Subject: Re: longjmp
To: None <downsj@CSOS.ORST.EDU, port-hp300@sun-lamp.cs.berkeley.edu>
From: None <mycroft@gnu.ai.mit.edu>
List: port-hp300
Date: 07/26/1994 17:34:49
The current implementation of longjmp() makes a bad assumption about
d0 being preserved.  On the first call in a dynamically linked
executable, the jump to sigreturn() goes through the binder, and d0 is
corrupted.  (Actually, it's set to the entry point of sigreturn().)

Obviously this needs to be fixed.


------------------------------------------------------------------------------