From: Phillip Wood <phillip.wood123@gmail.com>
Date: Mon, 09 Jun 2025 09:42:17 GMT
Subject: Re: [PATCH v3 0/2] stash: fix and improve "git stash -p <pathspec>"
Message-ID: <a66483fb-5bc4-42b5-b361-c900a69015ed@gmail.com>
In-Reply-To: <CAN0heSotWpNmqd905aknVTfk6WEcYifAwbXBKYfAWkhzxua3ZA@mail.gmail.com>

Hi Martin

On 07/06/2025 13:56, Martin Ågren wrote:
> Hi Phillip,
> 
> On Sat, 7 Jun 2025 at 11:45, Phillip Wood <phillip.wood123@gmail.com> wrote:
> [...]
> This range-diff matches what I'd expect. Now this test makes sure we
> really pick up the `-p`. On that note ... I just realized that all of
> these would keep the test passing:
> 
>   test_write_lines s y n | git stash -p file # what you have
>   test_write_lines s y n | git stash -p file otherfile
>   test_write_lines s y n | git stash -p .
>   test_write_lines s y n | git stash -p
> 
> So the implementation under test could bungle the pathspec, query the
> user for both `file` and `otherfile` (in that order!), get EOF from
> stdin while handling `otherfile`, leave it out of the stash, and end up
> passing the test. We could try to protect against this by providing
> another "y": if git wants to read something after our "s y n" sequence,
> we'll give it a "y" in the hopes that it will trip things up. We do want
> to test the handling of pathspecs here, so maybe tighten this?

Junio has merged this to next now. I was hoping that we would already 
have coverage for this with other tests but I couldn't see anything so 
I'll look at improving the coverage for "git stash push -p <pathspec>" 
in the next release cycle.

Thanks

Phillip