From: alainm AT news DOT RISQ DOT QC DOT ca (Alain Magloire)
Subject: Re: Problems with MALLOC and FREE
Newsgroups: comp.os.msdos.djgpp
References: <Pine DOT SUN DOT 3 DOT 91 DOT 991209160149 DOT 22560D-100000 AT is> <82rh3j$sfu AT cantine DOT wu-wien DOT ac DOT at> <385EC088 DOT 8403BF22 AT ibm DOT net> <7GM74.500$PK3 DOT 5161 AT dfiatx1-snr1 DOT gtei DOT net> <385FE52F DOT E2F5572B AT east DOT sun DOT com> <RIX74.3450$PK3 DOT 29375 AT dfiatx1-snr1 DOT gtei DOT net> <38609B74 DOT 75D7B734 AT acm DOT org>
X-Newsreader: TIN [version 1.2 PL2]
Lines: 61
Message-ID: <W%784.11$36.1990@carnaval.risq.qc.ca>
Date: Wed, 22 Dec 1999 17:37:58 GMT
NNTP-Posting-Host: 132.206.63.174
X-Complaints-To: abuse AT mcgill DOT ca
X-Trace: carnaval.risq.qc.ca 945884278 132.206.63.174 (Wed, 22 Dec 1999 12:37:58 EST)
NNTP-Posting-Date: Wed, 22 Dec 1999 12:37:58 EST
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com


Very nicely said, Eric.  It is worth a repost, see below the hole discussion ;-)
I don't know why, specially newbies, they all blame the library or the
compiler, when the programs crash. And 99.99%, it's in the user code.
How many times, we've heard : "it works when I do this, but crash when I
do that",  or the true classic ..... "It used to work fine!!!!!"

Well guess what, this what "undefined behaviour" is all about ;-)

Was not a tool call YAMD, to help somewhere ?

--
alain

Eric Sosman (esosman AT acm DOT org) wrote:
: Damian Yerrick wrote:
: > 
: > "Eric Sosman" <eric DOT sosman AT east DOT sun DOT com> wrote in message
: > news:385FE52F DOT E2F5572B AT east DOT sun DOT com...
: > > Damian Yerrick wrote:
: > > > [...]
: > > > Oh, by the way, does anyone know what the standard
: > > > says should happen when you free(NULL)?
: > >
: > >     "If _ptr_ is a null pointer, no action occurs."
: > 
: > Simplifies things a bit.
: > 
: > #define Free(ptr)      free(ptr); ptr = NULL

:     This is a poor idea on two counts.  First, the macro
: expansion is faulty and will fail badly in situations like

: 	if (throw_it_away)
: 	    Free(ptr);

: or (harder to fix)

: 	Type **ptr;
: 	...
: 	Free (*ptr++);

:     The second drawback is that NULLing a free'd pointer may
: give an illusion of security which is completely false.  If it
: is to be effective, `ptr' must be the one and only copy of the
: free'd pointer value anywhere in the program; setting `ptr' to
: NULL will not affect any other now-invalid copies which may
: reside elsewhere.  Note, for example, that if `ptr' is a
: function argument, the proposed Free() macro (in a corrected
: version) would only change the function's local copy, not the
: original in the function's caller.

:     It is much better -- in fact, it is essential -- to use
: proper discipline in managing dynamic memory.  Write your
: programs so that the responsibility for acquiring and releasing
: memory (and other resources, for that matter) is well controlled,
: and don't count on unreliable tricks to cover up carelessness.

: -- 
: Eric Sosman
: esosman AT acm DOT org