Use the desktop GL core profile on Linux when ES3 support is requested.

Use the desktop GL core profile on Linux when ES3 support is requested.

Request an OpenGL 4.2 context (the minimum version Chromium's WebGL
2.0 prototype currently requires) on Linux when
--enable-unsafe-es3-apis is specified on the command line.

Clean up the shader translator's selection of the GLSL output format.

BUG=471960
Committed: https://crrev.com/0b17446f9591af9377d18c9ccfe50471651d6bc7
Cr-Commit-Position: refs/heads/master@{#323658}

[Ken Russell (switch to Gerrit), 2015-04-02 04:48:47]

kbr@chromium.org changed reviewers:
+ piman@chromium.org

[Ken Russell (switch to Gerrit), 2015-04-02 04:48:48]

Please review. Thanks.

Tested with WebGL 2.0 content and gpu_unittests.

[piman, 2015-04-02 17:35:16]

On 2015/04/02 04:48:48, Ken Russell wrote:
> Please review. Thanks.
> 
> Tested with WebGL 2.0 content and gpu_unittests.

Why 4.2?
Many NVIDIA GPUs only support 3.3 - and that should be enough, no?

[Ken Russell (switch to Gerrit), 2015-04-02 17:41:00]

On 2015/04/02 17:35:16, piman (Very slow to review) wrote:
> On 2015/04/02 04:48:48, Ken Russell wrote:
> > Please review. Thanks.
> > 
> > Tested with WebGL 2.0 content and gpu_unittests.
> 
> Why 4.2?
> Many NVIDIA GPUs only support 3.3 - and that should be enough, no?

FeatureInfo::IsES3Capable currently requires desktop GL 4.2 or greater in order
to enable the ES3 prototype, since it's only that version that contains all of
the needed extensions in the core. We'll certainly revisit this later. MANGLE
will do a better job of supporting ES3 on earlier versions of desktop GL.

https://code.google.com/p/chromium/codesearch#chromium/src/gpu/command_buffer...

[piman, 2015-04-02 19:24:35]

On 2015/04/02 17:41:00, Ken Russell wrote:
> On 2015/04/02 17:35:16, piman (Very slow to review) wrote:
> > On 2015/04/02 04:48:48, Ken Russell wrote:
> > > Please review. Thanks.
> > > 
> > > Tested with WebGL 2.0 content and gpu_unittests.
> > 
> > Why 4.2?
> > Many NVIDIA GPUs only support 3.3 - and that should be enough, no?
> 
> FeatureInfo::IsES3Capable currently requires desktop GL 4.2 or greater in
order
> to enable the ES3 prototype, since it's only that version that contains all of
> the needed extensions in the core. We'll certainly revisit this later. MANGLE
> will do a better job of supporting ES3 on earlier versions of desktop GL.
> 
>
https://code.google.com/p/chromium/codesearch#chromium/src/gpu/command_buffer...

But passing the --enable-unsafe-es3-apis flag will disable the GPU altogether if
you're on a pre-Fermi gpu (context creation will fail). That seems counter
intuitive.

If the plan is to move everything to the core profile, requesting 4.2 is not
acceptable.

[Ken Russell (switch to Gerrit), 2015-04-02 19:39:20]

On 2015/04/02 19:24:35, piman (Very slow to review) wrote:
> On 2015/04/02 17:41:00, Ken Russell wrote:
> > On 2015/04/02 17:35:16, piman (Very slow to review) wrote:
> > > On 2015/04/02 04:48:48, Ken Russell wrote:
> > > > Please review. Thanks.
> > > > 
> > > > Tested with WebGL 2.0 content and gpu_unittests.
> > > 
> > > Why 4.2?
> > > Many NVIDIA GPUs only support 3.3 - and that should be enough, no?
> > 
> > FeatureInfo::IsES3Capable currently requires desktop GL 4.2 or greater in
> order
> > to enable the ES3 prototype, since it's only that version that contains all
of
> > the needed extensions in the core. We'll certainly revisit this later.
MANGLE
> > will do a better job of supporting ES3 on earlier versions of desktop GL.
> > 
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/gpu/command_buffer...
> 
> But passing the --enable-unsafe-es3-apis flag will disable the GPU altogether
if
> you're on a pre-Fermi gpu (context creation will fail). That seems counter
> intuitive.
> 
> If the plan is to move everything to the core profile, requesting 4.2 is not
> acceptable.

Yes, I understand and agree with that, and will address the TODO I added in
gl_context_glx. I don't yet know how to request a core profile context that
auto-upgrades to the latest available version. I need to do more testing on
various platforms. Landing this patch will clean up the issue I had left over
from the previous one and will enable easier testing on more drivers.

[piman, 2015-04-02 20:08:00]

On 2015/04/02 19:39:20, Ken Russell wrote:
> On 2015/04/02 19:24:35, piman (Very slow to review) wrote:
> > On 2015/04/02 17:41:00, Ken Russell wrote:
> > > On 2015/04/02 17:35:16, piman (Very slow to review) wrote:
> > > > On 2015/04/02 04:48:48, Ken Russell wrote:
> > > > > Please review. Thanks.
> > > > > 
> > > > > Tested with WebGL 2.0 content and gpu_unittests.
> > > > 
> > > > Why 4.2?
> > > > Many NVIDIA GPUs only support 3.3 - and that should be enough, no?
> > > 
> > > FeatureInfo::IsES3Capable currently requires desktop GL 4.2 or greater in
> > order
> > > to enable the ES3 prototype, since it's only that version that contains
all
> of
> > > the needed extensions in the core. We'll certainly revisit this later.
> MANGLE
> > > will do a better job of supporting ES3 on earlier versions of desktop GL.
> > > 
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/gpu/command_buffer...
> > 
> > But passing the --enable-unsafe-es3-apis flag will disable the GPU
altogether
> if
> > you're on a pre-Fermi gpu (context creation will fail). That seems counter
> > intuitive.
> > 
> > If the plan is to move everything to the core profile, requesting 4.2 is not
> > acceptable.
> 
> Yes, I understand and agree with that, and will address the TODO I added in
> gl_context_glx. I don't yet know how to request a core profile context that
> auto-upgrades to the latest available version.

We could try 4.2 first and if that fails, try again requesting 3.2.

> I need to do more testing on
> various platforms. Landing this patch will clean up the issue I had left over
> from the previous one and will enable easier testing on more drivers.

Ok, LGTM if you add a bug.

[Ken Russell (switch to Gerrit), 2015-04-02 22:38:00]

On 2015/04/02 20:08:00, piman (Very slow to review) wrote:
> On 2015/04/02 19:39:20, Ken Russell wrote:
> > On 2015/04/02 19:24:35, piman (Very slow to review) wrote:
> > > On 2015/04/02 17:41:00, Ken Russell wrote:
> > > > On 2015/04/02 17:35:16, piman (Very slow to review) wrote:
> > > > > On 2015/04/02 04:48:48, Ken Russell wrote:
> > > > > > Please review. Thanks.
> > > > > > 
> > > > > > Tested with WebGL 2.0 content and gpu_unittests.
> > > > > 
> > > > > Why 4.2?
> > > > > Many NVIDIA GPUs only support 3.3 - and that should be enough, no?
> > > > 
> > > > FeatureInfo::IsES3Capable currently requires desktop GL 4.2 or greater
in
> > > order
> > > > to enable the ES3 prototype, since it's only that version that contains
> all
> > of
> > > > the needed extensions in the core. We'll certainly revisit this later.
> > MANGLE
> > > > will do a better job of supporting ES3 on earlier versions of desktop
GL.
> > > > 
> > > >
> > >
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/gpu/command_buffer...
> > > 
> > > But passing the --enable-unsafe-es3-apis flag will disable the GPU
> altogether
> > if
> > > you're on a pre-Fermi gpu (context creation will fail). That seems counter
> > > intuitive.
> > > 
> > > If the plan is to move everything to the core profile, requesting 4.2 is
not
> > > acceptable.
> > 
> > Yes, I understand and agree with that, and will address the TODO I added in
> > gl_context_glx. I don't yet know how to request a core profile context that
> > auto-upgrades to the latest available version.
> 
> We could try 4.2 first and if that fails, try again requesting 3.2.
> 
> > I need to do more testing on
> > various platforms. Landing this patch will clean up the issue I had left
over
> > from the previous one and will enable easier testing on more drivers.
> 
> Ok, LGTM if you add a bug.

Done: crbug.com/473427

[Ken Russell (switch to Gerrit), 2015-04-02 22:38:04]

The CQ bit was checked by kbr@chromium.org

[Ken Russell (switch to Gerrit), 2015-04-02 22:38:04]

The patchset sent to the CQ was uploaded after l-g-t-m from piman@chromium.org
Link to the patchset: https://codereview.chromium.org/1052153002/#ps40001
(title: "Added link to bug in TODO.")

[commit-bot: I haz the power, 2015-04-03 00:54:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1052153002/40001

[commit-bot: I haz the power, 2015-04-03 07:43:39]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-04-03 20:33:54]

Patchset 3 (id:??) landed as
https://crrev.com/0b17446f9591af9377d18c9ccfe50471651d6bc7
Cr-Commit-Position: refs/heads/master@{#323658}

[marcheu, 2015-04-04 01:41:17]

A revert of this CL (patchset #3 id:40001) has been created in
https://codereview.chromium.org/1054883004/ by marcheu@chromium.org.

The reason for reverting is: This causes Chrome rendering to break on some
Chrome OS devices, see crbug.com/473891 for details
.