From: Alexandre Oliva (oliva@dcc.unicamp.br)
Date: Mon Feb 08 1999 - 13:28:34 EST
I've got a very slow host running SunOS 4.1.3. The good news is that
dlopen is now working on it. The bad news is that nothing else is :-(
Something is randomly breaking the state saved by setjmp() at thread
initialization; it usually ends up core dumping, but the core is
unusable because the stack is completely trashed.
Within gdb, the problem occurs quite frequently, always in thread
creation, in the second return from setjmp(). My guess was that the
problem was related with the signal handler being invoked while
setjmp() or longjmp() were running, which might cause the register
window to fill up, so that registers would have to be saved in the
stack frame. However, even when I blocked signals before setjmp() and
unblocked them after it and in the beginning of start_this...(), the
problems remained... Any suggestions? Should we increase the size of
the initially copied stack frame?
-- Alexandre Oliva http://www.dcc.unicamp.br/~oliva aoliva@{acm.org} oliva@{dcc.unicamp.br,gnu.org,egcs.cygnus.com,samba.org} Universidade Estadual de Campinas, SP, Brasil
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:58:01 EDT