Re: [PATCH v4 0/3] completion of 'git [-C <dir>] diff'

On Fri, Aug 7, 2026 at 2:32 AM Elijah Newren <[email protected]> wrote:
>
> On Thu, Aug 6, 2026 at 6:38 PM Junio C Hamano <[email protected]> wrote:
> >
> > The primary motivation for this topic is that the command-line
> > completion of 'git diff' does not handle paths (unlike 'git status'
> > and 'git add') and instead relies on the default behavior of Bash
> > command-line completion, which completes files in $PWD; this does
> > not work at all with the '-C <directory>' option.
> >
> > This series teaches the completion machinery to complete revisions
> > (unless '--' exists), then tracked paths, and then untracked paths,
> > before letting the Bash default kick in.  This way, we correctly
> > complete 'git diff' command line even when '-C <directory>' is in
> > effect.
>
> Nice.  :-)
>
> > The tests are the only changes relative to v2.  In the step where
> > tracked paths are completed, v2 did not demonstrate that untracked
> > ones are *not* completed at the same time.  Now we do by having
> > untracked 'file3' next to 'file1' and 'file2' that are tracked.  In
> > the last step, we demonstrate untracked paths that do not share
> > prefix with refs or tracked paths are completed, with or without the
> > "-C <dir>" option.
>
> I found this paragraph slightly hard to parse.  I think this means the same as:
>
> Only the tests changed since v2.  The tracked-paths step now also
> shows that untracked paths are not mixed in: an untracked 'file3'
> next to tracked 'file1'/'file2' means "git diff f" completes only the
> latter two.  The untracked-paths step shows that an untracked path
> which shares no prefix with a ref or tracked path still completes,
> with or without "-C ".
>
> (Not that it matters, since this is just the cover letter, but...)
>
> >  1/3: completion: no-op refactoring of diff completion
> >  2/3: completion: complete tracked paths for 'git diff'
> >  3/3: completion: 'git diff' completes untracked paths as a last
> >         resort
>
> I found two minor nits in the first two patches, and didn't spot any
> issues with the third patch.

Yep, modulo Elijah's comments (many eyes, thanks!) I'm happy with this round.

-- 
D. Ben Knoble