Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!mcvax!eurifb!wietse
From: wietse@eurifb.UUCP (Wietse Venema)
Newsgroups: comp.unix.wizards
Subject: Re: Problems with pointers
Message-ID: <214@eurifb.UUCP>
Date: Tue, 29-Sep-87 15:37:23 EDT
Article-I.D.: eurifb.214
Posted: Tue Sep 29 15:37:23 1987
Date-Received: Sat, 3-Oct-87 07:49:03 EDT
References: <9519@brl-adm.ARPA>
Organization: EUR/Bedrijfskunde,PObox 1738,Rotterdam,Neth.
Lines: 26
Summary: don't worry

In article <9519@brl-adm.ARPA>, GLASGOW@NOSC-TECR.arpa writes:
> int    *p;
> p = (some_big_structure *)malloc(sizeof(some_big_structure));
> 
> My problem arises when I try to release this memory back to the system. If
> I use free(), will it assume that p points to an integer and release only
> two bytes, or will it know that p is a larger structure and release the
> proper number of bytes.  Also, if it releases only two bytes, is there a
> way around this.  All of the different structures are large and memory is
> tight, so this is definitely a problem.

The calling sequence of free() is (from /usr/lib/llib-lc):

	free(p) char *p; 

All pointer values should be converted to char * before passing them to
free(). Hence, from the pointer value alone free() can't tell what block
of memory a pointer is pointing to. Malloc() keeps its own administration 
of the size of blocks of memory it doles out so that free() knows how much 
memory is being released. 

A nice example of an implementation of malloc() can be found in "The C 
programming language" by Kernighan & Ritchie.

	Wietse Venema		(uucp:		mcvax!eutwc1!wietse)
				(bitnet:	wswietse@heithe5)