1232302 - [e10s] Java game flickers while scrolling

Status: RESOLVED INVALID

(Core :: Graphics: Layers, defect)

Description

[Camelia Badau [:cbadau], Desktop QA]

Attachment: Nightly.gif

Reproducible on the latest Nightly 45.0a1(BuildID: 20151214030220)
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:34.0) Gecko/20100101 Firefox/34.0
Reproducible on Firefox 43 RC (BuildID: 20151208100201)
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:43.0) Gecko/20100101 Firefox/43.0
Reproducible on Firefox ESR 38.5.0(BuildID: 20151210074901)
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101 Firefox/38.0

Steps to reproduce: 
1. Launch Firefox.
2. Go to http://jgames.com/schnauser/ and play the Java game. 
3. Scroll up and down. 

Actual results: The game area flickers when scroll up and down is performed. Please see the attachments. 

Expected results: No flickers occur.

Comment 1

[Camelia Badau [:cbadau], Desktop QA]

Here is the gif for Firefox 43 RC: https://goo.gl/dQHLYK
Here is the gif for ESR 38.5.0: https://goo.gl/uOwo3s

Comment 2

[Loic]

I think it appears only with e10s enabled. Could you test on Nightly with e10s disabled too.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

Comment on attachment 8698015 [details]
Nightly.gif

This gif shows the expected behavior on Nightly - see https://bugzilla.mozilla.org/show_bug.cgi?id=1212813#c4.

(In reply to Camelia Badau, QA [:cbadau] from comment #1)
> Here is the gif for Firefox 43 RC: https://goo.gl/dQHLYK

This looks like the bug you're reporting.

> Here is the gif for ESR 38.5.0: https://goo.gl/uOwo3s

This gif doens't load for me.

Also, please add these as attachments to the bug so they're not dependent on the availability of a third-party service.

Comment 4

[Ryan VanderMeulen [:RyanVM]]

As Loic mentioned, hunting down a regression range with e10s disabled would be helpful.

Comment 5

[Loic]

With e10s enabled, regression range:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=2a51d8766dfbfd272518ab3a47d2ca5f452dc971&tochange=e1e281f04f2706c55d017ca55e35ada2e1d05aaf

Comment 6

[Ryan VanderMeulen [:RyanVM]]

That confirms comment 3 then - the behavior on Nightly is expected per bug 1212813. The more-interesting case is e10s disabled, which should behave more like the behavior seen the 43RC gif from comment 1 and in theory is more true to what our release users are seeing.

Comment 8

[Jim Mathies [:jimm]]

sorry, didn't see the e10s-disabled comments.

Comment 9

[Jamie Nicol [:jnicol]]

For non-e10s, I can reproduce on windows but not linux. It is extremely noticeable when resizing the window, even if the applet is not obscured. I'm not sure how to get mozregression-downloaded firefoxes load the java plugin, so am unable to get a regression range.

This happens on several random java applets, but not all. The ones for which this happens on firefox also flicker on internet explorer in the same way. And the ones which don't flicker on firefox also do not flicker on internet explorer.

I think it's possible that this is due to the applets unnecessarily repainting rather than a browser bug.

Comment 10

[Ryan VanderMeulen [:RyanVM]]

I would think that mozregression would pick up the installed Java plugin on its own without any extra effort since they're installed at the system level. Was that not working?

Anyway, interesting observations RE: IE. I think that we should take a look at older releases (like ESR31) and just close this as invalid if the behavior remains consistent. Especially since bug 1232181 looks like it'll fix this for the e10s case, which is more important in the long run.

Comment 11

[Ryan VanderMeulen [:RyanVM]]

Forgot to say - if for some reason the mozregression-downloaded builds aren't picking up the Java plugin for some reason, you can try playing with the --profile and --profile-persistence options to get it working.

Comment 12

[Loic]

Builds from mozreg used the Java plugin installed on the machine, as I tested with e10s enabled.

Comment 13

[Jim Mathies [:jimm]]

(In reply to Jamie Nicol [:jnicol] from comment #9)
> For non-e10s, I can reproduce on windows but not linux. It is extremely
> noticeable when resizing the window, even if the applet is not obscured. I'm
> not sure how to get mozregression-downloaded firefoxes load the java plugin,
> so am unable to get a regression range.
> 
> This happens on several random java applets, but not all. The ones for which
> this happens on firefox also flicker on internet explorer in the same way.
> And the ones which don't flicker on firefox also do not flicker on internet
> explorer.
> 
> I think it's possible that this is due to the applets unnecessarily
> repainting rather than a browser bug.

Yeah I think you're just experiencing bad resize paint performance in java, which isn't surprising.

Comment 14

[Jamie Nicol [:jnicol]]

I got the same behaviour on windows with 31 ESR, so closing this as INVALID for non-e10s case, and other tickets already track the e10s fix.