Red Hat Bugzilla – Bug 372391
Firefox crashes on some fonts
Last modified: 2007-11-30 17:12:21 EST
Description of problem:
Firefox crashes on some sites. The problem seems related to some fonts, just
before the crash ff prints out this:
(gecko:2883): Pango-WARNING **: failed to create cairo scaled font, expect ugly
output. the offending font is 'Lucida Sans 8.25'
Gdb tells me this (first 13 items from backtrace in attachment):
0x00f1c6bd in nsHTMLReflowState::CalcLineHeight (aPresContext=0xaecb988,
When I don't allow websites to use other fonts then I've defined in the
preferences the websites causing trouble work fine.
Version-Release number of selected component (if applicable):
But ff 126.96.36.199 alsa had the problem. The version that can be downloaded from
mozilla.org works fine.
Steps to Reproduce:
1. Make sure that you have Lucida Sans installed. It's included in the JVM's
from SUN. When installing the rpms and the compat packages from jpackage these
fonts are also available.
2. Make sure that "Allow pages to use their own fonts ..." is set in the preferences
3. Go to a website using this font. Like my own http://blog.zoeloelip.be
Firefox crashes (segmentation fault)
I can view the website!
Created attachment 252511 [details]
Backtrace from crash
Created attachment 253011 [details]
screenshot of unsuccesful attempt to reproduce
Cannot reproduce here with FF3alpha neither with 188.8.131.52. One possibility might
be something about the conflict between bitmap version of Lucida Sans from
xorg-fonts and the TTF ones from Java package?
Please, try to describe more details about your system to enable us to reproduce
this bug here.
Moreover, the backtrace is almost useless because it is the one only from one
thread of multithreaded firefox. You have to run gdb command
thread apply all backtrace
instead to get all threads.
The symlinks in /usr/share/fonts/java seemed to be dangling symlinks. I removed
them and the problem was gone. Afterwards I recated the dangling symlinks by I
can't reproduce the crash anymore.
OK, closing this bug per comment 4.