Bugzilla Bug 192310

i'm running gentoo unstable + gnome-overlay. i've applied the upstream tooltips
patch to gtk.
java-1.6: javaws and eclipse-3.2 segfault at startup, G_SLICE=always-malloc
changes nothing. stacktrace is not very helpfull (no symbols) although i tried
to compile gtk,glib and pango with -g, without -fomit-frame-pointer and
nostrip. i've also disabled xinerama.
java-1.5: javaws and eclipse-3.2 start if i set G_SLICE=always-malloc but
eclipse crashes after some usage (not related to tooltips imho) -> same thing
as 1.6
using XToolkit or MToolkit has no effect.


Reproducible: Always

Steps to Reproduce:
1.install gentoo ~x86
2.install gnome-overlay
3.run eclipse-3.2

Actual Results:  
eclipse, javaws segfaults and bug buddy appears but has no usefull informations
(i guess its debugging the wrong thread)

#0  0xb7fd5410 in __kernel_vsyscall ()
#1  0xb7fa7658 in pthread_join () from /lib/libpthread.so.0
#2  0x0804dce8 in ContinueInNewThread ()
#3  0x080497f6 in main ()

if i use gdb to debug the actual crash it's there are no symbols.



Portage 2.1.3.9 (default-linux/x86/2007.0/desktop, gcc-4.2.0, glibc-2.6.1-r0,
2.6.22-gentoo-r6 i686)
=================================================================
System uname: 2.6.22-gentoo-r6 i686 AMD Athlon(tm) XP 3000+
Timestamp of tree: Wed, 12 Sep 2007 13:00:01 +0000
app-shells/bash:     3.2_p17-r1
dev-java/java-config: 1.3.7, 2.0.33-r1
dev-lang/python:     2.5.1-r2
sys-apps/baselayout: 1.12.10-r4
sys-apps/sandbox:    1.2.18.1
sys-devel/autoconf:  2.13, 2.61-r1
sys-devel/automake:  1.5, 1.7.9-r1, 1.8.5-r3, 1.9.6-r2, 1.10
sys-devel/binutils:  2.18
sys-devel/gcc-config: 1.4.0-r2
sys-devel/libtool:   1.5.24
virtual/os-headers:  2.6.22-r2
ACCEPT_KEYWORDS="x86 ~x86"
CBUILD="i686-pc-linux-gnu"
CFLAGS="-mtune=athlon-xp -march=athlon-xp -O2 -fomit-frame-pointer
-frename-registers -fweb -ftracer -fgcse-after-reload -pipe"
CHOST="i686-pc-linux-gnu"
CONFIG_PROTECT="/etc /usr/share/X11/xkb"
CONFIG_PROTECT_MASK="/etc/env.d /etc/env.d/java/ /etc/gconf /etc/revdep-rebuild
/etc/terminfo /etc/texmf/web2c /etc/udev/rules.d"
CXXFLAGS="-mtune=athlon-xp -march=athlon-xp -O2 -fomit-frame-pointer
-frename-registers -fweb -ftracer -fgcse-after-reload -pipe"
DISTDIR="/usr/portage/distfiles"
EMERGE_DEFAULT_OPTS="--with-bdeps y"
FEATURES="distlocks fixpackages metadata-transfer parallel-fetch sandbox
sfperms strict unmerge-orphans userfetch userpriv"
GENTOO_MIRRORS="http://gentoo.inode.at
ftp://sunsite.informatik.rwth-aachen.de/pub/Linux/gentoo/ 
http://gentoo.osuosl.org http://mirror.phy.olemiss.edu/mirror/gentoo
http://csociety-ftp.ecn.purdue.edu/pub/gentoo/ http://distfiles.gentoo.org
http://www.ibiblio.org/pub/Linux/distributions/gentoo"
LANG="en_US.utf8"
LDFLAGS="-Wl,-O1"
LINGUAS="en en_GB en_US de"
PKGDIR="/usr/portage/packages"
PORTAGE_RSYNC_OPTS="--recursive --links --safe-links --perms --times --compress
--force --whole-file --delete --delete-after --stats --timeout=180
--exclude=/distfiles --exclude=/local --exclude=/packages
--filter=H_**/files/digest-*"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY="/usr/portage/local/layman/gnome-experimental"
SYNC="rsync://rsync.gentoo.org/gentoo-portage"
USE="3dnow 3dnowext 7zip S3TC X a52 aac aalib acl acpi aiglx alsa amr aoss
applet artworkextra async audiofile avahi bash-completion bcp berkdb
bitmap-fonts bjam bonobo boo boost browserplugin bzip2 c++ cairo caps cddb cdr
cli cluster cracklib crypt cscope css cups curl curlwrappers d daap dbus
deprecated device-mapper dga dia disk-partition divx dmi dnd dpms dri dts dv
dvb dvd dvdnav dvdr dvdread dvi dynamic eds effects elf emboss encode erandom
evo evolution exif expat extraengine faad fat fbcon fbdev festival ffmpeg fftw
flac fluidsynth fontconfig foomaticdb fortran fpx freetype ftp fuse galago gb
gcc-libffi gconf gd gdbm gedit gif gimp gimpprint glade glib glibc-omitfp glut
gmedia gnome gnomecanvas gnomedb gnustep gnutls grammar graphviz gs gsf gsnd
gstreamer gtk gtk2 gtkhtml guile hal hbci hdri hfs iconv icu idn inkjar intl
ipod ipv6 irda isdnlog java java5 javamail javascript jfs joystick jpeg jpeg2k
jsch jta keyring kqemu lcd lcms ldap libcaca libedit libgda libnotify
libsamplerate lirc log4j logrotate lzo lzw mad magic mailwrapper math matroska
mbox md5sum mdb midi mikmod mixer mmap mmx mmxext mng mono mozbranding
moznocompose moznoirc moznomail mozsvg mp2 mp3 mp4 mp4live mpeg mpeg2 mpi
mplayer mpm-prefork mpm-threadpool mpm-worker mudflap musepack musicbrainz
mythtv native nautilus ncurses net network network-cron networking nls
no-old-linux no-seamonkey nodod nogecko-sdk nomalloccheck nomotif notfc nptl
nptlonly nsplugin ntfs nvidia oav objc objc++ odk offensive ofx ogg ole openal
opengl openmp oss pam pam_chroot pam_console pam_timestamp pch pcre pdf perl
pic plugin plugins png pnm posix postgres postscript pppd pthreads pulseaudio
pyste python qt-static quicktime quotes radeon radio rar readline real
realmedia reflection regex regexp reiser4 reiserfs rtc samba sasl session slang
slp sndfile snmp soap socks5 speex spell spl sql sqlite sqlite3 srt sse ssl
startup-notification stencil-buffer svg t1lib tagwriting tcpd teletext tetex
theora thesaurus threads threadsafe tiff timidity tk toolkit-scroll-bars
tracker truetype truetype-fonts type1 type1-fonts unicode usb userlocales utf8
v4l v4l2 vdr vidix virus-scan vnc vorbis vorbis-psy wideband win32codecs wma
wmf wmp wordperfect wxwindows x264 x86 xattr xcb xerces xforms xfs xml xorg
xosd xpm xprint xrandr xscreensaver xulrunner xv xvid xvmc yv12 zlib zvbi"
ALSA_CARDS="loopback mpu401 seq-dummy usb-audio via82xx virmidi"
ALSA_PCM_PLUGINS="adpcm alaw asym copy dmix dshare dsnoop empty extplug file
hooks iec958 ioplug ladspa lfloat linear meter mulaw multi null plug rate route
share shm softvol" ELIBC="glibc" INPUT_DEVICES="dmc evdev joystick keyboard
mouse void" KERNEL="linux" LCD_DEVICES="bayrad cfontz cfontz633 glk hd44780
lb216 lcdm001 mtxorb ncurses text" LINGUAS="en en_GB en_US de"
LIRC_DEVICES="devinput" USERLAND="GNU" VIDEO_CARDS="dummy fbdev nv nvidia v4l
vesa vga none"
Unset:  CTARGET, INSTALL_MASK, LC_ALL, MAKEOPTS, PORTAGE_COMPRESS,
PORTAGE_COMPRESS_FLAGS, PORTAGE_RSYNC_EXTRA_OPTS

it seems bug-buddy 2.20.0 is causing the problem, as uninstalling it fixes it
(a warning about loading gnomebreakpad appears, but the appps work). i'll try
with 2.18.1 later ..

bug-buddy-2.18.1 works (it doesn't come with breakpad), so i guess there's an
fault in google breakpad included in the new bug-buddy.

maybe someone could reassign this to the gnome team?

Thanks for the pointer from the gnome bug, this is *exactly* what I've been
experiencing.  Have you filed a bug upstream yet?

Also, I found the attachment to this bug:

http://bugzilla.gnome.org/show_bug.cgi?id=479507

Which I'm going to apply locally and test out.  It should hopefully tell us
whether it's breakpad in particular causing the problem, or something else in
bug buddy...

Interestingly, I disabled the google-breakpad and the issue still occurred, so
it appears it's built into the changes that went into bug-buddy-2.20.0.  I'm
gonna do some more digging and see if I can isolate the issue a bit better. 
I'll attach the ebuild for disabling googles-breakpad, in case it's of
interest.

Created an attachment (id=131702) [details]
bug-buddy-2.20.0.ebuild - applies patch and adds USE flag

This also requires the patch from the gnome bug mentioned above to be added to
${FILESDIR}/bug-buddy-2.20.0-no-breakpad.patch

(In reply to comment #3)
> Thanks for the pointer from the gnome bug, this is *exactly* what I've been
> experiencing.  Have you filed a bug upstream yet?

no, i didn't. i was not sure where to file it (breakpad vs bug-buddy) and i was
afraid it had something to do with my toolchain (gcc,glibc,cflags,...)

Ok, after reading some *very* interesting information over on the opensolaris
bugs page:

http://bugs.opensolaris.org/view_bug.do?bug_id=6600538

it seems to explain a bit about what's going on.  Luckily it suggests a
workaround.  Unfortunately their workaround doesn't work.  Happily, it gives
enough information to figure out a different workaround.  5:)

For any apps that appear to be breaking because of breakpad, it's possible to
start the app as long as the GTK_MODULES environment variable doesn't contain
gnomebreakpad.

This should probably be assigned to bug-buddy directly upstream with references
to our bug and the opensolaris one (and it doesn't appear it's a toolchain
issue)  5;)

Thanks again for pointing this out to me, I can now at least still use eclipse!

Just to keep everybody on track, the upstream bug is Bug 479929.

It can be found at:

http://bugzilla.gnome.org/show_bug.cgi?id=479929

(In reply to comment #0)
> i'm running gentoo unstable + gnome-overlay. i've applied the upstream tooltips
> patch to gtk.

I _think_ this references the patch at
http://bugzilla.gnome.org/show_bug.cgi?id=460194 which I have now included in
gtk+-2.12.0-r1.

I personally had no plans to put bug-buddy-2.20 into portage tree until
google-breakpad can be fully disabled and the work for all that has no ill side
effects, plus debian has raised some licensing concerns. Breakpad as is is
completely useless to us due to debug binaries being completely impossible to
be created for breakpad/socorro to use together with the minidumps due to the
nature of Gentoo Linux package system. I had some chats on this with the
bug-buddy developers but I'll need to poke them again with my ideas on how to
approach systems like Gentoo with this new infrastructure.

The idea was to basically something like this:
Socorro(?) infrastructure upstream receives minidumps from users and then
processes them with debug packages and goes from there with automatic
classification and duplication tracking and so on. As it was with 2.18, the bug
reports get reported to upstream bugzilla and the automatic duplication and
other check system wouldn't be deployed. So the idea is to basically do the
step of combining minidump with debug packages at the users system and send the
results of THAT for automated duplication classification and other steps; so as
the next step after minidump is the one that we can't do due to everyones
packages being different (CFLAGS, et cetera), we could do that step on the
users machine instead and send those results as input to the step after that in
upstream infrastructure.

Mart, I've been using the no-google-breakpad patch (linked in comment 3) with
the ebuild attached in comment 5 without problem now for a couple of days, so
that should be a problem, most of the work's already been done.  Also, Fernando
just added a patch to this problem in the gnome bug (added to the URL field). 
It works fine, as long as google-breakpad is turned off...

(In reply to comment #9)
> (In reply to comment #0)
> > i'm running gentoo unstable + gnome-overlay. i've applied the upstream tooltips
> > patch to gtk.
> 
> I _think_ this references the patch at
> http://bugzilla.gnome.org/show_bug.cgi?id=460194 which I have now included in
> gtk+-2.12.0-r1.
> 

yes, thats the one.

> ..other check system wouldn't be deployed. So the idea is to basically do the
> step of combining minidump with debug packages at the users system and send the
> results of THAT for automated duplication classification and other steps; so as
> the next step after minidump is the one that we can't do due to everyones
> packages being different (CFLAGS, et cetera), we could do that step on the
> users machine instead and send those results as input to the step after that in
> upstream infrastructure.
> 

don't you need the debug builds (or unstripped binaries) to do this?

(In reply to comment #11)
> > ..other check system wouldn't be deployed. So the idea is to basically do the
> > step of combining minidump with debug packages at the users system and send the
> > results of THAT for automated duplication classification and other steps; so as
> > the next step after minidump is the one that we can't do due to everyones
> > packages being different (CFLAGS, et cetera), we could do that step on the
> > users machine instead and send those results as input to the step after that in
> > upstream infrastructure.
> > 
> 
> don't you need the debug builds (or unstripped binaries) to do this?
> 

Yes. It should employ similar throw-away rules as in socorro. Not enough debug
symbols means throw away the report and possibly notify user to get them with
CFLAGS="${CFLAGS} -g" FEATURES=splitdebug if he wants to help further. Bugs
without any information getting filed to upstream bugzilla right now with 2.18
only the be marked INCOMPLETE isn't helping anyone. There of course needs some
coordination with upstream how they want it for statistics - do they want to
know that a crash happened without a good backtrace to keep track of overall
count statistics or not, etc.
As a side note I'm not even sure something like this isn't already implemented,
but I haven't read anything anywhere yet about it (I'm still on 2.18 myself,
gradually going towards 2.20)
Our debug reports can and are most valueable in the case all of the debug stuff
is set up fully - now if we could just filter out the useless ones with this
infrastructure.
The whole system compiled with -g in CFLAGS and FEATURES="${FEATURES}
splitdebug installsources" really does give very full reports - with
installsources (requires debugedit package) often better than any other
distribution with debug packages.

ah .. i didn't know about the splitdebug feature, thats really nice ... but
installing the sources for every package sounds like doubling (or trippling?)
the hd requirements of my gentoo system.

There's now a set of patches to solve this for the case where another program
installs it's own sighandler (both with and without google-breakpad), and it's
been checked in upstream.  They can be found at:

http://svn.gnome.org/viewcvs/bug-buddy/trunk/gnome-breakpad/gnome-breakpad.cc?view=patch&r1=2340&r2=2431&pathrev=2436
http://svn.gnome.org/viewcvs/bug-buddy/trunk/gnome-breakpad/gnome-breakpad.cc?view=patch&r1=2436&r2=2435&pathrev=2436

The second patch corrects a failing in the first patch which stops it compiling
under gcc-4.2...

i can confirm the patches fix the problem, the second patch is also needed for
gcc-4.1.
hmm .. looking at the code, i realize there would have been a simple solution
to disable signal hooking with the environment varible
"GNOME_DISABLE_CRASH_DIALOG"

*** Bug 193756 has been marked as a duplicate of this bug. ***

I've added the patches from comment #14 to bug-buddy-2.20.0-r1.  I didn't have
any crashes before; can someone confirm that these fix the issue?

bug-buddy-2.20.0-r1 works for me.

works for me too

Fixed, then.  Thanks.