whisenhu@addamax.UUCP (05/22/89)
Well, this may be a bug in how the kernel does signal handling.
The BSD verison of /bin/sh allocated memory by just walking off the end
of memory, getting a segmentation violation, allocating more memory, then
returning to re-execute the instruction that caused the original fault.
What may be happening is that when the hardware takes the trap, the
PC points to the *next* instruction to be executed. This is probably
what the signal handler returns to. This was fixed for UTX2.1 (PowerNode)
by backing up the PC before storing so that the proper behavior was emulated.
I currently don't have access to the source for NP or for ksh-i.
Gary Whisenhunt
UUCP: {uunet}!uiucuxc!addamax!whisenhu Phone 1-217-359-0700
INTERNET: whisenhu%addamax@uxc.cso.uiuc.edu
BITENET: whisenhu%addamax@uxc.cso.uiuc.edu
MILNET: whisenhu%addamax@uiucuxc.arpa
CSNET: whisenhu%addamax%uxc@uiuc.csnet
US Mail: Addamax Corp, 2009 Fox Drive, Champaign, IL 61820