Bug 90993

Summary:	[m32 bisected] arb_gpu_shader5.execution.*array_indexing.vs-nonuniform-control-flow
Product:	piglit	Reporter:	Mark Janes <mark.a.janes>
Component:	tests	Assignee:	Piglit Mailing List <piglit>
Status:	RESOLVED FIXED	QA Contact:	Piglit Mailing List <piglit>
Severity:	normal
Priority:	medium	CC:	currojerez, mark.a.janes
Version:	unspecified
Hardware:	x86 (IA32)
OS:	Linux (All)
Whiteboard:
i915 platform:		i915 features:

Description Mark Janes 2015-06-16 16:38:01 UTC

Two new tests fail to pass on m32 builds:

spec.arb_gpu_shader5.execution.ubo_array_indexing.vs-nonuniform-control-flow
spec.arb_gpu_shader5.execution.sampler_array_indexing.vs-nonuniform-control-flow

IVB and later fails these tests.  The tests were introduced by:

piglit 2d4fe1a0fba128ea3189485f3b19b7a894b1d611
piglit d615a588e9487647e89bada9a6120855784f65e9

It is possible that the tests are correct, indicating a bug in mesa.

Comment 1 Mark Janes 2015-06-16 17:43:27 UTC

further testing has shown that these tests also fail for 10.5, or master with the non-NIR code path.  This is for both m32/m64.

Testing the non-NIR path also triggers a third (fs) failure for both m32/m64:

arb_gpu_shader5.execution.sampler_array_indexing.fs-struct-nonconst-sampler-nonconst

Comment 2 Mark Janes 2015-06-16 19:08:31 UTC

correction: there is no non-NIR path on master.  The non-NIR failures are on 10.5 only.

Comment 3 Francisco Jerez 2015-06-17 14:21:22 UTC

I'm unable to reproduce on mesa master, both tests pass for me on HSW GT2.

10.5 is expected to fail because the fix (e1ae0c3bc37be7b1de21ee248d674671d01da8e6 and b234537cc3e513ded9b5385d876e4c531f72af94) was applied after the branching point.

Comment 4 Mark Janes 2015-06-17 21:30:30 UTC

This reproduces 100% of the time on the build system, and the output is always the same:

jenkins@otc-gfxtest-hswgt2-01:~$ PIGLIT_PLATFORM=gbm LIBGL_DRIVERS_PATH=/tmp/build_root/m32/lib/dri S2TC_DITHER_MODE=NONE LD_LIBRARY_PATH=/tmp/build_root/m32/lib:/tmp/build_root/m32/lib/i386-linux-gnu:/tmp/build_root/m32/lib/dri:/tmp/build_root/m32/lib/piglit/lib GBM_DRI\
VERS_PATH=/tmp/build_root/m32/lib/dri /tmp/build_root/m32/lib/piglit/bin/shader_runner /tmp/build_root/m32/lib/piglit/tests/spec/arb_gpu_shader5/execution/ubo_array_indexing/vs-nonuniform-control-flow.shader_test -auto
Probe color at (174,0)
  Expected: 0.400000 0.600000 0.800000 0.200000
  Observed: 0.200000 0.200000 0.200000 0.200000
Probe color at (174,50)
  Expected: 0.400000 0.600000 0.800000 0.200000
  Observed: 0.200000 0.200000 0.200000 0.200000
Probe color at (174,75)
  Expected: 0.400000 0.600000 0.800000 0.200000
  Observed: 0.200000 0.200000 0.200000 0.200000
Probe color at (174,125)
  Expected: 0.400000 0.600000 0.800000 0.200000
  Observed: 0.200000 0.200000 0.200000 0.200000
PIGLIT: {"result": "fail" }


Curro, I can give you time on a build system if necessary, so you can debug it.

Comment 5 Francisco Jerez 2015-07-29 16:31:28 UTC

This should be fixed now.

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.