[Brass]

After a couple of beta testers found a weird behaviour in a new beta version of an application, I eventually tracked it down to the problem only occurring with PPC Macs, and then down to the following line of code:

if ( [[[thing amount] num2] floatValue] > 0.0 )

This worked fine on Intel Macs, but not on PPC Macs.
[[thing amount] num2] returns an NSNumber and the problem appears to occur when it returns nil. So a simple check for nil added to the logic fixed it completely, as shown below:

if ( [[thing amount] num2] && [[[thing amount] num2] floatValue] > 0.0 )

Surely this is absurd? Surely [<nil expression> floatValue] > 0.0 should always return NO/FALSE???? And therefore the addition should do nothing?

But I can assure you that in this case the addition fixes the problem completely.

Can anyone enlighten me as to why this behaviour occurs?

[Brass]

OK answered my own question, after further research.

Here's a discussion of the problem on the Cocoa mailing list... Cocoabuilder - (M. Carlson) Bug or feature?

Which led me to Apples official documentation of the behaviour: Universal Binary Programming Guidelines, Second Edition: Objective-C: Messages to nil

And I was naive enough to think that messages to a nil object would always return 0/nil/FALSE?
In short, that is only guarrenteed on Intel Macs, or if the message has an integer/object return type!