[Bug 92578] [BDW HSW dEQP] deqp-gles2 functional.texture.vertex.cube.filtering is failing in 4 of 36 test cases

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Tue Mar 13 06:11:37 UTC 2018


Kenneth Graunke <kenneth at whitecape.org> changed:

           What    |Removed                     |Added
             Status|REOPENED                    |RESOLVED
         Resolution|---                         |WONTFIX

--- Comment #9 from Kenneth Graunke <kenneth at whitecape.org> ---
I suspect this is an issue with the test.  GLES 3.0 enabled seamless cube map
filtering by default, while GLES 2.0 didn't.  But since GLES 2.0 can be legally
and silently promoted to 3.0, it's basically undefined whether it's enabled. 
There are comments in some of the tests that they try and avoid corner/edge
pixels that would hit this undefined behavior, but...the failures are all a
single pixel at the very corner.

Notably, these two tests begin to pass if we disable seamless cubemap filtering
(which we can't reasonably do):


These two continue failing:


Notably, i965 passes the dEQP-GLES3 versions of the same tests.  I'm not aware
of us doing something different for GLES2 and GLES3, so my guess is some bug
fix to the tests didn't get copy and pasted back to the older GLES2 test.

More importantly, these are not in the Android CTS mustpass list, which means
Google doesn't require them.  If Google doesn't think it's reasonable to
require graphics vendors to pass these tests, and neither does Khronos, I'm
inclined to agree and not spend any effort fixing them.  I think you should
change how you're running the test suite to only run tests in the mustpass

You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20180313/7d538aba/attachment.html>

More information about the intel-3d-bugs mailing list