From: Jeff King <peff@peff.net>
Date: Sat, 18 Oct 2025 09:42:45 GMT
Subject: Re: Regression in `git diff --quiet HEAD` when a new file is staged
Message-ID: <20251018094245.GB1060824@coredump.intra.peff.net>
In-Reply-To: <918E56B8-7009-4E8E-A98E-AC5B9CE4DD7C@gmail.com>

On Sat, Oct 18, 2025 at 09:04:40AM +0800, Lidong Yan wrote:

> I believe we can move Peff’s code from diff_flush() to diff_flush_patch_quiet().
> However, I'm unsure whether we should remove the dry-run logic. In dry-run
> mode, we would halt as early as possible in xdl_diff by using quick_consume().

Yeah, exactly.

I am OK to put the /dev/null code into diff_flush_patch_quiet(). That
would give all callers the same belt-and-suspenders protection.

The patch I posted put it where it was because that's where it was prior
to b55e6d36eb. It is essentially a revert (because I only wanted one
hunk I didn't call "git revert", but rather did a reversed patch
application).

But after that revert, I think it would be reasonable to move the code
on top (with the justification that it is helping the other caller of
the _quiet function).

-Peff