[3527] in Kerberos
Re: RS/6000 configuration file for Version 4 Keywords:
daemon@ATHENA.MIT.EDU (John Gilmore)
Fri Jul 8 20:08:47 1994
To: tytso@MIT.EDU (Theodore Ts'o)
Cc: eichin@MIT.EDU ("Mark W. Eichin"), jjorgens@wsuaix.csc.wsu.edu,
kerberos@MIT.EDU, gnu@cygnus.com
In-Reply-To: Your message of "Fri, 08 Jul 1994 13:41:58 EDT."
<9407081741.AA12400@tsx-11.MIT.EDU>
Date: Fri, 08 Jul 1994 16:43:50 -0700
From: John Gilmore <gnu@cygnus.com>
> compile with the -O flag --- there's a bug in the native compiler which
> screws up the optimization of the "left shift" operator.
Yeah. Doing the Mac port of V4 found two compiler bugs (one in MPW,
one in THINK) that the Ferguson DES code tickled. In one, they lost
the "unsigned"ness of a number that had been &'ed and then shifted
right. I forget the other. (Both were easy to circumvent once we
pried apart enough of the code to find them.) It's clear that few
compiler vendors are regression-testing the shift and logical
operators.
John
