Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!rutgers!ames!ucbcad!ucbvax!decvax!tektronix!sequent!mntgfx!dclemans
From: dclemans@mntgfx.MENTOR.COM (Dave Clemans)
Newsgroups: comp.sys.atari.st
Subject: Re: Mark Williams ver 2.0 info
Message-ID: <639@dclemans.mntgfx.MENTOR.COM>
Date: Tue, 12-May-87 11:34:34 EDT
Article-I.D.: dclemans.639
Posted: Tue May 12 11:34:34 1987
Date-Received: Sat, 16-May-87 01:28:16 EDT
References: <8704302259.AA00661@ucbvax.Berkeley.EDU> <4722@utcsri.UUCP> <692@ttidca.TTI.COM>
Organization: Mentor Graphics, Beaverton OR
Lines: 12
Summary: another known bug


Another Mark Williams C 2.0 bug;

there is as least one case where the compiler will, in an expression
that has generated a 16-bit signed temporary result and that temporary result
needs to be extended to 32 bits (again signed), generate the incorrect
code to do the sign extension.  Instead of the correct "ext.l" instruction,
it generates "ext.w", which clobbers the top byte of the 16 bit temporary
result.  Assuming that their assembler is written in C, the previously
reported bug in the assembler might even be caused by this.

dgc