Re: [PATCH v2 0/8] Debug merge-recursive.[ch]

From: Elijah Newren <newren@gmail.com>
Date: Apr 7, 2025, 22:23 UTC
Message-ID: <CABPp-BGN2JCt9BJ0wQqtgNLhD6Qf19PMzYWpjgpJ6O2OcyRCXg@mail.gmail.com>
In-Reply-To: <xmqqiknfn36y.fsf@gitster.g>

On Mon, Apr 7, 2025 at 1:10 PM Junio C Hamano <gitster@pobox.com> wrote:

Show 8 quoted lines

>
> "Elijah Newren via GitGitGadget" <gitgitgadget@gmail.com> writes:
>
> > This series does some preparation, then moves the code shared between
> > merge-recursive and merge-ort from the former to the latter, and then debugs
> > the remainder of merge-recursive.[ch].
>
> Help unconfusing me.

It was an attempt at humor, which importantly relied on the sentence you stripped out immediately before this part you quoted, namely:

> > As a wise man once told me, "Deleted code is debugged code!"

With this sentence as context, "debug the code" was a funny way of saying "delete the code". Because if it's deleted, we are no longer affected by any bugs contained within it (and that'd be true for both known bugs and latent ones we hadn't triggered yet).

Show 14 quoted lines

> I've already named the topic with "debug" in its name while queuing
> the original iteration of this series, as I was on vacation and did
> not want to spend more than minimum braincycles on naming, but now I
> am back, I sense that the use of the word, and the proposed log
> message for 6/8, are overly suboptimal.  If you are referring to
> fixing remaining bugs, "Debug the remainder of merge-recursive.[ch]"
> is not how we usually describe our fixes.
>
> I suspect that the overall sentiment behind this series is ...
>
>         Such and such bugs existed in the older backend, but now the
>         newer backend is used when the older one is asked, and the
>         newer backend does not share these bugs, we can simply
>         remove the buggy code specific to the older backend.

I'd say the sentiment is more:

   merge-ort was always meant to be a replacement for merge-recursive
   (and has various advantages -- worktree-less and index-less
   operation, faster, fixes some bugs); let's convert the rest of the
   callers over and then clean up by deleting merge-recursive (as
   well as removing the extra test scaffolding added long ago to aid
   in the conversion).

In other words the bug fixes, while real, were not the thrust of the story. I showed a handful of people my existing commit message and cover letter, all of whom found it a humorous way of stating that we're finally replacing merge-recursive. But, since this intent wasn't obvious to everyone, I can re-roll with some clarification.