From: Bello Olamide <belkid98@gmail.com>
Date: Tue, 10 Feb 2026 08:40:16 GMT
Subject: Re: [Outreachy PATCH v6 1/3] environment: stop storing `core.attributesFile` globally
Message-ID: <CAD=f0L8jeLE_Zy-5aW+DZrM6JNgbwRcgWCdW1f7SJ5uG_2R-yA@mail.gmail.com>
In-Reply-To: <88570b6c-2ff3-416f-b63f-2fb7c80c5af8@gmail.com>

On Sun, 8 Feb 2026 at 12:14, Phillip Wood <phillip.wood123@gmail.com> wrote:
>
>
>
> On 07/02/2026 01:14, Junio C Hamano wrote:
> > Olamide Caleb Bello <belkid98@gmail.com> writes:
> >
> >> Suggested-by: Phillip Wood <phillip.wood123@gmail.com>
> >> Mentored-by: Christian Couder <christian.couder@gmail.com>
> >> Mentored-by: Usman Akinyemi <usmanakinyemi202@gmail.com>
> >> Helped-by: Junio C Hamano <gitster@pobox.com>
> >> Signed-off-by: Olamide Caleb Bello <belkid98@gmail.com>
> >> ---
> >>   attr.c        |  7 ++++---
> >>   environment.c | 12 +++++++++---
> >>   environment.h | 11 ++++++++++-
> >>   repository.c  | 12 ++++++++++++
> >>   repository.h  |  7 +++++++
> >>   5 files changed, 42 insertions(+), 7 deletions(-)
> >
> > I bisected our recent CI failures that break fuzz smoke test down to
> > this change.
>
> The documentation for the LibFuzzer [1] notes
>
>      * The fuzzing engine will execute the fuzz target many times with
>        different inputs in the same process.
>
> and the first thing that the callback in oss-fuzz/fuzz-commit-graph.c
> does is
>
>         initialize_repository(the_repository);
>
> so I think the problem is that the assumption that a process will only
> initialize "the_repository" once is incompatible with the way LibFuzzer
> works. Maybe we should add
>
>         memset(the_repository, 0, sizeof(*the_repository));
>
> before the call in initialize_repository()?
>
> Thanks
>
> Phillip
>
> [1] https://llvm.org/docs/LibFuzzer.html
>

Hello Phillip,
thank you for your reviews and assistance so far.

So I moved the code
 memset(the_repository, 0, sizeof(*the_repository))
into the fuzz-commit-graph.c Fuzzer test before the call to
initialize_repository().
This made the fuzzer smoke tests pass.

But the line below,
`if (repo != the_repository)
    BUG("trying to read config from wrong repository instance")`,
in the repo_config_values() accessor function
to make sure we are reading the config_values only for the_repository ,
makes many tests fail.

I believe this is because repo_init() calls initialize_repository() and
repo_init() is also called in repo_submodule_init() when creating a
subrepo from a super project, and this subrepo passed to repo_init() is
not the_repository.