Re: [GSoC Patch v7 4/8] repo: add path.hooks with absolute and relative suffixes
- From
K Jayatheerth <jayatheerthkulkarni2005@gmail.com>
- Date
- Nov 3, 2026, 02:31 UTC
- Message-ID
- <CA+rGoLeSfwSQs6hOyGxjRhyCLqjWBMQ4yfXUYWiCz+MHqsgWAA@mail.gmail.com>
- In-Reply-To
- <aukHlEd5udC1gLYZ@denethor>
On Tue, Nov 3, 2026 at 4:02 AM Justin Tobler <jltobler@gmail.com> wrote:
Show 53 quoted lines
> > On 26/09/27 05:14PM, K Jayatheerth wrote: > > Hooks are an integral part of a repository's configuration and are > > commonly used by tooling to automate repository-specific workflows. > > Currently, scripts typically retrieve the hooks directory by invoking > > `git rev-parse --git-path hooks`. > > > > Introduce `path.hooks.absolute` and `path.hooks.relative` keys to > > `git repo info`. This exposes the hooks directory as a scriptable > > config-like key using standard format rules, allowing scripts to > > retrieve it through the same interface as other repository path > > information. > > > > Mentored-by: Justin Tobler <jltobler@gmail.com> > > Mentored-by: Lucas Seiki Oshiro <lucasseikioshiro@gmail.com> > > Signed-off-by: K Jayatheerth <jayatheerthkulkarni2005@gmail.com> > > --- > > Documentation/git-repo.adoc | 9 +++++++++ > > builtin/repo.c | 22 ++++++++++++++++++++++ > > t/t1900-repo-info.sh | 28 ++++++++++++++++++++++++++-- > > 3 files changed, 57 insertions(+), 2 deletions(-) > > > > diff --git a/Documentation/git-repo.adoc b/Documentation/git-repo.adoc > > index e524a07f53..20836cf8f6 100644 > > --- a/Documentation/git-repo.adoc > > +++ b/Documentation/git-repo.adoc > > @@ -119,6 +119,15 @@ values that they return: > > `path.gitdir.relative`:: > > The path to the Git repository directory relative to the current working directory. > > > > +`path.hooks.absolute`:: > > + The canonical absolute path to the repository's hooks directory. > > + Respects the `core.hooksPath` configuration. If `core.hooksPath` is > > + set to `/dev/null`, that value is returned unchanged. > > + > > +`path.hooks.relative`:: > > + The path to the repository's hooks directory relative to the current > > + working directory. Respects the `core.hooksPath` configuration. > > So if a repository has `core.hooksPath` set to "/dev/null", the relative > path output here would be something like "../../../../dev/null". I > recognize that this matches the current behavior: > > $ git -c core.hooksPath=/dev/null rev-parse --path-format=relative --git-path hooks > ../../../../dev/null > > ...but I'm still not really convinced that is is a behavior we would > want to carry over as it makes it harder to users to figure out when > hooks are disabled. > > I would suggest we print the "/dev/null" here even when the relative > path is requested and document the behavior. >
I am more inclined with you, and I see how returning "/dev/null" for the relative key makes it much easier for scripts to detect when hooks are disabled.
One thing I wanted to check: in a follow-up series, Junio suggested extracting a common helper library to share implementations between `git rev-parse` and `git repo info`. If `git repo info` special-cases "/dev/null" for `path.hooks.relative` while `rev-parse --git-path hooks` continues to relative-format it, would that divergence be an issue for sharing the underlying helper, or should we special-case it only in the `path.hooks` getter?
Show 38 quoted lines
> [snip]
> > diff --git a/t/t1900-repo-info.sh b/t/t1900-repo-info.sh
> > index eec576a1d9..1da5db4942 100755
> > --- a/t/t1900-repo-info.sh
> > +++ b/t/t1900-repo-info.sh
> > @@ -174,7 +174,11 @@ test_repo_info_path () {
> > cd repo/sub &&
> > ROOT="$(test-tool path-utils real_path ..)" && export ROOT &&
> > eval "$init_command" &&
> > - echo "path.$field_name.absolute=$ROOT/$expected_dir" >expect &&
> > + case "$expected_dir" in
> > + /*) EXPECT_ABS="$expected_dir" ;;
> > + *) EXPECT_ABS="$ROOT/$expected_dir" ;;
> > + esac &&
> > + echo "path.$field_name.absolute=$EXPECT_ABS" >expect &&
> > git repo info "path.$field_name.absolute" >actual &&
> > test_cmp expect actual
> > )
> > @@ -188,7 +192,11 @@ test_repo_info_path () {
> > cd repo/sub &&
> > ROOT="$(test-tool path-utils real_path ..)" && export ROOT &&
> > eval "$init_command" &&
> > - echo "path.$field_name.relative=../$expected_dir" >expect &&
> > + case "$expected_dir" in
> > + /*) EXPECT_REL="$(test-tool path-utils relative_path "$expected_dir" "$PWD")" ;;
> > + *) EXPECT_REL="../$expected_dir" ;;
> > + esac &&
> > + echo "path.$field_name.relative=$EXPECT_REL" >expect &&
> > git repo info "path.$field_name.relative" >actual &&
> > test_cmp expect actual
> > )
>
> From what I can gather, the code changes to this test helper is really
> only required for the "/dev/null" test case. IMO, this helper function
> has now become quite unwieldy and rather difficult to parse.
>
> I think we would be better off writing the "/dev/null" test case by hand.
>Agreed, reverting the changes to `test_repo_info_path` and writing the `/dev/null` test as a standalone test block will be much cleaner.
Show 26 quoted lines
> > @@ -213,6 +221,22 @@ test_repo_info_path 'gitdir with explicit GIT_DIR' 'gitdir' \ > > '.git' \ > > 'GIT_DIR="../.git" && export GIT_DIR' > > > > +test_repo_info_path 'hooks standard' 'hooks' '.git/hooks' > > + > > +test_repo_info_path 'hooks with core.hooksPath override' 'hooks' \ > > + 'custom-hooks' \ > > + 'git config core.hooksPath "$ROOT/custom-hooks" && mkdir -p "$ROOT/custom-hooks"' > > + > > +# /dev/null is not a real, canonicalizable filesystem path on Windows, > > +# so path resolution for core.hooksPath=/dev/null cannot be expected to > > +# produce a literal "/dev/null" the way it does on POSIX systems. > > +if ! test_have_prereq MINGW > > +then > > + test_repo_info_path 'hooks with core.hooksPath=/dev/null' 'hooks' \ > > + '/dev/null' \ > > + 'git config core.hooksPath /dev/null' > > +fi > > I'm not quite sure I follow here. Are we saying that the git-repo-info > command won't work on windows if the hooks are set to "/dev/null"? Or is > it just some issue with how the test is written? If its the latter, > maybe this is something we can work around if the test is written > without the use of the `test_repo_info_path` helper? >
It was both the command and the test helper on Windows, `/dev/null` is not an actual filesystem path, so passing it through `format_path()` (and `test-tool path-utils relative_path` in the test) does not resolve the same way as on POSIX systems (where `$PWD` on Windows has a drive letter like `C:/...`).
However, if we special-case `/dev/null` in the C code so that both `path.hooks.absolute` and `path.hooks.relative` output `/dev/null` directly without passing it through `format_path()`, and write the test by hand, it will work consistently on Windows too and we can drop the `!MINGW` prerequisite entirely.
> -Justin
Thank you, - K Jayatheerth