709479 - [@ hang] KiFastSystemCallRet <- NtUserWaitMessage <- nsBaseAppShell <- nsThread

Status: RESOLVED INCOMPLETE

(Firefox :: General, defect)

Description

[Rob]

Attachment: Possibly a Windbg Log that was collected when [@ hang] occurred.

Sometimes Nightly will hang and if I either mouse-over or watch "Window's Task Manager" I can see that I get control back for less than a second, at which point it hangs again for a few minutes. 

If I let this continue the "Mozilla Error Reporter" _might_ pop up or I get a BSOD (and lose the WinDbg Log (or the chance to use the 'MER')).


When I regained control during one of these hangs I quickly used WinDbg's "Break" and tried to create a decent WinDbg Log to submit.

I also have a couple of "Socorro Crash Reports" and another (Bug 651202 Comment 14) Report with the _same_ "Top Of Stack". I tried to search for Dupes and found Bug 558647 Comment 9 which is a 'meta' for "KiFastSystemCallRet".



Firefox 11.0a1 Crash Report [@ hang ] (Browser) bp-45b4bd32-41ee-4d96-bedb-509ce2111210

Firefox 11.0a1 Crash Report [@ hang ] (Plugin) bp-3772ee80-b847-44a8-ba9f-1aae92111210

Please note: Those two "bp-'s" have _exactly_ the same "Signature" yet one is Browser-based and the other is Plugin-based. it would be MUCH better if the Signatures differed (and were longer; had a bit of the Stack).

Comment 1

[Wayne Mery (:wsmwk)]

BSOD and related issues are not firefox problems

Comment 2

[Rob]

When I reported this the "about:crashes" link to bp-3772ee80-b847-44a8-ba9f-1aae92111210 showed this Crash as "[@ hang]" and now it is called "hang" or
"Signature 	hang 	More Reports 	Search" .

Strangely bp-3772ee80-b847-44a8-ba9f-1aae92111210 is now processed by Socorro as a "Null Signature" and the "Crash Signature:" in _this_ Bug Report does not show up in the "Related Bugs" Section of Socorro's Report, instead it says "Bugzilla - Report this Crash".

I guess differentiation between the two sources (Browser or Plugin) of this "Crash Signature" would help also.

Comment 3

[Matthias Versen [:Matti]]

Both crash IDs are useless and the [@hang] is no real signature.
The reason is apparent: "No thread was identified as the cause of the crash"

I don't think that breakpad can identify hangs in the application (that would be new) and I think both hangs are basically plugin related.

Which addons are installed and have you tried it without addons ?
I hope that you don't use addons with binary components.

Comment 4

[[Baboo]]

(In reply to Matthias Versen (Matti) from comment #3)
> I don't think that breakpad can identify hangs in the application (that
> would be new) and I think both hangs are basically plugin related.

Well, if the "process type" row says "Filename: NPSWF32.dll" then it should be clear that it is an Flash-plugin related issue.

Here a crash report where even the version is evident as the row says "Filename: NPSWF32_11_2_202_160.dll": bp-9bee5161-6a47-4e94-90cf-ec95a2120121

Comment 5

[Scoobidiver (away)]

Although two crash reports are generated (one per side), this "[@ hang]" plugin crash signature is the equivalent of the "[@ EMPTY: no crashing thread identified; corrupt dump ]" application crash signature.
See bp-cbc18eb7-a3e4-4780-a4bd-849042120119 that has the same structure.

It should be closed as incomplete.

Comment 6

[Matthias Versen [:Matti]]

I agree, there is not enough information in this bug report.
closing as incomplete

Comment 7

[Rob]

(In reply to Matthias Versen (Matti) from comment #6)
> I agree, there is not enough information in this bug report.
> closing as incomplete

OK.

My _main_ concern (from Comment 0) was:

'Please note: Those two "bp-'s" have _exactly_ the same "Signature" yet one is Browser-based and the other is Plugin-based. it would be MUCH better if the Signatures differed (and were longer; had a bit of the Stack).'


If the "Browser" (firefox.exe) and the "plugin-container.exe" had different "Signatures" (except where they _actually_ share "Code") if would be better than having two identical "Signatures" for _different_ Code.

I thought there was enough info to make clear the value of this.


Thanks for reading my Report,
Rob