I'm seeing a similar issue in Windows when using internal format 
RGB10_A2 (and UNSIGNED_INT_2_10_10_10_REV_EXT as type) on nVidias
(crrev.com/c/868172); searching for this around the net I'm getting
the feeling that 10 bit rendering must be force-enabled via control
panel (or perhaps connecting a 10bpc+ monitor)...

kbr@ any suggestion how to work around these issues on test infra?

Also possible: GeForce cards don't support 10bit modes, Quadro do:

> These programs use OpenGL 10-bit per color buffers which require an NVIDIA 
> Quadro GPU with DisplayPort connector.  A small number of monitors support 
> 10-bit per color with Quadro graphics cards over DVI.

https://nvidia.custhelp.com/app/answers/detail/a_id/3011/kw/30%20bit

In Linux it seems to be linked to full screen :
https://mediapc.org/video/10-bit-color-with-nvidia-geforce/

mcasas@: this bot is a MacBook Pro with NVIDIA GPU running 10.13.x and using its internal display. If your test is failing on it then this code path won't be working on customer machines of this configuration.

It seems that the problem is that somewhere in the path the R and B
components got interchanged (maybe nvidia interprets GL_RGB10_A2 as
RGBA and not as BGRA ?):

[ RUN      ] GLImageIOSurface/GLImageBindTest/1.BindTexImage
../../ui/gl/test/gl_test_helper.cc:77: Failure
The difference between expected and actual is 32, which exceeds error, where
expected evaluates to 48,
actual evaluates to 16, and
error evaluates to 0.
 at 0, 0 channel 0
../../ui/gl/test/gl_test_helper.cc:77: Failure
The difference between expected and actual is 32, which exceeds error, where
expected evaluates to 16,
actual evaluates to 48, and
error evaluates to 0.
 at 0, 0 channel 2

Gonna disable the test entry to flip the bot green while I get my hands
on an MBP + nVidia GeForce and investigate more. 

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/68721e13803faa18eb82d02f40244e54af52a084

commit 68721e13803faa18eb82d02f40244e54af52a084
Author: Miguel Casas-Sanchez <mcasas@chromium.org>
Date: Thu Jan 18 21:43:24 2018

gl_unittests mac: disable test that fails in MBP + nVidia

This CL disables GLImageIOSurface/GLImageBindTest/1.BindTexImage (i.e.
with gfx::BufferFormat::BGRX_1010102) that seems to be failing on
Mac Book Pro 10.13 bots with nVidia GeForce 750M "(4.0.8 [3.2.8])".

Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I817f489ef47a387a0d174fb084ed6717fe8c86d1
Reviewed-on: https://chromium-review.googlesource.com/874795
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#530281}
[modify] https://crrev.com/68721e13803faa18eb82d02f40244e54af52a084/ui/gl/gl_image_io_surface_unittest.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/eec77727c1debdbe5c5b0206f0274b0a091eaa2a

commit eec77727c1debdbe5c5b0206f0274b0a091eaa2a
Author: Miguel Casas-Sanchez <mcasas@chromium.org>
Date: Fri Jan 19 03:28:15 2018

GLImageIOSurface: moar testing and error logging

This CL extends gl_image_io_surface_unittest.cc to try
a few more formats and addresses the strange situation
of RGBA_8888 being handled as an BGRA by writing a
significant comment.

Also: better logging and comments in
gl_image_io_surface.mm and gl_image_test_support.cc.

Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I630a4bee5cd1003d4b10da8dece4dadcced63bed
Reviewed-on: https://chromium-review.googlesource.com/874898
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Reviewed-by: Antoine Labour <piman@chromium.org>
Cr-Commit-Position: refs/heads/master@{#530417}
[modify] https://crrev.com/eec77727c1debdbe5c5b0206f0274b0a091eaa2a/ui/gl/gl_image_io_surface.mm
[modify] https://crrev.com/eec77727c1debdbe5c5b0206f0274b0a091eaa2a/ui/gl/gl_image_io_surface_unittest.cc
[modify] https://crrev.com/eec77727c1debdbe5c5b0206f0274b0a091eaa2a/ui/gl/test/gl_image_test_support.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/5f8509786648215431ad301fa2ae9ed6c8fc09f8

commit 5f8509786648215431ad301fa2ae9ed6c8fc09f8
Author: Miguel Casas-Sanchez <mcasas@chromium.org>
Date: Sat Jan 20 01:34:40 2018

GLImageIOSurface: enable BGRX GLImageBindTest

This CL extends gl_image_io_surface_unittest.cc to run
GLImageBindTest with BGRX, which is treated internally as
BGRA, by extending the value-to-compare-to and writing a
significant comment to explain the situation.

Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: Ia9cfc20bb1be054da10cd0bd2526bc3f7f66d144
Reviewed-on: https://chromium-review.googlesource.com/876222
Reviewed-by: Erik Chen <erikchen@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Daniele Castagna <dcastagna@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#530721}
[modify] https://crrev.com/5f8509786648215431ad301fa2ae9ed6c8fc09f8/ui/gl/gl_image_io_surface.mm
[modify] https://crrev.com/5f8509786648215431ad301fa2ae9ed6c8fc09f8/ui/gl/gl_image_io_surface_unittest.cc

I got my hands on a MBPro Retina, Late 2013, w/ 10.13.1 and an nVidia 
GeForce GT 750M, alongside an Intel Iris Pro. Went to the Preferences,
Energy Save and disabled the "Automatic Graphics Switching", compiled
chromium and gl_unittests.  Chromium's chrome://gpu indentified the
nVidia as being used. With that:

- I could repro adding the BGRX_1010102 test case to gl_unittests' 
 GLImageBindTest/?.BindTexImage, and it fails like in the description
 of this bug, i.e. swapping the values read back from R and B channels.

- I patched in a CL making use of XR30: crrev.com/c/862326, which uses
 GpuMemoryBuffers backed by IOSurface being translated to GL_RGB10_A2
 GLImages: this patch works, in the sense that I can repro VP9.2 videos
 without any color swizzles. 

- Moreover, both the mentioned patch and the BGRX_1010102 gl_unittests
 works perfectly in my MBP 13' w/ intel only, and my iMac 2017 at home
 with the AMD GPU.

So... are we doing something in Chrome we aren't doing in the tests ?
I feel tempted to land the test again with a specific provision for
nVidia to swap the channels. Any comments ?

+ccameron

Chrome runs through the entire command buffer infrastructure, where gl_unittests is a much slimmer software stack going right to the macOS GL driver. It's certainly possible there is code somewhere in the command buffer which sets up a texture swizzle which would be correcting the swap of the B and R channels. I don't know where it would be though.

I'd hazard a guess based on the CGLTexImageIOSurface2D behavior ...

    Supported IOSurface texture format/type/internal format combinations in Mac OS X 10.6
    Format                          Type                                    Internal Format         Sized Internal Format
    GL_BGRA                         GL_UNSIGNED_INT_2_10_10_10_REV          GL_RGBA                 GL_RGBA8

... that there are bugs where some drivers treat this as BGRA for ReadPixels and others don't.

https://github.com/phracker/MacOSX-SDKs/blob/HEAD/MacOSX10.12.sdk/System/Library/Frameworks/OpenGL.framework/Versions/A/Headers/CGLIOSurface.h#L102 :
    Format                          Type                                    Internal Format         Sized Internal Format
    GL_BGRA                         GL_UNSIGNED_INT_2_10_10_10_REV          GL_RGBA                 GL_RGB10_A2

CL in crrev.com/c/879331 ,
verified on my borrowed MBP w/ nVidia.

Repro'ed with dcastagna@s code modified, see attached picture, will upload
code shortly and file with Apple & link here the bug for ref.

https://bugreport.apple.com/web/?problemID=36824694 is the bug report.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8eb4ed812199b6a8c0ea677b5e6243eb54248f94

commit 8eb4ed812199b6a8c0ea677b5e6243eb54248f94
Author: Miguel Casas-Sanchez <mcasas@chromium.org>
Date: Wed Jan 24 16:43:56 2018

Mac gl_unittests: reenable GLImageBindTest BGRX_1010102

This CL reenables the GLImageBindTest for BGRX_1010102, adding a provision
to swap the B and R components on nVidia platforms to avoid some mysterious
bug -- this is likely afflicting only the glReadPixels() since other tests
(e.g. crrev.com/c/862326 render colors without swapping), filed as
http://www.openradar.me/36824694, rdar://36824694.

Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I73dbe09b1c1d26cc8ac195808545de63da082ac7
Reviewed-on: https://chromium-review.googlesource.com/879331
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Reviewed-by: ccameron <ccameron@chromium.org>
Cr-Commit-Position: refs/heads/master@{#531557}
[modify] https://crrev.com/8eb4ed812199b6a8c0ea677b5e6243eb54248f94/ui/gl/gl_image_io_surface_unittest.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/560bf05aa22f8e8f04f95c6312f1c3eb94734e46

commit 560bf05aa22f8e8f04f95c6312f1c3eb94734e46
Author: Miguel Casas <mcasas@chromium.org>
Date: Wed Jan 24 19:53:16 2018

Revert "Mac gl_unittests: reenable GLImageBindTest BGRX_1010102"

This reverts commit 8eb4ed812199b6a8c0ea677b5e6243eb54248f94.

Reason for revert: Broke Mac Retina Release (NVIDIA) under
chromium.gpu.fyi:
https://ci.chromium.org/buildbot/chromium.gpu.fyi/Mac%20Retina%20Release%20%28NVIDIA%29/7282

Original change's description:
> Mac gl_unittests: reenable GLImageBindTest BGRX_1010102
> 
> This CL reenables the GLImageBindTest for BGRX_1010102, adding a provision
> to swap the B and R components on nVidia platforms to avoid some mysterious
> bug -- this is likely afflicting only the glReadPixels() since other tests
> (e.g. crrev.com/c/862326 render colors without swapping), filed as
> http://www.openradar.me/36824694, rdar://36824694.
> 
> Bug:  803473 
> Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
> Change-Id: I73dbe09b1c1d26cc8ac195808545de63da082ac7
> Reviewed-on: https://chromium-review.googlesource.com/879331
> Commit-Queue: Miguel Casas <mcasas@chromium.org>
> Reviewed-by: ccameron <ccameron@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#531557}

TBR=ccameron@chromium.org,mcasas@chromium.org

Change-Id: Ica55ba2ed4b1ed64d11e72bb6c254e155cf14fc7
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Reviewed-on: https://chromium-review.googlesource.com/883642
Reviewed-by: Miguel Casas <mcasas@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#531654}
[modify] https://crrev.com/560bf05aa22f8e8f04f95c6312f1c3eb94734e46/ui/gl/gl_image_io_surface_unittest.cc

Sorry the trybots didn't catch that. If we should be running gl_unittests on mac_optional_gpu_tests_rel (which is the only trybot that runs tests on Mac NVIDIA, due to capacity constraints), can you consider adding the following:

diff --git a/content/test/gpu/generate_buildbot_json.py b/content/test/gpu/generate_buildbot_json.py
index 34121960d5a5..f1fa7b648c87 100755
--- a/content/test/gpu/generate_buildbot_json.py
+++ b/content/test/gpu/generate_buildbot_json.py
@@ -1452,6 +1452,11 @@ COMMON_GTESTS = {
      ]
   },
   'gl_unittests': {
+    'tester_configs': [
+      {
+        'predicate': Predicates.DEFAULT_AND_OPTIONAL,
+      },
+    ],
     'disabled_tester_configs': [
       {
         'names': [


and re-running that script? Feel free to send me a CL. Thanks.

So, the CL got reverted due the same c5 error:

[----------] 1 test from GLImageIOSurface/GLImageBindTest/4, where TypeParam = <type>
[ RUN      ] GLImageIOSurface/GLImageBindTest/4.BindTexImage
../../ui/gl/test/gl_test_helper.cc:77: Failure
The difference between expected and actual is 32, which exceeds error, where
expected evaluates to 48,
actual evaluates to 16, and
error evaluates to 0.
 at 0, 0 channel 0
../../ui/gl/test/gl_test_helper.cc:77: Failure
The difference between expected and actual is 32, which exceeds error, where
expected evaluates to 16,
actual evaluates to 48, and
error evaluates to 0.
 at 0, 0 channel 2

so I made a CL with the swizzling and activated the bot (crrev.com/c/886543),
and o_0 mac_optional_gpu_tests_rel passes _without_ the swizzling. 

So what's the difference, if any, between the gpu.fyi bot [1] and the
optional trybot [2] ?

[1] https://chromium-swarm.appspot.com/task?id=3b42735cf3597110&refresh=10&request_detail=true&show_raw=1&wide_logs=true
[2] https://chromium-swarm.appspot.com/task?id=3b474365c0e24f10&refresh=10&request_detail=true&show_raw=1&wide_logs=true

There is no difference – or there should be no difference, anyway. The physical hardware is in the Swarming pool, and jobs run on whatever machine is free.

https://chromium-swarm.appspot.com/task?id=3b474365c0e24f10&refresh=10&show_raw=1

Where are the logging statements that were added? Can you change the command line arguments for this test step (in generate_buildbot_json.py) so that you can see which of your code paths are being taken?

#21: When the test pass, they don't show the verbose log, but in PS 5
https://chromium-review.googlesource.com/c/chromium/src/+/886543/5
we can see a failing build:
https://ci.chromium.org/buildbot/tryserver.chromium.mac/mac_optional_gpu_tests_rel/20555
and a failing log:
https://chromium-swarm.appspot.com/task?id=3b473d0d5cb40610&refresh=10&show_raw=1&wide_logs=true

which says |is_nvidia| is true :-/

Can you please explain the situation to me clearly? Are you seeing some of your tryjobs on mac_optional_gpu_tests_rel fail and some pass?

If so then can you please point to two tryjobs on mac_optional_gpu_tests_rel that have different behavior? That would indicate that the hardware isn't set up completely identically.

What we know:

The CL that landed in #c17, which had a B R channel swizzle, was reverted 
in #c18, because it failed with the swarming task page:
https://chromium-swarm.appspot.com/task?id=3b42735cf3597110&refresh=10&request_detail=true&show_raw=1&wide_logs=true
-- which shows a failure explained in #c20.

I tried to put up a new CL, enabling the mac_optional_gpu_tests_rel bot 
(crrev.com/c/886543). In this CL:
- PS5 has a channel swizzle, 
https://chromium-review.googlesource.com/c/chromium/src/+/886543/5
 and gl_unittests fail:
https://chromium-swarm.appspot.com/task?id=3b473d0d5cb40610&refresh=10&show_raw=1&wide_logs=true

- PS6 has no channel swizzle, (color components are just copied thru)
https://chromium-review.googlesource.com/c/chromium/src/+/886543/6
and the said gl_unittests pass:
https://chromium-swarm.appspot.com/task?id=3b474365c0e24f10&refresh=10&show_raw=1

The situation is 100% reproducible in a MBP + nVidia I borrowed here.

It sounds then like on Mac/NVIDIA you shouldn't try to do the swizzle. I don't know what driver bugs, etc. you might be facing or what the behavior is on other GPUs on macOS, or on other platforms.

If the test passes on mac_optional_gpu_tests_rel, it will pass on the waterfall. If there is *any* discrepancy in behavior, that is a bug. Please file it and I will investigate it.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/62633a07473a374967e85495bbcd30042f3a4d08

commit 62633a07473a374967e85495bbcd30042f3a4d08
Author: Miguel Casas-Sanchez <mcasas@chromium.org>
Date: Sat Jan 27 02:47:15 2018

Mac gl_unittests: enable GLImageBindTest BGRX_1010102

This CL adds gl_unittests to chromium.gpu.fyi, hence making it
run, among others, on the Mac nVidia bot that caused the original
CL to be reverted.

There's more information on the bug but, in a nutshell, the
original CL was reverted due to a specific nVidia bug that
manifested itself on the chromium.gpu.fyi mac bot; however,
it doesn't show up in the mac_optional_gpu_rel. So, this CL
has no specific nVidia provisions, just reenables the test,
and we'll see what happens in the chromium.gpu.fyi waterfall :-)


Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I71c0451d2e3b53baa3bd24bafdf1f7a5f1210363
Reviewed-on: https://chromium-review.googlesource.com/886543
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#532136}
[modify] https://crrev.com/62633a07473a374967e85495bbcd30042f3a4d08/content/test/gpu/generate_buildbot_json.py
[modify] https://crrev.com/62633a07473a374967e85495bbcd30042f3a4d08/testing/buildbot/chromium.gpu.fyi.json
[modify] https://crrev.com/62633a07473a374967e85495bbcd30042f3a4d08/ui/gl/gl_image_io_surface_unittest.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8c1f8bc8f96d1612f2f799bdf7b47a53c0421eb1

commit 8c1f8bc8f96d1612f2f799bdf7b47a53c0421eb1
Author: Miguel Casas <mcasas@chromium.org>
Date: Sun Jan 28 00:50:19 2018

Revert "Mac gl_unittests: enable GLImageBindTest BGRX_1010102"

This reverts commit 62633a07473a374967e85495bbcd30042f3a4d08.

Reason for revert: 
Broke 
 chromium.gpu.fyi
  Mac Experimental Retina Release (NVIDIA) 

https://ci.chromium.org/buildbot/chromium.gpu.fyi/Mac%20Experimental%20Retina%20Release%20%28NVIDIA%29/3232

Original change's description:
> Mac gl_unittests: enable GLImageBindTest BGRX_1010102
> 
> This CL adds gl_unittests to chromium.gpu.fyi, hence making it
> run, among others, on the Mac nVidia bot that caused the original
> CL to be reverted.
> 
> There's more information on the bug but, in a nutshell, the
> original CL was reverted due to a specific nVidia bug that
> manifested itself on the chromium.gpu.fyi mac bot; however,
> it doesn't show up in the mac_optional_gpu_rel. So, this CL
> has no specific nVidia provisions, just reenables the test,
> and we'll see what happens in the chromium.gpu.fyi waterfall :-)
> 
> 
> Bug:  803473 
> Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
> Change-Id: I71c0451d2e3b53baa3bd24bafdf1f7a5f1210363
> Reviewed-on: https://chromium-review.googlesource.com/886543
> Reviewed-by: Kenneth Russell <kbr@chromium.org>
> Commit-Queue: Miguel Casas <mcasas@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#532136}

TBR=kbr@chromium.org,mcasas@chromium.org

Change-Id: I5988cab9dcb3fe2da1f22750ec76c1ffb0cf4555
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Reviewed-on: https://chromium-review.googlesource.com/889771
Reviewed-by: Miguel Casas <mcasas@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#532257}
[modify] https://crrev.com/8c1f8bc8f96d1612f2f799bdf7b47a53c0421eb1/content/test/gpu/generate_buildbot_json.py
[modify] https://crrev.com/8c1f8bc8f96d1612f2f799bdf7b47a53c0421eb1/testing/buildbot/chromium.gpu.fyi.json
[modify] https://crrev.com/8c1f8bc8f96d1612f2f799bdf7b47a53c0421eb1/ui/gl/gl_image_io_surface_unittest.cc

Ah, I'm sorry. I just realized the difference between this bot:

https://ci.chromium.org/buildbot/chromium.gpu.fyi/Mac%20Experimental%20Retina%20Release%20%28NVIDIA%29/

and this bot:

https://ci.chromium.org/buildbot/chromium.gpu.fyi/Mac%20Retina%20Release%20%28NVIDIA%29/

(mac_optional_gpu_tests_rel mimics the configuration of the latter)

is that the Mac Experimental Release (NVIDIA) bot is running 10.13. The other bots are running 10.12.

There must be a behavioral difference in 10.13. I'm sorry we don't have trybots for that configuration yet; can you make progress? Locally, we have set up an external Thunderbolt drive with 10.13 stable and beta releases so that we can boot multiple machines into 10.12, 10.13, and 10.13 betas.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f237d4c82eb90f438e51331dc896b95fa4cadd29

commit f237d4c82eb90f438e51331dc896b95fa4cadd29
Author: Miguel Casas-Sanchez <mcasas@chromium.org>
Date: Wed Jan 31 17:35:23 2018

Mac gl_unittests: enable GLImageBindTest BGRX_1010102 (2nd go)

This CL relands crrev.com/c/886543, enabling the GLImageBindTest for
GLImageIOSUrface with the BufferFormat BGRX_1010102.

The CL also adds a provision to swizzle the color channels when running
on nVidia on High Sierra to workaround some bug that can be repro'ed on
the bots and stand alone laptops.

Bug:  803473 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I5472606a82b393fe1d2375b3c644e836ea7f12cf
Reviewed-on: https://chromium-review.googlesource.com/890007
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Miguel Casas <mcasas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#533313}
[modify] https://crrev.com/f237d4c82eb90f438e51331dc896b95fa4cadd29/ui/gl/gl_image_io_surface_unittest.cc

Seems like the bot is happy now:
https://chromium-swarm.appspot.com/task?id=3b668d04457a0710&refresh=0&show_raw=1&wide_logs=true

[47/51] GLImageIOSurface/GLImageBindTest/0.BindTexImage (34 ms)
[48/51] GLImageIOSurface/GLImageBindTest/1.BindTexImage (27 ms)
[49/51] GLImageIOSurface/GLImageBindTest/2.BindTexImage (13 ms)
[50/51] GLImageIOSurface/GLImageBindTest/3.BindTexImage (38 ms)
[51/51] GLImageIOSurface/GLImageBindTest/4.BindTexImage (15 ms)

Great, Miguel. Sorry for my confusion about the configurations earlier and glad you have everything working.

#31: no worries, let's keep the waterfalls green !