[Drawkit] Important Bug Fix

[Brad Larson]

Sure enough, that does the trick with the crashy sheets.  I figured it  
was something bizarre when I did a quick look-through to see what was  
causing that behavior.

Earlier this year, I spent a little under a month tracking a random  
crashing bug all the way to an sprintf call that looked correct, but  
was corrupting memory on Leopard (lesson: avoid sprintf if you can).

On Jun 5, 2008, at 7:37 PM, Graham Cox wrote:

> I'm putting this on the list ahead of a downloadable update because  
> it's fairly important.
>
> In DKGradient.m, change the following function:
>
> static int cmpColorStops (DKColorStop* lh, DKColorStop* rh, void  
> *context)
> {
> 	#pragma unused(context)
> 	
> 	float lp = [lh position];
> 	float rp = [rh position];
> 	
> 	if (lp < rp)
> 		return NSOrderedAscending;
> 	else if (lp > rp)
> 		return NSOrderedDescending;
> 	else
> 		return NSOrderedSame;
> }
>
>
> Previously the lh and rh parameters were passed as type id, now  
> they're explicit. (Also update the prototype or remove it).
>
> This change allows you to compile against the 10.5 SDK without any  
> runtime problems. This might seem counter-intuitive and it is - took  
> me 11 hours to track this one down - but the call to the -position  
> method was corrupting memory when using the 10.5 SDK because new  
> Leopard classes implement -position, returning larger data types  
> than a float, which was overwriting memory leading to bizarre  
> runtime bugs like sheets failing to work. The strong typing of lh  
> and rh make sure the code is compiled correctly.
>
> cheers, Graham
> _______________________________________________
> Drawkit mailing list
> Drawkit at lists.apptree.net
> http://lists.apptree.net/listinfo.cgi/drawkit-apptree.net

______________________
Brad Larson
SonoPlot, Inc.
3030 Laura Lane, Suite 120
Middleton, WI 53562