[110141] trunk/dports/math/fftw-3/Portfile

[Michael Dickens]

I think the multiple variants maybe did not work because they were not
"eval"d. Which is what the recent patch does, and thus almost succeeds. 
It does not handle -gcc48 (alone, no other variants) correctly. I don't
have time to fix it right now, but I can look at it tonight if there's
still an issue. I'm also going back and looking through the "Fortran
recipe" thread to see what I missed.  I do think it would be good to
have all Fortran-variant ports to use +- the same structure. - MLD

On Tue, Aug 27, 2013, at 02:23 PM, Joshua Root wrote:
> I think it actually declares multiple variants with the same name, but
> only one gets found when iterating through all a port's variants for
> display or checking conflicts. I think all of the code in the variant
> bodies would get executed though (if there was any)?
> 
> The attached patch should work.
> 
> On 2013-8-28 03:55 , Jeremy Huddleston Sequoia wrote:
> > So that clobbers the previous rather than appending?  That seems counter intuitive since the block is executed in each case (and it seemed to work fine for me with the original version).
> > 
> > Do you have a suggestion for how to do what is intended?  The approach used my mpich (constructing a variable listing all conflicts) doesn't work.
> > 
> > On Aug 27, 2013, at 10:51, Joshua Root <jmr at macports.org> wrote:
> > 
> >> I'd say the problem is declaring each variant gcc${ver_no_dot} multiple
> >> times.