1189817 - large talos e10s regressions on windows July 24th on mozilla-central

Status: RESOLVED WORKSFORME

(Testing :: Talos, defect)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

46 commits in a merge to m-c:
https://hg.mozilla.org/mozilla-central/pushloghtml?changeset=9bca608ab65a

yields a lot of windows regressions on e10s only:
http://alertmanager.allizom.org:8080/alerts.html?rev=9bca608ab65a&showAll=1&testIndex=0&platIndex=0

In summary these regressions are:
tresize - 15.8-39.7%
tpaint - 4.8% (win7 only)
tsvg_opacity - 2.9-3.2% (win7/win8)
tart - 6.8 - 7.1% 
tsvg - 4.8-6.5% (win7/win8)
tscroll - 40-51% (win7/win8)
** winxp has a 15% win
cart - 3.8% (win7)
** winxp has a 3.8% win
tp5o_scroll - 136-250%
damp - 4.6% - 6.8%

Comment 1

[Brad Lassey [:blassey] (use needinfo?)]

not tracking for e10s because this is APZ related

Comment 2

[Benoit Girard (:BenWa)]

Bug 1191539 caused several large improvements to tscroll-asap and tp5 on windows which were the some of the worse offenders here.

tresize should be improved by bug 1186662 but it will only improve if we're doing a real scroll (we get the windows sizemove message) so we might still have a large regression during synthetic resize.

Comment 4

[Kartikaya Gupta (email:[e-mail])]

We have made a bunch of improvements over time in various bugs, and there are a couple other still open for more specific things. The remaining regressions are just going to be accepted. There's no value in leaving this bug open anymore, I think.