Re: [PATCH v2 0/5] Some improvements to safe.directory on Windows

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 8, 2022, 16:38 UTC
Message-ID: <xmqqzggesnbh.fsf@gitster.g>
In-Reply-To: <pull.1286.v2.git.1659965270.gitgitgadget@gmail.com>

"Johannes Schindelin via GitGitGadget" <gitgitgadget@gmail.com> writes:

Show 20 quoted lines

> Due to the semantics being substantially different from Unix, the
> safe.directory feature presents its own set of problems on Windows. One
> particular issue would have prevented it from working in GitHub Actions'
> build agents, which we definitely rely on in the Git project itself. This
> was addressed via the fifth patch, which had made it (in a slightly
> different form) already into Git for Windows v2.35.2, and they are ready to
> be applied to core Git, too.
>
> The FAT32 patch came in later, and was released as part of Git for Windows
> v2.37.0, so I also have confidence that it is stable and ready to be
> integrated into core Git, too.
>
> Changes since v1:
>
>  * Restructured the patch series.
>  * Instead of an environment variable to turn on debugging, we now always
>    show the platform-dependent information together with the error message
>    about the dubious ownership (iff it is shown, that is), based on an idea
>    by Junio.
>  * Rebased onto gc/bare-repo-discovery to avoid a merge conflict.

I actually had to rebase it back so that we could merge it to 'maint' for further 2.37.x releases. I'll refer to the original patches in this thread when I merge the result to 'seen', of course, to make sure the results do match. It would have been slightly less convenient if you did not do this rebase, but it would have allowed me to have much better confidence in the result that may eventually go to 'maint'. After all, mistakes in resolving merge conflicts on 'seen' can be corrected before the topic hits 'next'.

Thanks. I do not know about the API calls mingw.c part of these patches make, but the overall structure looks sensible to me.