434236 - Frequent Freezing (Hang, Very high CPU usage, beach ball) in Firefox

Status: RESOLVED WORKSFORME

(Core :: JavaScript Engine, defect)

Description

[Andre-John Mas]

User-Agent:       Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_5_2; en-us) AppleWebKit/525.18 (KHTML, like Gecko) Version/3.1.1 Safari/525.18
Build Identifier: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-GB; rv:1.9) Gecko/2008051202 Firefox/3.0

I have just upgraded from 3.0 beta 5 to 3.0 RC1 and I am experiencing frequent beach balling (freezing/non-responsiveness). This forces me to force quit the application.

Reproducible: Sometimes

Steps to Reproduce:
1.
2.
3.

Comment 1

[Andre-John Mas]

Attachment: Log from force-quit

I have attached a log generated last time this happened and I force quit the application.

Comment 2

[Andre-John Mas]

Attachment: Second force-quit log

This has happened to me a few times since. I see CPU usage spike, getting beyond 160% (dual core CPU). In all instance I can only force-quit as Firefox does not seem to recover. 

I don't know whether it is significant but I do see a lot of recursion of 'JS_FloorLog2' in the log.

Comment 3

[Andre-John Mas]

Changing to Major, as this is happening too often for my liking.

Comment 4

[Andre-John Mas]

This is still happening with the release version of Firefox. In the report what I am seeing is the following being repeated:

 1 _JS_FloorLog2 + 21627 (in libmozjs.dylib) [0xe2346]
    1 _JS_CallTracer + 478 (in libmozjs.dylib) [0xca3f9]
      1 _JS_FloorLog2 + 21627 (in libmozjs.dylib) [0xe2346]
        1 _JS_CallTracer + 478 (in libmozjs.dylib) [0xca3f9]
          1 _JS_FloorLog2 + 21627 (in libmozjs.dylib) [0xe2346]
            1 _JS_CallTracer + 478 (in libmozjs.dylib) [0xca3f9]
              1 _JS_FloorLog2 + 21627 (in libmozjs.dylib) [0xe2346]
                1 _JS_CallTracer + 478 (in libmozjs.dylib) [0xca3f9]
                  1 _JS_FloorLog2 + 21627 (in libmozjs.dylib) [0xe2346]
                    1 _JS_CallTracer + 478 (in libmozjs.dylib) [0xca3f9]
                      1 _JS_FloorLog2 + 21627 (in libmozjs.dylib) [0xe2346]
                        1 _JS_CallTracer + 478 (in libmozjs.dylib) [0xca3f9]
 

Comment 5

[Philip flip Kromer]

I'm getting the same thing.
Looking at activity monitor the mem usage is increasing, increasing, increasing -- right now I see RSIZE @ 167MB, when it kicks in I'll see it climb uncontrollably past 1GB (dep. on how fast I can force-quit.)

I'll attach a force-quit log and a spin control hang log.

Please note that since the app doesn't crash Talkback doesn't kick in -- so I (still) don't know how to refer to a talkback, sorry.

Comment 6

[Philip flip Kromer]

Attachment: Apple Crash logs + system Prof -- please delete, sorry for spam.

an apple crash log, several spin ctl logs from various hangs in the same session, and system profile.

Comment 7

[Philip flip Kromer]

Attachment: Apple Crash logs + system Prof -- please delete (dupe)

an apple crash log, several spin ctl logs from various hangs in the same session, and system profile.

Comment 8

[Philip flip Kromer]

Attachment: Apple Crash logs + system Prof

an apple crash log, several spin ctl logs from various hangs in the same session, and system profile.

Comment 9

[Philip flip Kromer]

Shoot, sorry about the multiple attachment spam.  My browser was, ha ha,
undergoing this selfsame crash and I didn't know they went through.

I got the following when it finally crashed out:
  [Exception... "'[JavaScript Error: "script stack space quota is exhausted"
{file:
"file:///Applications/Network/Firefox.app/Contents/MacOS/modules/JSON.jsm"
line: 176}]' when calling method: [nsISessionStore::getBrowserState]" 
nsresult: "0x80570021 (NS_ERROR_XPC_JAVASCRIPT_ERROR_WITH_DETAILS)"  location:
"JS frame :: chrome://sessionmanager/content/sessionmanager.js :: anonymous ::
line 1919"  data: yes]

I'm attaching the corresponding sessionstore.js, which is humunguous.  I don't
know if that's because the SessionManager is causing the crash, or whether it's
humungous because it's saving the runaway memory pool.

Comment 10

[Philip flip Kromer]

Attachment: sessionstore.js from a runaway hang that managed to crash

[Exception... "'[JavaScript Error: "script stack space quota is exhausted" {file: "file:///Applications/Network/Firefox.app/Contents/MacOS/modules/JSON.jsm" line: 176}]' when calling method: [nsISessionStore::getBrowserState]"  nsresult: "0x80570021 (NS_ERROR_XPC_JAVASCRIPT_ERROR_WITH_DETAILS)"  location: "JS frame :: chrome://sessionmanager/content/sessionmanager.js :: anonymous :: line 1919"  data: yes]

Comment 11

[Philip flip Kromer]

Consider adding 
  keywords: memory, stack, beachball, OSX, FloorLog2, CallTracer, _JS_CallTracer, JS_FloorLog2, libmozjs.dylib, sessionstore, session manager, runaway
  version 3.x branch


(I would add these but I'm either not allowed to or not smart enough -- please ping me directly if it's the latter.)

Comment 13

[Andre-John Mas]

Is there anything particular about the JS_FloorLog2 that could cause it to recurse so deeply?

Comment 14

[Juriy "kangax" Zaytsev]

I have just experienced a crash that seems to have similar recursion (FF 3.0.11). Attaching a log from "problem details".

Comment 15

[Juriy "kangax" Zaytsev]

Attachment: crash report with stack trace

Comment 16

[Wayne Mery (:wsmwk)]

(In reply to comment #15)
> Created an attachment (id=382963) [details]
> crash report with stack trace

unfortunately lacks symbols

Comment 17

[Juriy "kangax" Zaytsev]

@Wayne Mery

Which symbols are you talking about?

Comment 18

[Wayne Mery (:wsmwk)]

Juriy in comment #17
> Which symbols are you talking about?

https://developer.mozilla.org/en/Using_the_Mozilla_symbol_server

Comment 19

[Martin]

Is this issue still reproducible?

Comment 20

[Andre-John Mas]

I am using 3.6.3 on MacOS X 10.6.3 and I haven't experienced it in a while.

Comment 21

[Wayne Mery (:wsmwk)]

WFM per reporter