Re: dist-upgrade broke Opera
Chris McCraw
>Did a dist-upgrade today and now Opera won't start.
>
>ERROR: ld.so: object 'libjvm.so' from LD_PRELOAD cannot be preloaded: ignored.
>ERROR: ld.so: object 'libawt.so' from LD_PRELOAD cannot be preloaded: ignored.
>./opera: line 325: 20749 Segmentation fault "${OPERA_BINARYDIR}opera" "$@"
>
>Tried the static versions of
>Opera 9.10
>Opera beta 9.20 build 617
>Opera weekly beta 9.20 build 631
>and all fail to start
>
>It looks like the culprit is an upgrade to libx11-6 from 1.0.3-6 to
>1.0.3-7, but I
>could easily be wrong.
I think you're right--we build X11R6.9 from source here and applied the
(essential!) security patch in line with this update and no matter which
version of opera we try we just get a segfault. Unfortunately we get
nothing useful in the gdb backtrace either :/
This is on an ubuntu system (dapper drake) with X11R6.9.
Answers eagerly sought! We'll have 30 or so people without opera tomorrow
morning when the security patch is spread around to the desktops...
Balaam's Miracle
wrote:
>> It looks like the culprit is an upgrade to libx11-6 from 1.0.3-6 to
>> 1.0.3-7, but I
>> could easily be wrong.
>
> I think you're right--we build X11R6.9 from source here and applied the
> (essential!) security patch in line with this update and no matter which
> version of opera we try we just get a segfault. Unfortunately we get
> nothing useful in the gdb backtrace either :/
>
> This is on an ubuntu system (dapper drake) with X11R6.9.
I've no solution or answers, but i'm running Feisty Faun (for the
uninitiated, this is Ubuntu 7.04, Dapper is 6.06) in VMWare and the
segfault problem exists here as well.
The version of libx11-6 is 2:1.1.1ubuntu3 as reported by Synaptic.
This makes me worry about something: Ubuntu Feisty Faun is very close to
its releasedate (april 19th) which will be using X11R7.2. This could mean
that if this is not solved, then a lot of people might be locked out of
Opera after 4/19 since they will be upgrading to the new Ubuntu release.
Hopefully, this will be resolved soon.
Knowing Opera, it will be solved ASAP, but it never hurts to voice my
worries... :-)
--
Balaam's Miracle
Jack Malmostoso
> It looks like the culprit is an upgrade to libx11-6 from 1.0.3-6 to
> 1.0.3-7, but I could easily be wrong.
Confirmed here on a Debian Sid system, both static and dynamic versions.
I'll try to downgrade to see if it helps.
--
Best Regards, Jack
Linux User #264449
Powered by Debian GNU/Linux on AMD64
Jack Malmostoso
> I'll try to downgrade to see if it helps.
Confirmed. libx11-6_1.0.3-6 works just fine.
Rijk van Geijtenbeek
> On Wed, 04 Apr 2007 13:57:55 -0500, Vince <nob...@home.invalid> wrote:
>
>> Tried the static versions of
>> Opera 9.10
>> Opera beta 9.20 build 617
>> Opera weekly beta 9.20 build 631
>> and all fail to start
>
> Is anybody from Opera paying attention?
Oh yes. A dozen duplicate bug reports have been received already by us,
and then there a Forum thread:
http://my.opera.com/community/forums/topic.dml?id=183923
... and mentions in the Desktopteam blog.
This has also been reported by users to Debian and Ubuntu:
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=417816
https://bugs.launchpad.net/ubuntu/+source/libx11/+bug/102851
> I don't expect a fix within one business day,
Good... problem is, first business day in Norway will be Tuesday. They
have lots of days off around Easter, starting wednesday afternoon. I hope
and think there will be progress before that though.
> but some sort of response saying you're aware of the problem and are
> working on a fix would be a common courtesy.
>
> Meanwhile this is just the tip of the iceberg, as bleeding edge distros
> trickle this library down to the more stable distros.
>
> Using Debian Sid here and have seen where Etch users are already
> affected. And using FireFox is torture when you're used to Opera.
>
> <whine /> I want my Opera.
--
Rijk van Geijtenbeek
Opera Software ASA, Documentation & QA
Tweak: http://my.opera.com/Rijk/blog/
Whiskers
> On Thu, 05 Apr 2007 22:16:56 +0200, "Rijk van Geijtenbeek"
> <ri...@opera-dot-com.invalid> wrote:
>
>>Good... problem is, first business day in Norway will be Tuesday. They
>>have lots of days off around Easter, starting wednesday afternoon. I hope
>>and think there will be progress before that though.
Thanks for the feed-back anyway :))
> Oy! I had no idea.
>
> Almost a week of FireFox??!!
> Think I'm going to break down and cry.
>
> Thanks for letting us know what's going on at Opera.
I'm giving Galeon a go. And there's always Lynx.
--
-- ^^^^^^^^^^
-- Whiskers
-- ~~~~~~~~~~
Jens Schüßler
> On Thu, 5 Apr 2007 22:04:28 +0100, Whiskers <catwh...@operamail.com> wrote:
>
>>I'm giving Galeon a go. And there's always Lynx.
>
> <http://my.opera.com/desktopteam/blog/>
Works on Debian Etch too.
Thanks for the quick response to the Opera-team!
Jens
Rijk van Geijtenbeek
> Did a dist-upgrade today and now Opera won't start.
..
> It looks like the culprit is an upgrade to libx11-6 from 1.0.3-6 to
> 1.0.3-7, but I could easily be wrong.
>
> Would much prefer to NOT downgrade to libx11-6 1.0.3-6.
> Is there anything I can do to get Opera working again?
You've probably all read this already, but yesterday a fixed build 633 was
made available:
http://my.opera.com/desktopteam/blog/2007/04/06/hotfix
Three cheers for Axel!
See https://bugs.freedesktop.org/show_bug.cgi?id=10536 for some
behind-the-scenes insights.
--
Rijk
Opera Software, QA etc...
Whiskers
I installed the hotfix on my Mandriva system this evening, and it works
:))
I'm also not unfavourably impressed by the new 'speed dial' feature; I can
see myself removing a few things from the 'Personal Bar'.
Remco Lanting
> On Thu, 5 Apr 2007 22:04:28 +0100, Whiskers <catwh...@operamail.com>
> wrote:
>
>> I'm giving Galeon a go. And there's always Lynx.
>
> There's a new Opera weekly that works.
> <http://my.opera.com/desktopteam/blog/>
>
>
> Thank you very much Opera.
>
You can also get it working on 9.10. This was just added to the
desktopteam blog:
"If you do not want to install a weekly release but continue running
version 9.1, you can change the 'DefaultDepth' option to 16 in your
/etc/X11/xorg.conf file, as this will also work around the problem."
--
Remco Lanting
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/
borda
>
> <http://my.opera.com/desktopteam/blog/>
>
>
> Thank you very much Opera.
>
after the upgrade to the weekly build, opera started but crashed
random... then i set the defaultdepth to 16bit and i switched back to
opera 9.1 and now it works again, but again, it crash random... oh well,
not so random after all, it crash when i scroll a page with some strange
characters (e.g.: japanese etc, and NOT when i load up the page... only
when opera render those particular characters)
:(
Rijk van Geijtenbeek
<bordaund...@alice.it>:
Try today's release at the URL above.
borda
>
> Try today's release at the URL above.
>
> --Rijk
> Opera Software, QA etc...
a huge thanks to all the opera team :)
happy holiday, you deserve it :)
sinistra
I can hear it stirring on the HD but it will not appear into the fluxbox
window manager
I can't re-image back in time, it's too far to go, will have to wait and see
perhaps. Is this newest problem with Opera and Debian testing happened to
others?
I need it back.
--
ne di sirint
S.
Eirik Byrkjeflot Anonsen
If you try to start it from a command line, do you get any error
messages?
eirik
sinistra
>If you try to start it from a command line, do you get >any error messages?
Thanks for your letter, this is the message:
opera: xcb_xlib.c:52 xcb_xlib_unlock: Assertion
'c->xlib.lock' failed.Aborted
What to do?
--
S.
Eirik Byrkjeflot Anonsen
Ouch! This is a bug in xlib. Some people have been adding a new
interface to X with some improvements over the old xlib stuff. The
error message above says that the sequence of calls into xcb (the new
interface) is wrong. Opera doesn't use xcb directly, so this is
almost certainly a bug in xlib.
In the long term, this should be fixed by upgrading the x libraries to
a version without this bug. In the short term, I don't know what the
solution is.
eirik
sinistra
Alright Eirik thanks for your help, I'll file a bug report and send along
your analysis to the xlib maintainer.
--
S.
sinistra
news:87wszuw...@opera.com:
> Ouch! This is a bug in xlib.
I received a letter today from Branden Robinson who was involved in X11
but less so now..he said:
As I understand it, the XCB-linked Xlib has not yet propagated into
unstable, so you must be running experimental libx11-6 packages.
Then I kindly received another letter from him saying:
I am informed by Michel Dänzer of the X Strike Force that there is a
workaround for this problem:
LIBXCB_SLOPPY_LOCK=1
This is a process environment setting, so you'll want to use "export" or
"setenv" or the appropriate command for your shell to get it into the
environment.
I post this should it help anyone on the edge with Debian Lenny.
--
S.
sinistra
Forward from Josh Triplett for the record: Bug#421854: Not a bug in libx11
Hello,
Despite previous comments forwarded to this bug report from
Eirik Byrkjeflot Anonsen, this bug does not lie in libx11 or libxcb.
The reported assertion means that some caller of libx11 has broken locking;
specifically, it calls UnlockDisplay without calling LockDisplay. (Note that
the functions XLockDisplay and XUnlockDisplay nest, but LockDisplay and
UnlockDisplay do not.)
If opera does not call libx11 directly, the bug likely lies in a library
between opera and libx11, such as a toolkit or an X extension library.
Please try to reproduce this problem under gdb, and get a backtrace. This
will help us track down where the broken locking exists. In the meantime,
since no opera package exists in Debian to reassign to, I'll leave this bug
open on libx11 for lack of somewhere better to track it.
Please see
<http://lists.debian.org/debian-devel-announce/2006/11/msg00010.html> for more
information.
- Josh Triplett
--
s
Eirik Byrkjeflot Anonsen
I was sure I'd read on the xorg list that these were typically errors
in xlib that got caught by xcb. Ah well, in that case, I would assume
it is qt that does it. While we do some direct X calls, I doubt we do
any locking ourselves.
I'll reply to the bug.
eirik
Eirik Byrkjeflot Anonsen
> Forward from Josh Triplett for the record: Bug#421854: Not a bug in libx11
>
> Please try to reproduce this problem under gdb, and get a backtrace. This
> will help us track down where the broken locking exists.
Oh, and that would be a good idea. Any chance you can do that?
eirik
sinistra
Highly unlikely Eirik right now everything on 9.21.641 with Debian Lenny is
the one stable thing in a changing universe.
--
s.