417101 - Firefox crashes on first exit after an update or fresh install

Status: VERIFIED DUPLICATE of bug 417115

(Toolkit :: Application Update, defect, P1)

Description

[Carsten Book [:Tomcat]]

Reproduced multiple times on Windows XP with Updates from alpha2, alpha3, alpha4, beta1de etc.

Steps to reproduce:
-> Clean Profile
-> Update from Versions before Firefox 3 Beta 3 (like Beta 1) to Beta 3 via Check for Updates
-> Firefox restarts to apply the update
-> When you exit now Firefox after the Update is applied, Firefox crashes on exit

Note, this is a one time crash, a later session is okay

http://crash-stats.mozilla.com/report/index/669bb7ca-d9be-11dc-8f5e-001a4bd46e84

Comment 1

[[:Cww]]

Ditto.  Totally confirming.  From Beta2 to Beta3.

Comment 2

[Al Billings [:abillings - ex-MoCo]]

I'm getting this ONLY on "complete" installs after I force fallback. This is on
Intel OS X 10.5:

Beta 2 to Beta 3: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; ko; rv:1.9b2)
Gecko/2007121014 Firefox/3.0b2 -> Mozilla/5.0 (Macintosh; U; Intel Mac OS X
10.5; ko; rv:1.9b3) Gecko/2008020511 Firefox/3.0b3

Beta 1 to Beta 3: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; fr; rv:1.9b1)
Gecko/2007110903 Firefox/3.0b1 -> Mozilla/5.0 (Macintosh; U; Intel Mac OS X
10.5; fr; rv:1.9b3) Gecko/2008020511 Firefox/3.0b3

This is done with new profiles. The crash doesn't happen every time but the
symptom on OS X is a Firefox hang on first complete shutdown/exit after the
update is finished and you have the Beta 3 firstrun, etc.

Comment 3

[Al Billings [:abillings - ex-MoCo]]

I tried five times with the beta 1 french updates and hung three times.

I tried four times with beta 2 korean and hung twice.

Comment 4

[Mike Beltzner [:beltzner, not reading bugmail]]

To be clear, this happens on the beta update channel, but didn't happen on the betatest or releasetest update channels?

What about the channel could cause a crash?

Comment 5

[Al Billings [:abillings - ex-MoCo]]

We did not see this on releastest earlier.

Comment 6

[Carsten Book [:Tomcat]]

(In reply to comment #4)
> To be clear, this happens on the beta update channel, but didn't happen on the
> betatest or releasetest update channels?
> 
> What about the channel could cause a crash?
> 

Yes, in my case the Windows XP release test updates were okay, the updates from
beta crash, also as example on a normal update from DE beta 1 to Beta 3 via
check for updates.

Comment 7

[Mike Schroepfer]

Are we positive the bits are the same in both channels?  Does this happen with full or partials?  Would be great to see those crash stacks..

Comment 8

[Al Billings [:abillings - ex-MoCo]]

I just went back and checked on releasetest with the same Korean build and I just got it. I'm wondering if somehow it is certain mirrors and not others?

Since I get a hang, I get no crash data. OS X just tells me that the app is not responding after I quit.

Comment 9

[Robert Helmer [:rhelmer]]

(In reply to comment #7)
> Are we positive the bits are the same in both channels?  Does this happen with
> full or partials?  Would be great to see those crash stacks..

Checking this now, but it looks the same to me. I'm wondering if it's a problem with one of our mirrors.

Comment 10

[Carsten Book [:Tomcat]]

Another crash stack is
http://crash-stats.mozilla.com/report/index/bd785402-d9c0-11dc-8abe-001a4bd43ef6?date=2008-02-12-23
btw.

Note: the Profile seems to be working after this crash. All bookmarks and
settings are still there and no data is lost.

also just to point this out, its a crash after exit, so it does not crash
during user operation.

Comment 11

[Mike Schroepfer]

We've got a much smaller user base for B3.  Can we pull all mirrors from bouncer and self-host as we debug?

Comment 12

[Justin Fitzhugh]

Yes - we have 2x mirrors at ISC and a few in San Jose.  They can cover the load.  Just let us know when...

Comment 13

[Carsten Book [:Tomcat]]

Attachment: last update log

the update log from (in this last crash) crash from DE beta 1 -> beta 3 DE

Comment 14

[Robert Helmer [:rhelmer]]

(In reply to comment #9)
> (In reply to comment #7)
> > Are we positive the bits are the same in both channels?  Does this happen with
> > full or partials?  Would be great to see those crash stacks..
> 
> Checking this now, but it looks the same to me. I'm wondering if it's a problem
> with one of our mirrors.

Bits and channels look consistent. From the description, I can't think of a way this would be a mirror problem, though; sorry for the noise.

Comment 15

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to comment #9)
> (In reply to comment #7)
> > Are we positive the bits are the same in both channels?  Does this happen with
> > full or partials?  Would be great to see those crash stacks..
> 
> Checking this now, but it looks the same to me. I'm wondering if it's a problem
> with one of our mirrors.
> 

ummm...if the shasums were invalid / corrupted, wouldn't the updater not accept the updates?

Comment 16

[[:Cww]]

Ok, I forgot to attach a user agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b3) Gecko/2008020514 Firefox/3.0b3

Comment 17

[Carsten Book [:Tomcat]]

also crash with the installer with no Firefox Profile before on Firefox 3 Beta3

->
http://crash-stats.mozilla.com/report/index/ea1365a8-d9c8-11dc-827a-001a4bd43e5c?date=2008-02-13-00

so this affects also the installer.

Comment 18

[Michael Morgan [:morgamic]]

(In reply to comment #10)
> Another crash stack is
> http://crash-stats.mozilla.com/report/index/bd785402-d9c0-11dc-8abe-001a4bd43ef6?date=2008-02-12-23

Hey guys, since we have a stack overflow here, the web server can't report on it (the raw dump is too large...).  Here is the raw output from the dumps table in the db:
http://people.mozilla.com/~morgamic/bd785402-d9c0-11dc-8abe-001a4bd43ef6.rawdump.txt

Aravind is pulling the processed dump from 669bb7ca-d9be-11dc-8f5e-001a4bd46e84
and we'll have that in a few as well.

Comment 19

[Michael Morgan [:morgamic]]

And the next...
http://people.mozilla.com/~aravind/669bb7ca-d9be-11dc-8f5e-001a4bd46e84_db_dump.txt

Comment 20

[Mike Beltzner [:beltzner, not reading bugmail]]

Updating summary.

We've now seen this bug in all instances, and can generalize it to being a crash upon exit after the first time the beta is run when doing:

 - fresh install
 - pave-over install
 - using binary installers
 - from updates

Comment 21

[Mike Beltzner [:beltzner, not reading bugmail]]

This was caused by an animated favicon on the firstrun page.

Comment 22

[Stephen Donner [:stephend] Not actively reading bugmail]

Verified dup