Bug 89819

Summary:	WebGL Conformance swrast failure in conformance/uniforms/uniform-default-values.html
Product:	Mesa	Reporter:	Luke <lukebenes>
Component:	Mesa core	Assignee:	mesa-dev
Status:	RESOLVED FIXED	QA Contact:	mesa-dev
Severity:	normal
Priority:	medium	CC:	sroland
Version:	unspecified
Hardware:	Other
OS:	All
Whiteboard:
i915 platform:		i915 features:

Description Luke 2015-03-30 00:56:09 UTC

The swrast renderer requires DRAW_USE_LLVM=false to pass the OGLES 2.0 Uniform Default Values test.

Steps to reproduce:
1. set LIBGL_ALWAYS_SOFTWARE=1
2. visit https://www.khronos.org/registry/webgl/sdk/tests/conformance/uniforms/uniform-default-values.html?webglVersion=1
3. set DRAW_USE_LLVM=false 
4. https://www.khronos.org/registry/webgl/sdk/tests/conformance/uniforms/uniform-default-values.html?webglVersion=1


llvmpipe results:
testing: samplerCube
fragment shaderFAIL Error in program linking:null
FAIL uniform is not zero
default value should be zero
FAIL at (0, 0) expected: 0,255,0,255 was 0,0,0,0
test test by setting value
FAIL at (0, 0) expected: 255,0,0,255 was 0,0,0,0
re-linking should reset to defaults
FAIL at (0, 0) expected: 0,255,0,255 was 0,0,0,0
FAIL getError expected: NO_ERROR. Was CONTEXT_LOST_WEBGL : should be no GL errors
...

See Bug 78875  for the i965

Comment 1 Luke 2015-06-04 04:33:48 UTC

Added Roland as he's been fixing up llvmpipe recently.

Comment 2 Roland Scheidegger 2015-06-04 12:28:12 UTC

wrong bogus default uniform value doesn't sound like something which could be done anything about in the driver.
That said, I tested this with llvmpipe and it seemed to work (I did see a crash though when loading that page but it didn't seem related, I'll fix that).

Comment 3 Luke 2015-06-06 18:52:25 UTC

Roland,
Thanks for looking into this! If you need the tests in a piglit form, see:

Bug 89754 -> http://cgit.freedesktop.org/piglit/commit/?id=aa97ac98dd8f1eab8f288f9509fdf3f91ffbecd6

and 

Bug 89759 -> http://cgit.freedesktop.org/piglit/commit/?id=b4497c06b5d7733357cbad7399e5855dd69a81e0

Comment 4 Luke 2015-06-07 04:47:18 UTC

> I did see a crash though when loading that page but it didn't seem related

Roland,
I just retested and only google-chrome is giving the results above. Firefox is crashing like you reported. To test with chrome:
chrome://flags/ -> Override software rendering list -> Enable

Comment 5 Roland Scheidegger 2015-06-08 23:57:38 UTC

FWIW I've fixed the crash.
I don't see any failures anymore with neither chromium nor firefox in that test now.

Comment 6 Luke 2015-06-15 18:47:23 UTC

Roland,
Did I miss the commit or is your fix still in the process of being reviewed? I'd be glad to help test out your patch.

Comment 7 Roland Scheidegger 2015-06-15 18:54:54 UTC

(In reply to Luke from comment #6)
> Roland,
> Did I miss the commit or is your fix still in the process of being reviewed?
> I'd be glad to help test out your patch.

The crash fix was 6e5970ffee0129fb94d8b7f0ebd4fac3992e7dce.

Comment 8 Luke 2015-06-16 20:17:26 UTC

Verified fixed by 6e5970ffee0129fb94d8b7f0ebd4fac3992e7dce in both Firefox and Chrome.

Great detective work Roland. Thanks!

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.