Re: [PATCH] revision: make get_commit_action() a pure predicate
- From
Michael Montalbo <mmontalbo@gmail.com>
- Date
- Jul 25, 2026, 19:25 UTC
- Message-ID
- <CAC2QwmKP16cyw0get3hEWP8GjcFkUHB3uXxcQi9hBCCM-B+ECw@mail.gmail.com>
- In-Reply-To
- <xmqqjyqk3w7d.fsf@gitster.g>
On Fri, Jul 24, 2026 at 2:38 PM Junio C Hamano <gitster@pobox.com> wrote:
> > Sorry, but I completely lost track and I do not recall suggesting a > change that amounts to 100+ lines of new lines. Are we doing any > code clean-up? Bugfix? A new feature?
A latent bug fix, but I understand why this was confusing.
This was the discussion I should have linked to:
https://lore.kernel.org/git/xmqqtsqxfdl4.fsf@gitster.g/.
I had the link in my GGG PR description but accidentally deleted it without re-adding when I remembered GGG PRs shouldn't use a description for one commit series.
The linked discussion refers to a new graph feature that invokes get_commit_action() under the assumption the function will not modify any commit state. The graph feature in question uses a configuration that just happens to avoid the branch of get_commit_action() that modifies a commit's line range state, so a bug isn't ultimately surfaced in the linked topic feature, but it remains a potential issue for future callers.
Unfortunately, I couldn't figure out a way to make a test that validates if the change is effective without creating a bespoke test-tool that calls the function with the "right" options set.