[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
"_thread_sys__exit": debugging clue req'd
- To: misc_(_at_)_openbsd_(_dot_)_org
- Subject: "_thread_sys__exit": debugging clue req'd
- From: "Bruce M. Walker" <bmw_(_at_)_borderware_(_dot_)_com>
- Date: Tue, 19 Sep 2000 07:49:01 -0400 (EDT)
[Gee whiz, I went off and did non-developer stuff for a few years
and they changed everything on me! ;-) ]
I'm trying to debug why pppoe (usr.sbin version) dies under load
for me. Summary: I am launching it from ppp using an appropriate
set device "!/usr/sbin/pppoe -i fxp0"
line, and it works for light network activity, but if I copy a 100K
file in either direction, pppoe exits causing ppp to recycle the
connection.
Anyway, when I connect to the running pppoe process with gdb and
break on _exit, all I ever get for a stack trace is
_thread_sys__exit
_exit
Argh! I wanna see where it came from! I tried linking the whole
thing -static: no help. I now see that there appears to be two
libc's: ordinary libc and "reentrant" libc_r where presumably
everything is thread-safe.
But I just want a normal old-fashioned non-threaded version of the
code so I can get a straight-forward stack trace. Is that too much
to ask? :-)
Can anyone enlighten an old war-horse on "how it's done now"?
Thanks!
-bmw
Visit your host, monkey.org