807741 - Blacklist WebGL antialiasing on Mac on NVIDIA and Intel driver

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Benoit Jacob [:bjacob] (mostly away)]

Thanks to our friends at Google for reporting this and making a testcase.

http://code.google.com/p/chromium/issues/detail?id=137303

Testcase:
http://www.khronos.org/registry/webgl/sdk/tests/conformance/rendering/multisample-corruption.html

There is no alternative to blacklisting MSAA on Mac/NVIDIA until this is fixed.

Apple Radar 12557679: Corruption of OpenGL Multisample Renderbuffers with NVIDIA chipsets

Comment 1

[Benoit Jacob [:bjacob] (mostly away)]

Attachment: patch

Doing this with a compiled blacklist as it's tedious to list all Mac OS versions in separate XML entries. Also with 17 we get an opportunity to deliver this to users rather soon.

Comment 2

[Joe Drew (not getting mail)]

Comment on attachment 677498 [details] [diff] [review]
patch

Review of attachment 677498 [details] [diff] [review]:
-----------------------------------------------------------------

::: widget/cocoa/GfxInfo.mm
@@ +328,5 @@
>        (nsAString&) GfxDriverInfo::GetDeviceVendor(VendorATI), GfxDriverInfo::allDevices,
> +      nsIGfxInfo::FEATURE_WEBGL_MSAA, nsIGfxInfo::FEATURE_BLOCKED_DEVICE);
> +    IMPLEMENT_MAC_DRIVER_BLOCKLIST(DRIVER_OS_ALL,
> +      (nsAString&) GfxDriverInfo::GetDeviceVendor(VendorNVIDIA), GfxDriverInfo::allDevices,
> +      nsIGfxInfo::FEATURE_WEBGL_MSAA, nsIGfxInfo::FEATURE_BLOCKED_DEVICE);

Add a comment referring to this bug.

Comment 3

[Benoit Jacob [:bjacob] (mostly away)]

Now that sucks: Google has disclosed this bug publicly on public-webgl and also the Chromium bug seems publicly accessible: http://code.google.com/p/chromium/issues/detail?id=137303

Can this wait for Firefox 17 or do we want to deploy this on the downloadable blacklist?

Comment 4

[Benoit Jacob [:bjacob] (mostly away)]

Also, is there still a point in hiding this bug?

Comment 5

[Benoit Jacob [:bjacob] (mostly away)]

The Chromium bug seems to have morphed into NVIDIA + Intel.

Joe: can you prepare the downloadable blacklist updates or teach me how to do that?

Comment 6

[Joe Drew (not getting mail)]

Getting a downloadable blacklist entry added is very easy. Simply file a bug like bug 806991, in that component, and specify what sections of the XML you want, and what you want them set to (literal strings). In this case it'll be all affected versions of OS X, so there will be several entries necessary. (eg Darwin 9, Darwin 10, Darwin 11, Darwin 12)

Comment 7

[Andrew McCreight [:mccr8]]

Thanks for rating it, but the rating is a keyword not a whiteboard entry. :)

Comment 8

[Benoit Jacob [:bjacob] (mostly away)]

Thanks Joe for the great documentation and ready-to-use bug request link. Filed bug 809550.

Comment 9

[Alex Keybl [:akeybl]]

Is the patch here the right path forward for ESR10 (given bug 809550, and our inability to remote block in FF10)? We're going to build after Wednesday, so we'd need this to land asap.

Comment 10

[Benoit Jacob [:bjacob] (mostly away)]

Attachment: block MSAA on Mac for ESR10

No, but _this_ patch is it. On try:
https://tbpl.mozilla.org/?tree=Try&rev=cee5ddc245fb

Comment 11

[Benoit Jacob [:bjacob] (mostly away)]

Given the timeframe, please land this patch when you approve it.

Try looks red, but it's unrelated. I seem to remember that pushing ESR10 to Try doesn't work anymore anyways.

Comment 12

[Alex Keybl [:akeybl]]

(In reply to Benoit Jacob [:bjacob] from comment #11)
> Given the timeframe, please land this patch when you approve it.

You've got all of today and tomorrow to land - we don't land on behalf of engineers because we don't have the expertise to investigate/resolve tree bustage.

Comment 13

[Benoit Jacob [:bjacob] (mostly away)]

OK! I misunderstood, I thought that this had to land yesterday.

https://hg.mozilla.org/releases/mozilla-esr10/rev/d5505b3df3c3

Comment 14

[Benoit Jacob [:bjacob] (mostly away)]

Also, this bug is FIXED by 809550.