564298 - crash [@ hang | mozilla::plugins::PPluginInstanceParent::CallUpdateWindow() ]

Status: RESOLVED INCOMPLETE

(Core Graveyard :: Plug-ins, defect)

Description

[juan becerra [:juanb]]

While testing 3.6.4build3 and Silverlight 4.

http://crash-stats.mozilla.com/report/index/bp-7269b4df-783f-4974-9b89-73fac2100506

I was trying to reproduce bug 563936 when I encountered this hang. I had an
existing Silverglight 3 installation, and I uninstalled it using the option in
the Control Panel. Then I ran a Silverlight 4 installer I had previously
downloaded. Then I launched Fx 3.6.4build3 and tried to play one of the videos
from the link below, then while it seemed to be loading I tried right clicking
to bring up the context menu. Generally that's how I got into that hang. The
browser window froze, but I was able to launch another window by double
clicking on my Fx desktop icon.

http://www.silverlight.net/learn/videos/silverlight-4-videos/

Comment 1

[juan becerra [:juanb]]

I am getting this crash shortly after escaping full screen Flash on a very old machine, sort of consistently (10.0 or 10.1 Flash versions).

Comment 2

[Daniel T]

While I was testing the new beta 4.0b6. Soon after the update, I started to get problems with flash 10.1.82.76 on youtube.

Here's a crash report from me: http://crash-stats.mozilla.com/report/index/ee49bab4-53dc-448c-81c6-90c622100919

As I counted from the relese of 4.0b6 I have got this same crash 7 times by now. And from time to time when I close a youtube tab it just won't close. But if the tab closes it won't close flash. Cause I still can hear the youtube clip playing in my headphones. After awhile it stops but it can take up to 1-5mins after I have closed the tab.

My machine:
Version	4.0b6
Build ID    20100914083612
Branch    2.0
OS    Windows NT
OS Version    5.1.2600 Service Pack 3
CPU    x86
CPU Info    AuthenticAMD family 15 model 75 stepping 2

Comment 3

[Marcia Knous [:marcia]]

Nominating, #17 top crash on Beta 7 and trended up. Also has been around since May.

Comment 4

[Johnny Stenback (:jst)]

Benjamin, this sounds like a blocker. Can you have a look, or reassign appropriately?

Comment 5

[Benjamin Smedberg]

We need the other half of the hang report in order to do anything useful. This is basically "Flash isn't responding".

Comment 6

[Daniel T]

(In reply to comment #5)
> We need the other half of the hang report in order to do anything useful. This
> is basically "Flash isn't responding".

Which other half do you mean? And how can I find it? I thought I could provied the report, cause I had four crashes with this perticilar error with 4.0b6.

Comment 7

[Benjamin Smedberg]

These are hang reports where Flash doesn't respond. Each time you "crash", you should get *two* crash reports, one which lists the Firefox stack (this bug) and one which lists the plugin stack. It is necessary to have both reports to do any diagnosis here.

Comment 8

[Benjamin Smedberg]

Here are some of the matching plugin reports:

http://crash-stats.mozilla.com/report/index/7d923b57-566d-47da-bd2d-bd5e12101206
[hang | memmove | nsTArray_base::ShiftData(unsigned int, unsigned int, unsigned int, unsigned int)] (flash)

http://crash-stats.mozilla.com/report/index/27229b2f-9ba5-437d-acb6-b8cec2101206
[hang | KiFastSystemCallRet] (from CreateFileW, silverlight)

http://crash-stats.mozilla.com/report/index/bb16547d-f8b1-42f1-98cb-edc642101206
[hang | F_596512829______________________________] (flash)

I don't see a lot of commonality here yet.

Comment 9

[Benjamin Smedberg]

Data from bug 617421 indicates that this is a bunch of different issues, many of which appear to be plugins actually not responding for lengthy periods of time. Quicktime has a very common case where it's calling ShellExecute which is then hanging.

http://crash-stats.mozilla.com/report/index/c847d5c0-28a7-49fb-910c-161a32101213
http://crash-stats.mozilla.com/report/index/ac56c9e0-3cb5-4eb4-a512-0320c2101213

Acrobat has a case where it's hanging *during* UpdateWindow while in 	NtUserCallHwndLock: 
http://crash-stats.mozilla.com/report/index/a62f0e98-20a1-4dba-944e-df2c52101213
http://crash-stats.mozilla.com/report/index/5843b17f-806c-4751-84f6-f7f4a2101213
http://crash-stats.mozilla.com/report/index/ff400400-1dd3-4c78-bc6f-1a0072101214

Flash is calling Sleep (this may be a debug version of Flash):
http://crash-stats.mozilla.com/report/index/3305dc26-5743-41cf-9602-a0e662101213
http://crash-stats.mozilla.com/report/index/3305dc26-5743-41cf-9602-a0e662101213

There are a set where Flash is running but it appears that we're just "behind" far enough that it's slow to respond.

Flash is hanging while printing(?) to a postscript driver which is making some sort of RPC call:
http://crash-stats.mozilla.com/report/index/7a8f2ef5-aa25-4099-a8d3-7c2672101213

WMP is hanging creating a file mapping? Or this is another "plugin is active but slow" case:
http://crash-stats.mozilla.com/report/index/48100173-e495-4074-afa3-0e5b92101213

malware is crashing:
http://crash-stats.mozilla.com/report/index/a6d4808f-9098-424b-bfe1-3011e2101214

Acrobat hangs with SendMessage:
http://crash-stats.mozilla.com/report/index/7a722202-00ab-4d8b-81fc-cffdb2101215
http://crash-stats.mozilla.com/report/index/4e132da7-652e-48bf-8209-668f92101215

Of these, the only ones which are suspicious are the two Acrobat ones (SendMessage and NtUserCallHwndLock). Everything else is either noise or Firefox behaving as designed. I'll file those two, and this bug can become a tracker or be resolved INCOMPLETE. I'll also get a few more things added to the skiplist to make the child-side signatures more interesting.

Comment 10

[chris hofmann]

wonder if we should also spin off partner bugs to see if adobe and MS can do something about the bad behavior thats turning up in comment 9.

Comment 11

[Kev Needham [:kev]]

I'll bring Adobe in on this, and will work with my MS contacts to see if we can get the right folks with WMP (if anyone knows folks from the WMP team, please let us know).

Comment 12

[Benjamin Smedberg]

No, we don't have STR and most of these are in the very low-noise range.

Comment 13

[Robert Kaiser]

I'm closing this as INCOMPLETE as those browser-side hang reports didn't get us anywhere and we are reworking hang reporting. Those signatures are not seen any more as browser sides of hang pairs are not recorded any more. We will have dumps/stacks of browser and other processes right in one report together with their plugin hang parts in the future (product-wise starting with FF18, we're still working on proper support on the server side, please contact bsmedberg if you want to help).