Date: Thu, 09 Nov 1995 14:48:32 -0800
   From: Greg Earle <earle@isolar.Tujunga.CA.US>

   ...
     1108 netscape-2.0beta^->o0!W[ CALL  connect(0x7,0x450d38,0x10)
     1108 netscape-2.0beta^->o0!W[ RET   connect 0
     1108 netscape-2.0beta^->o0!W[ CALL  old.send(0x7,0xf7ffccd0,0x28,0)
     1108 netscape-2.0beta^->o0!W[ GIO   fd 7 wrote 40 bytes
	  "\0\^A\^A\0\0\^A\0\0\0\0\0\0     netbsd4me\^Cjpl\^Dnasa\^Cgov\0\0\^A\0\
	   \^A"
     1108 netscape-2.0beta^->o0!W[ RET   old.send 40/0x28
     1108 netscape-2.0beta^->o0!W[ CALL  gettimeofday(0xf7ffc828,0)
   ...
     1108 netscape-2.0beta^->o0!W[ CALL  old.sigblock(0)
     1108 netscape-2.0beta^->o0!W[ RET   old.sigblock 8192/0x2000

All normal up to here...

     1108 netscape-2.0beta^->o0!W[ CALL  old.sigstack(0,0xf7ffc734)
     1108 netscape-2.0beta^->o0!W[ RET   old.sigstack 0

Program registers signal stack information...

     1108 netscape-2.0beta^->o0!W[ CALL  obs_4.2(0x45d158)
     1108 netscape-2.0beta^->o0!W[ PSIG  SIGALRM caught handler=0x4400c28 mask=0x0 code=0x0

Looks like a call to sleep(xxx); in SunOS libc... I could be wrong...

     1108 netscape-2.0beta^->o0!W[ NAMI  "netscape-2.0beta.core"

Appears as if it dies before returning from the sigalarm handler in
user space.  I would bet any amount of money the sigill happened
within the userland signal trampoline code.  With the output of:

bash$ adb /path/to/netscape/bin/netscape ./core
$c
0xXXXXXXXX,20?ai

Where '0xXXXXXXXX' is the instruction address where it died.  I could
tell you much more.

Later,
David S. Miller
davem@caip.rutgers.edu