Bug 1244

Summary: Epiphany crash while executing java script code
Product: Mageia Reporter: Herbert Poetzl <herbert>
Component: RPM PackagesAssignee: Mageia Bug Squad <bugsquad>
Status: RESOLVED WONTFIX QA Contact:
Severity: major    
Priority: Normal CC: cjw, dmorganec, fundawang, jani.valimaa, mageia, marja11, reinout
Version: 1Keywords: UPSTREAM
Target Milestone: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Source RPM: epiphany-2.30.6-2.mga1.src.rpm, webkit-1.2.7-4.mga1.src.rpm CVE:
Status comment:
Attachments: Crash on Recovery
Second Recovery Run
Third Recovery Run
Fourth Recovery Run
Fifth Recovery Run
Trace from the "lady gaga" search
Trace from the first "lady gaga" search recovery

Description Herbert Poetzl 2011-05-11 14:43:57 CEST
Description of problem:

Epiphany crashes randomly on executing java script code.
Note that plugins are disabled to avoid flash or similar.

Version-Release number of selected component (if applicable):
GNOME Web Browser 2.30.6

How reproducible:
Happens here quite often when simply browsing but also
when recovering from a previous crash.

Steps to Reproduce:
1. Browse the Internet :)
2.
3.

Find Stack traces attached ...
Comment 1 Herbert Poetzl 2011-05-11 14:46:28 CEST
Created attachment 381 [details]
Crash on Recovery
Comment 2 Herbert Poetzl 2011-05-11 14:54:59 CEST
Created attachment 382 [details]
Second Recovery Run
Comment 3 Herbert Poetzl 2011-05-11 14:55:25 CEST
Created attachment 383 [details]
Third Recovery Run
Comment 4 Herbert Poetzl 2011-05-11 14:55:56 CEST
Created attachment 384 [details]
Fourth Recovery Run
Comment 5 Herbert Poetzl 2011-05-11 14:56:21 CEST
Created attachment 385 [details]
Fifth Recovery Run
Comment 6 Remco Rijnders 2011-08-23 08:17:00 CEST
Hi Herbert, is this still an issue? If yes, do you have an URL that consistently results in a crash that we can test with?
Comment 7 Herbert Poetzl 2011-08-23 13:40:29 CEST
(In reply to comment #6)
> Hi Herbert, is this still an issue? 

yes, still an issue, it crashes at least 5 times a day, sometimes several times in a row on recovery ...


> If yes, do you have an URL that consistently results in a crash that we can test with?

I don't think there is such thing, it seems to me like a race condition, but I can easily test and/or debug this if you tell me how/what to do because if I start browsing/opening pages it usually takes less than a minute to trigger and often takes several 'recoveries' to get it up and running again ...

I just did a test and it worked like a charm:

 - start clean epiphany
 - open www.google.com
 - search for "lady gaga"
 - middle mouse open the first 10-15 links
 - close some of the tabs while they are loading
 - do another search on google

that was the session, epiphany exited with the 'usual' segfault ...
Comment 8 Herbert Poetzl 2011-08-23 13:42:42 CEST
Created attachment 733 [details]
Trace from the "lady gaga" search
Comment 9 Herbert Poetzl 2011-08-23 13:46:24 CEST
Created attachment 734 [details]
Trace from the first "lady gaga" search recovery
Comment 10 Herbert Poetzl 2011-08-27 04:53:05 CEST
Just to confirm, still an issue ... i.e. epiphany crashing after a few pages/tabs, recovery usually takes 3-10 attempts because epiphany crashes on the recovery as well ...
Comment 11 Manuel Hiebel 2011-10-02 00:40:05 CEST
Can you reproduce with a new user (xguest for example) ?

Version: Cauldron => 1
Severity: critical => major

Comment 12 Herbert Poetzl 2011-10-02 12:32:02 CEST
(In reply to comment #11)
> Can you reproduce with a new user (xguest for example) ?

Yes, switched to xguest, took about 30 seconds to crash Epiphany and cause BugBuddy to go crazy about it ...

Couldn't capture a debug trace as gdb doesn't work in the rbash environment
Comment 13 Manuel Hiebel 2011-10-04 15:01:03 CEST
Seems nobody can reproduce this 'bug'

Whiteboard: (none) => unconfirmed

Comment 14 Herbert Poetzl 2011-10-04 21:44:24 CEST
Sounds to me like nobody is trying, because I can trigger it on at least 3 different systems here, one has been 'upgraded' from Mandriva, 2 are new installs from scratch (default install with gnome selected), and all of them show the described crash within a few seconds of loading and clicking.

I'm willing to do some debugging but last time I checked msot involved 'debug' packages didn't even contain debug symbols ...
Comment 15 Manuel Hiebel 2011-10-04 22:28:36 CEST
Indeed I can comfirm that it crash. I will check tomorrow for installing debug, but I don't know much think about that. 

As there is no maintainer of epiphany (for now), I add the three most commiter of the package.

CC: (none) => cjw, dmorganec, fundawang

Comment 16 Herbert Poetzl 2011-10-04 22:44:00 CEST
from glancing at the traces, it looks to me like webkit (webkit-1.2.7-4.mga1, webkit1.0-1.2.7-4.mga1, libwebkitgtk1.0_2-1.2.7-4.mga1) might be to blame, in general, I presume that the java script code is either compiled incorrectly or overwritten/freed before executed which in turn leads to the crashes at dynamic addresses .. but that's just me guessing here ...
Comment 17 Reinout van Schouwen 2011-10-16 15:44:56 CEST
(In reply to comment #16)
> from glancing at the traces, it looks to me like webkit (webkit-1.2.7-4.mga1,
> webkit1.0-1.2.7-4.mga1, libwebkitgtk1.0_2-1.2.7-4.mga1) might be to blame, 

I agree. The webkitgtk version in Mageia 1 is just really outdated (the most recent stable version is 1.6.1). I'm afraid there's no other option than to live with the situation until the next Mageia release that will contain the latest Epiphany and WebkitGTK unless someone backports Gnome 3.2 to Mageia 1.

CC: (none) => reinout

Comment 18 Marja Van Waes 2011-12-11 15:27:39 CET
(In reply to comment #17)
> (In reply to comment #16)
> > from glancing at the traces, it looks to me like webkit (webkit-1.2.7-4.mga1,
> > webkit1.0-1.2.7-4.mga1, libwebkitgtk1.0_2-1.2.7-4.mga1) might be to blame, 
> 
> I agree. The webkitgtk version in Mageia 1 is just really outdated (the most
> recent stable version is 1.6.1). I'm afraid there's no other option than to
> live with the situation until the next Mageia release that will contain the
> latest Epiphany and WebkitGTK unless someone backports Gnome 3.2 to Mageia 1.

removing "unconfirmed" because it was confirmed by Manuel.

Adding two cauldron committers of webkit to the cc of this bug, who weren't in it yet

CC: (none) => jani.valimaa, mageia, marja11
Source RPM: epiphany-2.30.6-2.mga1.src.rpm => epiphany-2.30.6-2.mga1.src.rpm, webkit-1.2.7-4.mga1.src.rpm
Whiteboard: unconfirmed => (none)

Comment 19 Marja Van Waes 2012-01-29 19:43:06 CET
Sorry, Herbert /o\ 

maybe these things get fixed faster when you become a Mageia packager.

I took your silence as a message that you don't want to become a Mageia packager anymore. Did I understand correctly?

Well, the way things are developing now, there is hope there'll be a maintainer for webkit some day, look here for the statistics:

http://kamil.mageia.org.pl/MageiaStats/
Comment 20 Funda Wang 2012-01-29 19:52:44 CET
I would suggest you report this bug to https://bugs.webkit.org
Comment 21 Herbert Poetzl 2012-01-29 22:14:29 CET
(In reply to comment #19)
> Sorry, Herbert /o\ 
nothing to be sorry about on your part ... or did I miss something?

> maybe these things get fixed faster when you become a Mageia packager.

> I took your silence as a message that you don't want to become a Mageia
> packager anymore. Did I understand correctly?

well, I finally decided to step up as mageia maintainer (for some still unmaintained packages I'm using personally), but I don't think that I want to maintain webkit (no idea why such a central part doesn't have a maintainer yet, after all most drak* tools depend on webkit)
 
> Well, the way things are developing now, there is hope there'll be a maintainer
> for webkit some day, look here for the statistics:

> http://kamil.mageia.org.pl/MageiaStats/

yeah, let's hope ...
Comment 22 Marja Van Waes 2012-01-30 08:18:47 CET
(In reply to comment #21)
> (In reply to comment #19)
> > Sorry, Herbert /o\ 
> nothing to be sorry about on your part ... or did I miss something?
> 

No, I just dream of bug reports being triaged and assigned to the proper maintainer within a week after they're reported ;)

> 
> well, I finally decided to step up as mageia maintainer (for some still
> unmaintained packages I'm using personally), but I don't think that I want to
> maintain webkit (no idea why such a central part doesn't have a maintainer yet,
> after all most drak* tools depend on webkit)

Great, welcome Herbert!
Please add your name etc. here:
https://wiki.mageia.org/en/Becoming_a_Mageia_Packager#Packaging_apprentice_candidates



(In reply to comment #20)
> I would suggest you report this bug to https://bugs.webkit.org

Herbert, do you mind looking upstream to see whether there's already a bug report for this issue and if not, file one?

Please give a link to the report once you're done. Thanks :)

Keywords: (none) => UPSTREAM

Comment 23 Herbert Poetzl 2012-01-30 08:49:22 CET
(In reply to comment #22)

> No, I just dream of bug reports being triaged and assigned to the proper
> maintainer within a week after they're reported ;)

sounds too good to be true :)

> > well, I finally decided to step up as mageia maintainer (for some still
> > unmaintained packages I'm using personally), but I don't think that I want to
> > maintain webkit (no idea why such a central part doesn't have a maintainer yet,
> > after all most drak* tools depend on webkit)

> Great, welcome Herbert!
> Please add your name etc. here:
> https://wiki.mageia.org/en/Becoming_a_Mageia_Packager#Packaging_apprentice_candidates

no need I guess, I've been already mentored and started to work as packager/maintainer.
  
> (In reply to comment #20)
> > I would suggest you report this bug to https://bugs.webkit.org

> Herbert, do you mind looking upstream to see whether there's already a bug
> report for this issue and if not, file one?

Don't know if it was reported (yet) but I do not see those crashes on Cauldron (which uses a newer version of webkit), so I presume it is already fixed upstream ... 

so, either updating Mageia 1 or identifying and backporting the fix(es) from Cauldron would be the way to go here, but no chance without a maintainer ...

> Please give a link to the report once you're done. Thanks :)

will do some search when I find the time ...
Comment 24 Marja Van Waes 2012-01-30 09:09:34 CET
(In reply to comment #23)

> 
> no need I guess, I've been already mentored and started to work as
> packager/maintainer.
> 

Totally missed that, but even better so :)

I see, you already maintain 6 packages including util-vserver now :)

(I think I mixed up some vserver packages in the past, if you see me mixing up names again, please correct me. I won't be offended, I even mix up the names of my children with those of my brothers, so I know I do such things)



> 
> Don't know if it was reported (yet) but I do not see those crashes on Cauldron
> (which uses a newer version of webkit), so I presume it is already fixed
> upstream ... 
> 
> so, either updating Mageia 1 or identifying and backporting the fix(es) from
> Cauldron would be the way to go here, but no chance without a maintainer ...
> 
> > Please give a link to the report once you're done. Thanks :)
> 
> will do some search when I find the time ...

Thanks :)
Comment 25 Funda Wang 2012-01-30 09:37:49 CET
(In reply to comment #23)
> Don't know if it was reported (yet) but I do not see those crashes on Cauldron
> (which uses a newer version of webkit), so I presume it is already fixed
> upstream ... 
> 
> so, either updating Mageia 1 or identifying and backporting the fix(es) from
> Cauldron would be the way to go here, but no chance without a maintainer ...
webkitgtk updated its libmjaor since 1.4.0 (which is latest stable version of 1.2.7), backporting it will require rebuilding all the packages depending it, which is not suggested.
Comment 26 Marja Van Waes 2012-03-02 15:36:12 CET
(In reply to comment #25)
> (In reply to comment #23)
> > Don't know if it was reported (yet) but I do not see those crashes on Cauldron
> > (which uses a newer version of webkit), so I presume it is already fixed
> > upstream ... 
> > 
> > so, either updating Mageia 1 or identifying and backporting the fix(es) from
> > Cauldron would be the way to go here, but no chance without a maintainer ...
> webkitgtk updated its libmjaor since 1.4.0 (which is latest stable version of
> 1.2.7), backporting it will require rebuilding all the packages depending it,
> which is not suggested.

@ Funda

In other words: "wontfix" ?
Comment 27 Funda Wang 2012-03-03 05:42:00 CET
(In reply to comment #26)
> @ Funda
> 
> In other words: "wontfix" ?
Yes, it is difficult to fix, as there is no code branch upstream at svn.webkit.org
Comment 28 Marja Van Waes 2012-03-03 06:31:26 CET
(In reply to comment #27)
> (In reply to comment #26)
> > @ Funda
> > 
> > In other words: "wontfix" ?
> Yes, it is difficult to fix, as there is no code branch upstream at
> svn.webkit.org

OK, thx for replying, Funda

Closing

Status: NEW => RESOLVED
Resolution: (none) => WONTFIX