1169956 - It takes a long time to paint the mozilla.org page because of scaled tiled images

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Mike Conley (:mconley) (:⚙️)]

STR (non-e10s);

1) Visit mozilla.org in a tab
2) Open a second tab, and visit about:blank in it
3) Switch back to the mozilla.org tab

ER:

Nearly-instantaneous switching to the mozilla.org tab

AR:

~2s delay before the tab switches, before which the whole browser janks.

Here's a profile (ignore the content process - again, this was done in a non-e10s window):

https://people.mozilla.org/~sguo/cleopatra/#report=3e45f228f6c9e9d9deb92111715d900db6797825&selection=0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,153,154,155,156,214,215,216,217,218,219,220,221,420,1037,1038,1039,1039,1040,1041,1042,1043,1044,1045,237,238,239,240,766,242,243,244,767,1046,1047,1048,1049,1050,1051,1052,1053,1054,1055,1056,1057,773,229,230,231,329,331,332,333,249,250,251,252,253,254,255,1513

Comment 1

[Mike Conley (:mconley) (:⚙️)]

Note that attempting this with e10s enabled causes the content process to hang for approximately the same amount of time. As one would expect, the main browser UI remains responsive, but we get big fat spinners in the content area.

Comment 2

[Mike Conley (:mconley) (:⚙️)]

Last good nightly: 2015-04-20
First bad nightly: 2015-04-21

Pushlog:
http://hg.mozilla.org/mozilla-central/pushloghtml?startdate=2015-04-20&enddate=2015-04-21

jrmuizel puts his bet squarely on http://hg.mozilla.org/mozilla-central/rev/2d97da8feb72.

I can try to confirm this tomorrow.

Comment 3

[Jeff Muizelaar [:jrmuizel]]

[Tracking Requested - why for this release]: This is a major unexpected performance regression.

Comment 4

[Mason Chang [Inactive] [:mchang]]

Attachment: Backout bug 1073209 - Reinstate CSRD on OS X

Since I haven't had time to invest in bug 1154311, and removing CSRD on OSX has caused this plus bug 1164601, add CSRD back to OS X for now to fix the regressions.

Comment 5

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 8615483 [details] [diff] [review]
Backout bug 1073209 - Reinstate CSRD on OS X

Review of attachment 8615483 [details] [diff] [review]:
-----------------------------------------------------------------

:(

Comment 6

[Mike Conley (:mconley) (:⚙️)]

I can confirm that this backout makes things better on my machine.

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/baeb37ae48d2

Comment 8

[Guilherme Lima]

Landed yesterday - https://hg.mozilla.org/mozilla-central/rev/baeb37ae48d2

Comment 9

[jen strickland :jenstrickland]

I tracked for 40 and 41, due to the performance regression, but if I understand correctly this is now fixed.

Comment 10

[Mike Conley (:mconley) (:⚙️)]

Hey mchang - can I assume we'll want this to land on Aurora as well?

Comment 11

[Mason Chang [Inactive] [:mchang]]

Attachment: aurora backout patch

Backout patch rebased for aurora.

Comment 12

[Mason Chang [Inactive] [:mchang]]

Comment on attachment 8621144 [details] [diff] [review]
aurora backout patch

Approval Request Comment
[Feature/regressing bug #]: Bug 1073209 on OS X.
[User impact if declined]: Performance regression with scaled background images.
[Describe test coverage new/current, TreeHerder]: Mochitests / reftests  / manual testing.
[Risks and why]: Low: CreateSamplingRestrictedDrawable has been in firefox for a very long time. Bug 1073209 deleted it on OS X, which caused performance regressions. This reinstates CreateSamplingRestrictedDrawable 
[String/UUID change made/needed]: None

Comment 13

[Sylvestre Ledru [:Sylvestre]]

Comment on attachment 8621144 [details] [diff] [review]
aurora backout patch

Important regression, taking it.

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/c65ac58ea75e