Re: [PATCH 4/5] merge-ll: use read_mmfile() to read external merge results

From: Jeff King <peff@peff.net>
Date: Sep 29, 2026, 20:41 UTC
Message-ID: <20260929204157.GA1733321@coredump.intra.peff.net>
In-Reply-To: <20260929201134.GA1713437@coredump.intra.peff.net>

On Tue, Sep 29, 2026 at 04:11:34PM -0400, Jeff King wrote:

Show 5 quoted lines

> I had imagined just fixing this in ll_ext_merge(), like:
> [...]
> which reduces the weirdness coming out of that function. But it wouldn't
> help with other drivers (which may or may not have similar problems? I'd
> guess not, since they are all operating internally).

So here are patches to do that, including a cleaned-up version of the reproduction I posted.

I think ll_ext_merge() is the only driver that has this weird error case, so it should be sufficient. Your patch would protect a potential future driver, but I'd be surprised if we had one that introduced the same NULL-but-not-an-error behavior.

  [6/5]: merge-ll: handle external driver status before reading result
  [7/5]: merge-ll: report an error when reading external merge results fails

 merge-ll.c        | 14 ++++++-------
 t/t4200-rerere.sh | 51 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 58 insertions(+), 7 deletions(-)

-Peff