1204076 - MultiTiledContentClient on B2G doesn't produce the same results as SingleTiledContentClient on this SVG with transforms

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect)

Description

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

Attachment: Test case

While tracking down some reftest failures from my patches in bug 1201529 I discovered the root cause appears to be some sort of a bug in the MultiTiledContentClient code (specifically the DrawTargetTiled code which is enabled on B2G). The simple STR are:

1) Flash a B2G device with a build that contains the patches on bug 1201529. (The relevant effect is to cause the WantAsyncScroll() function in nsGfxScrollFrame.cpp to return true for the root content scrollframe).
2) Load the attached file. This file is the gfx/tests/reftest/1143303-1.svg file, minus the meta-viewport tag that I'm adding to it in bug 1201529.

Expected results:
The SVG loads as completely green, the way it does on desktop

Actual results:
A red circle appears

Something in the DrawTargetTiled code doesn't like something about the transforms in the SVG.

Note that you can also reproduce in the B2G emulator by running the reftest directly, assuming you remove the meta-viewport tags I'm adding in bug 1201529:

B2G-emulator$ ./mach reftest-remote gecko/gfx/tests/reftest/11433030-1.svg

A similar problem occurs with a second reftest, although that one is not as easy to reproduce on a standalone device since it requires an async zoom. The reftest reproduces it though:

B2G-emulator$ ./mach reftest-remote gecko/layout/reftests/reftest-sanity/async-zoom-2.html

Comment 1

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

The patch on bug 1201937 seems to fix this.

Comment 2

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

Attachment: Remove workarounds

Comment 3

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

Try push: https://treeherder.mozilla.org/#/jobs?repo=try&revision=75c7c3c268a2

Comment 4

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

Hm according to the try push the async-zoom-2 failure is still there on the B2G emulator R-1 job. The 11433030-1.svg reftest failure is gone though.

Comment 5

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

Attachment: Remove workaround

Well I guess I can remove this one. I'll clone this bug for the other test.

Comment 6

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/085253afab58

Comment 7

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/085253afab58