Re: [PATCH 0/4] remote-hg: more improvements

From: Junio C Hamano <gitster@pobox.com>
Date: May 9, 2014, 17:16 UTC
Message-ID: <xmqq61le27p5.fsf@gitster.dls.corp.google.com>
In-Reply-To: <536c15cf8594b_6f3bb152ecb@nysa.notmuch>

Felipe Contreras <felipe.contreras@gmail.com> writes:

Show 13 quoted lines

> Junio C Hamano wrote:
>> Felipe Contreras <felipe.contreras@gmail.com> writes:
>> 
>> > I already said this multiple times, but let me be clear once more:
>> >
>> > MASTER HAS A REGRESSION (for all versions of Mercurial).
>> 
>> As you said, that is not a regression, isn't it?  It is an old
>> breakage that existed even before 1.9 (was it 1.8.3 or something?)
>
> No. It does't happen in 1.9, it will happen in 2.0.
>
> That's a REGRESSION.

You earlier said in our exchange:

    > Wrong. The code in question was not recent, it was introdued in 1.8.3,
    > more than one year ago.

and "git blame -L870,880 contrib/remote-helpers/git-remote-hg" does show that 68d4f4f3 (remote-hg: custom method to write tags, 2013-04-22) is the culplit that has introduced ManifestLookupError.

And it is in since 1.8.3. How does it break 2.0 and without breaking 1.9? The dormant and otherwise unused code got its first caller, or something?

If this were a core part of the system the only sensible thing we can do at this point is to revert at this late stage of the cycle, but I do not think I have time to bisect and find that culprit, as today is when -rc3 has been scheduled to happen.