[Pythonmac-SIG] First experiences with the python24-fat tree.

[Bob Ippolito]

On Feb 4, 2006, at 12:26 PM, Ronald Oussoren wrote:

>
> On 4-feb-2006, at 21:02, Ronald Oussoren wrote:
>>
>> What I don't quite understand is that the latter patch doesn't  
>> quite work: the i386 contains weak links to fstatvfs and other  
>> 10.4-only symbols, but the ppc version does not. A small test  
>> program does work correctly.
>
> I hope nobody noticed this e-mail :-(. You have to set  
> MACOSX_DEPLOYMENT_TARGET=10.4 (or any value except 10.1) before  
> building, else gcc ignores weak linking instructions for ppc  
> builds. I've solved this more permantently by adding these lines to  
> Makefile.pre.in (just after settings CONFIGURE_MACOSX_DEPLOYMENT):
>
> #
> # Ensure that the deployment target doesn't change between  
> configure and
> # build time.
> MACOSX_DEPLOYMENT_TARGET=@CONFIGURE_MACOSX_DEPLOYMENT_TARGET@
>
> This ensures that make and distutils will use the same setting of  
> MACOSX_DEPLOYMENT_TARGET. A better fix might be to drop  
> CONFIGURE_MACOSX_DEPLOYMENT_TARGET entirely, I'll do that later.

Maybe what we should do is assume MACOSX_DEPLOYMENT_TARGET=10.3  
unless it's set manually, and make sure to use it in the makefiles  
any time a compiler is invoked.  distutils already ensures  
MACOSX_DEPLOYMENT_TARGET is set for anything it does.

I'm also thinking about having distutils.util.get_platform() return  
macosx-$MACOSX_DEPLOYMENT_TARGET-arch rather than the actual current  
version, since for all intents and purposes that's what distutils is  
going to be producing binaries for.

-bob