Path: chuka.playstation.co.uk!news
From: "pal" <pal_news@hotmail.com>
Newsgroups: scee.yaroze.freetalk.english
Subject: Re: Ten bytes bug in gcc?
Date: 15 Aug 2001 09:56:14 GMT
Organization: PlayStation Net Yaroze (SCEE)
Lines: 31
Message-ID: <01c12570$15068c00$a708e4d5@pal-s-omnibook>
References: <01c124b1$16cc8cc0$LocalHost@pal-s-omnibook> <9lbdjv$ds71@www.netyaroze-europe.com> <01c1250d$76401bc0$501ce4d5@pal-s-omnibook> <3B7A41C2.B4C5098E@127.0.0.1>
NNTP-Posting-Host: nas-cbv-6-8-167.dial.proxad.net
X-Newsreader: Microsoft Internet News 4.70.1155

Majik <majik@127.0.0.1> wrote in article <3B7A41C2.B4C5098E@127.0.0.1>...
<snip>
> mybuffer:	[reserve 4 bytes]
> 		[reserve 4 bytes]
> 		[reserve 2 bytes]
> dummy:		[reserve 2 bytes]
> main:		[some code]
> 
> the problem is, as far as I remember, is that neither of the labels
> dummy: or main: occur on a 4 byte
> boundary

Ok so I agree that the problem must be caused by bad alignment of main.
However, in your example, doesn't main == mybuffer + 12bytes? Then main is
4byte-aligned. dummy is not, but as it's not used, it's okay.

If main is not 4-byte aligned, then there have to be some filling bytes
between mybuffer and dummy (or dummy and main), but then these bytes cannot
have been reserved other than for alignment, can they? Then dummy would
have undergone alignment, but main would not, which is rather unfair
behaviour (esp. considering that code is usually more sensitive to
alignment).

Anyway, I can't understand how such a bug would not cause much more
havoc... any data not properly sized should cause the following functions
to crash, shouldn't it (as I said, it occurred to me with a struct, and I
guess there are a lot of structs out there that are not 4byte multiples)?


pal