Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!apple!apple.com!rpd
From: rpd@apple.com (Rick Daley)
Newsgroups: comp.sys.mac.programmer
Subject: Re: Stale Pointers
Message-ID: <2551@internal.Apple.COM>
Date: 28 Jun 89 21:53:41 GMT
References: <2394@ur-cc.UUCP>
Sender: usenet@Apple.COM
Distribution: usa
Organization: Apple Computer
Lines: 21

In article <2394@ur-cc.UUCP> giaccone@uhura.cc.rochester.edu (Tony 
Giaccone) writes:
>         (**myImage).rasI_image = (SIPtr) NewPtr(size);
>         if  ((**myImage).rasI_image == NULL) {
...
> it seems odd that Lightspeed C should do part of the dereference
> before the call, and the other part after the call.  What do you think?

This is a nasty problem that bites people every once in a while.  However, 
C does not define the order of evaluation in an assignment.  Your code 
might work with some compilers, but it's not safe.  The best thing to do 
is to use a temporary variable:
          tmp = (SIPtr) NewPtr(size);
          if  (((**myImage).rasI_image = tmp)== NULL) {
This is better than HLocking the block because it avoids fragmenting the 
heap.  The reason you saw the bug is that the Memory Manager had to 
compact the heap to make room for the new block.  If you lock the 
relocatable block, you might cause the NewPtr to fail.

                                            Rick Daley
                                            rpd@apple.com