Ask driver for driver specific extension

Ask driver for driver specific extension

EGL_KHR_fence_sync is an EGL only extension that is not returned by the
cross-platform glGetString(GL_EXTENSIONS).

Add a workaround for qualcomm devices before 4.4, which had a bug with
the EGL_KHR_fence_sync extension.

BUG=375365, 278606, 371530
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=272261

[no sievers, 2014-05-21 22:24:53]

Actually I think we need to figure out now if EGL_KHR_fence_sync fence testing
works on <4.4, for example JB MR2.

For async readpixels it does not matter, because there should be no PBO support
in ES2, so we wouldn't hit that (I guess that's why I removed the workaround in
https://codereview.chromium.org/285523002/ regardless of Android version).

But for the sync query stuff the question is whether the fences actually ever
clear on the Qualcomm ES2 drivers.

[boliu, 2014-05-21 23:39:27]

added a workaround for qualcomm < 4.4 as discussed offline

[no sievers, 2014-05-21 23:45:18]

lgtm.

So although the bug is not fixed in all 4.4 Qualcomm drivers out there, because
we now prefer ES3 fences we are good (assuming all Qualcomm 4.4 drivers are ES3
which should be the case).

[boliu, 2014-05-22 00:21:59]

The CQ bit was checked by boliu@chromium.org

[commit-bot: I haz the power, 2014-05-22 00:25:34]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/boliu@chromium.org/292153009/20001

[boliu, 2014-05-22 04:12:02]

Failing compile on windows due to C2026:
http://msdn.microsoft.com/en-us/library/dddywwsc.aspx

The json string is longer than 16380 characters, and my c++ foo is not good
enough to figure out how LONG_STRING_CONST actually works to fix this...trying
random things now

[mkisilev, 2014-05-22 09:14:32]

On 2014/05/22 04:12:02, boliu wrote:
> Failing compile on windows due to C2026:
> http://msdn.microsoft.com/en-us/library/dddywwsc.aspx
> 
> The json string is longer than 16380 characters, and my c++ foo is not good
> enough to figure out how LONG_STRING_CONST actually works to fix this...trying
> random things now

split json onto two strings by inserting:

)
LONG_STRING_CONST(

for example:

{
  "id": 73,
  "description": "Using D3D11 causes browser crashes on certain Intel GPUs",
  "cr_bugs": [310808],
  "os": {
	"type": "win"
  },
  "vendor_id": "0x8086",
  "features": [
	"disable_d3d11"
  ]
},
)
LONG_STRING_CONST(
{
  "id": 74,
  "cr_bugs": [278606],
  "description": "Testing EGL sync fences is broken on Qualcomm before Android
4.4",

[boliu, 2014-05-22 14:09:04]

On 2014/05/22 09:14:32, mkisilev wrote:
> On 2014/05/22 04:12:02, boliu wrote:
> > Failing compile on windows due to C2026:
> > http://msdn.microsoft.com/en-us/library/dddywwsc.aspx
> > 
> > The json string is longer than 16380 characters, and my c++ foo is not good
> > enough to figure out how LONG_STRING_CONST actually works to fix
this...trying
> > random things now
> 
> split json onto two strings by inserting:
> 
> )
> LONG_STRING_CONST(

Thanks! That doe seem to work. Just had to be careful where I insert the comment
to document this...preprocessor is hard :)

[boliu, 2014-05-22 15:41:05]

sievers, you probably should look over this again

[no sievers, 2014-05-22 17:18:44]

+Mo to advise with the long string const problem

[Zhenyao Mo, 2014-05-22 18:30:00]

On 2014/05/22 17:18:44, sievers wrote:
> +Mo to advise with the long string const problem

the long string const solution looks correct.

[boliu, 2014-05-22 18:35:19]

The CQ bit was checked by boliu@chromium.org

[boliu, 2014-05-22 18:38:23]

On 2014/05/22 18:30:00, Zhenyao Mo wrote:
> On 2014/05/22 17:18:44, sievers wrote:
> > +Mo to advise with the long string const problem
> 
> the long string const solution looks correct.

Thanks. cq-ing

[commit-bot: I haz the power, 2014-05-22 18:40:26]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/boliu@chromium.org/292153009/100001

[commit-bot: I haz the power, 2014-05-22 19:09:09]

Change committed as 272261