676458 - [@ hang | google_breakpad::ExceptionHandler::WriteMinidump ]

Status: RESOLVED WONTFIX

(Core Graveyard :: Plug-ins, defect)

Description

[Vic]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:6.0) Gecko/20100101 Firefox/6.0
Build ID: 20110729080751

Steps to reproduce:

Visited this page: http://issuu.com/abbotsfordnews/docs/abbotsford-news_thu_aug4_2011


Actual results:

Firefox stopped responding.


Expected results:

I should be allowed to read that newspaper.

Comment 1

[Vic]

http://crash-stats.mozilla.com/report/index/bp-79cb039c-e23c-4ce0-b3d3-f035a2110803

http://crash-stats.mozilla.com/report/index/bp-3f89e67e-e89b-460e-827a-339852110803

Comment 2

[George Carstoiu]

Mozilla/5.0 (X11; Linux i686; rv:8.0a1) Gecko/20110808 Firefox/8.0a1

I've tested this under Win7 x86 and x86_64 and Mac OS X 10.6 and everything seems in order, although there's a different behavior on Win 7 vs Mac OS X when pressing "Click to read". On Mac OS X the pluginwrapper is displayed in fullscreen, while on Win 7 more of a pop-out window is displayed (not Fullscreen). In all cases, Fx does not crash or freeze.

Under Ubuntu 11.04 x86, Fx freezes when trying to read the articles using the "Click to read" feature. Also, I found it difficult to go through all the pages of the newspaper by using the left and right animated arrows (without switching to fullscreen by clicking "Click to read").

What is strange is that almost the same behavior is noticeable on other browsers on Ubuntu - Opera 11.5 and Chrome 13, so I am not sure if this is a Fx related issue.

Comment 4

[Vic]

(In reply to George Carstoiu from comment #2)
> Mozilla/5.0 (X11; Linux i686; rv:8.0a1) Gecko/20110808 Firefox/8.0a1
> 
> I've tested this under Win7 x86 and x86_64 and Mac OS X 10.6 and everything
> seems in order, although there's a different behavior on Win 7 vs Mac OS X
> when pressing "Click to read". On Mac OS X the pluginwrapper is displayed in
> fullscreen, while on Win 7 more of a pop-out window is displayed (not
> Fullscreen). In all cases, Fx does not crash or freeze.
> 
> Under Ubuntu 11.04 x86, Fx freezes when trying to read the articles using
> the "Click to read" feature. Also, I found it difficult to go through all
> the pages of the newspaper by using the left and right animated arrows
> (without switching to fullscreen by clicking "Click to read").
> 
> What is strange is that almost the same behavior is noticeable on other
> browsers on Ubuntu - Opera 11.5 and Chrome 13, so I am not sure if this is a
> Fx related issue.

Don't press "click to read", press the Next or Previous button instead. So it should not become full screen, but just change the page.

This will cause the whole Firefox to be not responsive.

Whereas Chrome doesn't crash, Chrome responds by saying the plugin crashed, but it is still responsive.

Comment 5

[George Carstoiu]

(In reply to Vic from comment #4)
> Don't press "click to read", press the Next or Previous button instead. So
> it should not become full screen, but just change the page.
> 
> This will cause the whole Firefox to be not responsive.
> 
> Whereas Chrome doesn't crash, Chrome responds by saying the plugin crashed,
> but it is still responsive.

I tried that on Windows and Mac and everything worked perfectly. The problem appeared when trying this on Ubuntu - the page changed ar first, but then I was unable to move to the following page. Sometimes, the browser froze for a few seconds while the content was being loaded, but as I said previously, the same behavior is present on Chrome and Opera on Ubuntu.

Comment 6

[Vic]

Please refer to Bug 566062 - OOPP hang in mozilla::plugins::PPluginInstanceParent::CallPBrowserStreamConstructor 

That causes Firefox to be not responsive.

Browser crash reports: 

https://crash-stats.mozilla.com/report/index/4ca8b230-ce05-41cd-855e-ffdfb2110803

https://crash-stats.mozilla.com/report/index/095727b9-3c65-4e63-921b-a5f942110803

Comment 7

[Vic]

Edited summary because Bug 676492 was marked as a duplicate of this bug, but that has more crashes.

Comment 8

[Vic]

Socorro please pick up https://crash-stats.mozilla.com/report/index/bp-fe95ae88-925a-43c3-ac45-fb51e2110820 ... it been 36 hours since I added that to Crash Signature here... I thought you ran every 6 hours...maybe you need the URL which I have pasted this time :)

Comment 9

[Vic]

fix grammar: *it has been* ;)

Comment 10

[Bill Gianopoulos [:WG9s]]

(In reply to George Carstoiu from comment #5)
> I tried that on Windows and Mac and everything worked perfectly. The problem
> appeared when trying this on Ubuntu - the page changed ar first, but then I
> was unable to move to the following page. Sometimes, the browser froze for a
> few seconds while the content was being loaded, but as I said previously,
> the same behavior is present on Chrome and Opera on Ubuntu.

It has been my experience that flash hangs that occur using flash version 10 under Linux, that do not hang under using Flash version 10 under Windows can be fixed by running Flash 11 beta.

Since the original report here was about an issue under windows, I don't think what you are seeing is the same as what the original reporter experienced.

Comment 11

[Vic]

Edited title so hopefully Socorro associates with this bug. 

https://crash-stats.mozilla.com/report/index/bp-fe95ae88-925a-43c3-ac45-fb51e2110820

Comment 15

[Warren]

Here are some observations if it helps anyone. 
These crashes pop up when playing a flash video (can be windowed or full screen) and I hit the back button. Most of the time it's not a problem but occasionally (5%?) all of the following happens:
1. It stays on the flash page and doesn't navigate to the previous page
2. The video stops playing
3. The audio continues playing
4. All Firefox stops responding to any input
5. After about 30 seconds, all instances of the flash plugin crash including those in other tabs. It navigates to the page I was trying to go back top, and may respond to other commands I had queued up. 

It only seems to happen when playing an embedded video, though that accounts for probably 95% of my flash usage. I have been trying to find a way to successfully repro the problem, but it's been occurring at random. All other functionality in Firefox and other apps has been stable so I don't think I'm encountering hardware instability. I've updated the video driver since then as well with the same result.

I'll keep searching for a consistent repro and post if I can figure it out.

Comment 16

[Wayne Mery (:wsmwk)]

(In reply to Warren from comment #15)
> I'll keep searching for a consistent repro and post if I can figure it out.

warren, any luck?

Comment 17

[(not currently active) Ted Mielczarek]

Sounds like this is either a bug in our plugin code, or in Flash.

Comment 18

[Sergey Sokoloff]

I've just managed to hang Firefox 9 on Windows by mere visiting http://issuu.com/rootua/docs/fossnews24

(I've even hadn't to press anything like «Next» button. I've just opened the URL and seen the almost instantly frozen Firefox.)

I hope this may help in your testing, if that can be reproduced.

Flash Plugin is 10.1.51.66 (aka 10.1 d51).

Comment 19

[Benjamin Smedberg]

What exactly is this bug about? hang | google_breakpad::ExceptionHandler::WriteMinidump is a generic signature which should be fixed in Socorro, bug 709209. This bug appears to have initially been about specific STR a hang using Flash, but later comments have veered off, it seems.

It looks from comment 2 and 4 that either Flash or the particular .swf file in question has a bug where it never returns control. In Firefox, that will show up as a hang signature, which in this case means that the hang detector is working correctly and this is either INVALID or WORKSFORME ;-)

smooney, what does "topcrash" mean here?

Comment 20

[Scoobidiver (away)]

(In reply to Benjamin Smedberg  [:bsmedberg] from comment #19)
> what does "topcrash" mean here?
The hang | google_breakpad::ExceptionHandler::WriteMinidump crash signature is #3 top crasher in 8.0 and #1 in 7.0.1 on Linux.

Comment 21

[Benjamin Smedberg]

As noted that is a generic signature caused by Socorro brokenness, it is a useless marker until bug 709209 is fixed.

Comment 23

[Scoobidiver (away)]

File a new bug with only the plugin side crash signature if you don't want it is hijacked by other issues (see comment 19).
I removed the plugin side crash signatures to make it clear.