Closed Bug 1009642 Opened 10 years ago Closed 10 years ago

Hundreds of webgl_conformance failures on OSX 10.9

Categories

(Core :: Graphics: CanvasWebGL, defect)

Product:
Core
Component:
Graphics: CanvasWebGL
Platform:
x86
macOS
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED WORKSFORME

People

(Reporter: jgriffin, Assigned: u480271)

References

Details

When run on OSX 10.9, the test test_webgl_conformance_test_suite.html reports hundreds of failure of this type:

 19:13:17     INFO -  2785 INFO TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/webgl-conformance/test_webgl_conformance_test_suite.html | [conformance/extensions/webgl-compressed-texture-s3tc.html] Test failed - at (5, 0) expected: 0,255,0,255 was 0,0,255,255
19:13:17     INFO -  FAIL: at (5, 0) expected: 0,255,0,255 was 0,0,255,255
19:13:17     INFO -  STACK TRACE:
19:13:17     INFO -  compareRect@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:594:1
19:13:17     INFO -  testDXTTexture@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:521:13
19:13:17     INFO -  testDXTTextures@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:257:9
19:13:17     INFO -  testDXT1_RGB@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:197:5
19:13:17     INFO -  runTestExtension@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:176:5
19:13:17     INFO -  @http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:108:9
19:13:17     INFO -  2785 INFO TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/webgl-conformance/test_webgl_conformance_test_suite.html | [conformance/extensions/webgl-compressed-texture-s3tc.html] Test failed - at (5, 0) expected: 0,255,0,255 was 0,0,255,255
19:13:17     INFO -  FAIL: at (5, 0) expected: 0,255,0,255 was 0,0,255,255
19:13:17     INFO -  STACK TRACE:
19:13:17     INFO -  compareRect@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:594:1
19:13:17     INFO -  testDXTTexture@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:521:13
19:13:17     INFO -  testDXTTextures@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:257:9
19:13:17     INFO -  testDXT1_RGB@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:197:5
19:13:17     INFO -  runTestExtension@http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:176:5
19:13:17     INFO -  @http://mochi.test:8888/tests/content/canvas/test/webgl-conformance/conformance/extensions/webgl-compressed-texture-s3tc.html:108:9

full log:  https://tbpl.mozilla.org/php/getParsedLog.php?id=39531720&tree=Cedar&full=1
Milan, does anyone on your team have time to take a look at this?  This is probably the biggest blocker to getting tests running on OSX 10.9 on all trunk trees.
Flags: needinfo?(milan)
I actively develop on OSX 10.9 don't see these errors on my rMBP. I'm checking against my Haswell Mac Mini.
It's possible that there is some configuration needed on the OSX 10.9 slaves, in that case.
(In reply to Dan Glastonbury :djg :kamidphish from comment #2)
> I actively develop on OSX 10.9 don't see these errors on my rMBP. I'm
> checking against my Haswell Mac Mini.

We can give you access to an OSX 10.9 test slave; would you be able to test it there and see if there's some kind of configuration problem on those slaves that's causing these issues?
I don't recall 10.9 swapping green and blue channels :), so if we can get a test slave (send it Dan's way), we can take a look.
Assignee: nobody → dglastonbury
Flags: needinfo?(milan)
Nothing in the output is dumping renderer info.. I thought we had that in the test framework.  Would be useful to have.

Also, from looking at the test failures -- we have a 2-3 webgl tests failing (s3tc, a glsl gradient function, and one other thing I can't remember), and then also test-canvas.html, which is just 2D canvas.  So this isn't just a webgl-only thing.

(Each webgl test may have hundreds of test cases inside it.. the s3tc one itself has a few hundred. So the actual "areas" that are failing are much smaller than it appears from the output.)
Depends on: 1019135
From what I can ascertain from the test machine, there is a bug in glCompressedTexSubImage2D on Intel HD 3000 drivers. The failing tests are ones that update a sub-image of an image. I don't see a work around for this, so I'm going to blacklist s3tc textures on OSX 10.9 with Intel HD 3000.

I've tested Intel HD 4000 and Intel HD 5000/Iris and those drivers don't exhibit the same issue.
Depends on: 1024256
This will require uplift to Aurora and Beta. I'll check that the patch applies cleanly to both of those.
This seems to have been resolved.
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → WORKSFORME
Dan, let's see if this problem is fixed on the Apple side in 10.10, since our code only guards it against 10.9?  If it isn't, we can open another bug...
Flags: needinfo?(dglastonbury)
(In reply to Milan Sreckovic [:milan] from comment #10)
> Dan, let's see if this problem is fixed on the Apple side in 10.10, since
> our code only guards it against 10.9?  If it isn't, we can open another
> bug...

Milan, I'd need access to HD 3000 + 10.10. I don't have that hardware.
Flags: needinfo?(dglastonbury)
Re: Comment 11.
Flags: needinfo?(milan)
(In reply to Dan Glastonbury :djg :kamidphish from comment #11)
> (In reply to Milan Sreckovic [:milan] from comment #10)
> > Dan, let's see if this problem is fixed on the Apple side in 10.10, since
> > our code only guards it against 10.9?  If it isn't, we can open another
> > bug...
> 
> Milan, I'd need access to HD 3000 + 10.10. I don't have that hardware.

Got it.
Flags: needinfo?(milan)
Looks like this is an issue on 10.10 too. (Bug 1124996)
You need to log in before you can comment on or make changes to this bug.