numpy & non-Apple gcc?

[Michael Dickens]

Does anyone know or remember why MP's version of numpy (I use py26-numpy) requires non-Apple gcc?  Why -not- use Apple's GCC?  In playing around with various installs (+gcc44, +universal), reading through debug outputs, and testing builds, I do not find that fortran is required to compile (except for tests, which IMHO can be ignored); numpy uses the fortran compiler to link only, but any GCC can do that just as well.  Further, Apple's GCC can easily do multiple -arch's, making a universal install simple compared with (e.g.) gcc44 -- which requires the use of the muniversal portgroup and some other Portgroup tweaking.  I'm not against using gcc44 (or 43 or 45 or whatever), just wondering if anyone knows.  Thanks! - MLD