659348 - Flash from previously closed tab reappears when firefox hangs

Status: RESOLVED WORKSFORME

(Core Graveyard :: Plug-ins, defect)

Description

[Nochum Sossonko [:Natch]]

Attachment: Screenshot of Bug

After closing a tab (about ~2 minutes later) and reopening/closing others, my Firefox hung on the new yahoomail beta (they're onunload script was hanging) and the flash video reappeared. Very strange.

Picture attached.

about:support - graphics:

        Adapter Description
        NVIDIA GeForce 8600M GT

        Vendor ID
        10de

        Device ID
        0407

        Adapter RAM
        256

        Adapter Drivers
        nvd3dum nvwgf2um,nvwgf2um

        Driver Version
        8.17.12.7061

        Driver Date
        4-7-2011

        Direct2D Enabled
        true

        DirectWrite Enabled
        true (7.0.6002.18409, font cache 13.54 MB)

        WebGL Renderer
        Google Inc. -- ANGLE -- OpenGL ES 2.0 (ANGLE 0.0.0.611)

        GPU Accelerated Windows
        1/1 Direct3D 10

Comment 1

[Nochum Sossonko [:Natch]]

Ftr, this is the script causing the problem:

http://mail.yimg.com/zz/combo?nq/979/yui/yui-min.js&nq/979/oop/oop-min.js&nq/979/dom/dom-min.js&nq/979/event/event-min.js&nq/979/event-custom/event-custom-min.js&nq/979/base/base-base-min.js&nq/979/plugin/plugin-min.js&nq/979/pluginhost/pluginhost-min.js&nq/979/node/node-min.js&nq/979/attribute/attribute-min.js&nq/979/json/json-min.js&nq/979/intl/intl-min.js&nq/979/datatype/lang/datatype-date.js&nq/979/datatype/datatype-date-min.js&nq/979/datatype/datatype-xml-min.js&nq/979/cookie/cookie-min.js&nq/979/async-queue/async-queue-min.js&nq/979/collection/array-extras-min.js&nq/979/querystring/querystring-parse-simple-min.js&nq/979/loader/loader-min.js:1

This might just be a js hang bug with weird side effects.

Comment 2

[Nochum Sossonko [:Natch]]

This is actually a bigger issue then I originally thought. If you open a PB session and then exit it, flash from the PB session will do the same exact thing.

Simple STR:

1) Enter PB mode (ctrl-shif-p).
2) Visit some youtube video.
3) Wait until the video is fully buffered (this *might* not be necessary).
4) Exit PB Mode.
5) Type |javascript: while(true);| in the location bar.
6) You'll see the PB mode flash show up when the browser freezes up.

Comment 3

[(no longer active)]

Hmm, don't we kill the plugin-container process when we navigate away from a tab containing the plugin?  This seems really bad...

Comment 4

[Benjamin Smedberg]

No, we never unload plugins. They aren't per-tab anyway.

Comment 5

[(no longer active)]

(In reply to comment #4)
> No, we never unload plugins. They aren't per-tab anyway.

So, is it expected for a plugin to continue painting when a tab containing one of its instances has been closed?

Comment 6

[Benjamin Smedberg]

No, of course that's not expected.

Comment 7

[Nochum Sossonko [:Natch]]

Actually, in the flash case it isn't as bad, with the reader plugin embedded it's far worse as the whole page reappears.

I'll attach an example when I have a minute.

Comment 8

[Benjamin Smedberg]

This is not a new regression, release drivers don't need to track it for a particular release.

Comment 10

[Nochum Sossonko [:Natch]]

1. This bug might need to be hidden from public view (it more or less defeats some common uses for pb mode).

2. This should really get fixed asap (and get backported as well), it's a very bad bug when considering that it can happen at any moment (should the browser freeze), can't be dismissed or hidden, and can potentially reveal embarrassing browser usage to any onlooker.

I haven't been able to find a (non-private) pdf page that exhibits the bug mentioned in comment 7, which is why I haven't uploaded a test case for that yet.

Comment 11

[(mostly gone) XtC4UaLL [:xtc4uall]]

Natch, is this still an Issue?
I fail to repro using the Steps in Comment 2 - in Step 6 I see no PB Content, tried PDF and Flash.

Comment 12

[Nochum Sossonko [:Natch]]

Can't seem to repro either, I'll mark WORKSFORME for now.