On Sat, 2002-07-06 at 17:31, Gwenole Beauchesne wrote:
Hi,
> Hi,
> Could someone please tell me how to reliably reproduce that? The crash
> looks nice and would be great to get the actual params chain given to
> the gmp library. I will try to have a look on that one too.
>
> Bye,
> Gwenole
>
>
I did some experimentation and found a way to reproducce the crash
systematically. First off, I started a bare X session with xinit (to
preclude the possibility of complex interactions with the GNOME
desktop). In the xterm, I then started sawfish with errors logging as
in: sawfish 2>errors.log &
After on, starting three instances of "gedit" in succession always
produces a crash on the third instance. The same thing happens with the
galeon browser. In this case however, it doesn't matter if the
instances are started from the xterm or by the File->New window menu:
the crash always happens on the third instance. I also experimented
with the mozilla browser. In this case, five windows need to be created
to produce a crash.
Attempts to create the crash with simple X apps like xterm or xman have
not been successful though. Starting a dozen instances of a KDE app
like kedit doesn't make sawfish freak either...
Quite strange! Some apps make sawfish crash but not others...
Anyway, here is the content of the error log:
--------------------------------------------
randraw.c:118: GNU MP assertion failed: ((mp_limb_t) 1 << m2exp) > c
rep: received fatal signal: Aborted
struct debug_buf common:
Backtrace in `fatal_signal_handler':
<(null)+4385352>
<abort+395>
<(null)+1948234>
<(null)+1951180>
<__gmp_rand+299>
<__gmpz_urandomm+221>
<(null)+1410273>
<(null)+1355253>
Lisp backtrace:
#2 random ...
#1 place-window-randomly ...
#0 place-window (#<window e00020>)
-------------------------------------
All the crashes produce invariably the same log entry except for the
window number (in the last line).
Bye,
Eric
