gmp3 with 4.08.x

Michael Weber michael.weber@post.rwth-aachen.de
Wed, 25 Apr 2001 12:52:47 +0200


On Wed, Apr 25, 2001 at 09:42:51 +0100, Simon Marlow wrote:
> > On Tue, Apr 24, 2001 at 15:55:59 +0000, Marcin 'Qrczak' Kowalczyk wrote:
> > > Use -O -fasm
> > 
> > Option order matters?!?!  That's evil!
> 
> In this case, the option order mattered in 4.08 but not for 5.00.
> However, option order matters in lots of other ways: -I order matters
> (search path ordering), -l order matters (library link order), -O vs.
> -Onot order matters (the latest one wins) etc etc.

hmm, yeees, but AFAICT the ordering just affects options of the same
type (-Onot just affects other -O* options, or -lfoo just affects -l*
type options)...  Well, the point is moot and since it doesn't matter
for 5.00, I'll leave it as it is...

> Ok... so, where are we going with all this?  The changes to the NCG to
> support GMP 3 are pretty small and can be picked up from the HEAD easily
> enough.  

I was more thinking of forthcoming libgmp versions... it's quite
possible that again they use evil #define tricks.

> I assume your motivation is for easier packaging - GMP 2 packages aren't
> easily available, or whatever.  I don't think we'll switch the 4.08
> branch over to GMP 3, though: it's just too big a change to apply to the
> branch at this stage.

Ok, this is not the best situation, but I have to live with it then.


Cheers,
Michael