t_after_or_end(WHERE_DEFAULT, 0), "reject WHERE_END");
+
+ TEST(t_token_len_without_separator("Signed-off-by:", 13),
+ "token with trailing punctuation (colon)");
+ TEST(t_token_len_without_separator("Signed-off-by", 13),
+ "token without trailing punctuation");
+ TEST(t_token_len_without_separator("Foo bar:", 7),
+ "token with spaces with trailing punctuation (colon)");
+ TEST(t_token_len_without_separator("Foo bar", 7),
+ "token with spaces without trailing punctuation");
+ TEST(t_token_len_without_separator("-Foo bar:", 8),
+ "token with leading non-separator punctuation");
+ TEST(t_token_len_without_separator("- Foo bar:", 9),
+ "token with leading non-separator punctuation and space");
+ TEST(t_token_len_without_separator("F:", 1),
+ "one-letter token");
+ TEST(t_token_len_without_separator("abc%de#f@", 8),
+ "token with punctuation in-between");
+ TEST(t_token_len_without_separator(":;*%_.#f@", 8),
+ "token with multiple leading punctuation chars");
+
+ /*
+ * These tests fail unexpectedly. They are probably bugs, although it may be
+ * the case that these bugs never bubble up to the user because of other
+ * checks we do elsewhere up the stack.
+ */
+ TEST(t_same_token("", "foo", 0),
+ "empty trailer_item");
+ TEST(t_same_token("foo", "", 0),
+ "empty arg_item");
+
+ return test_done();
+}
diff --git a/trailer.c b/trailer.c
index a2c3ed6f28..9f59d8d7a6 100644
--- a/trailer.c
+++ b/trailer.c
@@ -13,35 +13,8 @@
* Copyright (c) 2013, 2014 Christian Couder <chriscool@tuxfamily.org>
*/
-struct conf_info {
- char *name;
- char *key;
- char *command;
- char *cmd;
- enum trailer_where where;
- enum trailer_if_exists if_exists;
- enum trailer_if_missing if_missing;
-};
-
static struct conf_info default_conf_info;
-struct trailer_item {
- struct list_head list;
- /*
- * If this is not a trailer line, the line is stored in value
- * (excluding the terminating newline) and token is NULL.
- */
- char *token;
- char *value;
-};
-
-struct arg_item {
- struct list_head list;
- char *token;
- char *value;
- struct conf_info conf;
-};
-
static LIST_HEAD(conf_head);
static char *separators = ":";
@@ -62,7 +35,7 @@ static const char *git_generated_prefixes[] = {
pos != (head); \
pos = is_reverse ? pos->prev : pos->next)
-static int after_or_end(enum trailer_where where)
+int after_or_end(enum trailer_where where)
{
return (where == WHERE_AFTER) || (where == WHERE_END);
}
@@ -73,14 +46,14 @@ static int after_or_end(enum trailer_where where)
* 13, stripping the trailing punctuation but retaining
* internal punctuation.
*/
-static size_t token_len_without_separator(const char *token, size_t len)
+size_t token_len_without_separator(const char *token, size_t len)
{
while (len > 0 && !isalnum(token[len - 1]))
len--;
return len;
}
-static int same_token(struct trailer_item *a, struct arg_item *b)
+int same_token(struct trailer_item *a, struct arg_item *b)
{
size_t a_len, b_len, min_len;
diff --git a/trailer.h b/trailer.h
index 795d2fccfd..b2031eb305 100644
--- a/trailer.h
+++ b/trailer.h
@@ -146,4 +146,36 @@ int trailer_iterator_advance(struct trailer_iterator *iter);
*/
void trailer_iterator_release(struct trailer_iterator *iter);
+int after_or_end(enum trailer_where where);
+size_t token_len_without_separator(const char *token, size_t len);
+
+struct conf_info {
+ char *name;
+ char *key;
+ char *command;
+ char *cmd;
+ enum trailer_where where;
+ enum trailer_if_exists if_exists;
+ enum trailer_if_missing if_missing;
+};
+
+struct trailer_item {
+ struct list_head list;
+ /*
+ * If this is not a trailer line, the line is stored in value
+ * (excluding the terminating newline) and token is NULL.
+ */
+ char *token;
+ char *value;
+};
+
+struct arg_item {
+ struct list_head list;
+ char *token;
+ char *value;
+ struct conf_info conf;
+};
+
+int same_token(struct trailer_item *a, struct arg_item *b);
+
#endif /* TRAILER_H */
--
2.41.0.162.gfafddb0af9-goog
^ permalink raw reply related [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-06-21 15:57 ` Linus Arver
@ 2023-06-26 13:15 ` Phillip Wood
2023-06-28 21:17 ` Linus Arver
2023-06-29 5:52 ` Oswald Buddenhagen
0 siblings, 2 replies; 29+ messages in thread
From: Phillip Wood @ 2023-06-26 13:15 UTC (permalink / raw)
To: Linus Arver, phillip.wood
Cc: git, Ævar Arnfjörð Bjarmason, Josh Steadmon,
Calvin Wan
Hi Linus
On 21/06/2023 16:57, Linus Arver wrote:
> Hello,
>
> Phillip Wood <phillip.wood123@gmail.com> writes:
>> I'd be interested to hear what you think of the alternative
>> approach in the patch below. The patch can be fetched with
>>
>> git fetch https://github.com/phillipwood/git unit-tests
>
> I tried out your unit testing framework to see how it works in practice.
> It has most of what I'd expect in a test framework, e.g. showing file
> and line number in a test failure. Here are some changes I'd like to
> see:
Thanks for taking the time to try this out.
> - Make the 'TEST' macro accept the test description first. Or, keep the
> 'TEST' macro but also name a new macro 'IT' that accepts the
> description first, to encourage usage that reads in a
> behavior-driven-development (BDD) style, like 'IT("should accept foo",
> t_bar(...))'. I find some test descriptions easier to write this way.
The test description is a printf style format string followed by
arguments. This allows parameterized tests to include the parameter
values in the description to aid debugging but it means the test
function must be the first parameter. We could have IT("should accept
%d", t(), i) but that would be a bit weird.
I hope that we will end up with some guidelines for writing unit tests
that recommend testing the expect behavior and to avoid testing
implementation details. I'm don't think we necessarily need an IT()
macro to do that.
> - The 'check_int(result, ==, expected)' usage would be better without
> the commas, like 'check_int(result == expected)'. Is this possible?
I don't particularly like the comma's either but in order to have decent
diagnostic messages we need to pass 'result' and 'expected' as separate
parameters so we can print them. We could have separate macros for
different comparisons check_int_eq(a, b), check_int_lt(a, b), ... I
don't have a strong preference either way.
> - It would be nice to report the test name for failing tests, so that
> this output can be parsed to check for the failing test name.
I'm not quite sure what you mean here. The test name in printed in the
"not ok" lines for failing tests as shown in the output from your tests
below.
> There are other smaller changes I'd like (such as optional
> colorization), but I think these bits can be sorted out much later.
Sure
> My patch can be found at
>
> git fetch https://github.com/listx/git trailer-unit-tests
>
> and also inline at the bottom of this email message.
>
> Build and test with
>
> make unit-tests && t/unit-tests/t-trailer
>
> Sample output:
>
> $ make -j47 unit-tests && t/unit-tests/t-trailer
> CC t/unit-tests/t-trailer.o
> LINK t/unit-tests/t-trailer
> ok 1 - empty trailer_item and arg_item
> ok 2 - identical
> ok 3 - case should not matter
> ok 4 - arg_item is longer than trailer_item
> ok 5 - trailer_item is longer than arg_item
> ok 6 - no similarity
> ok 7 - accept WHERE_AFTER
> ok 8 - accept WHERE_END
> ok 9 - reject WHERE_END
> ok 10 - token with trailing punctuation (colon)
> ok 11 - token without trailing punctuation
> ok 12 - token with spaces with trailing punctuation (colon)
> ok 13 - token with spaces without trailing punctuation
> ok 14 - token with leading non-separator punctuation
> ok 15 - token with leading non-separator punctuation and space
> ok 16 - one-letter token
> ok 17 - token with punctuation in-between
> ok 18 - token with multiple leading punctuation chars
> # check "result == expected" failed at t/unit-tests/t-trailer.c:25
> # left: 1
> # right: 0
> not ok 19 - empty trailer_item
This test fail because same_token() tests that the longer of its two
arguments starts with the shorter argument, not that the two arguments
are equal. As the shorter argument is empty in this case it returns
true, not false.
> # check "result == expected" failed at t/unit-tests/t-trailer.c:25
> # left: 1
> # right: 0
> not ok 20 - empty arg_item
> 1..20
>
> Note that some of the tests I expect to fail are passing.
Which ones are those?
> These may be
> bugs but I am not sure because I don't fully understand the
> interpret-trailers machinery yet.
Thanks for sharing these tests, it is helpful to see examples of the
test framework being used. In this case the tests are checking static
functions from trailer.c. This mean that you have to export a number of
functions and structs that are really implementation details. Where
possible I think we want to concentrate our testing effort on the public
API as (a) that is what matters to callers and (b) it avoids having to
export internal functions and encourages people to test behavior rather
than implementation details. There may be some cases where there are
internal helpers that would benefit from tests. In those cases we can
either hide the export behind a #define so it is only visible to the
test code or just #include the file being tested in the test file (gross
but works). An alternative to both of those is for the tests to live in
the implementation files and have them conditionally compiled (a bit
like rust).
Thanks again for your comments, Best Wishes
Phillip
> Linus
>
> ---- >8 ----
> From e5907b328d35eb37cfd1feb3a2431cb672beb4c8 Mon Sep 17 00:00:00 2001
> From: Linus Arver <linusa@google.com>
> Date: Thu, 15 Jun 2023 18:25:47 -0700
> Subject: [PATCH] add some unit tests for interpret-trailers
>
> Signed-off-by: Linus Arver <linusa@google.com>
> ---
> Makefile | 5 +++
> t/unit-tests/t-trailer.c | 77 ++++++++++++++++++++++++++++++++++++++++
> trailer.c | 33 ++---------------
> trailer.h | 32 +++++++++++++++++
> 4 files changed, 117 insertions(+), 30 deletions(-)
> create mode 100644 t/unit-tests/t-trailer.c
>
> diff --git a/Makefile b/Makefile
> index f0ca5bae4b..9d68518168 100644
> --- a/Makefile
> +++ b/Makefile
> @@ -2668,6 +2668,7 @@ scalar-objs: $(SCALAR_OBJS)
>
> UNIT_TEST_PROGRAMS += t-basic
> UNIT_TEST_PROGRAMS += t-strbuf
> +UNIT_TEST_PROGRAMS += t-trailer
> UNIT_TEST_PROGS = $(patsubst %,t/unit-tests/%$X,$(UNIT_TEST_PROGRAMS))
> UNIT_TEST_OBJS = $(patsubst %,t/unit-tests/%.o,$(UNIT_TEST_PROGRAMS))
> UNIT_TEST_OBJS += t/unit-tests/test-lib.o
> @@ -3848,5 +3849,9 @@ t/unit-tests/t-strbuf$X: t/unit-tests/t-strbuf.o t/unit-tests/test-lib.o $(GITLI
> $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
> $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
>
> +t/unit-tests/t-trailer$X: t/unit-tests/t-trailer.o t/unit-tests/test-lib.o builtin/interpret-trailers.o $(GITLIBS) GIT-LDFLAGS
> + $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
> + $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
> +
> .PHONY: unit-tests
> unit-tests: $(UNIT_TEST_PROGS)
> diff --git a/t/unit-tests/t-trailer.c b/t/unit-tests/t-trailer.c
> new file mode 100644
> index 0000000000..150b5606fa
> --- /dev/null
> +++ b/t/unit-tests/t-trailer.c
> @@ -0,0 +1,77 @@
> +#include "test-lib.h"
> +#include "trailer.h"
> +
> +static void t_token_len_without_separator(const char *token, size_t expected)
> +{
> + size_t result;
> + result = token_len_without_separator(token, strlen(token));
> + check_uint(result, ==, expected);
> +}
> +
> +static void t_after_or_end(enum trailer_where where, int expected)
> +{
> + size_t result;
> + result = after_or_end(where);
> + check_int(result, ==, expected);
> +}
> +
> +
> +static void t_same_token(char *a, char *b, int expected)
> +{
> + struct trailer_item trailer_item = { .token = a };
> + struct arg_item arg_item = { .token = b };
> + size_t result;
> + result = same_token(&trailer_item, &arg_item);
> + check_int(result, ==, expected);
> +}
> +
> +int cmd_main(int argc, const char **argv)
> +{
> + TEST(t_same_token("", "", 1),
> + "empty trailer_item and arg_item");
> + TEST(t_same_token("foo", "foo", 1),
> + "identical");
> + TEST(t_same_token("foo", "FOO", 1),
> + "case should not matter");
> + TEST(t_same_token("foo", "foobar", 1),
> + "arg_item is longer than trailer_item");
> + TEST(t_same_token("foobar", "foo", 1),
> + "trailer_item is longer than arg_item");
> + TEST(t_same_token("foo", "bar", 0),
> + "no similarity");
> +
> + TEST(t_after_or_end(WHERE_AFTER, 1), "accept WHERE_AFTER");
> + TEST(t_after_or_end(WHERE_END, 1), "accept WHERE_END");
> + TEST(t_after_or_end(WHERE_DEFAULT, 0), "reject WHERE_END");
> +
> + TEST(t_token_len_without_separator("Signed-off-by:", 13),
> + "token with trailing punctuation (colon)");
> + TEST(t_token_len_without_separator("Signed-off-by", 13),
> + "token without trailing punctuation");
> + TEST(t_token_len_without_separator("Foo bar:", 7),
> + "token with spaces with trailing punctuation (colon)");
> + TEST(t_token_len_without_separator("Foo bar", 7),
> + "token with spaces without trailing punctuation");
> + TEST(t_token_len_without_separator("-Foo bar:", 8),
> + "token with leading non-separator punctuation");
> + TEST(t_token_len_without_separator("- Foo bar:", 9),
> + "token with leading non-separator punctuation and space");
> + TEST(t_token_len_without_separator("F:", 1),
> + "one-letter token");
> + TEST(t_token_len_without_separator("abc%de#f@", 8),
> + "token with punctuation in-between");
> + TEST(t_token_len_without_separator(":;*%_.#f@", 8),
> + "token with multiple leading punctuation chars");
> +
> + /*
> + * These tests fail unexpectedly. They are probably bugs, although it may be
> + * the case that these bugs never bubble up to the user because of other
> + * checks we do elsewhere up the stack.
> + */
> + TEST(t_same_token("", "foo", 0),
> + "empty trailer_item");
> + TEST(t_same_token("foo", "", 0),
> + "empty arg_item");
> +
> + return test_done();
> +}
> diff --git a/trailer.c b/trailer.c
> index a2c3ed6f28..9f59d8d7a6 100644
> --- a/trailer.c
> +++ b/trailer.c
> @@ -13,35 +13,8 @@
> * Copyright (c) 2013, 2014 Christian Couder <chriscool@tuxfamily.org>
> */
>
> -struct conf_info {
> - char *name;
> - char *key;
> - char *command;
> - char *cmd;
> - enum trailer_where where;
> - enum trailer_if_exists if_exists;
> - enum trailer_if_missing if_missing;
> -};
> -
> static struct conf_info default_conf_info;
>
> -struct trailer_item {
> - struct list_head list;
> - /*
> - * If this is not a trailer line, the line is stored in value
> - * (excluding the terminating newline) and token is NULL.
> - */
> - char *token;
> - char *value;
> -};
> -
> -struct arg_item {
> - struct list_head list;
> - char *token;
> - char *value;
> - struct conf_info conf;
> -};
> -
> static LIST_HEAD(conf_head);
>
> static char *separators = ":";
> @@ -62,7 +35,7 @@ static const char *git_generated_prefixes[] = {
> pos != (head); \
> pos = is_reverse ? pos->prev : pos->next)
>
> -static int after_or_end(enum trailer_where where)
> +int after_or_end(enum trailer_where where)
> {
> return (where == WHERE_AFTER) || (where == WHERE_END);
> }
> @@ -73,14 +46,14 @@ static int after_or_end(enum trailer_where where)
> * 13, stripping the trailing punctuation but retaining
> * internal punctuation.
> */
> -static size_t token_len_without_separator(const char *token, size_t len)
> +size_t token_len_without_separator(const char *token, size_t len)
> {
> while (len > 0 && !isalnum(token[len - 1]))
> len--;
> return len;
> }
>
> -static int same_token(struct trailer_item *a, struct arg_item *b)
> +int same_token(struct trailer_item *a, struct arg_item *b)
> {
> size_t a_len, b_len, min_len;
>
> diff --git a/trailer.h b/trailer.h
> index 795d2fccfd..b2031eb305 100644
> --- a/trailer.h
> +++ b/trailer.h
> @@ -146,4 +146,36 @@ int trailer_iterator_advance(struct trailer_iterator *iter);
> */
> void trailer_iterator_release(struct trailer_iterator *iter);
>
> +int after_or_end(enum trailer_where where);
> +size_t token_len_without_separator(const char *token, size_t len);
> +
> +struct conf_info {
> + char *name;
> + char *key;
> + char *command;
> + char *cmd;
> + enum trailer_where where;
> + enum trailer_if_exists if_exists;
> + enum trailer_if_missing if_missing;
> +};
> +
> +struct trailer_item {
> + struct list_head list;
> + /*
> + * If this is not a trailer line, the line is stored in value
> + * (excluding the terminating newline) and token is NULL.
> + */
> + char *token;
> + char *value;
> +};
> +
> +struct arg_item {
> + struct list_head list;
> + char *token;
> + char *value;
> + struct conf_info conf;
> +};
> +
> +int same_token(struct trailer_item *a, struct arg_item *b);
> +
> #endif /* TRAILER_H */
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-06-26 13:15 ` Phillip Wood
@ 2023-06-28 21:17 ` Linus Arver
2023-06-29 5:52 ` Oswald Buddenhagen
1 sibling, 0 replies; 29+ messages in thread
From: Linus Arver @ 2023-06-28 21:17 UTC (permalink / raw)
To: Phillip Wood, phillip.wood
Cc: git, Ævar Arnfjörð Bjarmason, Josh Steadmon,
Calvin Wan
Phillip Wood <phillip.wood123@gmail.com> writes:
>> - Make the 'TEST' macro accept the test description first. Or, keep the
>> 'TEST' macro but also name a new macro 'IT' that accepts the
>> description first, to encourage usage that reads in a
>> behavior-driven-development (BDD) style, like 'IT("should accept foo",
>> t_bar(...))'. I find some test descriptions easier to write this way.
>
> The test description is a printf style format string followed by
> arguments. This allows parameterized tests to include the parameter
> values in the description to aid debugging but it means the test
> function must be the first parameter. We could have IT("should accept
> %d", t(), i) but that would be a bit weird.
I see. I agree that
IT("should accept %d", t(), i)
would look rather odd.
> I hope that we will end up with some guidelines for writing unit tests
> that recommend testing the expect behavior and to avoid testing
> implementation details. I'm don't think we necessarily need an IT()
> macro to do that.
The IT() macro idea was only suggested as a stylistic choice. I also
agree that an IT() macro is not necessary to test the expected behavior.
>> - The 'check_int(result, ==, expected)' usage would be better without
>> the commas, like 'check_int(result == expected)'. Is this possible?
>
> I don't particularly like the comma's either but in order to have decent
> diagnostic messages we need to pass 'result' and 'expected' as separate
> parameters so we can print them. We could have separate macros for
> different comparisons check_int_eq(a, b), check_int_lt(a, b), ... I
> don't have a strong preference either way.
I see. If it's a one-time cost to do this (enable comma-less usage) then I
strongly favor doing so. But I guess it's the kind of thing where we
could easily have a mechanical conversion from one style to the other in
the future, so it can certainly wait.
>> - It would be nice to report the test name for failing tests, so that
>> this output can be parsed to check for the failing test name.
>
> I'm not quite sure what you mean here. The test name in printed in the
> "not ok" lines for failing tests as shown in the output from your tests
> below.
It would be nice if the failing output (in my example) also printed the
"t_same_token" function name.
>> There are other smaller changes I'd like (such as optional
>> colorization), but I think these bits can be sorted out much later.
>
> Sure
>
>> My patch can be found at
>>
>> git fetch https://github.com/listx/git trailer-unit-tests
>>
>> and also inline at the bottom of this email message.
>>
>> Build and test with
>>
>> make unit-tests && t/unit-tests/t-trailer
>>
>> Sample output:
>>
>> $ make -j47 unit-tests && t/unit-tests/t-trailer
>> CC t/unit-tests/t-trailer.o
>> LINK t/unit-tests/t-trailer
>> ok 1 - empty trailer_item and arg_item
>> ok 2 - identical
>> ok 3 - case should not matter
>> ok 4 - arg_item is longer than trailer_item
>> ok 5 - trailer_item is longer than arg_item
>> ok 6 - no similarity
>> ok 7 - accept WHERE_AFTER
>> ok 8 - accept WHERE_END
>> ok 9 - reject WHERE_END
>> ok 10 - token with trailing punctuation (colon)
>> ok 11 - token without trailing punctuation
>> ok 12 - token with spaces with trailing punctuation (colon)
>> ok 13 - token with spaces without trailing punctuation
>> ok 14 - token with leading non-separator punctuation
>> ok 15 - token with leading non-separator punctuation and space
>> ok 16 - one-letter token
>> ok 17 - token with punctuation in-between
>> ok 18 - token with multiple leading punctuation chars
>> # check "result == expected" failed at t/unit-tests/t-trailer.c:25
>> # left: 1
>> # right: 0
>> not ok 19 - empty trailer_item
>
> This test fail because same_token() tests that the longer of its two
> arguments starts with the shorter argument, not that the two arguments
> are equal. As the shorter argument is empty in this case it returns
> true, not false.
To me this behavior is surprising because this means that we could have,
for example,
1. "" and "foo"
2. "" and "bar"
and in both cases the empty string would be considered the "same token"
as "foo" and "bar". I think we should rename the function to something
other than same_token(), perhaps same_prefix_token() or similar.
>> # check "result == expected" failed at t/unit-tests/t-trailer.c:25
>> # left: 1
>> # right: 0
>> not ok 20 - empty arg_item
>> 1..20
>>
>> Note that some of the tests I expect to fail are passing.
>
> Which ones are those?
Oops. Disregard my comment here (it was from an earlier draft of the
email).
>> These may be
>> bugs but I am not sure because I don't fully understand the
>> interpret-trailers machinery yet.
>
> Thanks for sharing these tests, it is helpful to see examples of the
> test framework being used. In this case the tests are checking static
> functions from trailer.c. This mean that you have to export a number of
> functions and structs that are really implementation details.
Indeed. I had a feeling that what I was doing was probably not the best
idea (exposing previously private internals of the trailer.c logic). I
agree with your analysis.
> Where
> possible I think we want to concentrate our testing effort on the public
> API as (a) that is what matters to callers and (b) it avoids having to
> export internal functions and encourages people to test behavior rather
> than implementation details.
Agreed. This is the "libification" work that needs to be done here so
that we can test the public API behaviors, not the internal functions. I
merely chose to sidestep that work in my example to keep the patch small.
> There may be some cases where there are
> internal helpers that would benefit from tests. In those cases we can
> either hide the export behind a #define so it is only visible to the
> test code or just #include the file being tested in the test file (gross
> but works).
My C is rusty so it would be great if such an example can be contributed
to the codebase in the future.
> An alternative to both of those is for the tests to live in
> the implementation files and have them conditionally compiled (a bit
> like rust).
Ditto.
> Thanks again for your comments, Best Wishes
You're welcome. Thank you for showing us what a minimal test framework
could look like.
Linus
> Phillip
>
>
>> Linus
>>
>> ---- >8 ----
>> From e5907b328d35eb37cfd1feb3a2431cb672beb4c8 Mon Sep 17 00:00:00 2001
>> From: Linus Arver <linusa@google.com>
>> Date: Thu, 15 Jun 2023 18:25:47 -0700
>> Subject: [PATCH] add some unit tests for interpret-trailers
>>
>> Signed-off-by: Linus Arver <linusa@google.com>
>> ---
>> Makefile | 5 +++
>> t/unit-tests/t-trailer.c | 77 ++++++++++++++++++++++++++++++++++++++++
>> trailer.c | 33 ++---------------
>> trailer.h | 32 +++++++++++++++++
>> 4 files changed, 117 insertions(+), 30 deletions(-)
>> create mode 100644 t/unit-tests/t-trailer.c
>>
>> diff --git a/Makefile b/Makefile
>> index f0ca5bae4b..9d68518168 100644
>> --- a/Makefile
>> +++ b/Makefile
>> @@ -2668,6 +2668,7 @@ scalar-objs: $(SCALAR_OBJS)
>>
>> UNIT_TEST_PROGRAMS += t-basic
>> UNIT_TEST_PROGRAMS += t-strbuf
>> +UNIT_TEST_PROGRAMS += t-trailer
>> UNIT_TEST_PROGS = $(patsubst %,t/unit-tests/%$X,$(UNIT_TEST_PROGRAMS))
>> UNIT_TEST_OBJS = $(patsubst %,t/unit-tests/%.o,$(UNIT_TEST_PROGRAMS))
>> UNIT_TEST_OBJS += t/unit-tests/test-lib.o
>> @@ -3848,5 +3849,9 @@ t/unit-tests/t-strbuf$X: t/unit-tests/t-strbuf.o t/unit-tests/test-lib.o $(GITLI
>> $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
>> $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
>>
>> +t/unit-tests/t-trailer$X: t/unit-tests/t-trailer.o t/unit-tests/test-lib.o builtin/interpret-trailers.o $(GITLIBS) GIT-LDFLAGS
>> + $(QUIET_LINK)$(CC) $(ALL_CFLAGS) -o $@ $(ALL_LDFLAGS) \
>> + $(filter %.o,$^) $(filter %.a,$^) $(LIBS)
>> +
>> .PHONY: unit-tests
>> unit-tests: $(UNIT_TEST_PROGS)
>> diff --git a/t/unit-tests/t-trailer.c b/t/unit-tests/t-trailer.c
>> new file mode 100644
>> index 0000000000..150b5606fa
>> --- /dev/null
>> +++ b/t/unit-tests/t-trailer.c
>> @@ -0,0 +1,77 @@
>> +#include "test-lib.h"
>> +#include "trailer.h"
>> +
>> +static void t_token_len_without_separator(const char *token, size_t expected)
>> +{
>> + size_t result;
>> + result = token_len_without_separator(token, strlen(token));
>> + check_uint(result, ==, expected);
>> +}
>> +
>> +static void t_after_or_end(enum trailer_where where, int expected)
>> +{
>> + size_t result;
>> + result = after_or_end(where);
>> + check_int(result, ==, expected);
>> +}
>> +
>> +
>> +static void t_same_token(char *a, char *b, int expected)
>> +{
>> + struct trailer_item trailer_item = { .token = a };
>> + struct arg_item arg_item = { .token = b };
>> + size_t result;
>> + result = same_token(&trailer_item, &arg_item);
>> + check_int(result, ==, expected);
>> +}
>> +
>> +int cmd_main(int argc, const char **argv)
>> +{
>> + TEST(t_same_token("", "", 1),
>> + "empty trailer_item and arg_item");
>> + TEST(t_same_token("foo", "foo", 1),
>> + "identical");
>> + TEST(t_same_token("foo", "FOO", 1),
>> + "case should not matter");
>> + TEST(t_same_token("foo", "foobar", 1),
>> + "arg_item is longer than trailer_item");
>> + TEST(t_same_token("foobar", "foo", 1),
>> + "trailer_item is longer than arg_item");
>> + TEST(t_same_token("foo", "bar", 0),
>> + "no similarity");
>> +
>> + TEST(t_after_or_end(WHERE_AFTER, 1), "accept WHERE_AFTER");
>> + TEST(t_after_or_end(WHERE_END, 1), "accept WHERE_END");
>> + TEST(t_after_or_end(WHERE_DEFAULT, 0), "reject WHERE_END");
>> +
>> + TEST(t_token_len_without_separator("Signed-off-by:", 13),
>> + "token with trailing punctuation (colon)");
>> + TEST(t_token_len_without_separator("Signed-off-by", 13),
>> + "token without trailing punctuation");
>> + TEST(t_token_len_without_separator("Foo bar:", 7),
>> + "token with spaces with trailing punctuation (colon)");
>> + TEST(t_token_len_without_separator("Foo bar", 7),
>> + "token with spaces without trailing punctuation");
>> + TEST(t_token_len_without_separator("-Foo bar:", 8),
>> + "token with leading non-separator punctuation");
>> + TEST(t_token_len_without_separator("- Foo bar:", 9),
>> + "token with leading non-separator punctuation and space");
>> + TEST(t_token_len_without_separator("F:", 1),
>> + "one-letter token");
>> + TEST(t_token_len_without_separator("abc%de#f@", 8),
>> + "token with punctuation in-between");
>> + TEST(t_token_len_without_separator(":;*%_.#f@", 8),
>> + "token with multiple leading punctuation chars");
>> +
>> + /*
>> + * These tests fail unexpectedly. They are probably bugs, although it may be
>> + * the case that these bugs never bubble up to the user because of other
>> + * checks we do elsewhere up the stack.
>> + */
>> + TEST(t_same_token("", "foo", 0),
>> + "empty trailer_item");
>> + TEST(t_same_token("foo", "", 0),
>> + "empty arg_item");
>> +
>> + return test_done();
>> +}
>> diff --git a/trailer.c b/trailer.c
>> index a2c3ed6f28..9f59d8d7a6 100644
>> --- a/trailer.c
>> +++ b/trailer.c
>> @@ -13,35 +13,8 @@
>> * Copyright (c) 2013, 2014 Christian Couder <chriscool@tuxfamily.org>
>> */
>>
>> -struct conf_info {
>> - char *name;
>> - char *key;
>> - char *command;
>> - char *cmd;
>> - enum trailer_where where;
>> - enum trailer_if_exists if_exists;
>> - enum trailer_if_missing if_missing;
>> -};
>> -
>> static struct conf_info default_conf_info;
>>
>> -struct trailer_item {
>> - struct list_head list;
>> - /*
>> - * If this is not a trailer line, the line is stored in value
>> - * (excluding the terminating newline) and token is NULL.
>> - */
>> - char *token;
>> - char *value;
>> -};
>> -
>> -struct arg_item {
>> - struct list_head list;
>> - char *token;
>> - char *value;
>> - struct conf_info conf;
>> -};
>> -
>> static LIST_HEAD(conf_head);
>>
>> static char *separators = ":";
>> @@ -62,7 +35,7 @@ static const char *git_generated_prefixes[] = {
>> pos != (head); \
>> pos = is_reverse ? pos->prev : pos->next)
>>
>> -static int after_or_end(enum trailer_where where)
>> +int after_or_end(enum trailer_where where)
>> {
>> return (where == WHERE_AFTER) || (where == WHERE_END);
>> }
>> @@ -73,14 +46,14 @@ static int after_or_end(enum trailer_where where)
>> * 13, stripping the trailing punctuation but retaining
>> * internal punctuation.
>> */
>> -static size_t token_len_without_separator(const char *token, size_t len)
>> +size_t token_len_without_separator(const char *token, size_t len)
>> {
>> while (len > 0 && !isalnum(token[len - 1]))
>> len--;
>> return len;
>> }
>>
>> -static int same_token(struct trailer_item *a, struct arg_item *b)
>> +int same_token(struct trailer_item *a, struct arg_item *b)
>> {
>> size_t a_len, b_len, min_len;
>>
>> diff --git a/trailer.h b/trailer.h
>> index 795d2fccfd..b2031eb305 100644
>> --- a/trailer.h
>> +++ b/trailer.h
>> @@ -146,4 +146,36 @@ int trailer_iterator_advance(struct trailer_iterator *iter);
>> */
>> void trailer_iterator_release(struct trailer_iterator *iter);
>>
>> +int after_or_end(enum trailer_where where);
>> +size_t token_len_without_separator(const char *token, size_t len);
>> +
>> +struct conf_info {
>> + char *name;
>> + char *key;
>> + char *command;
>> + char *cmd;
>> + enum trailer_where where;
>> + enum trailer_if_exists if_exists;
>> + enum trailer_if_missing if_missing;
>> +};
>> +
>> +struct trailer_item {
>> + struct list_head list;
>> + /*
>> + * If this is not a trailer line, the line is stored in value
>> + * (excluding the terminating newline) and token is NULL.
>> + */
>> + char *token;
>> + char *value;
>> +};
>> +
>> +struct arg_item {
>> + struct list_head list;
>> + char *token;
>> + char *value;
>> + struct conf_info conf;
>> +};
>> +
>> +int same_token(struct trailer_item *a, struct arg_item *b);
>> +
>> #endif /* TRAILER_H */
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-06-26 13:15 ` Phillip Wood
2023-06-28 21:17 ` Linus Arver
@ 2023-06-29 5:52 ` Oswald Buddenhagen
2023-06-30 9:48 ` Phillip Wood
1 sibling, 1 reply; 29+ messages in thread
From: Oswald Buddenhagen @ 2023-06-29 5:52 UTC (permalink / raw)
To: phillip.wood
Cc: Linus Arver, git, Ævar Arnfjörð Bjarmason,
Josh Steadmon, Calvin Wan
On Mon, Jun 26, 2023 at 02:15:39PM +0100, Phillip Wood wrote:
>On 21/06/2023 16:57, Linus Arver wrote:
>> - Make the 'TEST' macro accept the test description first. Or, keep
>> the
>> 'TEST' macro but also name a new macro 'IT' that accepts the
>> description first, to encourage usage that reads in a
>> behavior-driven-development (BDD) style, like 'IT("should accept foo",
>> t_bar(...))'. I find some test descriptions easier to write this way.
>
>The test description is a printf style format string followed by
>arguments. This allows parameterized tests to include the parameter
>values in the description to aid debugging but it means the test
>function must be the first parameter. We could have IT("should accept
>%d", t(), i) but that would be a bit weird.
>
with some minor preprocessor magic [1], you could make that
IT(("should accept %d", i), t(i))
which would be somewhat more noisy, but arguably even somewhat clearer.
notably,
IT("should accept foo", t())
would still work with the same macro.
[1] https://stackoverflow.com/a/62984543/3685191
somewhat on a tangent: it's also possible to overload macros on argument
count [2], which may also come in handy.
[2] https://stackoverflow.com/a/24028231/3685191
regards,
ossi
^ permalink raw reply [flat|nested] 29+ messages in thread
* Re: [RFC PATCH 1/2] Add C TAP harness
2023-06-29 5:52 ` Oswald Buddenhagen
@ 2023-06-30 9:48 ` Phillip Wood
0 siblings, 0 replies; 29+ messages in thread
From: Phillip Wood @ 2023-06-30 9:48 UTC (permalink / raw)
To: Oswald Buddenhagen, phillip.wood
Cc: Linus Arver, git, Ævar Arnfjörð Bjarmason,
Josh Steadmon, Calvin Wan
On 29/06/2023 06:52, Oswald Buddenhagen wrote:
> On Mon, Jun 26, 2023 at 02:15:39PM +0100, Phillip Wood wrote:
>> On 21/06/2023 16:57, Linus Arver wrote:
>>> - Make the 'TEST' macro accept the test description first. Or, keep the
>>> 'TEST' macro but also name a new macro 'IT' that accepts the
>>> description first, to encourage usage that reads in a
>>> behavior-driven-development (BDD) style, like 'IT("should accept
>>> foo",
>>> t_bar(...))'. I find some test descriptions easier to write this way.
>>
>> The test description is a printf style format string followed by
>> arguments. This allows parameterized tests to include the parameter
>> values in the description to aid debugging but it means the test
>> function must be the first parameter. We could have IT("should accept
>> %d", t(), i) but that would be a bit weird.
>>
> with some minor preprocessor magic [1], you could make that
>
> IT(("should accept %d", i), t(i))
>
> which would be somewhat more noisy, but arguably even somewhat clearer.
> notably,
>
> IT("should accept foo", t())
>
> would still work with the same macro.
>
> [1] https://stackoverflow.com/a/62984543/3685191
Thanks, I'd not come across that trick before. As you say it is a it
noisy though.
> somewhat on a tangent: it's also possible to overload macros on argument
> count [2], which may also come in handy.
>
> [2] https://stackoverflow.com/a/24028231/3685191
When I was writing my original reply to Linus I did wonder if we could
count the arguments. I didn't pursue it as I don't really want to create
a dozen different macros for different argument counts. I think TEST()
is understandable by anyone reading the code whereas IT() seems a bit
odd unless one is used to BDD.
Best Wishes
Phillip
^ permalink raw reply [flat|nested] 29+ messages in thread
end of thread, other threads:[~2023-06-30 9:49 UTC | newest]
Thread overview: 29+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-04-27 17:50 [RFC PATCH 0/2] add an external testing library for unit tests Calvin Wan
2023-04-27 17:50 ` [RFC PATCH 1/2] Add C TAP harness Calvin Wan
2023-04-27 18:29 ` SZEDER Gábor
2023-04-27 18:38 ` Calvin Wan
2023-04-27 20:15 ` Phillip Wood
2023-04-28 16:31 ` Calvin Wan
2023-05-02 15:46 ` Felipe Contreras
2023-05-10 15:46 ` Phillip Wood
2023-05-11 23:16 ` Glen Choo
2023-05-18 10:04 ` Phillip Wood
2023-06-21 15:57 ` Linus Arver
2023-06-26 13:15 ` Phillip Wood
2023-06-28 21:17 ` Linus Arver
2023-06-29 5:52 ` Oswald Buddenhagen
2023-06-30 9:48 ` Phillip Wood
2023-05-02 15:54 ` Felipe Contreras
2023-05-02 16:39 ` Ævar Arnfjörð Bjarmason
2023-05-02 18:11 ` Felipe Contreras
2023-05-02 16:34 ` Ævar Arnfjörð Bjarmason
2023-05-10 8:18 ` Phillip Wood
2023-04-27 17:50 ` [RFC PATCH 2/2] unit test: add basic example Calvin Wan
2023-04-27 18:47 ` Junio C Hamano
2023-05-02 15:58 ` Felipe Contreras
2023-04-27 18:39 ` [RFC PATCH 0/2] add an external testing library for unit tests Junio C Hamano
2023-04-27 18:46 ` Calvin Wan
2023-04-27 21:35 ` brian m. carlson
2023-05-02 4:18 ` Felipe Contreras
2023-05-02 13:52 ` Ævar Arnfjörð Bjarmason
2023-05-02 15:28 ` Felipe Contreras
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).