[MASTER] firefox-3.0b5 received an X Window System error: 'BadIDChoice'

[U. B.]

Binary package hint: firefox-3.0

Starting a few days ago, maybe a week or so, FF3 started crashing randomly,
after 1h or so, usually when there are tens of windows and tabs open. It
probably happened after some update of the X server, come to think of it.

Calling it from a terminal and not from the window manager yields the following
trace:

The program 'firefox' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 73442980 error_code 14 request_code 153 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)

I am not sure what to do with that? Is firefox the program I should run with
the --sync command line option?

Relevant packages (hardy up-to-date):
ii xserver-xorg 1:7.3+10ubuntu10 the X.Org X server
ii xserver-xorg-core 2:1.4.1~git20080131-1ubuntu9 Xorg X server - core server
ii kdm-kde4 4:4.0.3-0ubuntu2 X display manager for KDE 4
ii firefox-3.0 3.0~b5+nobinonly-0ubuntu3 safe and easy web browser from Mozilla

Tell me what to do to help you investigate this issue.

[U. B. (u710bts)]

Update. No core file is produced, and running firefox under gdb gives nothing. The "apt-cache search firefox debug" command does not show any package related to firefox-3.
The problem reproduces fairly regularly. Example of a gdb session:

$ ulimit -c
10000
$ gdb
[...]
(gdb) file /usr/lib/firefox-3.0b5/firefox
Reading symbols from /usr/lib/firefox-3.0b5/firefox...(no debugging symbols found)...done.
(gdb) break gdk_x_error
Function "gdk_x_error" not defined.
Make breakpoint pending on future shared library load? (y or [n]) y
Breakpoint 1 (gdk_x_error) pending.
(gdb) run --sync
[...]
[New Thread 0x40c5d950 (LWP 24772)]
[Thread 0x40c5d950 (LWP 24772) exited]
[New Thread 0x40c5d950 (LWP 24773)]
The program 'firefox' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 147095966 error_code 14 request_code 153 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)

Program exited with code 01.
(gdb) quit

I'm still confused as to what I could or should do to help pin down the source of the trouble.

[Gabriel de Perthuis (g2p)]

You can install debug packages and get a backtrace by following the instructions there: https://wiki.ubuntu.com/MozillaTeam/Bugs#head-c576e78d92cb3c959c271158b6ace98be835de83
Beware: the packages are in an external repository, and unsigned :-(

[Kimbotha (r-launchpad-covil-net)]

I have this issue too.

ii firefox 3.0.1+build1+nobinonly-0ubuntu0.8.04.3 meta package for the popular mozilla web browser
ii firefox-3.0 3.0.1+build1+nobinonly-0ubuntu0.8.04.3 safe and easy web browser from Mozilla
ii kdm-kde4 4:4.0.3-0ubuntu2 X display manager for KDE 4
ii xserver-xorg 1:7.3+10ubuntu10.2 the X.Org X server
ii xserver-xorg-core 2:1.4.1~git20080131-1ubuntu9.2 Xorg X server - core server

[Kimbotha (r-launchpad-covil-net)]

The program 'firefox' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 16193363 error_code 14 request_code 55 minor_code 0)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)
Locking assertion failure. Backtrace:
#0 /usr/lib/libxcb-xlib.so.0 [0xb653f767]
#1 /usr/lib/libxcb-xlib.so.0(xcb_xlib_lock+0x2e) [0xb653f81e]
#2 /usr/lib/libX11.so.6 [0xb6acb518]
#3 /usr/lib/libX11.so.6(XFreeCursor+0x25) [0xb6aa78f5]
#4 /usr/lib/libqt-mt.so.3(_ZN11QCursorDataD1Ev+0x3f) [0xb4ee3389]
#5 /usr/lib/libqt-mt.so.3(_ZN7QCursorD1Ev+0x5a) [0xb4ee358e]
#6 /usr/lib/libqt-mt.so.3 [0xb4ee35d5]
#7 /lib/tls/i686/cmov/libc.so.6(exit+0xd4) [0xb7d0c084]
#8 /usr/lib/libgdk-x11-2.0.so.0 [0xb66b1637]
#9 /usr/lib/libbonoboui-2.so.0 [0xb62ca665]
#10 /usr/lib/libX11.so.6(_XError+0xfe) [0xb6ac473e]
#11 /usr/lib/libX11.so.6 [0xb6acbe5c]
#12 /usr/lib/libX11.so.6(_XEventsQueued+0x4f) [0xb6acc71f]
#13 /usr/lib/libX11.so.6(XPending+0x62) [0xb6ab59d2]
#14 /usr/lib/libgdk-x11-2.0.so.0 [0xb66a604d]
#15 /usr/lib/libgdk-x11-2.0.so.0 [0xb66a6141]
#16 /usr/lib/libglib-2.0.so.0(g_main_context_prepare+0x190) [0xb6bb53e0]
#17 /usr/lib/libglib-2.0.so.0 [0xb6bb5b09]
#18 /usr/lib/libglib-2.0.so.0(g_main_context_iteration+0x6c) [0xb6bb63ac]
#19 /usr/lib/xulrunner-1.9.0.1/libxul.so [0xb782287c]

[a-r-k-i-b-o-t-t (arkibott-ray)]

(firefox:16185): GLib-GObject-CRITICAL **: g_object_unref: assertion `G_IS_OBJECT (object)' failed

(firefox:16185): Gtk-CRITICAL **: gtk_drag_set_icon_pixbuf: assertion `GDK_IS_DRAG_CONTEXT (context)' failed
The program 'firefox' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 183275268 error_code 14 request_code 154 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)
Segmentation fault

This is also with 8.04.1 LTS Ubuntu and the up to date Packages. Firefox3.0.1 etc.

Only ONE Window will be restored after a new start. Other Windows and Tabs are lost.

[Alexander Sack (asac)]

Do you use firefox in kde?

On Tue, Jul 29, 2008 at 08:47:27PM -0000, Kimbotha wrote:
> The program 'firefox' received an X Window System error.
> This probably reflects a bug in the program.
> The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
> (Details: serial 16193363 error_code 14 request_code 55 minor_code 0)
> (Note to programmers: normally, X errors are reported asynchronously;
> that is, you will receive the error a while after causing it.
> To debug your program, run it with the --sync command line
> option to change this behavior. You can then get a meaningful
> backtrace from your debugger if you break on the gdk_x_error() function.)
> Locking assertion failure. Backtrace:
> #0 /usr/lib/libxcb-xlib.so.0 [0xb653f767]
> #1 /usr/lib/libxcb-xlib.so.0(xcb_xlib_lock+0x2e) [0xb653f81e]
> #2 /usr/lib/libX11.so.6 [0xb6acb518]
> #3 /usr/lib/libX11.so.6(XFreeCursor+0x25) [0xb6aa78f5]
> #4 /usr/lib/libqt-mt.so.3(_ZN11QCursorDataD1Ev+0x3f) [0xb4ee3389]
> #5 /usr/lib/libqt-mt.so.3(_ZN7QCursorD1Ev+0x5a) [0xb4ee358e]
> #6 /usr/lib/libqt-mt.so.3 [0xb4ee35d5]
> #7 /lib/tls/i686/cmov/libc.so.6(exit+0xd4) [0xb7d0c084]
> #8 /usr/lib/libgdk-x11-2.0.so.0 [0xb66b1637]
> #9 /usr/lib/libbonoboui-2.so.0 [0xb62ca665]
> #10 /usr/lib/libX11.so.6(_XError+0xfe) [0xb6ac473e]
> #11 /usr/lib/libX11.so.6 [0xb6acbe5c]
> #12 /usr/lib/libX11.so.6(_XEventsQueued+0x4f) [0xb6acc71f]
> #13 /usr/lib/libX11.so.6(XPending+0x62) [0xb6ab59d2]
> #14 /usr/lib/libgdk-x11-2.0.so.0 [0xb66a604d]
> #15 /usr/lib/libgdk-x11-2.0.so.0 [0xb66a6141]
> #16 /usr/lib/libglib-2.0.so.0(g_main_context_prepare+0x190) [0xb6bb53e0]
> #17 /usr/lib/libglib-2.0.so.0 [0xb6bb5b09]
> #18 /usr/lib/libglib-2.0.so.0(g_main_context_iteration+0x6c) [0xb6bb63ac]
> #19 /usr/lib/xulrunner-1.9.0.1/libxul.so [0xb782287c]
>

 - Alexander

[Kimbotha (r-launchpad-covil-net)]

Hi Alexander,

Yes, I run Firefox on Kubuntu using KDE 3.

Regards,

Kim

[a-r-k-i-b-o-t-t (arkibott-ray)]

I see this error using gnome 2.22.3.

As the Error Message suggests, i tried running firefox with the --sync parameter. It still crashes randomly with it.

A funny detail: Sometimes the icon of a bookmark shows the wrong icon. I thought this was related to this X --sync problem, but actually its another random glitch.. but not so serious.

[L. David Baron (dbaron)]

This may be the same as bug 454677, but it started happening quite often to me only in the last few days, so I suspect it's different.

In the past four days, I've crashed three times with a BadIdChoice X error:
2008-09-28 12:42:13
2008-09-29 18:47:44
2008-10-01 10:39:24

The first two times the error looked like this:

The program 'firefox-bin' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 158749545 error_code 14 request_code 155 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)

(different serial the second time)

But the third time it gave me more information:

The program 'firefox-bin' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 150818044 error_code 14 request_code 155 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)
Locking assertion failure. Backtrace:
#0 /usr/lib/libxcb-xlib.so.0 [0x7f833992297c]
#1 /usr/lib/libxcb-xlib.so.0(xcb_xlib_lock+0x15) [0x7f8339922a15]
#2 /usr/lib/libX11.so.6 [0x7f833d7e4323]
#3 /usr/lib/libXrender.so.1(XRenderFreePicture+0x46) [0x7f833abd87c6]
#4 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/libthebes.so [0x7f83327bd8f4]
#5 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/libthebes.so(_moz_cairo_surface_finish+0x3b) [0x7f83327a51cb]
#6 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/libthebes.so(_moz_cairo_surface_destroy+0x58) [0x7f83327a5248]
#7 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/libthebes.so(_ZN11gfxASurface7ReleaseEv+0x36) [0x7f83327676c6]
#8 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/components/libgkgfxthebes.so [0x7f832eed9c6f]
#9 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/components/libgkgfxthebes.so [0x7f832eeda2ba]
#10 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/components/libgkgfxthebes.so [0x7f832eee5544]
#11 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/components/libgkgfxthebes.so [0x7f832eee584f]
#12 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/libxpcom_core.so(_ZN15nsCOMArray_baseD2Ev+0x3b) [0x7f833fd9b0eb]
#13 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/components/libimglib2.so [0x7f832f30ed41]
#14 /home/dbaron/builds/mozilla-central/obj/firefox-debugopt/dist/bin/components/libimglib2.so [0x7f832f30cccf]
#15 /home/dbaron...

[L. David Baron (dbaron)]

Created attachment 341309
fixed version of stack, with line numbers

[Jmuizelaar (jmuizelaar)]

Was the stack trace from running with --sync?

[L. David Baron (dbaron)]

This happens about once a day, seemingly randomly, so I can't just rerun with --sync and make it crash again.

[Vladimir Vukicevic (vvuk)]

Likely not -- the minor number corresponds to RenderCreatePicture, not FreePicture. Might have to dig into the X code to see what generates BadID; owen was suggesting that it might be due due to IDs getting out of sync somehow:

13:03 < otaylor> vlad_: Trying to create two resources with the ID of the
                 second less than the ID of the first would cause that

But I have no idea how we'd get into that situation, unless the IDs wrapped around?

[L. David Baron (dbaron)]

Created attachment 341701
stack with --sync

[L. David Baron (dbaron)]

Created attachment 341790
second stack with --sync

This one has better symbols and is quite different from the first.

[Vladimir Vukicevic (vvuk)]

Assigning blocking+ and over to me until we figure out what's going on.

[L. David Baron (dbaron)]

Created attachment 342157
third stack with --sync

I've now installed better debug symbols for libcairo.so.2, so I should get a better stack next time.

[L. David Baron (dbaron)]

Created attachment 342987
fourth stack with --sync

Some of the Mozilla filenames / line numbers might be off, since I'd rebuilt some things since starting.

[U. B. (u710bts)]

It keeps failing randomly with 3.0.3+build1+nobinonly-0ubuntu0.8.04.1.
I guess this happens when there are many windows/tabs open and the computer is short on memory, but it's hard to tell.
I would test more if I got explicit and comprehensive steps, involving nothing but official packages.

[Vladimir Vukicevic (vvuk)]

*** Bug 454678 has been marked as a duplicate of this bug. ***

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

(In reply to comment #10)
> *** Bug 454678 has been marked as a duplicate of this bug. ***

Note that unlike the crashes mentioned in both this bug and bug 454677 (which have request_code 155 minor_code 4), bug 454678 has request_code 53 minor_code 0. Since the day when I reported bugs 454677 and 454678 (and actually maybe a week or more before), I'm getting them approximately once or twice a day (14/155/4 much more often than 14/53/0), and only after SeaMonkey has run some hours since last restart. I suspect bug 454677 is due to the same cause as this one but I have doubts about bug 454678. We'll see when (and if) this one gets fixed.

Currently running

Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1b2pre) Gecko/20081020 SeaMonkey/2.0a2pre - Build ID: 20081020000453

and I've had a case of this problem (14/155/4) on it. I'm keeping sysout/syserr logs (available on request) but I don't know how to catch a usable stack trace with gdb. (Breakpad is never triggered by these crashes.)

[Vladimir Vukicevic (vvuk)]

The requests will be different -- it's being triggered by anything that allocates an ID. dbaron has some good stack traces showing that happening inside this bug, but from reading the X source I don't know what could be causing it.

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

*** Bug 454677 has been marked as a duplicate of this bug. ***

[a-r-k-i-b-o-t-t (arkibott-ray)]

Is this a GTK bug?
Recently I got a similar error with thunderbird.

The Mozilla guys told me it's a GTK thing.. but I wonder why it only occurs with Mozilla programs.

[Alexander Sack (asac)]

most likely a gtk qt engine thing. have you tested if this still happens on kde4?

[Vladimir Vukicevic (vvuk)]

Gonna take this off the blocking list for now; I haven't been able to reproduce it in a VM, and David says he hasn't seen it for a while. From reading the X source and talking to X developers, the two things that could cause it are either a corrupt ID making it over, or running out of XIDs on the X side. The latter would be odd, because we'd probably have run out of memory well before we got to running out of XIDs. The former is possible, but again odd..

Useful information to get if it happens again would be the ID in _XIDHandler, and xrestop information.

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

(In reply to comment #14)
> Gonna take this off the blocking list for now; I haven't been able to reproduce
> it in a VM, and David says he hasn't seen it for a while. [...]

I used to get these almost every day; the latest case I see is recorded in a logfile with a datestamp of 2008-10-23 13:54 +0200.

[Karlt (karlt)]

Created attachment 347218
xrestop -b -m 1

Recorded shortly after OpenOffice received:

  X-Error: BadIDChoice (invalid resource ID chosen for this connection)
  Major opcode: 53 (X_CreatePixmap)

Then, ten minutes later the firefox process listed here (version 3.0.3)
received:

  BadIDChoice (invalid resource ID chosen for this connection)
  Details: serial 85482098 error_code 14 request_code 153

I don't see any large XID users.

Two different apps crashing the same way suggests that the bug is at a low
level. I have xorg-server-1.4.2, libxcb-1.0, libX11-1.1.4, cairo-1.6.4-r1.

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

(In reply to comment #16)
[...]
> Two different apps crashing the same way suggests that the bug is at a low
> level. I have xorg-server-1.4.2, libxcb-1.0, libX11-1.1.4, cairo-1.6.4-r1.

I'm on openSUSE 11.0, currently with:
xorg-x11-7.3-96.2
xorg-x11-server-7.3-110.9
xorg-x11-libxcb-7.3-48.1
xorg-x11-libX11-7.3-56.1
cairo-1.4.14-32.1

I just checked my SeaMonkey saved logs (20 of them, the oldest one dated October 24). Neither of the strings "BadIdChoice" and "X Window" happens in any of them, which means the bug (and its dupes bug 454677 and bug 454678, which I reported) have disappeared some time ago on my system. I haven't particularly noted if and when any "SuSE Online Updates" were applied to any Xorg-X11 packages, but I check at least once a day for any available upgrades (and apply them if there are).

However I had a "BadWindow (invalid window parameter)" X-error (error 3 request 20 minor 0) yesterday in BonEcho (Firefox 2 nightly). At the moment, this is an isolated case.

My current builds:
-- SeaMonkey:
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.1b2pre) Gecko/20081112 SeaMonkey/2.0a2pre - Build ID: 20081112033648
-- Firefox:
Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.1.19pre) Gecko/20081112 BonEcho/2.0.0.19pre - Build ID: 2008111203

[Karlt (karlt)]

FWIW, same issue with
firefox-3.0.4, xorg-server-1.5.2, libxcb-1.1.90.1, libX11-1.1.5.

[L. David Baron (dbaron)]

I haven't seen this for quite a while... perhaps it was fixed when I upgraded to Ubuntu 8.10 from 8.04.

[glepore70 (greg-rhobard)]

I just started getting this bug on KDE4.2 Beta 2. The problem occurs in Firefox when copying and pasting in an input box or the location bar. Selected style is QTCurve, although I believe it also happened with Oxygen style. GTK Styles is set to Use my KDE style.

[lessoffensive (lessoffensive)]

After a large amount of testing and finally correlating logs, I determined this to be caused by installing version 1.3.10a of the flashblock extension though apt. I use amd64, kubuntu intrepid, Kde 4.1 (and 4.2), nvidia drivers, and QtCurve. I think the last two looked like they may have been related, but probably aren't.

1.3.10a in the repositories is too old; I switched to version 1.5.7.1 from addons.mozilla.org and my problems have finally resolved.

Lesson: don't install Firefox extensions from apt.

[Thomas Riccardi (kiwiiii)]

I have the same issue on debian lenny 64bits with firefox (iceweasel) 3.0.4-1.
I also use the flashblock extension, but with the last version 1.5.7.1, and I have the issue...
So lessoffensive, flashblock version 1.5.7.1 does not solve the issue in every case. (I also use nvidia drivers from nvidia.com)

I tried without flashplugin-nonfree and without flashblock, but it crashes again.

[lessoffensive (lessoffensive)]

Thomas, you're right, flashblock wasn't it. It's strange though that it worked for me for about two weeks with that new upgraded version but now I have the same crash again. Are you using KDE 4.X? What may be involved is that I just changed nvidia drivers last night and it's started crashing since. I'm using the nvidia 180.22 drivers, upgraded from 180.16.

I still haven't ruled out QtCurve though, so I'm upgrading to the latest (0.60 from the notz ppa) and will see how that works. I'm shocked so few people have this problem and it's persisted as long as it has being as totally intermittent as it has been, as well.

It also seems to only happen when I get enough tabs to have to scroll through them on the tab bar on the window (the last crash was 2 windows, 140 tabs); that doesn't seem to cause the crash, but it never seems to be able to get off the ground after it either crashes unrelatedly or being closed.

I'm also using an nvidia geforce 8600 GT. What card and drivers are you using? I've had problems with the 8XXX series and the nvidia drivers for about the last year--maybe the bug is somewhere in the combination of bad driver version(s) and 8XXX series of cards.

[Thomas Riccardi (kiwiiii)]

lessoffensive, it is probably an issue related to the nvidia proprietary driver.
I use gnome 2.22 so it's probably not KDE related.
I once thought it was related to x86_64 : I made a chroot yesterday and used firefox 3.04 i686, and it also crashed... So it's probably neither 64b.
For information I use firefox with ~10 windows and a total of ~120 tabs.

I have a 9800GT (which is in fact a 8800GT renamed), and currently use the beta drivers 180.18.
As your firefox crashes with the new 180.22 and not with 180.16, I'll try to use 180.16 for a while to see if it crashes or not.

So now I don't really know what to do, maybe a bugreport to nvidia?

[lessoffensive (lessoffensive)]

I tested out the 180.16 drivers again and it crashed without effort. I also just tested firefox in safe mode and it crashed as well.

I'm very willing to believe this has to do with nvidia binary drivers and 8XXX series cards as I've only had continued, unaddressed problems with the combination over the past year. Looking over my install logs again it also occurs to me that it could be possible that it's my xorg server version. I'm running 1.5.2(-2ubuntu3) of xserver-xorg-core, which I upgraded to in mid october, which is around when these problems started occurring (a dramatic shift from firefox never crashing!). I think I'm going to try a 1.5.3 release from the xorg-edgers ppa and see if that fixes anything. Other than that, it seems like filing a bug report, buying a new card, or upgrading to jaunty are among my next options.

I'll keep you posted.

[zen0 (a-neuron)]

I am seeing this also, several times per day.
I have ATI video though, using proprietary driver on amd64.

[lessoffensive (lessoffensive)]

well xorg server 1.5.3 didn't fix it. Still debating upgrading to Jaunty.

zen0: so you're getting a BadIDChoice x server error as well? With how many windows and tabs? Are you running KDE? What widget style are you using (QtCurve?)? Do you have flash installed and if so, what version? Are you running Intrepid ibex?

[Thomas Riccardi (kiwiiii)]

So as it seems to be not specific to nvidia drivers, it is probably a Xorg bug...
I'll try to run firefox in gdb, it may be useful.

The bugreport is already in the mozilla bugtrack (https://bugzilla.mozilla.org/show_bug.cgi?id=430091) but it seems to be quite inactive.

Maybe a bugreport on Xorg would be useful.

[L. David Baron (dbaron)]

Saw one again today.

[zen0 (a-neuron)]

generally 10+ windows/ 100+ tabs. plenty of ram though.
not running KDE recently, been seeing this on gnome.
appears not to be flash trouble this time... currently using the 64bit flash beta, but I have tried to no avail the various combinations of nspluginwrapper, flashblock, removing flash.
running intrepid.

[Karlt (karlt)]

*** Bug 430091 has been marked as a duplicate of this bug. ***

[Nathaniel Smith (njs)]

I see this bug (or something that looks like it!) every few days using the Ubuntu 8.10 shipped Firefox 3.0.5 ("3.0.5+nobinonly-0ubuntu0.8.10.1", ATM), on a 2-core x86-64 (xlib 1.1.5, xcb 1.1.noidea). The last crash I captured was from CreatePixmap (request code 53), not sure if I'm seeing it on other requests or not. I've never seen it with any application besides firefox. (Though firefox is presumably pushing the X library stack a lot harder than anything else I use; I don't run multi-day OOo sessions.)

Restarting firefox is about the most frustrating computing experience I can think of (cf #177175), so I'm certainly eager to help...

...however, I suspect this was tracked down already in March (though ignored until November):
http://osdir.com/ml/freedesktop.xcb/2008-04/msg00071.html
http://cgit.freedesktop.org/xorg/lib/libX11/commit/?id=cc19618d2eb3ed92a0b574aee26a7da8b4aed5d2

[Karlt (karlt)]

(In reply to comment #22)
> ...however, I suspect this was tracked down already in March (though ignored
> until November):
> http://osdir.com/ml/freedesktop.xcb/2008-04/msg00071.html
> http://cgit.freedesktop.org/xorg/lib/libX11/commit/?id=cc19618d2eb3ed92a0b574aee26a7da8b4aed5d2

Thanks for that link. It looks related but I'm not sure that it's the fix we need, because I don't think Mozilla uses XCreate* functions from more than 1 thread.

I'd be interested to hear if anyone is seeing the crash with a libX11 built without xcb. (This can be checked with "ldd /usr/lib*/libX11.so | grep xcb".)
All the stacks that I've seen here have libX11 built using libxcb.

All the stacks that I've seen involve a Create function that uses XAllocID. Usually the result of XAllocID is assigned to local variables in the Create function called "pid" as well as "req->pid" (and would be returned on success).
It could be useful to know what value either of these variables hold if this information can be obtained using a debugger after breaking on gdk_x_error.

[Nathaniel Smith (njs)]

(In reply to comment #23)
> Thanks for that link. It looks related but I'm not sure that it's the fix we
> need, because I don't think Mozilla uses XCreate* functions from more than 1
> thread.

Ah man, and that was such a good guess... but I've just double-checked with gdb, and in about 10 minutes of browsing I wasn't able to catch any calls to _XAllocID, _XAllocIDs, or xcb_generate_id from anything but the main thread, so you're right. (Unfortunately, the conditional breakpoint is a little expensive to leave running in general, though.)

[Nathaniel Smith (njs)]

Just got another crash, and this time I saved a core file from inside gdk_x_error. It's 250 MB compressed and only useful if you have Ubuntu 8.10 debugging packages available, so I'm not attaching it, but I can provide it on request.

Relevant bit of backtrace:
#5 0x00007f7d774a71ab in _XSyncFunction (dpy=0x8c4c00)
    at ../../src/Synchro.c:37
#6 0x00007f7d774b398b in _XIDHandler (dpy=0x8c4c00) at ../../src/xcb_io.c:275
#7 0x00007f7d77775f77 in XRenderCreatePicture () from /usr/lib/libXrender.so.1

So in _XIDHandler, we have dpy->xcb->next_xid == 57206503. Not an obviously invalid value.

The xid comes from xcb_generate_id, so I peeked in there at the algorithm it uses, and noticed something funny. xcb_generate_id stores its state in the dpy->xcb->connection->xid struct. At the time of the crash, some members of this struct:
  xid.base == 56623104
  xid.inc == 1
  xid.last == 57206504
  xid.max == 57206499

By my reading, it should be impossible for xid.last to exceed xid.max, yet here we are.

[Nathaniel Smith (njs)]

Asked on the xcb list:
http://lists.freedesktop.org/archives/xcb/2009-January/004247.html

[Karlt (karlt)]

Thanks, Nathaniel. Please keep that core file and copies or at least the package versions of xcb xlib and libXrender.

dpy->xcb->next_xid is most likely the xid after the one that caused the problem.
More likely the problem xid is the one in the XRenderCreatePicture frame.
Can you switch to that frame ("f 7"), please, and print req->pid and pid?

[Nathaniel Smith (njs)]

I can't seem to get at req->pid or pid easily -- it turns out I didn't have the debug info installed for libXrender when I started firefox, so the gdb attached to the firefox process can't figure out where local variables like that are hidden, and that when I start gdb on the core file, it declines to load any debug symbols whatsoever ("i shar" claims that the only library mapped is ld-linux).

However, the resourceid field in the XErrorEvent we catch is 57206502, and that should be the same as what we'd see from req->pid or pid.

[Karlt (karlt)]

(In reply to comment #25)
> The xid comes from xcb_generate_id, so I peeked in there at the algorithm it
> uses, and noticed something funny. xcb_generate_id stores its state in the
> dpy->xcb->connection->xid struct. At the time of the crash, some members of
> this struct:
> xid.base == 56623104
> xid.inc == 1
> xid.last == 57206504
> xid.max == 57206499
>
> By my reading, it should be impossible for xid.last to exceed xid.max, yet here
> we are.

(In reply to comment #26)
> Asked on the xcb list:
> http://lists.freedesktop.org/archives/xcb/2009-January/004247.html

Nice detective work. Thanks!

[Jst (jst)]

FWIW, I'm seeing this once every 2-4 days on one of my linux boxes, and I'm pretty sure I've seen it on another box as well. The box where I see it for sure is running Firefox in a VNC X server on Fedora Core 10. The other box is running Fedora Core 8, also under VNC.

[Karlt (karlt)]

Nathaniel and Barton found a theoretical situation that would cause this crash
(http://lists.freedesktop.org/archives/xcb/2009-January/004248.html) but it hasn't actually been proven that this situation is happening.
If this situation is happening, it can be detected by

* getting debuginfo packages for libxcb, and
* setting a conditional breakpoint on the line in xcb_generate_id after the
  call to xcb_xc_misc_get_xid_range_reply, with condition "range->count == 1".

(It is not necessary to run with --sync.)
If the breakpoint is hit, then the program will crash soon afterwards.
If the BadIDChoice results without hitting the breakpoint, then there is also another bug.

[Nathaniel Smith (njs)]

Ding ding ding!

Breakpoint 2, xcb_generate_id (c=0x1450640) at xcb_xid.c:51
51 c->xid.last = range->start_id;
$559 = {response_type = 1 '\001', pad0 = 0 '\0', sequence = 29345, length = 0, start_id = 41385235, count = 1}

Breakpoint 1, gdk_x_error (display=0x144fc00, error=0x7ffff91140d0) at /build/buildd/gtk+2.0-2.14.4/gdk/x11/gdkmain-x11.c:613

I finally caught the crash in action, and indeed, it occurred immediately after the first time the X server sent back a xid_range_reply with count=1.

So my crashes are definitely due to this xcb bug.

[Iterator (michael-schakulat)]

Try to set another GTK Style in your System-Settings.
Use the option "Use another Style" instead of the option "Use my KDE style in GTK appllications".
This workaround should prevent that FF crashes randomly.
You could use "Human-Clearlooks" for example.

[Alexander Sack (asac)]

so, does this happen with other engines then gtk-qt-engine?

[Karl Tomlinson (bugs+launchpad)]

It happens with other engines also. It is a bug in libxcb.

http://lists.freedesktop.org/archives/xcb/2009-January/004248.html
http://lists.freedesktop.org/archives/xcb/2009-February/004287.html

[D. Hugh Redelmeier (hugh-mimosa)]

Nathaniel: what does this suggest should be done?
Is fixing xcb_generate_id in libxcb going to fix this problem or is some fix to firefox required as well?

PS: I'm getting these same crashes in Firefox on Fedora 10.

PPS: I wondered whether this bug report is related https://bugzilla.redhat.com/show_bug.cgi?id=458657

[Nathaniel Smith (njs)]

Yes, it looks like the solution is to fix xcb_generate_id.

The relevant patch hasn't been committed upstream yet, see
  http://article.gmane.org/gmane.comp.freedesktop.xcb/4192
for discussion and the patch.

Probably the most useful thing would be for those seeing the crash to test out that patch and see if it prevents the crash (and doesn't cause any new ones!).

The RH bug indeed looks like it might be related (I really hope there aren't *two* ways of getting BadIDChoice errors!), and if the commenters there are seeing the crash so often, then their testing feedback would probably be very useful.

[D. Hugh Redelmeier (hugh-mimosa)]

This appears to be a problem with a known fix.

Symptoms: firefox and other programs crash with BadIdChoice errors.
https://bugzilla.mozilla.org/show_bug.cgi?id=458092
https://bugzilla.redhat.com/show_bug.cgi?id=458657
https://bugs.launchpad.net/ubuntu/+source/firefox-3.0/+bug/220628
https://bugs.launchpad.net/firefox/+bug/255195

It looks as if this has been tracked down.
Start of thread: http://lists.freedesktop.org/archives/xcb/2009-January/004247.html
Note that this thread crosses a month boundary so that threading in the archive doesn't get you to postings in February.
Continue here: http://lists.freedesktop.org/archives/xcb/2009-February/004287.html

The latest proposed patch is attached to this: http://lists.freedesktop.org/archives/xcb/2009-February/004326.html
except that I don't see the attachment. Perhaps scrubbed by the archiver.
This version seems to have it:
http://article.gmane.org/gmane.comp.freedesktop.xcb/4192
Subsequent messages suggest that a little polishing is still going on.

[D. Hugh Redelmeier (hugh-mimosa)]

Nathaniel: thanks for all your work on this.

As I read it, Bart's patch still needs a couple of tweaks. I hope he comes up with a revision to reflect these.

I created a bugzilla entry on the freedesktop.org bugzilla since I think that this needs to be reported against libxcb, not (just) its clients. Please add to it.

https://bugs.freedesktop.org/show_bug.cgi?id=20254

[D. Hugh Redelmeier (hugh-mimosa)]

See the libxcb project bugzilla entry https://bugs.freedesktop.org/show_bug.cgi?id=20254
See the firefox project bugzilla entry https://bugzilla.mozilla.org/show_bug.cgi?id=458092

[lessoffensive (lessoffensive)]

Hugh have you tested this patch at all? I've upgraded to Jaunty and tried building a patched version and doing a quick and dirty library replacement job, but X wasn't pleased with my libxcb.so and would simply bail without error. I haven't been testing for this failure yet on Jaunty, but I haven't encountered it yet with some moderate usage.

[Thomas Riccardi (kiwiiii)]

lessoffensive >

I patched libxcb with the patch from the ML, and now the bug is gone for more than a week !

But I have another bug, it may be not related though : firefox ends with

"firefox-bin: Fatal IO error 11 (Resource temporarily unavailable) on X server :0.0."

In attached file there is the full output.
(I don't have the time of each error message so I don't know if they all append just before the crash, or not... And I did not had such error messages before patching libxcb).

[D. Hugh Redelmeier (hugh-mimosa)]

lessoffensive:

No, I have not tested the patch.

I'm actually experiencing the problem on Fedora 10 (my main desktop). I use Ubuntu on some computers but not long enough to experience this problem.

I am also concerned that the patch isn't in a final form yet. If I remember correctly, the mailing list contains a few messages suggesting changes to it (although they probably don't have much impact).

Finally, I'm lazy: figuring out how to try the patch will take several hours of my time.

BTW, even though most of my FireFox crashes seem to be this bug, there are other problems. Firefox crashed with a segfault a few days ago.

[R. Drew Davis (drewclist)]

It happened to me now and again on firefox on Iwhat I believe to be an up to date Intrepid Ibex installation. I have no idea how to reproduce the problem. I wasn't even working in the firefox window when that window crashed. I'd had enough mysterious firefox crashes that I'd started firefox from a term window command line so I could see any messages.

The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 26264659 error_code 14 request_code 152 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)

I tend to have many tabs open in my firefox window at any given time. I also have more than a few plugins in use:

Last updated: Sat, 14 Mar 2009 17:47:46 GMT
User Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.0.7) Gecko/2009030422 Ubuntu/8.10 (intrepid) Firefox/3.0.7
Extensions (enabled: 21, disabled: 1):

    * Adblock Plus 0.7.5.4
    * AutoCopy 0.9.1
    * Autofill Forms 0.9.5.2
    * CLICK! by snapshot.jkn.com 101
    * DownloadHelper 4.2
    * EmailTheWeb.com 2.08
    * Flashblock 1.5.8
    * Image Zoom 0.3.1
    * InfoLister 0.10
    * Interclue 1.5.8.1
    * Lazarus: Form Recovery 1.0.5
    * Link Alert 0.8.2.1
    * NoScript 1.6.9.3
    * Page Title Eraser 0.7.6
    * Plain Text to Link 1.5.20080618
    * QuickNote 0.6.0.4
    * ScrapBook 1.3.3.9
    * Tab Mix Plus 0.3.6.1.080416 [disabled]
    * Ubuntu Firefox Modifications 0.6
    * URL Fixer 1.5
    * Wired-Marker 3.1.09020600
    * Zotero 1.0.9

Themes (1):

    * Default 3.0.7 [selected]

Plugins (9):

    * Default Plugin
    * Demo Print Plugin for unix/linux
    * DivX® Web Player
    * Java(TM) Plug-in 1.6.0_10-b33
    * QuickTime Plug-in 7.2.0
    * Shockwave Flash
    * Totem Web Browser Plugin 2.24.3
    * Windows Media Player Plug-in 10 (compatible; Totem)
    * Xine Plugin

Attached for what it is worth is a list of all my installed packages from dpkg -l. If there's something more I should collect if/when this crash recurs, please let me know.

Drew

[L. David Baron (dbaron)]

I rebuilt xcb with the patch from comment 34 and then restarted Firefox, but it wouldn't start due to:

firefox-bin: xcb_xid.c:51: xcb_generate_id: Assertion `c->xid.last == c->xid.max' failed.

[Nathaniel Smith (njs)]

"I just need to get a final commit together. This weekend,
honest. Thanks much for the note." -- http://article.gmane.org/gmane.comp.freedesktop.xcb/4317

[Alexander Sack (asac)]

from upstream discussion it currently seems this is a libxcb issue - closing firefox and gtk-qt-engine task for now.

[Alexander Sack (asac)]

according to https://bugzilla.mozilla.org/show_bug.cgi?id=458092#c37 there is a new patch on its way ... we should keep an eye on that.

[BlueT - Matthew Lien - 練喆明 (bluet)]

I was on google-analytics.com viewing my reports, and it crashed.

[D. Hugh Redelmeier (hugh-mimosa)]

see (and consider testing) the proposed patch
  http://thread.gmane.org/gmane.comp.freedesktop.xcb/4317/focus=4324
  http://article.gmane.org/gmane.comp.freedesktop.xcb/4324

The conversation appears to be in the mozilla bugzilla even though this would seem to be a better forum. Here again is the moz bz link:
  https://bugzilla.mozilla.org/show_bug.cgi?id=458092

[Nathaniel Smith (njs)]

Here's the proposed patch, and he's looking for testers:
http://thread.gmane.org/gmane.comp.freedesktop.xcb/4317/focus=4324

[Karlt (karlt)]

*** Bug 484997 has been marked as a duplicate of this bug. ***

[lessoffensive (lessoffensive)]

I patched libxcb with the proposed patch and have been testing it since last night. I haven't received any BadIDChoice errors, which in the scenario I've been in basically ensured that it would crash before all the tabs finished reloading after it crashed (I had about 140 tabs open in 2 windows). I patched it against 1.1.93 and it seems to be working well.

This has been a serious problem for me and made firefox basically unusable in the fashion I typically use it in for the last 6 months; I hope this patch (or an improved one, if I find there to be problems with this one) makes it into the mainline asap. If it doesn't, I'll be running my own patched version of xcb because this problem is really very huge for me.

[Julien Danjou (jdanjou)]

commit beccb0be15f5699c942a0af33307d9e4bf797e2a
Author: Bart Massey <email address hidden>
Date: Tue Mar 24 16:24:04 2009 -0700

    kludgily hand-merged xid fixes

    Signed-off-by: Bart Massey <email address hidden>
    Signed-off-by: Julien Danjou <email address hidden>

[Nathaniel Smith (njs)]

Fix has been committed to xcb trunk:

http://lists.freedesktop.org/archives/xcb/2009-March/004479.html

http://cgit.freedesktop.org/xcb/libxcb/commit/?id=beccb0be15f5699c942a0af33307d9e4bf797e2a

[David Grossberg (davidgro)]

I placed the patch in libxcb1_1.1-1ubuntu1 (Hardy) and so far it seems to work just fine.

Before patching, this bug would be the usual mode of failure for any Firefox session that went longer than a day or two, I have not had a chance to stress test Firefox, but I had found another program that could produce this bug on demand:
In the game Clonk Rage ( http://www.clonk.de/cr.php ) if I attempted to change the screen resolution it would produce the BadIDChoice error every time. Now after the patch it correctly changes the screen resolution and the rest of the game is playable as well (so far as I can tell from a couple of the tutorial levels)

My system is Kubuntu Hardy x86, and my GPU is an ATI Radeon Xpress 1100 IGP, using the proprietary drivers from the standard repositories. (The game might act differently on other systems, in case anyone else wants to check my results)

[Wrnc Rdal (rrrgtr)]

fix committed https://bugzilla.mozilla.org/show_bug.cgi?id=458092#c40
looking forward to this, 3 more crashes today.

[Karlt (karlt)]

Marking FIXED based on comment 40 and
https://bugs.freedesktop.org/show_bug.cgi?id=20254#c2

[BlueT - Matthew Lien - 練喆明 (bluet)]

matthew@Priapus:~/bug-report$ Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
[FDLProc] canHandle()
The program 'firefox' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 65177197 error_code 14 request_code 152 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)

[Alexander Sack (asac)]

On Sat, Mar 28, 2009 at 05:44:44PM -0000, lessoffensive wrote:
> I patched libxcb with the proposed patch and have been testing it since
> last night. I haven't received any BadIDChoice errors, which in the
> scenario I've been in basically ensured that it would crash before all
> the tabs finished reloading after it crashed (I had about 140 tabs open
> in 2 windows). I patched it against 1.1.93 and it seems to be working
> well.
>
> This has been a serious problem for me and made firefox basically
> unusable in the fashion I typically use it in for the last 6 months; I
> hope this patch (or an improved one, if I find there to be problems with
> this one) makes it into the mainline asap. If it doesn't, I'll be
> running my own patched version of xcb because this problem is really
> very huge for me.
>

Can you please attach the patch you are currently using and provide
reference where you picked it from (e.g. mailing list message, etc.).

Thanks a lot for your work on this.

 - Alexander

[Alexander Sack (asac)]

nevermind. found the patch i think.

[Alexander Sack (asac)]

uploaded package libxcb_1.1.93-0ubuntu4~asac1 for jaunty to my PPA. please test that:

https://edge.launchpad.net/~asac/+archive/ppa

Do we need backports for intrepid hardy too?

[BlueT - Matthew Lien - 練喆明 (bluet)]

to Alexander Sack,
since jaunty is still in beta and not released yet, people might not upgraded to jaunty beta on their computers, so I think it would be better to have intrepid builds too.
(I have 2 PC and 1 laptop is still using intrepid for my job/works).

Is there any chance to have intrepid builds in your PPA?
That would be awsome if I (and WE) can update our firefox with the patch.

[BlueT - Matthew Lien - 練喆明 (bluet)]

matthew@Priapus:~/bug-report$ firefox -g --sync &
[1] 13228
matthew@Priapus:~/bug-report$ /home/matthew/bin/backtrace.sh /usr/lib/firefox-3.0.8/firefox -x /tmp/mozargs.o13230
outputting trace to '/tmp/gdb-firefox.txt'
FoxyProxy settingsDir = /home/matthew/.mozilla/firefox/hpelwq2p.default

matthew@Priapus:~/bug-report$
matthew@Priapus:~/bug-report$ Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
!!! [Hook] hook(): title not found
!!! [Hook] hook(): title not found
[FDLProc] canHandle()
[FDLProc] canHandle()
[FDLProc] canHandle()
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
[ConvConfHandler] isPreferred contentType: application/zip
Removing DOMNodeRemoved listener
[FDLProc] canHandle()
Removing DOMNodeRemoved listener
[FDLProc] canHandle()
The program 'firefox' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 40698879 error_code 14 request_code 152 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)

[1]+ Done firefox -g --sync

[John Vivirito (gnomefreak)]

I sort of agree with Blue. Intrepid fix would be good however I'm thinking Hardy is more important. More testing before it hitting Hardy to make sure no regression is introduced would be my opinion

[Matt Zimmerman (mdz)]

I have started seeing this only just recently. It happened for the first time this week but has happened 2-3 more times since then. I've recently installed a number of new extensions, and the latest updates to the Flash plugin.

Does anyone know how to reproduce this?

[Timeless-bemail (timeless-bemail)]

*** Bug 480179 has been marked as a duplicate of this bug. ***

[FredBezies (fredbezies-deactivatedaccount)]

Using Alexander Sack ppa and no more crash on startup. Wonderful. Any hope to get this in for jaunty final release ?

[Andrew Pollock (apollock)]

I think there are similar underlying problems with libxcb in Hardy, although not manifesting themselves in the same manner as described in this bug report. Please also patch libxcb in Hardy if possible.

[Perry E. Metzger (perry-piermont)]

A version of the fixes to libxcb for Intrepid would be highly appreciated. Anyone have a patched package for me to test out?

[Timo Aaltonen (tjaalton)]

I've seen this myself, and will make sure the patch is SRU'd to jaunty and likely to hardy. Not sure about intrepid, jaunty will be released this week anyway..

[Alexander Sack (asac)]

taking over as i already have the package in my ppa.

[Joel Ebel (jbebel)]

We've now discovered some crashes in emacs that this patch also fixes. This is under hardy, so I encourage this patch to be released to hardy as well.

[BlueT - Matthew Lien - 練喆明 (bluet)]

Keep experiencing crashes on Intrepid and Hardy we don't have a patched libxcb for Intrepid and Hardy LTS.

matthew@Priapus:~/bug-report$ firefox -g --sync &
[1] 809
matthew@Priapus:~/bug-report$ /home/matthew/bin/backtrace.sh /usr/lib/firefox-3.0.10/firefox -x /tmp/mozargs.WpE811
outputting trace to '/tmp/gdb-firefox.txt'

matthew@Priapus:~/bug-report$
matthew@Priapus:~/bug-report$ FoxyProxy settingsDir = /home/matthew/.mozilla/firefox/hpelwq2p.default
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
!!! [Hook] hook(): title not found

matthew@Priapus:~/bug-report$ [FDLProc] canHandle()
[FDLProc] canHandle()
[FDLProc] canHandle()
[FDLProc] canHandle()
[FDLProc] canHandle()
[FDLProc] canHandle()
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
Removing DOMNodeRemoved listener
** Message: NP_Initialize
** Message: NP_Initialize succeeded
** Message: totemPlugin [0x2147ecc0]
** Message: Init mimetype 'application/x-mplayer2' mode 1
** Message: Base URI is 'http://home.51.com/prettywinnie'
** Message: Real mimetype for 'application/x-mplayer2' is 'video/x-msvideo'
argv[0] src http://pic.51.com/mid/20040505143457.mid
argv[1] name Player
argv[2] type application/x-mplayer2
argv[3] pluginspage http://home.51.com/prettywinnie
argv[4] showcontrols 1
argv[5] showpositioncontrols 0
argv[6] showaudiocontrols 1
argv[7] showtracker 1
argv[8] showdisplay 0
argv[9] showstatusbar 1
argv[10] autosize 0
argv[11] showgotobar 0
argv[12] showcaptioning 0
argv[13] autostart 1
argv[14] autorewind 0
argv[15] animationatstart 0
argv[16] transparentatstart 0
argv[17] allowscan 1
argv[18] enablecontextmenu 1
argv[19] clicktoplay 0
argv[20] defaultframe datawindow
argv[21] invokeurls 0
argv[22] align baseline
argv[23] border 0
argv[24] height 68
argv[25] width 100%
** Message: Stream requested (force viewer: 0)
** Message: mSrcURI: http://pic.51.com/mid/20040505143457.mid
** Message: mCache: 0
** Message: mControllerHidden: 0
** Message: mShowStatusbar: 1
** Message: mHidden: 0
** Message: mAudioOnly: 0
** Message: mAutoPlay: 1, mRepeat: 0
** Message: Launching: /usr/lib/totem/gstreamer/totem-plugin-viewer --plugin-type gmp --user-agent Mozilla/5.0 (X11; U; Linux i686; zh-TW; rv:1.9.0.10) Gecko/2009042523 Ubuntu/8.10 (intrepid) Firefox/3.0.10 --mimetype video/x-msvideo --statusbar
** Message: Viewer spawned, PID 16952
** Message: Initial window set, XID 483f74e size 200x68
** Message: No viewer proxy yet, deferring SetWindow
** Message: GetScriptableNPObject [0x2147ecc0]
** Message: totemGMPPlayer [0x316df588]
** Message: Viewer DBus interface name is 'org.gnome.totem.PluginViewer_16952'
** Message: NameOwnerChanged old-owner '' new-owner ':1.539'
** Message: Viewer now connected to the bus
** Message: ViewerSetup
** Message: Calling SetWindow
Viewer: SetWindow XID 75757390 size 200:68
TotemEmbedded-Message: Viewer state: STOPPED
** Message: NewStream mimetype 'audio/midi' URL 'http://pic.51.com/mid/20040505143457.mid'
** Message: Not expe...

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

Any plans on porting this fix to the 1.9.0 branch? I've had it yesterday (2009-05-14) on a GranParadiso nightly (Fx 3.0.11pre).

[Zack Weinberg (zackw)]

The bug is in a system library, not our code; you need to talk to your distributor about getting the fix into their copy of libxcb.

[D. Hugh Redelmeier (hugh-mimosa)]

See Novell OpenSuSE bug report:
https://bugzilla.novell.com/show_bug.cgi?id=423740

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

(In reply to comment #44)
> The bug is in a system library, not our code; you need to talk to your
> distributor about getting the fix into their copy of libxcb.

I see. Well, my distributor is Novell/SuSE (but I get Mozilla products straight from Mozilla in .tar.bz2 form) and I diligently apply all "online updates" to their distribution (currently openSUSE 11.1). Let's hope they'll upgrade their X libraries some day.

[D. Hugh Redelmeier (hugh-mimosa)]

Tony Mechelynck:
Report this bug to Novell. Some other distro's have already released updates for it. https://bugs.freedesktop.org/show_bug.cgi?id=20254

[Antoine-mechelynck-gmail (antoine-mechelynck-gmail)]

(In reply to comment #46)
> Tony Mechelynck:
> Report this bug to Novell. Some other distro's have already released updates
> for it. https://bugs.freedesktop.org/show_bug.cgi?id=20254

I did, by commenting on what looked like an appropriate existing bug, see https://bugzilla.novell.com/show_bug.cgi?id=423740#c1

I did not search for duplicates, however. If you think that it is indeed the appropriate Novell bug, you may (or may not) want to add a crossref for it on your freedesktop bug.

[Timo Aaltonen (tjaalton)]

uploaded to jaunty-proposed.

[Martin Karpinski (mkarpinski)]

Can anyone give me a timeframe for when this patch might be available? I've switched to Ubuntu from Vista about two months ago and have had to deal with daily crashes from this bug.

[Martin Pitt (pitti)]

Please upload to karmic ASAP

[Martin Pitt (pitti)]

Accepted libxcb into jaunty-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Timo Aaltonen (tjaalton)]

Karmic already has 1.3-2, which includes the fix.

[Àlex Magaz (rivaldi8)]

It works for me (tested with Firefox 3.5).

[Damian Frick (dafrick)]

I've had the same problem and the patch works so far, however it seems, that it also introduced a new bug.
It occurs under the same conditions as the bug that is now fixed.
So when trying to reproduce it I opened a lot (200+) tabs from Digg an Youtube and at some point a tab openes in an Untitled window outside of firefox and I get the following errors:

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c71753 unexpectedly destroyed

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c71752 unexpectedly destroyed

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c71780 unexpectedly destroyed

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c7179f unexpectedly destroyed

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c7179e unexpectedly destroyed

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c71771 unexpectedly destroyed

(firefox:9845): Gdk-WARNING **: GdkWindow 0x4c71770 unexpectedly destroyed

(firefox:9845): Gdk-CRITICAL **: gdk_window_new: assertion `window != NULL' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_set_user_data: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_set_back_pixmap: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): GLib-GObject-CRITICAL **: g_object_set_data: assertion `G_IS_OBJECT (object)' failed

(firefox:9845): GLib-GObject-CRITICAL **: g_object_set_data: assertion `G_IS_OBJECT (object)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_move_resize: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_hide: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_new: assertion `window != NULL' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_set_user_data: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_set_back_pixmap: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): GLib-GObject-CRITICAL **: g_object_set_data: assertion `G_IS_OBJECT (object)' failed

(firefox:9845): GLib-GObject-CRITICAL **: g_object_set_data: assertion `G_IS_OBJECT (object)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_move_resize: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_hide: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_show_unraised: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_new: assertion `window != NULL' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_set_user_data: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_set_back_pixmap: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): GLib-GObject-CRITICAL **: g_object_set_data: assertion `G_IS_OBJECT (object)' failed

(firefox:9845): GLib-GObject-CRITICAL **: g_object_set_data: assertion `G_IS_OBJECT (object)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_move_resize: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_hide: assertion `GDK_IS_WINDOW (window)' failed

(firefox:9845): Gdk-CRITICAL **: gdk_window_show_unraised: assertion `GDK_IS_WINDOW (window)' failed

Whe I close that window firefox crashes, if I...

[Alexander Sack (asac)]

Damian, you sure this problem goes away when downgrading? maybe it was just hidden because the other bug was too frequent.

[Timo Aaltonen (tjaalton)]

Uploaded to hardy-proposed, including a bunch of other fixes.

[Damian Frick (dafrick)]

@Alexander
It is entirely possible, but there is no way for me to find out, since firefox will crash a lot sooner without the patch. So if any one of you doesn't get this kind of behaviour, please just ignore my comment as it probably isn't related to this bug, but if you have a little time try it out and if it comes up its possible they're related.

[Karl Tomlinson (bugs+launchpad)]

The bug Damian is seeing has existed for a long time.
It is more likely to occur in the same conditions (short XID range response) as lead to this bug.

The cause is described here:
http://bugzilla.gnome.org/show_bug.cgi?id=581526

[Damian Frick (dafrick)]

Thank you for pointing that out, Karl.

[Launchpad Janitor (janitor)]

This bug was fixed in the package libxcb - 1.1.93-0ubuntu3.1

---------------
libxcb (1.1.93-0ubuntu3.1) jaunty-proposed; urgency=low

  [ Alexander Sack ]
  * debian/patches/lp220628_xid_fixes.patch: add fix for LP: #220628 - fix
    BadIDChoice issues reported against firefox
  * debian/rules: in turn enable simple patchsys

 -- Timo Aaltonen <email address hidden> Wed, 10 Jun 2009 12:33:39 +0300

[Martin Pitt (pitti)]

verification-done tag removed for the hardy update. Does this really need to be fixed in hardy as well? For firefox 3.5 backports?

[Alexander Sack (asac)]

On Thu, Jul 09, 2009 at 01:22:11PM -0000, Martin Pitt wrote:
> verification-done tag removed for the hardy update. Does this really
> need to be fixed in hardy as well? For firefox 3.5 backports?

no. the verification is done. the other problem is a slightly
different bug and should be tracked differently imo.

 - Alexander

[Martin Pitt (pitti)]

Alexander Sack [2009-07-09 14:27 -0000]:
> On Thu, Jul 09, 2009 at 01:22:11PM -0000, Martin Pitt wrote:
> > verification-done tag removed for the hardy update. Does this really
> > need to be fixed in hardy as well? For firefox 3.5 backports?
>
> no. the verification is done.

Misunderstanding? The verification was done for jaunty, and I copied
jaunty-proposed to jaunty-updates. But there is no hardy-proposed
version for libxcb; since the task is still open, I assumed that you
want to fix it there as well. But we need to test the hardy update
separately.

> the other problem is a slightly different bug and should be tracked
> differently imo.

Yes, I agree.

[Martin Pitt (pitti)]

Accepted libxcb into hardy-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Bugmail-asutherland (bugmail-asutherland)]

This apparently landed on Ubuntu jaunty 9.04 (the current stable release) on July 9th (2009).

https://bugs.launchpad.net/ubuntu/+source/firefox-3.0/+bug/220628

[Ben-bucksch (ben-bucksch)]

Still happens for me with Firefox 3.5.2

The program 'firefox-bin' received an X Window System error.
This probably reflects a bug in the program.
The error was 'BadIDChoice (invalid resource ID chosen for this connection)'.
  (Details: serial 178785760 error_code 14 request_code 156 minor_code 4)
  (Note to programmers: normally, X errors are reported asynchronously;
   that is, you will receive the error a while after causing it.
   To debug your program, run it with the --sync command line
   option to change this behavior. You can then get a meaningful
   backtrace from your debugger if you break on the gdk_x_error() function.)
libxcb: WARNING! Program tries to lock an already locked connection,
        which indicates a programming error.
        There will be no further warnings about this issue.

[Ben-bucksch (ben-bucksch)]

Please ignore last comment, I use an old SuSE 10.3, and didn't realize that I need to update X11.

[Martin Pitt (pitti)]

Anyone who could test the packages in hardy-proposed?

[Joel Ebel (jbebel)]

Tested under hardy. Firefox fails to start with:

firefox: ../../src/xcb_lock.c:33: _XCBUnlockDisplay: Assertion `xcb_get_request_sent(dpy->xcb->connection) == dpy->request' failed.

[Alexander Sack (asac)]

On Thu, Oct 29, 2009 at 12:13:38AM -0000, Joel Ebel wrote:
> Tested under hardy. Firefox fails to start with:
>
> firefox: ../../src/xcb_lock.c:33: _XCBUnlockDisplay: Assertion
> `xcb_get_request_sent(dpy->xcb->connection) == dpy->request' failed.

So you tested hardy-proposed bits? or just confirmed that you see this bug
in hardy without having hardy-proposed enabled?

 - Alexander

[Joel Ebel (jbebel)]

That was with the package in hardy proposed. Firefox failing to start was not a symptom of the original bug.

[Martin Pitt (pitti)]

Joel, can you confirm that downgrading libxcb to hardy final fixes the assertion failure?

[Joel Ebel (jbebel)]

Correct. The problem doesn't occur with libxcb1 1.1-1ubuntu1. I did notice that this is exposed by the fact that we divert /usr/bin/firefox to /usr/bin/firefox.distrib. The problem doesn't occur if /usr/bin/firefox-3.0 is called directly. I also discovered a similar error starting gnome-screensaver with the new libxcb that is not present with the original version.

 $ gnome-screensaver
gnome-screensaver: ../../src/xcb_lock.c:33: _XCBUnlockDisplay: Assertion `xcb_get_request_sent(dpy->xcb->connection) == dpy->request' failed.
Aborted

[Martin Pitt (pitti)]

This proposed update has been in hardy-proposed for almost half a year without any positive feedback, and it was reported to cause a regression here. Therefore it was removed from hardy-proposed again.

[André Pirard (a.pirard)]

I'm surprised that I'm only the 6th person to tick the "me too".
I'm really pested with exactly the same problem.
Every half an hour or so, Firefox disappears when active with many windows and tabs.
I also experienced Thunderbird disappearing, but far less often.
Even Pidgin disappeared a couple of times.
At first I thought it was IP related.
Until I saw that same X BadIDChoice message and I came here.
Should I also mention that Gnome (GDM I suppose) most often cannot switch to another session because of an X failure? It rarely fails when a second attempt is made.
I didn't check any message except in the Firefox case, though.

Ubuntu 8.10 (Gnome). Firefox 3.0.15. libxcb1 is 1.1-1.1, which seems to be what matters.

What should I do to get rid of such a nasty problem?
Shouldn't bug summaries explain the procedure to follow in terms that can be understood by everybody?

How do I get to using 1.1-1ubuntu1 instead if it's true there's no problem with it?

If the importance is "unknown", I can help: it's high for active people using the Web a lot.

Thanks.

[Steven Shaw (steshaw)]

Having a fix for this bug was pretty important to me. I'm a heavy browser of the internet with multiple windows with multiple tabs. I was hit with this bug quite frequently (and over and over and over again when I tried to restore my previous browsing sessions). I switched to Google Chrome.