From eichin@ATHENA.MIT.EDU Tue Nov 6 20:37:44 1990 SUM: Mark W. Eichin ->marc@MIT.EDU,watchmakers, sipb SUB: Re: debugging malloc Received: by E40-PO.MIT.EDU (5.45/4.7) id AA28246; Tue, 6 Nov 90 18:48:42 EST Received: from E40-008-8.MIT.EDU by ATHENA.MIT.EDU with SMTP id AA07915; Tue, 6 Nov 90 18:47:30 EST Received: by E40-008-8.MIT.EDU (5.61/4.7) id AA12449; Tue, 6 Nov 90 18:47:27 -0500 Date: Tue, 6 Nov 90 18:47:27 -0500 From: Mark W. Eichin Message-Id: <9011062347.AA12449@E40-008-8.MIT.EDU> To: marc@MIT.EDU Cc: watchmakers@ATHENA.MIT.EDU, sipb@ATHENA.MIT.EDU In-Reply-To: Marc Horowitz Tue, 06 Nov 90 06:02:20 EST <9011061102.AA17346@BILL-THE-CAT.MIT.EDU> Subject: Re: debugging malloc SUM: Marc Horowitz ->watchmakers,sipb SUB: debugging malloc I need a debugging malloc. Specifically, something which will do bounds checking on each call to malloc and free. Right now, g++ on the vax is dumping code in malloc. Does anybody have any experience with this, or a debugging malloc I can try using? This is for a problem set, so I'd appreciate any fast response. Thanks in advance. Marc attach eichin; /mit/eichin/projects/malloc/malloc.c (.o file there is for the vax, I think.) If you build it, make sure scribblecheck is #def'd. After every malloc and free it checks to make sure nothing in the space after or before the block in question has been trashed; it also detects multiple free's and other types of corruption. It calls botch with a string argument which describes the problem, though the source is generally the best commentary on why it botched. If you define "parent" (I think, check the code) it will log what caller freed or malloced the block most recently (this works on the vax; I *may* have gotten it working on the RT, but I doubt it. It involves walking up the stack to find the return address.) It's based on an old gnuemacs malloc; I put this together for Foundation when I was working for wdc. Many people have used it over the years (I think, there's a previous entry in Lore about it too...) _Mark_