GNOME Bugzilla – Bug 318863
browser | clicking Hebrew Links leads to assersion
Last modified: 2006-02-21 11:38:08 UTC
Steps to reproduce: 1. open firefox 2. goto www.ynet.co.il click on one of the links in hebrew 3. the window is acting wierd, close and refresh a few times then the headers of all the windows in the gnome disapears the x log shows this Window manager warning: Log level 6: file hebrew-fc.c: line 314(hebrew_engine_shape): assertion failed: (face) Stack trace: i'm not sure which proccess crashes... i'd add a trace if i'll get any pointers Other information:
easy to reproduce via any browser (gleon etc') note: i'm using gentoo
I can confirm that this page makes epiphany extremely slow (well, or maybe it's going to hang forever, didn't take enough time to find out). It's using 100% CPU and when I interrupt it and it happens to be in pango, it is this (or below this):
+ Trace 63537
However it's far from clear that this is a pango problem and more likely a firefox problem. In firefox trunk it doesn't hang but is still extremely slow, proabably due to the ridiculously heavy flash usage on that page. Suggesing NOTGNOME.
i've reproduced it using epiphany as well (again the same behavior) i believe it's the window manager that is crashing (it worked fine till the last update i've had last night to the gnome... ) if i knew which proccess is the window manager i could produce a trace (p.s the firefox or any other broswer doesn't crash, only the window manager )
i've found why it crashed the metacity. i've copied into /usr/share/fonts/TTF the fonts i've been using in windows these files were masked with no read permission to regular users. in the perviouse versions of pango or metacity that wasn't a problem (it was handles correctly - it just didn't use them) with the current gnome/pango/metacity the metacity crashes i guess it's not critical but it is a regression from previous versions.
Quite possible that it's a Cairo bug.
Crashes for me with Epiphany inside Pango - the backtrace looks screwed up so some valgrind action would be useful. I'd tend to blame the Hebrew shaper rather than Cairo ... A cairo bug would be unlikely to show up just on this one page.
I actually hit this (or something like this) myself with an incoming mail in evo. It turned out to be a combination of: - Decreased robustness against missing fonts (sort of bug 314239) Pango was never good, but it seems to be considerably worse these days. - fontconfig problems. It never seems to correct out of date information in ~/.fonts.cache-2. rm ~/.fonts.cache-2 was how I "fixed" the problem for myself - it was pointing to a font that I moved to a different place months ago.
~/.fonts.cache-2 stuff is really new and experimental. Getting better everyday.
*** This bug has been marked as a duplicate of 314239 ***