From: Jeff King <peff@peff.net>
Date: Fri, 11 Aug 2023 15:02:44 GMT
Subject: Re: [PATCH 0/4] commit-graph: fsck zero/non-zero generation number fixes
Message-ID: <20230811150244.GD2303200@coredump.intra.peff.net>
In-Reply-To: <cover.1691699851.git.me@ttaylorr.com>

On Thu, Aug 10, 2023 at 04:37:32PM -0400, Taylor Blau wrote:

> This series expands on a patch that Peff sent earlier in this thread to
> remove a section of unreachable code that was noticed by Coverity in the
> `verify_one_commit_graph()` function.
> 
> The first couple of patches addresses the main issue, which is that we
> couldn't verify ancient commit-graphs written with zero'd generation
> numbers. The third patch adds additional tests to ensure our coverage in
> this area is complete, and the final patch is a cleanup.

Thanks for untangling (and explaining!) some of the history here. I
think this series is a definite improvement, including that final
cleanup. But I also think that the "two bits" approach mentioned by
Junio would be better still.  IMHO the intent of the code would be more
clear, and it would avoid the flip-flopping error case.

-Peff