1268096 - WebGL point sprites malfunctions in Firefox 46 on Windows D3D11 ANGLE

Status: RESOLVED FIXED

(Core :: Graphics: CanvasWebGL, defect)

Description

[AlteredQualia]

User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:46.0) Gecko/20100101 Firefox/46.0
Build ID: 20160421124000

Steps to reproduce:

Visit these pages:

http://alteredqualia.com/xg/examples/skylon.html
http://alteredqualia.com/xg/examples/deferred_particles_nebula.html


Actual results:

There are missing particle effects: stream of particles coming out of spacecraft, also particles nebula in second demo.


Expected results:

There should be particles.

---------

This is likely due to ANGLE regress. 

The same thing happened in Chrome few months ago (January-February 2016) - broken gl.POINTS primitives on Windows due to ANGLE changes. 

Chrome already got it fixed, but seems Firefox still uses older ANGLE version that still had this regress (both stable Firefox 46 and the most recent Firefox Nightly 49 have broken particles).

If this is indeed the case, fix should be relatively simple - use a newer ANGLE version.

See this Chrome issue for more details: https://bugs.chromium.org/p/chromium/issues/detail?id=586531

Comment 1

[Kohei Yoshino]

[Tracking Requested - why for this release]: A site-compat regression from Firefox 46, initially reported on Twitter: https://twitter.com/alteredq/status/725076739911737345

Comment 2

[Alice0775 White]

Regression window:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=1101f5bec98449a2c033c0f8df80b2d8885870d5&tochange=f982d05e0cde64865fe11e1e0f4c6799497ff437

Regressed by: 
	f982d05e0cde	Jeff Muizelaar — Bug 1232902. Update ANGLE to chromium/2592

Comment 3

[Marcia Knous [:marcia]]

Tracking for 46, 47, 48 and 49.

Comment 4

[Liz Henry (:lizzard) (relman/hg->git project)]

Jeff, how bad is this for users? If we do a dot release in the next week, and you have a fix and it is testable and not too risky feeling, it could still get into 46.

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/769357ba097f

Comment 6

[Jeff Muizelaar [:jrmuizel]]

Attachment: Port upstream change for point sprites.

Approval Request Comment
[Feature/regressing bug #]: Bug 1232902
[User impact if declined]: Some WebGL content will be broken.
[Describe test coverage new/current, TreeHerder]: Tested by upstream and landed in Chrome.
[Risks and why]: Should be pretty low.

Comment 7

[Kelsey Gilbert [:jgilbert]]

Thanks, jrmuizel.

Comment 8

[Kohei Yoshino]

Posted the site compatibility doc: https://www.fxsitecompat.com/en-CA/docs/2016/webgl-point-sprites-are-not-rendered-properly-in-certain-environments/

Comment 10

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/769357ba097f

Comment 11

[Liz Henry (:lizzard) (relman/hg->git project)]

I'm leaning towards letting this ride with 47 so we have more time for testing or finding any resulting issues. (rather than shipping as a ridealong with 46.0.1)

Comment 12

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Hello AlteredQualia, could you please verify this issue is fixed as expected on a latest Nightly build? Thanks!

Comment 13

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Comment on attachment 8746617 [details] [diff] [review]
Port upstream change for point sprites.

Recent regression, Aurora48+, Beta47+

Comment 14

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/cb7afc6a6450

Comment 15

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-beta/rev/199b4fd2bf1e10d423176dea7d6b25fb1d4c2072

Comment 16

[AlteredQualia]

(In reply to Ritu Kothari (:ritu) from comment #12)
> Hello AlteredQualia, could you please verify this issue is fixed as expected
> on a latest Nightly build? Thanks!

Yup, it now works ok here with the latest Nightly 49.0a1 (2016-05-02).

Thank you.

Comment 17

[Marcia Knous [:marcia]]

Wes - I want to mark this as fixed in 49, but just checking to make sure it landed. 

The bug is also tagged leave-open - jgilbert do you still want to leave it open?

Comment 18

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

It only didn't get marked resolved for 49 because of the leave-open. If that goes away, this can be resolved. So I'll defer to jgilbert. :)

Comment 19

[jazzzz]

Is there a workaround?
It breaks the markers on our map, it gives a really bad experience for Firefox users on our website, e.g.

https://www.bienici.com/recherche/achat/france?camera=18_-1.5113552_48.6358157

There should be POIs on the map, none is displayed.

Comment 20

[Alice0775 White]

Please use Google Chrome.

Comment 21

[Kelsey Gilbert [:jgilbert]]

(In reply to jazzzz from comment #19)
> Is there a workaround?
> It breaks the markers on our map, it gives a really bad experience for
> Firefox users on our website, e.g.
> 
> https://www.bienici.com/recherche/achat/france?camera=18_-1.5113552_48.
> 6358157
> 
> There should be POIs on the map, none is displayed.

Draw your own small quads.
FWIW, the way you use point-sprites is technically non-portable, since max point size can be 1.0.

Also, you should always test on Nightly/Dev Edition/Beta before the version becomes Release. Once a bug hits Release, it's generally going to be there for six weeks unless it's a major widespread problem, as stability of our Release channel is paramount.

Comment 22

[Kelsey Gilbert [:jgilbert]]

FWIW, under the hood, ANGLE is converting your point sprites to quads, so doing it yourself will generally be more efficient.

Also, if you use a library, file a bug with that library.

Comment 23

[Kelsey Gilbert [:jgilbert]]

I'll add a regression test for this.

Comment 24

[jazzzz]

(In reply to Jeff Gilbert [:jgilbert] from comment #22)
> FWIW, under the hood, ANGLE is converting your point sprites to quads, so
> doing it yourself will generally be more efficient.
> 
> Also, if you use a library, file a bug with that library.

Thanks, we'll convert our point sprites to quads.

Comment 25

[Kelsey Gilbert [:jgilbert]]

Attachment: 0001-Bug-1268096-r-jrmuizel-Add-regression-test.patch

Add a regression test.

Comment 26

[Kelsey Gilbert [:jgilbert]]

I'll take this to drive through the test.

Comment 27

[Kelsey Gilbert [:jgilbert]]

Comment on attachment 8754123 [details] [diff] [review]
0001-Bug-1268096-r-jrmuizel-Add-regression-test.patch

Approval Request Comment
[Feature/regressing bug #]: n/a
[User impact if declined]: lack of regression test
[Describe test coverage new/current, TreeHerder]: this is the test
[Risks and why]: none
[String/UUID change made/needed]: none

Comment 28

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/7502d9ef870f

Comment 29

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Comment on attachment 8754123 [details] [diff] [review]
0001-Bug-1268096-r-jrmuizel-Add-regression-test.patch

Test checkins do not need review by release management team. They are auto-approved.

Comment 30

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

This test is failing: https://treeherder.mozilla.org/logviewer.html#?job_id=28297027&repo=mozilla-inbound

Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/e58f50a265a1

Comment 31

[Kelsey Gilbert [:jgilbert]]

Should be simple, but to keep things clean, here's a try run:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=507ceac0e399

(In reply to Ritu Kothari (:ritu) from comment #29)
> Comment on attachment 8754123 [details] [diff] [review]
> 0001-Bug-1268096-r-jrmuizel-Add-regression-test.patch
> 
> Test checkins do not need review by release management team. They are
> auto-approved.

Awesome, great policy.

Comment 32

[Milan Sreckovic [:milan] (needinfo for best results)]

Lost track of this?

Comment 33

[Kelsey Gilbert [:jgilbert]]

(In reply to Milan Sreckovic [:milan] from comment #32)
> Lost track of this?

I've been lax about commenting. Adding the test creates a perma-orange due to leaking the WebGLContextLossHandler object. (allegedly)
Landing the test is blocked on fixing this.

Comment 34

[Kelsey Gilbert [:jgilbert]]

Comment on attachment 8754123 [details] [diff] [review]
0001-Bug-1268096-r-jrmuizel-Add-regression-test.patch

Moved to bug 1280508.