Re: Enable Guard Malloc (was Re: Isolating a bug?)

FROM : daniel
DATE : Wed Apr 06 18:47:18 2005

I was wrong: the guard malloc feature *is* mentioned in Mac OS X 
Debugging Magic.  I  guess I just missed it.

Daniel

On Apr 6, 2005, at 9:31 AM, daniel wrote:

> I thought random memory stompers were hard to track down, too.  At 
> least, until I noticed just yesterday that I've been glancing right 
> over the "Enable Guard Malloc" option in the Xcode debug menu.
>
> I had a random memory smasher that I was dreading.  The MallocHelp 
> related utilities did nothing for me, so I was looking forward to 
> slowly and painfully paring down my program in a way that would 
> preserve the bug but reduce the amount of examinable code.
>
> Then I did a web search where I noticed somebody mentioning the "guard 
> malloc" feature (man libgmalloc for more info).  This feature is like 
> a super-paranoid "guard edges" feature.  From the man page:
>
>     When memory is freed, libgmalloc deallocates its
>     virtual memory, causing reads or writes to the freed buffer cause a 
> bus
>     error.  Thus, nasty, hard-to-track-down bugs are immediately obvious, 
> and
>     you'll know exactly which code is causing the problem.
>
> I switched it on, ran the program, and located the memory stomping 
> code on the first try.  I'm surprised this isn't mentioned in the 
> "Mac OS X Debugging Magic" technote.  Perhaps since it's included in a 
> clearly visible Xcode menu, it's not considered "magic."  I feel 
> pretty embarrassed now for not knowing about it, but I will certainly 
> never forget it.
>
> I'm sure there are "memory stompers" of some kind that fall outside of 
> the scope of this tool's usefulness, but I am guessing that the vast 
> majority of memory stompers will be located by enabling this tool and 
> simply waiting.  Grab some coffee while your application launches.
>
> Daniel
>
>
> On Apr 6, 2005, at 4:41 AM, Ondra Cada wrote:
>

>> If that does not help, I kind of fear the problem *may* be elsewhere, 
>> in an uninitialized pointer or somrthing alike, whose result are 
>> mangled in-memory data. That's a bug alas very difficult to find, and 
>> assembly-level debugging might be needed. Well, hopefully, the cause 
>> would be something more trivial :)

>
> _______________________________________________
> Do not post admin requests to the list. They will be ignored.
> Cocoa-dev mailing list      (<email_removed>)
> Help/Unsubscribe/Update your Subscription:
> http://lists.apple.com/mailman/options/cocoa-dev/<email_removed>-
> sweater.com
>
> This email sent to <email_removed>