Bug 106907

Any chance you might be able to write a small piglit test
that shows the bug? For example:

https://cgit.freedesktop.org/piglit/commit/?id=f1dc46ddf8c1

FYI on my Fedora machine, there are failures with this test already with Mesa 17.3.6 so at least some of these are pretty old issues, before shader cache was turned on.

(In reply to Tapani Pälli from comment #2)
> FYI on my Fedora machine, there are failures with this test already with
> Mesa 17.3.6 so at least some of these are pretty old issues, before shader
> cache was turned on.

It sounds like this is possibly an interaction with
ARB_get_program_binary, and not the shader cache.

Hmm, except ARB_get_program_binary wasn't really
supported until 18.0. So, if failures go back to 17.3, then
maybe it is something else.

I've noticed that if I skip uniform removal in opt_dead_code (even builtin uniforms), then this test passes. I have no idea why this is though. I have been running chrome with following arguments: "--use-gl=egl --disable-gpu-program-cache --disable-gpu-shader-disk-cache".

(In reply to Tapani Pälli from comment #4)
> I've noticed that if I skip uniform removal in opt_dead_code (even builtin
> uniforms), then this test passes. I have no idea why this is though. I have
> been running chrome with following arguments: "--use-gl=egl
> --disable-gpu-program-cache --disable-gpu-shader-disk-cache".

Hmm it seems that these flags are not really honored by Chrome though, if I 'disable' program binary by just returning from program binary functions then there are far less failures ... so this is probably about program binary after all.

Fix proposal sent here:
https://lists.freedesktop.org/archives/mesa-dev/2018-June/197678.html

(In reply to Tapani Pälli from comment #6)
> Fix proposal sent here:
> https://lists.freedesktop.org/archives/mesa-dev/2018-June/197678.html

Hi, Tapani, thank you for your patch, seems that this patch could resolve the bug, I will double check it tomorrow.
Will you merge this patch to mesa master?

(In reply to xinghua from comment #7)
> (In reply to Tapani Pälli from comment #6)
> > Fix proposal sent here:
> > https://lists.freedesktop.org/archives/mesa-dev/2018-June/197678.html
> 
> Hi, Tapani, thank you for your patch, seems that this patch could resolve
> the bug, I will double check it tomorrow.
> Will you merge this patch to mesa master?

If others (reviewers) agree that this is correct, then yes.

(In reply to Jordan Justen from comment #1)
> Any chance you might be able to write a small piglit test
> that shows the bug? For example:
> 
> https://cgit.freedesktop.org/piglit/commit/?id=f1dc46ddf8c1

Hi, Jordan, Tapani had already submitted a patch for this issue. Do I still need to write a test in piglit.

(In reply to xinghua from comment #9)
> (In reply to Jordan Justen from comment #1)
> > Any chance you might be able to write a small piglit test
> > that shows the bug? For example:
> > 
> > https://cgit.freedesktop.org/piglit/commit/?id=f1dc46ddf8c1
> 
> Hi, Jordan, Tapani had already submitted a patch for this issue. Do I still
> need to write a test in piglit.

No its ok I made Tapani write a piglit test to go along with the patch :P

--- 8< ---
commit ab2643e4b06f63c93a57624003679903442634a8
Author: Tapani Pälli <tapani.palli@intel.com>
Date:   Thu Jun 14 11:10:20 2018 +0300

    glsl: serialize data from glTransformFeedbackVaryings
    
    While XFB has been enabled for cache, we did not serialize enough
    data for the whole API to work (such as glGetProgramiv).
    
    Fixes: 6d830940f7 "Allow shader cache usage with transform feedback"
    Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=106907
    Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>