GNOME Bugzilla – Bug 440514
Abort in pango_ot_tag_from_language
Last modified: 2007-05-26 01:18:59 UTC
I watch GNOME and gnumeric via svn, the latter through garnome. I rebuilt gnumeric today and had the following when I ran it: gnumeric Pango-CRITICAL **: pango_ot_tag_from_language: assertion `language != NULL' failed aborting... I also have a trace as many of the garnome libs were built with -g. Here it is: System: Linux 2.6.21.1 #59 SMP Mon Apr 30 11:14:00 EDT 2007 i686 X Vendor: The X.Org Foundation X Vendor Release: 60900000 Selinux: No Accessibility: Disabled GTK+ Theme: Gorilla Icon Theme: Gorilla Memory status: size: 73949184 vsize: 73949184 resident: 29188096 share: 24240128 rss: 29188096 rss_rlim: 4294967295 CPU usage: start_time: 1179856344 rtime: 48 utime: 43 stime: 5 cutime:0 cstime: 0 timeout: 0 it_real_value: 0 frequency: 100 Backtrace was generated from '/usr/bin/gnumeric' (no debugging symbols found) Using host libthread_db library "/lib/tls/libthread_db.so.1". [Thread debugging using libthread_db enabled] [New Thread -1232021824 (LWP 19827)] 0xb6b6704e in __waitpid_nocancel () from /lib/tls/libpthread.so.0
+ Trace 135554
Thread 1 (Thread -1232021824 (LWP 19827))
----------- .xsession-errors (300048696 sec old) --------------------- iceauth: creating new authority file /home/ronis/.ICEauthority --------------------------------------------------
It is only aborting because you have asked for criticals to abort. Anyway, something in very recent pango has changed. We could argue it is an ABI change. They would probably counter that we very doing something wrong. We would then say that maybe they should have had better docs. Etc. Try this patch. Index: src/item-bar.c =================================================================== --- src/item-bar.c (revision 15573) +++ src/item-bar.c (working copy) @@ -157,9 +157,11 @@ g_object_unref (layout); ib->pango.item->analysis.font = g_object_ref (ib->normal_font); + ib->pango.item->analysis.language = + pango_context_get_language (context); ib->pango.item->analysis.shape_engine = pango_font_find_shaper (ib->normal_font, - pango_context_get_language (context), + ib->pango.item->analysis.language, 'A'); ib->indent = ib_compute_pixels_from_indent (sheet, ib->is_col_header);
I updated in svn and rebuilt. I get past this abort, but still crash (I'm not sure that this is the same bug). Basically, when I do something that opens a window (e.g., opening a file). I crash. Here's a traceback, unfortunately, without gnumeric having debugging symbols (I'm rebuilding now). I'm pretty sure that this is not the same bug. I'm going to attach the traceback
Created attachment 88806 [details] Traceback Traceback for problem reported in preceding comment.
That is a totally different bug. Was that file created by Gnumeric? Something appears to be wrong with it. May we have a copy? If it is too senstive to attach here, perhaps you could email a copy to mortenw@gnome.org The crash appears to be inside code for a theme. Could you please try switching to the default theme and see if the problem persists? Further things to try... 1. Run with G_SLICE=debug-blocks 2. Run under valgrind with G_SLICE=always-malloc
I'm going to attach a log file below. The problem happens even if I don't have a file. Simply running gnumeric followed by File->open triggers the hang. One thing that is different is when I run under valgrind with G_SLICE=always-malloc; now it takes a while but ultimately the open-file dialog appears properly. Valgrind is complaining about several things though. As to theme, I'm running garnome-svn and there doesn't seem to be a default theme. (I'm using something called Gorilla). One last thing. Today garnome updated the versions of glib and gtk+ (to 2.13.2 and 2.11.0) respectively. Perhaps this is where the problem lies? E.g., the API may have changed.
Created attachment 88811 [details] More gdb and valgrind output Debugging output requested in earlier comment
That is bug 438456, fixed recently in gtk-themes. Note, that running with G_SLICE=always-malloc will hide the bug.
OK, but this is not bug 438456 correct?
This bug seems to be caused by gtk+. I'd upgraded to version 2.11.0 from 2.10.12. Downgrading back to 2.10.12 "fixes" the problem.
On more thing. I thought to downgrade because other apps were crashing when I tried to open a file. Turns out that there were lots of changes in gtkfilechooser in gtk+-2.11.0, clearly something is broken.
You probably need to upgrade gtk-engines too. That is a separate module. The fix from bug 438456 that you need is only in svn, not in any release. I suggest that you revert to the default theme. Closing -- the remainder is not a Gnumeric problem.