[PATCH 5 of 5 STABLE v2] revset: use _diffset in orderedlazyset.__sub__

Gregory Szorc gregory.szorc at gmail.com
Tue Sep 9 11:47:33 CDT 2014


On 9/9/14 9:36 AM, Augie Fackler wrote:
> On Mon, Sep 08, 2014 at 02:59:47PM -0700, Gregory Szorc wrote:
>> # HG changeset patch
>> # User Gregory Szorc <gregory.szorc at gmail.com>
>> # Date 1410200370 25200
>> #      Mon Sep 08 11:19:30 2014 -0700
>> # Node ID d2c3dd725cf9b64b564f13096102f61f6b0d06ba
>> # Parent  b8226fb4800578b0d7b05f255f4c26140e458256
>> revset: use _diffset in orderedlazyset.__sub__
>
> This series looks good to me, but did you intend this for stable?
>
> (I'll leave these flagged, and will consider dropping them into default
> - this seems like it might be too invasive for stable.)

The clone time regression for the Firefox repo in 3.0 is a *major* 
regression. It appears Mercurial is hung for several minutes. I would 
like to see this fix rolled out ASAP so the window of impacted hg 
versions is as narrow as possible. If this is accepted in stable, I 
would even go so far as to ask for an early 3.0.2 release to further 
reduce the window by a few weeks.

I have the benchmarks running now, but I believe b8226fb48005 (the 
gnarliest patch in the series) isn't required to fix the regression and 
omitting it would introduce no revset benchmark regressions. Would that 
make stable inclusion easier?


More information about the Mercurial-devel mailing list