Re: [BUG?] ls-files -o now traverses nested repo when given multiple pathspecs

[Elijah Newren <newren@gmail.com>]

Hi Kyle,

On Sat, Dec 7, 2019 at 9:31 PM Kyle Meyer <kyle@kyleam.com> wrote:
>
> Kyle Meyer <kyle@kyleam.com> writes:
>
> > Elijah Newren <newren@gmail.com> writes:
> >> [...]
> >> At least my changes in git-2.24.0 made the behavior consistent; it'll
> >> always traverse into a directory that matches a given pathspec.
> >
> > I might be getting mixed up, but the changes in 2.24.0 did introduce
> > some inconsistent behavior (in the no trailing slash case) with respect
> > to giving a single pathspec and giving multiple pathspecs, no?  Using
> > your example:
> >
> >     $ git --version
> >     git version 2.24.0
> >     $ git ls-files -o untracked_repo
> >     untracked_repo/
> >     $ git ls-files -o untracked_repo empty
> >     empty
> >     untracked_repo/
> >     untracked_repo/empty
>
> It looks like the "multiple pathspecs trigger traversal" change isn't
> limited to nested repositories.  It can also be observed with
> --directory and plain untracked directories.  Assume the tree layout
> from your example again.  With a single pathspec (and no slash),
> 'ls-files -o --directory' will not expand the untracked directory's
> contents:
>
>     $ git ls-files -o --directory untracked_dir
>     untracked_dir/
>
> But, as of 89a1f4aaf7, tacking on an additional pathspec will cause
> ls-files to traverse into the untracked directory:
>
>     $ git ls-files -o --directory untracked_dir empty
>     empty
>     untracked_dir/
>     untracked_dir/empty
>
> In contrast, on 89a1f4aaf7^ the same command shows
>
>     $ git ls-files -o --directory untracked_dir empty
>     empty
>     untracked_dir/

Yeah, I spotted that too.  You left out a case, a single pathspec with
the trailing slash:

   git ls-files -o --directory untracked_dir/

That will traverse into the directory before or after my changes.  I
also spotted a few other bugs, e.g. try out 'git ls-files -o .git/'
(with either git-2.23 or git-2.24).  Whoops.  We do correctly avoid
traversing into the .git directory if multiple pathspecs are provided.
Anyway, this whole area seems to be a bug factory.  Every time I think
I'm close to having some patches to send to the list to fix up the
issues I've found, I find the fix isn't where I thought it was and/or
find yet another bug.  Quite aggravating.

I'm thinking of just sending the patches I have, since they fix up all
the issues we've discussed so far (including the .git/ case I just
mentioned), and ignoring the 2-3 other bugs I found that are still
broken other than providing testcases documenting their breakage.