572571 - "ASSERTION: FBO not supported" and "ASSERTION: nsCARenderer::Render failure"

Status: RESOLVED FIXED

(Core Graveyard :: Plug-ins, defect)

Description

[Jesse Ruderman]

Attachment: testcase

Using a debug build on Mac OS X 10.6:

###!!! ASSERTION: FBO not supported: 'Error', file /Users/jruderman/mozilla-central/gfx/src/thebes/utils/nsCoreAnimationSupport.mm, line 541

###!!! ASSERTION: nsCARenderer::Render failure: 'Not Reached', file /Users/jruderman/mozilla-central/layout/generic/nsObjectFrame.cpp, line 3671

Comment 1

[Jesse Ruderman]

Attachment: gdb output (includes stacks for the assertions)

Comment 2

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

This is plugins, I believe -- don't think we use CA for anything other than OOPP in the browser.

Comment 3

[Benoit Girard (:BenWa)]

Yup, it's a problem with the 'Core Animation' drawing model for plug-ins. 

Jesse can you include your Graphic Card information?

Comment 4

[Jesse Ruderman]

QuickTime Plug-in 7.6.6

NVIDIA GeForce 9400

Comment 5

[Jesse Ruderman]

Attachment: detailed graphics card info

Comment 6

[Benoit Girard (:BenWa)]

I looked into the problem further and assumed that TEXTURE_RECTANGLE was not supported. From what I gather from the CGLTexImageIOSurface2D requires the use of TEXTURE_RECTANGLE so we might have no choice but to use it for IOSurface.

Jesse can you see if you have "OpenGL Driver Monitor" installed. Select 'Monitors', "Renderer Info" from the menu bar. Select 'Save As Text' and attach it? I'm interested in seeing what support is provided TEXTURE_RECTANGLE.

Comment 7

[Jesse Ruderman]

Attachment: output from OpenGL Driver Monitor

Here you go, BenWa.  (I clicked "Expand All" before clicking "Save As Text".)

Comment 8

[Benoit Girard (:BenWa)]

It just occurred to me that the Quicktime's object frame had a width of 20,000. Core Animation, or at least the CARenderer, doesn't even support anything this large: 
-[<OOPMovieLayer: 0x2f57b00> display]: Ignoring bogus layer size (20000.000000, 150.000000)
Chromium has the same error.

roc thinks that we should just render nothing, which I agree with, which is the behavior I currently get. The only problem is that the code keeps trying to initialize and this is draining the CPU so I will post a patch for this.

If someone thinks the expect behavior should be otherwise let me know.

Comment 9

[Benoit Girard (:BenWa)]

Attachment: Prevent Setup if size is unsupported

This patch prevents the caller from trying to initialize the nsCARenderer again if we know that this size is unsupported.

Comment 10

[Benoit Girard (:BenWa)]

Checked the patch for bit-rot, still working. Checkin-needed.

Comment 11

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

http://hg.mozilla.org/integration/mozilla-inbound/rev/946a55feeb02

Comment 12

[Mounir Lamouri (:mounir)]

Merged:
http://hg.mozilla.org/mozilla-central/rev/946a55feeb02

Comment 13

[Vlad [QA]]

How can this be verified? Can you please provide some STR.
Thanks