Okta chiclets load very slowly one by one and when done loading start over again for a total of two renders
Categories
(Core :: Graphics: Canvas2D, defect)
Tracking
()
People
(Reporter: alex_mayorga, Unassigned)
References
()
Details
(Keywords: nightly-community)
Basic information
Steps to Reproduce: Load https://<your-workplace>.okta.com
Expected Results: Okta chiclets load fast in one go
Actual Results: Okta chiclets load very slowly one by one and when done loading start over again for a total of two renders
More information
Screenshot: N/A
Profile URL: https://share.firefox.dev/2SFzSQ6
Basic systems configuration:
OS version: Windows 10.0.21376 Build 21376
GPU model: Intel(R) HD Graphics 620
Number of cores: 2
Amount of memory (RAM): Installed Physical Memory (RAM) 24.0 GB
Thanks so much for your help.
|
||
Comment 1•3 years ago
|
||
This appears to be caused by locking contention during readback on a canvas surface. Jeff, any idea who should look at this?
|
||
Comment 2•3 years ago
|
||
Alex, does it work better if you set gfx.canvas.remote=true and restart?
|
Reporter | |
Comment 3•3 years ago
|
||
¡Hola Jeff!
When I checked about:config that preference was already set.
I tried with gfx.canvas.remote=false as well and the experience is pretty much the same albeit a tad slower on my unscientific observation.
Please let me know if you need me to test anything else.
¡Gracias!
Alex
|
||
Comment 5•3 years ago
|
||
Hi Alex,
I don't think I have a login to be able to test this.
Would you be able to post two new profiles using the "Firefox Graphics" settings, one with gfx.canvas.remote set to false and one set to true.
Also, if this is a recent regression can you try https://mozilla.github.io/mozregression/ to narrow it down.
Thanks.
|
|
Reporter | |
Comment 6•3 years ago
|
||
¡Hola Bob!
Please see the profiles:
gfx.canvas.remote false https://share.firefox.dev/3v4llf6
gfx.canvas.remote true https://share.firefox.dev/3eYpziF
I will try to find a regression window.
¡Gracias!
Alex
|
||
Updated•3 years ago
|
|
|
||
Comment 7•3 years ago
|
||
(In reply to alex_mayorga from comment #6)
...
gfx.canvas.remote false https://share.firefox.dev/3v4llf6
gfx.canvas.remote true https://share.firefox.dev/3eYpziF
Thanks Alex, looks like this is down to reading back data out of the canvas.
Either in the content process main thread (with gfx.canvas.remote=false) or in the GPU process canvas thread otherwise.
It makes me wonder if they have changed the page, which might mean finding a regression range in Firefox is fruitless.
|
|
Reporter | |
Updated•3 years ago
|
|
|
Reporter | |
Updated•3 years ago
|
|
|
Reporter | |
Comment 8•3 years ago
|
||
¡Hola Bob!
Hope these lines find you well.
Can you please describe a happy path forward for this bug?
The issue is still seen in Nightly 93.
The issue does not happen on Chrome nor Edge.
Is this a Web compatibility bug perhaps?
¡Gracias!
Alex
|
|
||
Comment 9•3 years ago
|
||
(In reply to alex_mayorga from comment #8)
¡Hola Bob!
Hope these lines find you well.
Can you please describe a happy path forward for this bug?
The issue is still seen in Nightly 93.
The issue does not happen on Chrome nor Edge.
Is this a Web compatibility bug perhaps?
¡Gracias!
Alex
Hi Alex,
I don't really have any new suggestions.
Have you been able to find a regression range or was this a change on the okta site?
As I've said before this appears to be taking all the time pulling back data out of the canvas, I'm not sure why the site relies on this so heavily.
Loading mozilla.okta.com doesn't display much and seems fast, but maybe that's not a valid test.
Perhaps you could explain what the Okta chiclets are.
Description
•