1446667 - layout.display-list.retain cause a blink while scrolling

Status: RESOLVED WORKSFORME

(Core :: Web Painting, defect)

Description

[u614034]

Attachment: about:support.txt

User Agent: Mozilla/5.0 (Windows NT 6.1; rv:60.0) Gecko/20100101 Firefox/60.0
Build ID: 20180315232954

Steps to reproduce:

load any site
use mouse-wheel to scroll on screen at a time(run-> main.cpl ->wheel ->scroll one screen at a time in windows)
there is similar blink but fixed with layout.display-list.retain=false

Comment 1

[Alice0775 White]

I think the problem is gfx.webrender.all = true. The preferences is false by default in Beta60.

Comment 2

[Alice0775 White]

@reporter,
Could you try reset the pref(gfx.webrender.all)?

Comment 3

[u614034]

(In reply to Alice0775 White from comment #2)
> @reporter,
> Could you try reset the pref(gfx.webrender.all)?

Did same issue

Comment 4

[Alice0775 White]

(In reply to Emilio from comment #3)
> (In reply to Alice0775 White from comment #2)
> > @reporter,
> > Could you try reset the pref(gfx.webrender.all)?
> 
> Did same issue

Did you restart browser after reset the pref?  (Restarting browser is required for the settings to take effect.)

And Could you try with clean new profile?

Comment 5

[u614034]

(In reply to Alice0775 White from comment #4)
> (In reply to Emilio from comment #3)
> > (In reply to Alice0775 White from comment #2)
> > > @reporter,
> > > Could you try reset the pref(gfx.webrender.all)?
> > 
> > Did same issue
> 
> Did you restart browser after reset the pref?  (Restarting browser is
> required for the settings to take effect.)
> 
> And Could you try with clean new profile?

@Alice0775 
yes made a new profile, still the problem occurs!
toggling layout.display-list.retain=false fixes it or setting the mouse setting in windows to default.

Comment 6

[u614034]

(In reply to Emilio from comment #0)
> Created attachment 8959850 [details]
> about:support.txt
> 
> User Agent: Mozilla/5.0 (Windows NT 6.1; rv:60.0) Gecko/20100101 Firefox/60.0
> Build ID: 20180315232954
> 
> Steps to reproduce:
> 
> load any site
> use mouse-wheel to scroll on screen at a time(run-> main.cpl ->wheel
> ->scroll one screen at a time in windows)
> there is similar blink but fixed with layout.display-list.retain=false

It's more like a pause which causes the blink until the page is redrawn
more visible on heavy sites or multiple heavy sites opened at once.

Comment 7

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

Firefox 60 is also affected per bug #1440144.

Comment 8

[Matt Woodrow (:mattwoodrow)]

I can't reproduce this locally, how long does the flash usually last? Does it stay missing until you interact with the page again, or does it show up quickly after?

Comment 9

[u614034]

(In reply to Matt Woodrow (:mattwoodrow) from comment #8)
> I can't reproduce this locally, how long does the flash usually last? Does
> it stay missing until you interact with the page again, or does it show up
> quickly after?

Anywhere from 1sec to 3secs

See also Bug 1442573

Comment 10

[Ryan VanderMeulen [:RyanVM]]

I'm wondering if this is a dupe of bug 1447193. Maybe not if 60 is also affected, but it's also not clear to me if this has actually been reproduced on 60 Beta builds or not. Emilio, is this still reproducible for you?

Comment 11

[u614034]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #10)
> I'm wondering if this is a dupe of bug 1447193. Maybe not if 60 is also
> affected, but it's also not clear to me if this has actually been reproduced
> on 60 Beta builds or not. Emilio, is this still reproducible for you?

Firefox
Version 	60.0b11
Build ID 	20180409184545

still has the issues most visible if opening ten tabs at once and  using scrolling on one page it takes a pause then the page is rendered again
scroll one screen at a time in windows settings,smooth scroll is also disabled.

Comment 12

[u614034]

 	Firefox
Version 	61.0a1
Build ID 	20180410220129

Also show the issue
set
scroll one screen at a time in windows mouse settings,smooth scroll to disabled
visit while page loads scroll using scroll button

http://news.softpedia.com/

Comment 13

[u614034]

even here, please note it happens on first scroll only after that it cannot be reproduced except reloading again.

Comment 14

[u614034]

Attachment: small blinks-pauses.webm

Comment 15

[u614034]

Attachment: No pauses or blinks.webm

Set
browser.tabs.remote.autostart;false

and/or 

layout.display-list.retain;false

Comment 16

[Ryan VanderMeulen [:RyanVM]]

(In reply to Emilio from comment #11)
> still has the issues most visible if opening ten tabs at once and  using
> scrolling on one page it takes a pause then the page is rendered again
> scroll one screen at a time in windows settings,smooth scroll is also
> disabled.

Darn. Thanks for the update.

Comment 17

[u614034]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #16)
> (In reply to Emilio from comment #11)
> > still has the issues most visible if opening ten tabs at once and  using
> > scrolling on one page it takes a pause then the page is rendered again
> > scroll one screen at a time in windows settings,smooth scroll is also
> > disabled.
> 
> Darn. Thanks for the update.

Sure, hopefully can be fixed and merged to ESR60 soon

Comment 18

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

@ Emilio - Please abstain from adding bugs (especially completely unrelated) to "Depends on:" and "Blocks:". Thanks

Comment 19

[u614034]

(In reply to Virtual_ManPL [:Virtual] - (please needinfo? me - so I will see your comment/reply/question/etc.) from comment #18)
> @ Emilio - Please abstain from adding bugs (especially completely unrelated)
> to "Depends on:" and "Blocks:". Thanks

Sorry looked like same

Comment 20

[Matt Woodrow (:mattwoodrow)]

I still can't reproduce this, news.softpedia.com loads pretty much instantly and I can scroll the whole page without issues.

Could you try getting a performance profile for this, so we can see if it's just slow, or if it's waiting when it shouldn't be -  https://developer.mozilla.org/en-US/docs/Mozilla/Performance/Reporting_a_Performance_Problem

Comment 21

[u614034]

Old profile

https://perfht.ml/2Hmq5Ha

with webrender

https://perfht.ml/2qs1ebc

https://perfht.ml/2qt9BDJ

Comment 22

[Virtual_ManPL [:Virtual] 🇵🇱 - (please needinfo? me - so I will see your comment/reply/question/etc.)]

(In reply to Emilio from comment #21)
> Old profile
> 
> https://perfht.ml/2Hmq5Ha
> 
> with webrender
> 
> https://perfht.ml/2qs1ebc
> 
> https://perfht.ml/2qt9BDJ

FYI - WebRender isn't stable and enabled be default yet, disable it to minimize preference noise.

Comment 23

[Julien Cristau [:jcristau]]

layout.display-list.retain will be turned off for 60 in bug 1454509

Comment 24

[Jason Mechelynck]

Hi guys seem to have a similar problem with nightly61

so here is the performance profile

While 2 pages loading

https://perfht.ml/2qQXOhL

while 10 pages loading
https://perfht.ml/2qS5e4h

While completely loaded my session and seeing flicks while scrolling on every tab.
https://perfht.ml/2HK3419

Will Firefox web-render fix this

Should i keep running Nightly61 for telemetry?
Or revert to stable if you guys have the data.
Thank you for making Firefox


Mr.Ryan Mr.Matt can you look in this

Does not happen in Firefox60beta

Comment 25

[Ryan VanderMeulen [:RyanVM]]

Matt is the engineering owner for this feature.

Comment 26

[Jason Mechelynck]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #25)
> Matt is the engineering owner for this feature.

Thanks

Comment 27

[Ryan VanderMeulen [:RyanVM]]

Jason, it looks like the profiles from comment 24 no longer exist. Would you be willing to generate a new one on a Nightly build w/o WebRender enabled?

Comment 28

[Jason Mechelynck]

Sorry for replying late, was on a camping trip
just returned so

so here is a quick performance profile 
https://perfht.ml/2wOHioV

did not test with the original session

Comment 29

[Ryan VanderMeulen [:RyanVM]]

Thanks, Jason. We'll see if any of our engineers with expertise in looking at these notice anything obvious in it.

Comment 30

[Matt Woodrow (:mattwoodrow)]

(In reply to Jason Mechelynck from comment #28)
> Sorry for replying late, was on a camping trip
> just returned so
> 
> so here is a quick performance profile 
> https://perfht.ml/2wOHioV
> 
> did not test with the original session

Fascinating! Basically all your threads are idle, except for the compositor itself (powering the async scrolling), and that's running *super* slow (up to a full second per frame).

For some reason it looks like you're getting WARP, the software implementation of D3D11, which is supposed to be disallowed (since it's slow!).

Does switching the pref layers.acceleration.disabled to true make things better for you?

We're tracking this specific issue in bug 1401268.

Comment 31

[Jason Mechelynck]

(In reply to Matt Woodrow (:mattwoodrow) from comment #30)
> (In reply to Jason Mechelynck from comment #28)
> > Sorry for replying late, was on a camping trip
> > just returned so
> > 
> > so here is a quick performance profile 
> > https://perfht.ml/2wOHioV
> > 
> > did not test with the original session
> 
> Fascinating! Basically all your threads are idle, except for the compositor
> itself (powering the async scrolling), and that's running *super* slow (up
> to a full second per frame).
> 
> For some reason it looks like you're getting WARP, the software
> implementation of D3D11, which is supposed to be disallowed (since it's
> slow!).

Would be a Lie if said understand what you mean.

> Does switching the pref layers.acceleration.disabled to true make things
> better for you?


No seems to spike the CPU more

> We're tracking this specific issue in bug 1401268.

So updated windows and Nightly to latest

rebooted

re run
so can you look into these


https://perfht.ml/2Iy279d

Comment 32

[Jason Mechelynck]

look @ this too(especially toward the end/start)

https://perfht.ml/2IqCVFY

Comment 33

[Jason Mechelynck]

NI the wrong person

Comment 34

[Jason Mechelynck]

pref layers.acceleration.disabled to true

look at the scrolling towards the end

https://perfht.ml/2IuIhvP

Comment 35

[Ryan VanderMeulen [:RyanVM]]

While this whole situation still seems pretty interesting and odd, it also doesn't sound like something we'd expect to see much in the wild (and also not directly tied to RDL). Dropping Fx61 tracking status based on that, though I'd still consider a patch if the investigation leads to something.

Comment 36

[saproeno]

anyone landing a patch for this in 63?

Comment 37

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

Too late for 63/64. We could still take a patch for 65/66.

Comment 38

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

Too late to fix in 64. Marking this issue as fix-optional for 65; if you land a patch in nightly and think it's low-risk for beta, please request uplift.

Comment 39

[Alin Ilea, Desktop Test Engineering]

@Matt is there anything left to do here or this ticket can be closed since there are no new updates for the last 3 years and it might not be relevant anymore? I cannot reproduce the issue on my Win7 or Win10.
Thanks!

Comment 40

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee didn't login in Bugzilla in the last 7 months and this bug has severity 'major'.
:miko, could you have a look please?
For more information, please visit auto_nag documentation.