Re: Base additions on iOS 4?

[Richard Frith-Macdonald]

On 7 Feb 2011, at 21:17, David Wetzel wrote:

> Hi,
> 
> I made it work and would like to know if I should just commit the per file 
> changes back to trunk?
> The XCode project (with modified source files) is at
> 
> http://www.turbocat.net/GNUstepBase.zip
> 
> A example diff is attached.

Thanks, it's great that you have done this, but committing right now it may be 
a bit premature.

I looked at your example, and the first thing I noticed (since it's near the 
start) was the +initialize method.
I understand that NSMapTable is not available in iOS (though looking at the 
OSX10.6 documentation I don't see where it's deprecated from 10.6 on as your 
comment says), and I can see why you replace it with an NSMutableDictionary, 
but your code then goes on to do a load of calls like:
[attrNames setObject:(void*)@"XML_ATTRIBUTE_CDATA"  
forKey:(void*)XML_ATTRIBUTE_CDATA];
Now, the arguments here should be objects, not void*, and I would expect the 
key you are using (an integer constant from libxml2) to cause a crash when the 
dictionary tries to retain it.
So I don't understand how this could run (unless I've misread, or the iOS 
implementation of NSMutableDictionary is *very* different).

As you are presumably (by my standards at least) an iOS expert, having ported 
to it, can you please let me know:
a. how to get gnustep-make on a snow-leopard system to target the iOS 
Foundation rather than the Cocoa Foundation or,
b. if you only used xcode, do you know what options we might need to add to 
gnustep-make to target iOS?

The reason I ask is that I'm downloading the iOS SDK, and I'd like to try 
running the gnustep regression tests on it to see what the differences are.
If we can get gnustep-make and the testsuite to work for iOS on a mac running 
OSX, we can run the GSXML regression tests against your ported codebase, and we 
can add new regression tests for the features you are using (probably good for 
you and good for GNUstep).

Thanks.