Re: MATH_EMULATE

[Joerg Sonnenberger <joerg%britannica.bec.de@localhost>]

On Wed, Jan 23, 2008 at 12:43:46AM +1100, Simon Burge wrote:
> And then forgetting to add "-msoft-float" to programs
> (make configure; vi `find work*/ -name Makefile`; make install) and
> having everything compile but not work is good value too.

Yeah, that's what I mean with a subport -- the compiler default should
be -msoft-float to avoid exactly this problem. The problem why I want to
avoid MATH_EMUL is that correctly emulating the 387 is a pain and the
old GPL code was not perfect. The missing parts in userland are
relatively small as long as the compiler is dealt with, I think.

The point here is that -msoft-float is a userland problem and one we
deal with on other platforms anyway. MATH_EMUL is kind of like the VAX
port. After a release you get a number of complains that it doesn't work
as intended, but noone who is going to fix it.

Joerg