On a very intermittent basis, this test (in GLESEXT and GLES32 conformance tests) fails on skl, kbl, and cfl with:
Rendered data for [x=3 y=0 attachment=0 layer=0] [0, 0, 0, 0] are different from reference data [255, 255, 255, 255] !
I've sent patches which I think will fix it:
Seems like this test is flaky again on gen9 hardware:
(sort by 'hardware', see 2 failures in last few days on gen9)
Test output is a little different than last time:
Rendered data for [x=0 y=0 attachment=0 layer=0] [0, 0, 0, 0] are different from reference data [255, 255, 255, 255] !
Hello. Just a side note - tested this test in a loop (about 100-200 times) on latest mesa from git on these HW => CFL, KBL, SKL.
Didn't see any failures.