Build arch for configure based upon CPUTYPE, prototype (was: Re: cvs commit: ports/devel/glib20 Makefile)

[Kris Kennaway]

On Sat, Jan 13, 2007 at 07:40:22PM +0100, Alexander Leidinger wrote:
> Quoting Jean-Yves Lefort <jylefort at FreeBSD.org> (Fri, 12 Jan 2007 12:13:06 +0100):
> 
> > On Fri, 12 Jan 2007 11:17:26 +0100
> > Alexander Leidinger <Alexander at Leidinger.net> wrote:
> > 
> > > Quoting Jean-Yves Lefort <jylefort at FreeBSD.org> (from Thu, 11 Jan 2007
> > > 17:21:36 +0000 (UTC)):
> > >
> > > > jylefort    2007-01-11 17:21:36 UTC
> > > >
> > > >   FreeBSD ports repository
> > > >
> > > >   Modified files:
> > > >     devel/glib20         Makefile
> > > >   Log:
> > > >   On i386 (>= 486) and amd64, fix the arch passed to configure so that
> > > >   glib will implement atomic operations with assembler instructions
> > > >   rather than mutexes (improves performance).
> > >
> > > What about setting GNOME_ARCH based upon CPUTYPE instead of hardcoding
> > > it to i486?
> > 
> > The problem is not specific to GNOME. I prefer the solution below.
> > 
> > > Or much better: set --build in bsd.port.mk globally based upon CPUTYPE
> > > (to iX86 on IA32, I'm not sure if it is wise to change the amd64 one
> > > globally).
> > 
> > Yes. And do pass x86_64 for amd64, that's what a majority of configure
> > scripts expect since they have linux in mind.
> 
> Hi, the proof of concept is attached. Configure shows i686-portbld...
> when building glib20 on my machine (CPUTYPE set to athlon-xp).
> 
> For the final patch I want to sort the list of CPUTYPEs. I'm sending
> this version to get some testresults (failing ports?) and
> reviews/comments.

This is likely to break large numbers of ports, so we can only proceed
if you're willing to deal with that.

Kris
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-ports/attachments/20070113/615adfb4/attachment.pgp