Re: bug report - BUG: builtin/pack-objects.c:4310: should_include_obj should only be called on existing objects

From: Nik Garza <nikolas@applied.co>
Date: Apr 28, 2025, 22:55 UTC
Message-ID: <CA+ODqj34b2DsvjxGgVb=jSaTUNYRh_nNtaCzCx8QiJ8QFsRecA@mail.gmail.com>
In-Reply-To: <20250428192320.3595509-1-jonathantanmy@google.com>

Resending....

I haven't been able to define a repro case unfortunately. The error surfaces randomly in our CI infrastructure. I'll take a look at some of the failures we have seen and try to come up with a repro. Will add a test if I can find something.

On Mon, Apr 28, 2025 at 12:23 PM Jonathan Tan <jonathantanmy@google.com> wrote:

Show 20 quoted lines

>
> Elijah Newren <newren@gmail.com> writes:
> > Cc'ing the author of that commit for his comments.
>
> I took a look. I'm not sure why I made the assumption that
> should_include_obj() would only ever be called on objects in the repo
> - in process_tree() in list-objects.c, the case of a missing object is
> handled only after should_include_obj() is called. Looking back at the
> earliest mention of this on the list [1] I don't see any clues either.
>
> In any case, the fix is probably to change it so that
> should_include_obj() returns 0 if the object is absent.
>
> Having said that, I couldn't come up with a test that exercises this
> failure mode. Nik, do you have a minimal repo that reproduces this
> error? If yes, if you could contribute a test in the form of the 'after
> fetching descendants of non-promisor commits, gc works' test in t5616,
> that would help prevent regressions in the future.
>
> [1] https://lore.kernel.org/git/fb2c202591b466eea33b4585e47b70e9086603bb.1729549127.git.jonathantanmy@google.com/

-- 
Nikolas Garza

Software Engineer

nikolas@applied.co

(209) 499-1193

Applied Intuition, Inc.