From: Elijah Newren <newren@gmail.com>
To: Glen Choo <chooglen@google.com>
Cc: Elijah Newren via GitGitGadget <gitgitgadget@gmail.com>,
git@vger.kernel.org
Subject: Re: [PATCH 01/24] init-db: remove unnecessary global variable & document existing bug
Date: Thu, 11 May 2023 18:21:00 -0700 [thread overview]
Message-ID: <CABPp-BFT_aAmMcD7HKVuaZR3kkGND1mBF3R8OvisHimYWiX48g@mail.gmail.com> (raw)
In-Reply-To: <kl6lsfc2zjje.fsf@chooglen-macbookpro.roam.corp.google.com>
On Thu, May 11, 2023 at 1:43 PM Glen Choo <chooglen@google.com> wrote:
>
> "Elijah Newren via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
> > From: Elijah Newren <newren@gmail.com>
> >
> > This commit was prompted by a desire to move the functions which
> > builtin/init-db.c and builtin/clone.c share out of the former file and
> > into setup.c. One issue that made it difficult was the
> > init_is_bare_repository global variable.
> >
> > init_is_bare_repository is actually not very useful. It merely stores
> > the return value from is_bare_repository() and only for the duration of
> > a few additional function calls before its value is checked, and none of
> > those functions do anything that could change is_bare_repository()'s
> > return value. So, we can simply dispense with the global by replacing
> > it with is_bare_repository().
>
> I think the purpose of init_is_bare_repository is something different.
> But based off my different understanding, I can't reproduce any
> different behavior. I don't know if I'm just confused or not, but I'll
> leave some breadcrumbs here to check my understanding.
So, while my patch does not change behavior under any circumstances,
my explanation was slightly wrong. More on that below...
> Reordering the hunks for clarity,
>
> > @@ -422,8 +436,6 @@ int init_db(const char *git_dir, const char *real_git_dir,
> >
> > safe_create_dir(git_dir, 0);
> >
> > - init_is_bare_repository = is_bare_repository();
> > -
> > /* Check to see if the repository version is right.
> > * Note that a newly created repository does not have
> > * config file, so this will not fail. What we are catching
>
> Here, init_db() caches the value of is_bare_repository(), which itself
> reads the value of is_bare_repository_cfg, which can be modified by when
> we read "core.bare" via git_config(git_default_config) or similar
> (basically, any config callback that uses git_default_config). It is
> also modified in other places though, like setup.c.
Close. is_bare_repository() not only reads the value of
is_bare_repository_cfg, but it ANDs the result with
!get_git_work_tree(). This turns out to be important.
> IIUC, we haven't actually parsed "core.bare" at this point. The
> git_config() call just above this calls "plaform_core_config", which is
> either "mingw_core_config" (doesn't read "core.bare") or
> noop_core_config (noop).
I believe this is correct.
> > @@ -231,9 +230,24 @@ static int create_default_files(const char *template_path,
> > * We must make sure command-line options continue to override any
> > * values we might have just re-read from the config.
> > */
> > - is_bare_repository_cfg = init_is_bare_repository || !work_tree;
> > if (init_shared_repository != -1)
> > set_shared_repository(init_shared_repository);
> > + /*
> > + * TODO: heed core.bare from config file in templates if no
> > + * command-line override given
> > + *
> > + * Unfortunately, this location in the code is far too late to
> > + * allow this to happen; both builtin/init.db and
> > + * builtin/clone.c setup the new repository and call
> > + * set_git_work_tree() before this point. (Note that both do
> > + * that repository setup before calling init_db(), which in
> > + * turn calls create_default_files().) Fixing it would
> > + * require too much refactoring, and no one seems to have
> > + * wanted this behavior in 15+ years, so we'll continue
> > + * ignoring the config for now and just override
> > + * is_bare_repository_cfg unconditionally.
> > + */
> > + is_bare_repository_cfg = is_bare_repository() || !work_tree;
> >
> > /*
> > * We would have created the above under user's umask -- under
>
> Now, we're in the midst of the re-init. Expanding the context a little,
> we see:
>
> git_config(git_default_config, NULL);
>
> /*
> * We must make sure command-line options continue to override any
> * values we might have just re-read from the config.
> */
> is_bare_repository_cfg = init_is_bare_repository || !work_tree;
>
> So now we've read the new config of the re-inited repo, which might have
> "core.bare" set to a value other than what "git init-db [--bare]"
> started with
Correct.
> so we want to _intentionally_ ignore it.
That's what the code does, but not what it should do. If neither
`--bare` nor `--no-bare` is given on the command line, we ought to pay
attention to the config setting.
The fact that we do ignore the config regardless of command line flags
is the bug that this patch documents.
> We do this by
> reading out the cached value, _not_ by calling is_bare_repository()
> again.
Almost, but you missed part of the line you are commenting on. We
read out the cached value *and* OR it with !work_tree. The
distinction may look small, but it turns out to be critical.
> So it seems to me like this patch changes the intent.
Yeah, looks like it does.
But it doesn't change the result. Note the critical thing here is we
changed from computing:
init_is_bare_repository || !work_tree;
to
is_bare_repository() || !work_tree;
Further,
init_is_bare_repository = [cached value of] is_bare_repository_cfg
&& [cached value of] !get_git_work_tree()
is_bare_repository() = [current value of] is_bare_repository_cfg &&
[current value of] !get_git_work_tree()
However, nothing calls repo_init() or set_git_work_tree() between any
of these calls, so
[current value of] !get_git_work_tree() == [cached value of]
!get_git_work_tree() == !work_tree
So, both before and after what we are computing simplifies to
<something> && !work_tree || !work_tree
which further simplifies, regardless of whether <something> is true or
false, down to
!work_tree
So whether we are using a cached value of is_bare_repository_cfg or a
current value of is_bare_repository_cfg is irrelevant. In fact, from
the analysis above, we can simplify this line to just
is_bare_repository_cfg = !work_tree;
completely ignoring both is_bare_repository_cfg and
is_bare_repository(), and it won't change the behavior. I just did
it; it passes all tests.
> Where I struggling with is how to make this behave badly. The lines
> above seem to be defensive in nature - we never use
> is_bare_repository_cfg past this point, but we want to guard against
> unintentional behavior in the future.
Not quite true; there is another call to is_bare_repository() in the
same function about 60 lines later, which does pay attention to
is_bare_repository_cfg.
[...]
> If I'm right (which I'm not sure about), we might need to keep
> init_is_bare_repository around _somewhere_. Not a global, but maybe
> as a param.
This makes sense, despite the other bugs present. I'll make this
change, as well as split this patch into two as Calvin suggested, and
update the description to correct my errors and explain the bug
better.
next prev parent reply other threads:[~2023-05-12 1:21 UTC|newest]
Thread overview: 115+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-05-07 3:45 [PATCH 00/24] Header cleanups (final splitting of cache.h, and some splitting of other headers) Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 01/24] init-db: remove unnecessary global variable & document existing bug Elijah Newren via GitGitGadget
2023-05-11 17:24 ` Calvin Wan
2023-05-12 1:55 ` Elijah Newren
2023-05-11 20:43 ` Glen Choo
2023-05-12 1:21 ` Elijah Newren [this message]
2023-05-12 17:26 ` Glen Choo
2023-05-07 3:45 ` [PATCH 02/24] init-db, clone: change unnecessary global into passed parameter Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 03/24] setup: adopt shared init-db & clone code Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 04/24] read-cache: move shared commit and ls-files code Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 05/24] add: modify add_files_to_cache() to avoid globals Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 06/24] read-cache: move shared add/checkout/commit code Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 07/24] statinfo: move stat_{data,validity} functions from cache/read-cache Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 08/24] run-command.h: move declarations for run-command.c from cache.h Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 09/24] name-hash.h: move declarations for name-hash.c " Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 10/24] sparse-index.h: move declarations for sparse-index.c " Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 11/24] preload-index.h: move declarations for preload-index.c from elsewhere Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 12/24] diff.h: move declaration for global in diff.c from cache.h Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 13/24] merge.h: move declarations for merge.c " Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 14/24] repository.h: move declaration of the_index " Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 15/24] read-cache*.h: move declarations for read-cache.c functions " Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 16/24] cache.h: remove this no-longer-used header Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 17/24] log-tree: replace include of revision.h with simple forward declaration Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 18/24] repository: remove unnecessary include of path.h Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 19/24] diff.h: remove unnecessary include of oidset.h Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 20/24] list-objects-filter-options.h: remove unneccessary include Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 21/24] builtin.h: remove unneccessary includes Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 22/24] git-compat-util.h: remove unneccessary include of wildmatch.h Elijah Newren via GitGitGadget
2023-05-07 3:45 ` [PATCH 23/24] hash-ll, hashmap: move oidhash() to hash-ll Elijah Newren via GitGitGadget
2023-05-11 17:24 ` Calvin Wan
2023-05-11 18:31 ` Jeff King
2023-05-11 18:39 ` Felipe Contreras
2023-05-12 2:08 ` Elijah Newren
2023-05-12 4:07 ` Jeff King
2023-05-07 3:45 ` [PATCH 24/24] fsmonitor-ll.h: split this header out of fsmonitor.h Elijah Newren via GitGitGadget
2023-05-08 15:29 ` [PATCH 00/24] Header cleanups (final splitting of cache.h, and some splitting of other headers) Junio C Hamano
2023-05-09 1:31 ` Elijah Newren
2023-05-09 16:03 ` Junio C Hamano
2023-05-08 15:55 ` Junio C Hamano
2023-05-11 17:41 ` Calvin Wan
2023-05-12 6:58 ` Elijah Newren
2023-05-12 7:04 ` [PATCH v2 00/27] " Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 01/27] init-db: document existing bug with core.bare in template config Elijah Newren via GitGitGadget
2023-05-12 21:34 ` Jonathan Tan
2023-05-16 3:08 ` Elijah Newren
2023-05-12 7:04 ` [PATCH v2 02/27] init-db: remove unnecessary global variable Elijah Newren via GitGitGadget
2023-05-12 20:20 ` Glen Choo
2023-05-12 7:04 ` [PATCH v2 03/27] init-db, clone: change unnecessary global into passed parameter Elijah Newren via GitGitGadget
2023-05-12 21:28 ` Jonathan Tan
2023-05-16 2:57 ` Elijah Newren
2023-05-12 7:04 ` [PATCH v2 04/27] setup: adopt shared init-db & clone code Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 05/27] read-cache: move shared commit and ls-files code Elijah Newren via GitGitGadget
2023-05-12 21:25 ` Jonathan Tan
2023-05-12 7:04 ` [PATCH v2 06/27] add: modify add_files_to_cache() to avoid globals Elijah Newren via GitGitGadget
2023-05-15 21:44 ` Jonathan Tan
2023-05-12 7:04 ` [PATCH v2 07/27] read-cache: move shared add/checkout/commit code Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 08/27] statinfo: move stat_{data,validity} functions from cache/read-cache Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 09/27] run-command.h: move declarations for run-command.c from cache.h Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 10/27] name-hash.h: move declarations for name-hash.c " Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 11/27] sparse-index.h: move declarations for sparse-index.c " Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 12/27] preload-index.h: move declarations for preload-index.c from elsewhere Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 13/27] diff.h: move declaration for global in diff.c from cache.h Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 14/27] merge.h: move declarations for merge.c " Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 15/27] repository.h: move declaration of the_index " Elijah Newren via GitGitGadget
2023-05-15 21:50 ` Jonathan Tan
2023-05-12 7:04 ` [PATCH v2 16/27] read-cache*.h: move declarations for read-cache.c functions " Elijah Newren via GitGitGadget
2023-05-16 0:00 ` Jonathan Tan
2023-05-12 7:04 ` [PATCH v2 17/27] cache.h: remove this no-longer-used header Elijah Newren via GitGitGadget
2023-05-16 0:02 ` Jonathan Tan
2023-05-16 3:02 ` Elijah Newren
2023-05-12 7:04 ` [PATCH v2 18/27] log-tree: replace include of revision.h with simple forward declaration Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 19/27] repository: remove unnecessary include of path.h Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 20/27] diff.h: remove unnecessary include of oidset.h Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 21/27] list-objects-filter-options.h: remove unneccessary include Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 22/27] builtin.h: remove unneccessary includes Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 23/27] git-compat-util.h: remove unneccessary include of wildmatch.h Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 24/27] merge-ll: rename from ll-merge Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 25/27] khash: name the structs that khash declares Elijah Newren via GitGitGadget
2023-05-12 7:04 ` [PATCH v2 26/27] hash-ll, hashmap: move oidhash() to hash-ll Elijah Newren via GitGitGadget
2023-05-16 0:16 ` Jonathan Tan
2023-05-16 3:06 ` Elijah Newren
2023-05-12 7:04 ` [PATCH v2 27/27] fsmonitor-ll.h: split this header out of fsmonitor.h Elijah Newren via GitGitGadget
2023-05-16 0:17 ` Jonathan Tan
2023-05-16 3:07 ` Elijah Newren
2023-05-16 6:33 ` [PATCH v3 00/28] Header cleanups (final splitting of cache.h, and some splitting of other headers) Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 01/28] init-db: document existing bug with core.bare in template config Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 02/28] init-db: remove unnecessary global variable Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 03/28] init-db, clone: change unnecessary global into passed parameter Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 04/28] setup: adopt shared init-db & clone code Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 05/28] read-cache: move shared commit and ls-files code Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 06/28] add: modify add_files_to_cache() to avoid globals Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 07/28] read-cache: move shared add/checkout/commit code Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 08/28] statinfo: move stat_{data,validity} functions from cache/read-cache Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 09/28] run-command.h: move declarations for run-command.c from cache.h Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 10/28] name-hash.h: move declarations for name-hash.c " Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 11/28] sparse-index.h: move declarations for sparse-index.c " Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 12/28] preload-index.h: move declarations for preload-index.c from elsewhere Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 13/28] diff.h: move declaration for global in diff.c from cache.h Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 14/28] merge.h: move declarations for merge.c " Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 15/28] repository.h: move declaration of the_index " Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 16/28] read-cache*.h: move declarations for read-cache.c functions " Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 17/28] cache.h: remove this no-longer-used header Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 18/28] log-tree: replace include of revision.h with simple forward declaration Elijah Newren via GitGitGadget
2023-05-16 6:33 ` [PATCH v3 19/28] repository: remove unnecessary include of path.h Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 20/28] diff.h: remove unnecessary include of oidset.h Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 21/28] list-objects-filter-options.h: remove unneccessary include Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 22/28] builtin.h: remove unneccessary includes Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 23/28] git-compat-util.h: remove unneccessary include of wildmatch.h Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 24/28] merge-ll: rename from ll-merge Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 25/28] khash: name the structs that khash declares Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 26/28] object-store-ll.h: split this header out of object-store.h Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 27/28] hash-ll, hashmap: move oidhash() to hash-ll Elijah Newren via GitGitGadget
2023-05-16 6:34 ` [PATCH v3 28/28] fsmonitor-ll.h: split this header out of fsmonitor.h Elijah Newren via GitGitGadget
2023-05-23 19:49 ` [PATCH v3 00/28] Header cleanups (final splitting of cache.h, and some splitting of other headers) Calvin Wan
2023-05-24 0:11 ` Elijah Newren
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CABPp-BFT_aAmMcD7HKVuaZR3kkGND1mBF3R8OvisHimYWiX48g@mail.gmail.com \
--to=newren@gmail.com \
--cc=chooglen@google.com \
--cc=git@vger.kernel.org \
--cc=gitgitgadget@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).