Re: rev-list --ancestry-path with arg - bug or undocumented requirement

From: Elijah Newren <newren@gmail.com>
Date: Nov 20, 2024, 03:22 UTC
Message-ID: <CABPp-BE1C2izp1a0Xm8_0KU+kas8XKejDyix+AzXqbCOeVp2Dg@mail.gmail.com>
In-Reply-To: <CADYQcGpXm=RTEYyxqdSowQ4Vg9jmXuCzOOpd-TgDX8U814BReg@mail.gmail.com>

On Tue, Nov 19, 2024 at 1:53 PM Kai Koponen <kaikoponen@google.com> wrote:

Show 19 quoted lines

>
> What did you do before the bug happened? (Steps to reproduce your issue)
> Create a toy repo with commits in the following structure:
> A - B - C
>      \
>       D
>
> Run `git rev-list --ancestry-path=D A..C`.
>
> What did you expect to happen? (Expected behavior)
> Commits A and B should be listed.
>
> What happened instead? (Actual behavior)
> No commits listed.
>
> What's different between what you expected and what actually happened?
> --ancestry-path with a commit arg seems to return 0 results unless the
> specified commit is itself part of the range. The rev-list documentation
> does not mention that this is a requirement.

Yeah, the commit message that introduced --ancestry-path=COMMIT did document this in its commit message (see 257418c59040c13bfa839e01922e21833cda6a52257418c59040 (revision: allow --ancestry-path to take an argument, 2022-08-19)), and I think my original documentation patch was a little clearer on this point, but when Jonathan suggested an alternative for the documentation (https://lore.kernel.org/git/20220818222416.3567602-1-jonathantanmy@google.com/), I didn't think about someone attempting a commit outside the ancestry-path as the argument. My bad.

Based on the references above, would you like to suggest a documentation patch to ensure this requirement is documented?