Subject: Re: port-i386/2564: failure of i387 exp() function
To: None <netbsd-bugs@NetBSD.ORG>
From: John M Vinopal <firstname.lastname@example.org>
Date: 06/21/1996 20:10:44
I may take back my implication of the exp() function. Adding suitable
debug printf()s changes the execution of the code such that it works
with -O, but fails with -O -fcaller-saves. Removal of some of these
printf()s (specifically in the end of the patch() routine) produces
code that fails with -O but works with -O0.
So is it a library problem, a codegen issue, buggy code, or ???
Any help appreciated. FreeBSD running gcc 2.6.3 produces expected
results. My test case is: a.out 17 8 0