Re: crashes on intel

FROM : Shawn Erickson
DATE : Sat Jun 03 19:14:01 2006

On Jun 3, 2006, at 8:03 AM, Buddy Kurz wrote:

> FWIW: I experienced a similar problem when an application randomly 
> crashed at the same spot on Intel but worked OK on PPC.
> I was tracking a changing string value using something like if (!
> ([oldValue isEqualTo:newValue]) {...}
> The crash occurred on Intel when the oldValue had not been set.
> The problem was either -NSString isEqualTo:(NSString)arg crashing 
> on a nil object value (on intel)  or a non-nil un-initialized 
> object pointer in the method. I was unable to duplicate the problem 
> in a simpler test application.
>
> My solution was to initialize the oldValue to @"" (I was assuming a 
> nil value was the culprit)
> In retrospect, I'm wondering if there was just some random value in 
> the un-initialized variable.


That sounds like you had a bug in your code exposed when running on 
Intel.

If you attempt to send a message to an unintialized variable (stack 
based) you have the chance of attempting to send a message using an 
invalid pointer. The way the stack is managed on PowerPC and Intel is 
different so when running on Intel you likely had a better chance of 
hitting the issue.

Sending a message to nil will not cause a crash on either platform 
(unless you some how hit a bug in the objective-c runtime but I kinda 
doubt that).

-Shawn