Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uwm.edu!lll-winken!decwrl!orc!orc.olivetti.com!chase
From: chase@orc.olivetti.com
Newsgroups: gnu.gcc.bug
Subject: Alignment bug? on 386
Summary: Maybe this is just a misfeature
Message-ID: <48839@ricerca.UUCP>
Date: 12 Jan 90 01:09:28 GMT
Sender: news@orc.Olivetti.Com
Reply-To: chase@orc.olivetti.com (David Chase)
Distribution: gnu
Organization: Olivetti Research California (Menlo Park)
Lines: 20

This is a crappy bug report on two counts, but I'm reporting it anyway
and will try to provide more information in the future.  First, this
may be only a misfeature, and not a bug, and (second) I don't have
anything like a small program that provokes it (yet).

The bug:

On a 386, running Mach, and using gcc (v 1.36), gcc-as, and gcc-ld, it
is possible for a variable "x" declared

  "static void * x;"

to have an odd address.  This doesn't seem to happen usually.  I
realize that the 386 copes with this situation perfectly well, but it
doesn't appear to be what is intended to happen; pointers allocated
within records are aligned to a four-byte boundary, and pointers
declared in all my simple test programs are aligned to a four-byte
boundary.  Besides, it gives the garbage collector fits.

David