On Mon, 2009-10-05 at 12:58 +0200, Fabian Groffen wrote:
> On 05-10-2009 11:56:11 +0100, Alan Hourihane wrote:
> > On Mon, 2009-10-05 at 11:45 +0200, Fabian Groffen wrote:
> > > On 04-10-2009 10:02:07 +0200, Fabian Groffen wrote:
> > > > > What happened ?
> > > > 
> > > > Good questions.  I merged this change from gx86 already with the
> > > > comment that this would probably break (again ...).  Also in the ebuild,
> > > > internal libffi is removed conditional now, but that would mean
> > > > virtual/libffi would have to have keywords for ~mips and ~m68k now.
> > > > It does not.  Seems bugzie is down at the moment, so I can't check if
> > > > there are already bugs for this issue.
> > > 
> > > ~m68k and ~mips keywords were just recklessly dropped from the ebuild.
> > > 
> > > Given that python compiled fine at some point in the past, does Python's
> > > internal libffi compile on m68k-mint, or did it just skip libffi
> > > dependant modules?
> > 
> > python-2.6.2-r2 compiled just fine without libffi when the ebuild had
> > the ~m68k-mint keyword on virtual/libffi.
> 
> yeah, but back then it didn't rm -Rf internal/libffi

The Modules/_ctypes/libffi is the same as the external version. So no,
it won't compile.

Alan.