Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!think!ima!haddock!karl
From: karl@haddock
Newsgroups: net.lang.c
Subject: Re: Structure alignment question
Message-ID: <86900047@haddock>
Date: Mon, 8-Sep-86 16:48:00 EDT
Article-I.D.: haddock.86900047
Posted: Mon Sep  8 16:48:00 1986
Date-Received: Mon, 8-Sep-86 20:52:49 EDT
References: <3550@brl-smoke.ARPA>
Lines: 16
Nf-ID: #R:brl-smoke.ARPA:3550:haddock:86900047:000:841
Nf-From: haddock!karl    Sep  8 16:48:00 1986


rbbb@rice.EDU (David) writes:
>C compilers that I have seen supporting structure-valued functions pass a
>hidden parameter to the function containing the address to store the result.
>Can a compiler assume that the storage for the result is word-aligned?
>(That is, can a compiler make the same assumptions when returning a
>structure that it does when assigning a structure?)

Certainly.  It can make this assumption for *any* expression of type "pointer
to struct foo", whether it's explicit or (as in this example) generated by
the compiler.  *All* instances of a type (including those generated by a
pointer dereference) have the alignment of that type.  The user can sometimes
circumvent this (type punning with casts or unions), but the result is
undefined.

Karl W. Z. Heuer (ima!haddock!karl; karl@haddock.isc.com), The Walking Lint