[Piglit] Piglit Plans

[Daniel Kurtz]

+piglit@

Hi Chad,

I do not have much time to work on piglit these days, and when I do, I
usually just focus on trying to get piglit to work with the Chrome OS
ARM (EGL/GLES) platforms.
The last time I looked at it was about a month ago.

(1) To get stuff working, originally I planned to use libepoxy.
However, it has been 10 months since Eric Anholt first sent up
libepoxy conversion patches to piglit, and upstream piglit still
hasn't been converted yet.  So, I decided to just hack up piglit
dispatch to work with Mali's eglGetProcAddress() not accepting core
functions (glGetString) instead.
That patch also contains a fixes eglQueryString(EGL_NO_DISPLAY,
EGL_EXTENSIONS) when an EGL does not support
EGL_EXT_client_extensions.
* https://chromium-review.googlesource.com/#/c/223644/1

(2) I also have a set of piglit patches that fix some EGL tests, I've
been meaning to upstream them when I get a chance:
* https://chromium-review.googlesource.com/#/c/223645/

I have rebased these patches onto current upstream piglit commit [0]
and uploaded them to my newly created github repo [1].
[0] 88fcc67ab168ede4299effd2b0bac9ab7591c552
   registry: Update gl.xml
[1] https://github.com/djkurtz/piglit/tree/egl_work

With this patch set I get the following results on a "peach_pit"
Chromebook (Samsung ARM Chromebook 2:
http://www.samsung.com/us/computer/chrome-os-devices/XE503C12-K01US):
  # DISPLAY=:0 ./bin/piglit run lib/piglit/tests/quick.py results/q1
  [25492/25492] fail: 28, pass: 86, skip: 25378 Running Test(s): 25492
  Thank you for running Piglit!

112 tests total actually run.

(3) I also have a branch on github [2] that currently runs about 800
piglit tests for EGL/GLES.

[2] https://github.com/djkurtz/piglit/tree/egl_work_tom_dan

 # DISPLAY=:0 ./bin/piglit run lib/piglit/tests/quick.py results/q2
 [26446/26446] fail: 485, pass: 330, skip: 25631 Running Test(s):
26446
 Thank you for running Piglit!

The glsles shader/glslparser work is based heavily off of work
originally done by Tom Gall.  This work was also never upstreamed to
piglit, so, again, it gets stale quickly.
Also it takes a long time to go through all of shader test failures
and skips to see if they are real driver issues, bad tests, or due to
issues with the egl/gles2 piglit integration.   This is often a slow
manual process which involves lots of double checking against GLSL and
OpenGL ES specs.  This patch set is still very much WIP, and needs a
lot of work to clean it up and make it more upstream friendly.

Hope that helps.

Thanks,
-Dan

On Thu, Nov 13, 2014 at 5:59 AM, Chad Versace <chad.versace at intel.com> wrote:
> Daniel, what are your current plans for Piglit? And what are the big issues
> you are facing? I'm particularly interested in learning what you're doing
> with libepoxy.
>
> (If you're open to discussing this publicly, feel free to move this
> conversation to the Piglit list).
>
> ~Chad