Date: Mon, 3 Jul 1995 03:01:24 -0700 (PDT) From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> To: gena@NetVision.net.il (Gennady Sorokopud) Cc: hackers@FreeBSD.org Subject: Re: gcc-2.6.3 with bounds checking Message-ID: <199507031001.DAA18102@gndrsh.aac.dev.com> In-Reply-To: <XFMail.950703125210.gena@NetVision.net.il> from "Gennady Sorokopud" at Jul 3, 95 12:59:17 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > Hello! > > In the last 2 days i experimented with bounds-checking gcc. > I find that it's very usefull debugging tool but it also has it's > drawbacks: > > 1) it's too strict :-( > For example it reports "reference to undefined pointer" > when you try something like: if (!ptr) ... > and ptr was not initialized before (i know, it's a bad style, > but not an error) It _is_ an error, referencing an object that is undefined is an undefined operation, the above if (!ptr) is indeterminate and would depend on what ever trash was in ptr :-(. This is more than just ``bad style'', the bounds-checking gcc has indeed pointed to a program flaw. > 2) if it finds an error it aborts the program, and this really bothers > if error is in some system library :-) > (I modified libcheck.a to "recover" after such situation but > it's still unstable) Humm, well, in a critical application that must not ever use bad data I would want it to abort and call my exit routines so I could get a bug report about it, but in a debug situation I would just like to see the errors flagged and the program continue. When you added this ``recover'' function, did you make it possible to turn it on and off with some type of knob? > If you want to try it, you can ftp it from: > ftp://Burka.NetVision.net.il/pub/bounds/cc_bounds.tar.gz Humm... wonder if it would work on the kernel :-) :-). -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Reliable computers for FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199507031001.DAA18102>