[Distutils] GCC versions and binary egg names

P.J. Eby pje at telecommunity.com
Tue Jul 27 17:58:49 CEST 2010

At 10:22 AM 7/27/2010 -0400, Barry Warsaw wrote:
>On Jul 26, 2010, at 09:26 PM, P.J. Eby wrote:
> >At 02:17 PM 7/26/2010 -0400, Barry Warsaw wrote:
> >>I wonder if any distutils experts would care to comment on PEP 3149,
> >>and its applicability to this problem?
> >>
> >>http://www.python.org/dev/peps/pep-3149/
> >>
> >>Is there any way this PEP can help?
> >
> >The d/m/u flags look like what's needed, *within* a given platform.
> >The overall platform definition (OS, processor, gcc, etc.) is where
> >things are still hairy, though.  (Binary eggs are of course already
> >tagged with Python major/minor versions, so it's only the
> >OS/processor/etc. that remains an open issue for platform strings.)
>I'm disinclined to expand PEP 3149 to include other platform designations.

To be clear, I wasn't asking for it to.  I was just saying that egg 
platform strings could adopt the d/m/u flags convention, but it 
wasn't in itself sufficient for getting the egg platform string mess 
sorted out on non-Windows/non-OSX platforms.

>Ronald brought this up on python-dev regarding OS X, but it adds complexity
>that isn't necessary for my particular use case.  I'm in favor of keeping the
>spec narrow - unless you feel otherwise, and someone comes up with an elegant
>way to support the egg use case without making the file names really long and

Nope, I don't see a need here.  If an egg platform string is 
involved, it means that the egg's contents are already segregated, 
and the platform info then needn't be duplicated on the .so files within it. 

More information about the Distutils-SIG mailing list