Crash: (ga-client.c:237):ga_client_start: assertion failed: (priv->poll == NULL)

[Adam Buchbinder]

There is an interaction between epiphany-browser and avahi-daemon; epiphany will occasionally close for no reason with the message:

** ERROR:(ga-client.c:237):ga_client_start: assertion failed: (priv->poll == NULL)

Disabling avahi (/etc/init.d/avahi-daemon stop) appears to be a viable workaround, but I've only been running it for a day or so, so I don't know just yet. Upstream reports that the problem lies in how epiphany interfaces with avahi, so I'm filing this against epiphany, though it's been reported to both bugtrackers.

[Adam Buchbinder (adam-buchbinder)]

I am running avahi-daemon 0.6.22-2ubuntu4 and epiphany-browser 2.22.2-0ubuntu0.8.04.5 on Ubuntu Hardy. It may be possible to cherry-pick a patch from upstream SVN:

http://svn.gnome.org/viewvc/epiphany/trunk/src/bookmarks/ephy-bookmarks.c?r1=8270&r2=8317

I should also add that I use epiphany with -p; I don't know if that's required to reproduce the issue.

[Adam Buchbinder (adam-buchbinder)]

Pardon me; a better link for that patch is as follows.

http://svn.gnome.org/viewvc/epiphany?view=revision&revision=8317

[Sebastien Bacher (seb128)]

the issue is fixed in intrepid

[Sebastien Bacher (seb128)]

can you confirm that applying this change fix the issue on hardy?

[Adam Buchbinder (adam-buchbinder)]

I've built a patched package; I'll test it out when I get back to the affected machine.

How do you know it's fixed in Intrepid? Is the patch applied in the version there? I've only run into this on a laptop with a wireless card, as changes to the network seem to make Avahi do *something*; if epiphany runs smoothly for you, that doesn't necessarily mean that the problem is fixed.

[Sebastien Bacher (seb128)]

upstream rolled a new tarball this week which is in intrepid so if the change was in svn it should be in intrepid now

[Adam Buchbinder (adam-buchbinder)]

Ah, Sebastian, I understand. Thanks.

The patched version has been running without this crash for the better part of the day now, though, as I said, the bug is intermittent and difficult to trigger. (I've had crashes, but those were all bug 256495.) I'll continue using epiphany for everyday browsing (I'm running all instances under gdb now), and report back immediately if I see a crash in ga_client_start. Given that the crash happened at least once daily, I'll consider it a fix if the browser doesn't die (from this bug) by tonight. Does that sound reasonable? Should I set up a PPA for further testing? I haven't done it before, but it looks reasonably straightforward.

[Adam Buchbinder (adam-buchbinder)]

I had another crash; backtrace as follows:

#0 0xb7f9a424 in __kernel_vsyscall ()
#1 0xb721ed05 in sem_wait@@GLIBC_2.1 () from /lib/tls/i686/cmov/libpthread.so.0
#2 0xb05601a8 in ?? () from /usr/lib/flashplugin-nonfree/libflashplayer.so
#3 0xb72184fb in start_thread () from /lib/tls/i686/cmov/libpthread.so.0
#4 0xb6f7ae5e in clone () from /lib/tls/i686/cmov/libc.so.6

I think this was a Flash thing (again); the stack trace looks like the one in bug 173845, though I'm not running over LTSP. I'm still considering the patch good.