Subject: backtracing signals
To: None <email@example.com>
From: Emmanuel Dreyfus <firstname.lastname@example.org>
Date: 05/28/2001 11:01:02
Now Linux's gdb is working in emulation, I have a better idea of the bug
that breaks JVM native threads and Opera. It's occuring exactly at the
same place in both programs.
gdb says we are getting a SIGBUS on return from linux_sys_rt_sigsuspend.
I've been trying to find out where this SIGBUS was coming from for a few
days now, without success.
If I drop into ddb on signal delivery, ddb trace command is just able to
tell me I'm coming from trap. I had a look to trap() in
sys/arch/powerpc/powerpc/trap.c. A SIGBUS can be fired from there if
there is an alignement exception raised. I aded some printf to test
this, it's not the case.
The question are: where a SIGBUS can come from, and is there anty better
way of backtracing it than bloating the kernel if prints?