Message-Id:	<199811070306.TAA11240@geocities.com>
From:	cssl AT geocities DOT com
Sender:	cssl AT mail DOT geocities DOT com
To:	bowman <bowman AT montana DOT com>, djgpp-workers AT delorie DOT com
Date:	Sat, 7 Nov 1998 03:20:38 GMT0BST
MIME-Version:	1.0
Subject:	Re: V 2.02 'free' problem
X-mailer:	Pegasus Mail v3.31
Reply-To:	djgpp-workers AT delorie DOT com

 > Still playing with it, but nothing is jumping out to grab me. It
 > certainly looks like a problem with the block structure being
 > corrupted. MC is known to have memory leaks, but if it is overrunning
 > the buffer, i am in deep trouble, since the failure occurs from many
 > different sequences of calls. I've exercised it through the same
 > steps, and after 4 or 5 thousand invocations of 'free' something goes
 > sour.
Have you already tried FORTIFY or similar libraries ? They can help a 
lot... Some bugs become EVIDENT when (for example) the allocated memory is 
initialized with a certain pattern... Buffer overruns are detected simply 
allocating a bigger buffer with "guards" around. If at free() a guard is 
not the same as it was at malloc, an overrun happened...
Well, all this (and something more) is handled by (Z)FORTIFY...

BYtE,
 Diego
 

- Raw text -