904620 - WebGL hangs in SurfaceStream_TripleBuffer_Async::WaitForCompositor

Status: VERIFIED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

[Benoit Girard (:BenWa)]

STR:
1) Open: http://glecollinet.github.io/webgl-demo/

I get a hang with nightly on 10.8

Comment 1

[Guillaume Abadie]

Same 10.9, bisecting...

Comment 2

[Guillaume Abadie]

I'm having trouble to compile on 10.9, but here is the window I already have by the ftp:
 - ok:  ff0a372e3170 (2013-07-15-03-02-02)
 - ok:  c2b375f3a909 (2013-07-31-03-02-03)
 - ok:  76a944fa6b25 (2013-08-02-03-02-04)
 - bad: d0edf8086809 (2013-08-03-03-02-05)
 - bad: 0a63cd911b4f (2013-08-04-03-02-07)
 - bad: 1fb5d14e8348 (2013-08-07-03-02-16)

Comment 3

[mbtyke]

This appeared for me for the first time in Firefox Nightly 8-3-2013.

http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2013-08-03-03-02-05-mozilla-central/

I'm running on a Mac 10.8.

This is blocking partner contract work for Firefox OS.

Comment 4

[mbtyke]

Attachment: This is a sample on Mac OS X taken from Activity Monitor

I took a sample with Activity Monitor on Mac OS X to add more information about the hang.

Comment 5

[Kelsey Gilbert [:jgilbert]]

This does not appear to affect windows or linux.

Comment 7

[Kelsey Gilbert [:jgilbert]]

Odds are high that this is:
904611

Comment 8

[Kelsey Gilbert [:jgilbert]]

I suspect bug 900188. Mattwoodrow?
It looks like this patch makes it so we don't always SwapConsumer anymore, which will hang SwapProducer if SP is called twice before SC is called.

Comment 9

[Matt Woodrow (:mattwoodrow)]

Attachment: Always send the SurfaceStream across

Looks like we just cancel the transaction if it's empty, and never schedule a composite.

We could add an explicit 'composite using the existing surface' message or type, but I don't think it's worth it.

Comment 10

[Matt Woodrow (:mattwoodrow)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/de84e2d0c9c5

Comment 11

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/de84e2d0c9c5

Comment 12

[Chris Peterson [:cpeterson]]

btw, this bug was a regression from bug 900188 in Nightly 25 build 2013-08-03.

Comment 13

[Alex Keybl [:akeybl]]

Ready for uplift Matt?

Comment 14

[Matt Woodrow (:mattwoodrow)]

Comment on attachment 790987 [details] [diff] [review]
Always send the SurfaceStream across

[Approval Request Comment]
Bug caused by (feature/regressing bug #): Bug 900188
User impact if declined: Hangs with WebGL content
Testing completed (on m-c, etc.): Been on m-c for weeks.
Risk to taking this patch (and alternatives if risky): Very low risk, just disables a (broken) optimization.
String or IDL/UUID changes made by this patch: None

Comment 15

[Lukas Blakk [:lsblakk] use ?needinfo]

Comment on attachment 790987 [details] [diff] [review]
Always send the SurfaceStream across

[Triage Comment]
itym beta nomination since this is already marked fixed on 26 (Aurora).

Comment 16

[Ryan VanderMeulen (PTO, back 6-April)]

https://hg.mozilla.org/releases/mozilla-beta/rev/a111e4aa24ad

Comment 17

[Ryan VanderMeulen (PTO, back 6-April)]

Backed out for Android reftest-2 crashes.
https://hg.mozilla.org/releases/mozilla-beta/rev/6f8b48123dc4

https://tbpl.mozilla.org/php/getParsedLog.php?id=28351578&tree=Mozilla-Beta
https://tbpl.mozilla.org/php/getParsedLog.php?id=28350963&tree=Mozilla-Beta

Comment 18

[Matt Woodrow (:mattwoodrow)]

I think we need bug 912173 to be uplifted with it.

Comment 20

[Simona Badau, Desktop QA]

Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:25.0) Gecko/20100101 Firefox/25.0
Build ID: 20130926170421

The situation looks pretty bad on Mac OS X and Firefox 25 beta 3:
https://moztrap.mozilla.org/results/cases/?filter-run=2331

Will this get uplifted to Firefox 25?

Comment 21

[u279076]

(In reply to Simona B [QA] from comment #20)
> Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:25.0) Gecko/20100101
> Firefox/25.0
> Build ID: 20130926170421
> 
> The situation looks pretty bad on Mac OS X and Firefox 25 beta 3:
> https://moztrap.mozilla.org/results/cases/?filter-run=2331
> 
> Will this get uplifted to Firefox 25?

Still waiting on bug 912173 to be uplifted which was just requested yesterday.

Comment 22

[Alex Keybl [:akeybl]]

(In reply to Anthony Hughes, Mozilla QA (:ashughes) from comment #21)
> (In reply to Simona B [QA] from comment #20)
> > Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:25.0) Gecko/20100101
> > Firefox/25.0
> > Build ID: 20130926170421
> > 
> > The situation looks pretty bad on Mac OS X and Firefox 25 beta 3:
> > https://moztrap.mozilla.org/results/cases/?filter-run=2331
> > 
> > Will this get uplifted to Firefox 25?
> 
> Still waiting on bug 912173 to be uplifted which was just requested
> yesterday.

That patch is now approved for Beta 25.

Comment 23

[Ryan VanderMeulen (PTO, back 6-April)]

https://hg.mozilla.org/releases/mozilla-beta/rev/f72d3e3dc13c

Comment 24

[Bogdan Maris, Desktop Test Engineering]

Mozilla/5.0 (Macintosh; Intel Mac OS X 10.8; rv:25.0) Gecko/20100101 Firefox/25.0

Looks good now on Firefox 25 beta 4 (buildID: 20131001024718) and latest Aurora (buildID: 20131002004002) after running the tests from the testrun wrote in comment 20.