Hi Tobias,
On Sat, Nov 21, 2009 at 03:23:46PM +0100, Tobias Nygren wrote:
> On Sat, 21 Nov 2009 23:58:13 +1100
> Simon Burge <simonb%NetBSD.org@localhost> wrote:
>
> > I've just upgraded my netbsd-4/i386 machine to pkgsrc-2008Q3, which
> > included a jump from firefox-2.0.0.17 to firefox-3.5.3nb2.
> >
> > When I try to start firefox, xulrunner pops up a dialog box or three
> > which disappear to quickly to see what they say. If I start with no
> > ~/.mozilla directory I get three of these dialog boxes, and if I start
> > with an existing profile I get just one of these. Then xulrunner-bin
> > dumps core.
>
> This sounds vaguely familiar. I think it can happen when fontconfig
> doesn't find any fonts.
I've tried things like:
rm -rf ~/.font*
fc-cache
with no noticeable change in behaviour.
> > Note that if I start /usr/pkg/lib/firefox/firefox directly, instead of
> > using the /usr/pkg/bin/firefox script it does appear to start (after
> > popping up two dialog boxes that disappear like earlier) but then after
> > the initial firefox window starts it appears to wedge and the firefox
> > process has two threads, one blocked in sawait and the other in poll.
>
> The stub launcher is known to not work correctly. You really need
> to start it like done in the wrapper script (but see below...).
>
> > First question - has anyone successfully used firefox 3.5 on netbsd-4?
>
> Haven't used XFree86 in a long time, sorry.
Does xorg work on netbsd-4? I could try that, but it'll probably be a
couple of weekends away. This is my main workstation, and I'm trying
not to make it too unavailable. I guess a next step is upgrading to
netbsd-5, but I've been putting that off for as long as possible :-)
> > Second question - any suggestions on making this work?
>
> If you rebuild xulrunner with PKG_OPTIONS.xulrunner+=debug in mk.conf
> it should print some clues to the problem on stderr. There were some
> additional debugging instructions posted earlier:
> http://mail-index.netbsd.org/pkgsrc-users/2009/09/25/msg010827.html
I've attached two log files. The first was with no pre-existing
~/.mozilla directory. After this, the xulrunner back trace is:
#0 0xb997622f in kill () from /usr/lib/libc.so.12
#1 0xb9a12340 in abort () from /usr/lib/libc.so.12
#2 0xb99d4c06 in __assert13 () from /usr/lib/libc.so.12
#3 0xb9b4c207 in _cairo_hash_table_destroy () from
/usr/pkg/lib/libcairo.so.2
#4 0xb9b5c383 in _cairo_scaled_font_map_destroy () from
/usr/pkg/lib/libcairo.so.2
#5 0xb9b482b2 in _cairo_font_face_reset_static_data () from
/usr/pkg/lib/libcairo.so.2
#6 0xb9b48063 in cairo_debug_reset_static_data () from
/usr/pkg/lib/libcairo.so.2
#7 0xba4d3703 in gtk_moz_embed_get_title () from
/usr/pkg/lib/xulrunner/libxul.so
#8 0xba4dc1f3 in XRE_main () from /usr/pkg/lib/xulrunner/libxul.so
#9 0x0804a9e8 in _rtld_setup ()
#10 0x08049d84 in ___start ()
#11 0x08049ce7 in _start ()
When I run firefox a second time, the log file is reasonably different and the
back trace is quite different:
#0 0xb963a10b in FcPatternObjectPosition () from
/usr/pkg/lib/libfontconfig.so.1
#1 0xb963a169 in FcPatternObjectFindElt () from
/usr/pkg/lib/libfontconfig.so.1
#2 0xb963b284 in IA__FcPatternRemove () from /usr/pkg/lib/libfontconfig.so.1
#3 0xbb6e10f6 in gfxPangoFontGroup::GetFcFamilies () from
/usr/pkg/lib/xulrunner/libxul.so
#4 0xbb6e28a8 in gfxPangoFontGroup::NewFontEntry () from
/usr/pkg/lib/xulrunner/libxul.so
#5 0xbb6e73c0 in gfxPangoFontGroup::MakeFontSet () from
/usr/pkg/lib/xulrunner/libxul.so
#6 0xbb6e9183 in gfxPangoFontGroup::GetBaseFontSet () from
/usr/pkg/lib/xulrunner/libxul.so
#7 0xbb6e9d7d in gfxPangoFontGroup::GetBasePangoFont () from
/usr/pkg/lib/xulrunner/libxul.so
#8 0xbb6ea8a9 in gfxPangoFontGroup::GetFontAt () from
/usr/pkg/lib/xulrunner/libxul.so
#9 0xbb4f16a5 in non-virtual thunk to nsPrintSession::Release() () from
/usr/pkg/lib/xulrunner/libxul.so
#10 0x089e1280 in ?? ()
#11 0x00000000 in ?? ()
I've also tried (suggested in the thread you pointed me too) running:
/usr/pkg/lib/xulrunner/xulrunner-bin \
/usr/pkg/lib/firefox/application.ini -safe-mode -jsconsole
but that gives the same backtrace as the previous back trace.
Does the existance of cairo and pango in these traces point to general
X/fontish related problems? The only other "gui" type programs I really
use are amarok and digikam, and both of these appear to be working fine.
Cheers,
Simon.
Attachment:
firefox-a.out.bz2
Description: Binary data
Attachment:
firefox-b.out.bz2
Description: Binary data