Re: Possible regression in `git grep --untracked` pathspec matching in v2.27.0

From: Elijah Newren <newren@gmail.com>
Date: Jun 3, 2020, 14:14 UTC
Message-ID: <CABPp-BEPOn_JL1FDA-z1jfkfej4dEbbODkcmzNWwY7V-=WsMVw@mail.gmail.com>
In-Reply-To: <CAJmdR_pG74x0Zn43MSm7zXNcoitqjjOy+WnhyGBW+oFjVFLbRQ@mail.gmail.com>

Hi,

On Tue, Jun 2, 2020 at 8:30 PM John Millikin <jmillikin@stripe.com> wrote:

Show 26 quoted lines

>
> In v2.26.2 and earlier, `git grep --untracked` with negative pathspec patterns
> would filter out paths matching those patterns. In v2.27.0, this no longer works
> as expected.
>
> Example output:
>
> # v2.26.2 (good)
> $ ./git grep --untracked --files-with-matches -e '.' -- 'quote.*' ':!quote.c'
> quote.h
> $
>
> # v2.27.0 (bad)
> $ ./git grep --untracked --files-with-matches -e '.' -- 'quote.*' ':!quote.c'
> quote.c
> quote.h
> $
>
> I've bisected the change to commit 95c11ecc73f286e0a95d9591ae98f1221efe4633,
> which changed the handling of path matches in callers of `fill_directory()`.
>
> It's not clear from git-grep(1) whether the old or new behavior is "correct",
> so I request that the fix include a clarification of expected semantics.
>
> Thanks,
> - John Millikin

Oh, fun, an issue from _that_ series. At least the commit you bisected to wasn't 8d92fb292706 (dir: replace exponential algorithm with a linear one, 2020-04-01), though.

Thanks for the detailed report. I'll try to get some time to take a look later this week.

Elijah