I didn't notice it with your update so that's a good sign. The version that's in the distro right now (3.8.2-2; that's what I meant by 'existing') gets confused and decides that a core2duo machine is actually a pentium III:
http://akhmac.blogdns.net/finkbuilds/archives/2008-08.html#e2008-08-28T11_18_28.txt I'll see if the octave-atlas test case noted in that entry behaves properly or not. Jack Howarth wrote: > What exactly is the misbuild on i386 darwin? Unfortunately, all of > my machines are set up for x86_64 fink at the moment so it will take > a bit to test that. I did notice some apparently non-fatal issues with > the depreciated -mcpu option during the build. Is that what you are > referring to? > Jack > > On Fri, Jul 24, 2009 at 07:22:32AM -0400, Alexander Hansen wrote: > >> I'm testing this out now. The existing package misbuilds on core2duo >> machines on 10.5 (32-bit) and I've got a simple test case that I can >> try. >> >> On Jul 23, 2009, at 11:15 PM, Jack Howarth wrote: >> >> >>> I have uploaded test packaging for atlas-3.9.11-1 onto >>> fink tracking. This packaging for atlas (which is unmaintained >>> in fink) provides the changes... >>> >>> 1) Update to latest upstream 3.9.11 release. >>> 2) Switch build from gcc43 to gcc44. >>> 3) Adjust build to support x86_64 fink. >>> >>> The changes in the atlas.info file from the current release are... >>> >>> --- atlas.info 2008-08-28 18:49:53.000000000 -0400 >>> +++ /sw/fink/10.4/local/main/finkinfo/atlas.info 2009-07-23 >>> 22:22:13.000000000 -0400 >>> @@ -1,6 +1,6 @@ >>> Package: atlas >>> -Version: 3.8.2 >>> -Revision: 2 >>> +Version: 3.9.11 >>> +Revision: 1 >>> Description: Portably optimal linear algebra software >>> DescDetail: << >>> The current version provides a complete BLAS and LAPACK API. >>> @@ -32,14 +32,14 @@ >>> -L%p/lib -llapack -lcblas -lf77blas -latlas >>> << >>> Source: mirror:sourceforge:math-atlas/%n%v.tar.bz2 >>> -Source-MD5: dd888f5d066c1fafeaaf2ae6b37e0b85 >>> +Source-MD5: fa95dfd4a69e5614db1061a740bc5265 >>> Source2: http://www.netlib.org/lapack/lapack-3.1.1.tgz >>> Source2-MD5: 00b21551a899bcfbaa7b8443e1faeef9 >>> SourceDirectory: ATLAS >>> License: BSD >>> # since we're experimenting with building a dylib : >>> BuildDependsOnly: true >>> -BuildDepends: gcc43 >>> +BuildDepends: gcc44 >>> PatchScript: << >>> #!/bin/sh -ev >>> sed -i.bak -e '/PowerBook/,/RackMac/s,"c,"k,g' CONFIG/src/backend/ >>> archinfo_freebsd.c >>> @@ -71,7 +71,10 @@ >>> ln -s %p/bin/gcc-4 gcc >>> export PATH=`pwd`:$PATH >>> cd bld >>> - ../ATLAS/configure -v 2 --prefix=%p --with-netlib-lapack=%b/../ >>> LAPACK/lapack_LINUX.a -b 32 -F ic "$iflags" -F if "$iflags" $cflags >>> + if [ %m = 'x86_64' ] >>> + then ../ATLAS/configure -v 2 --prefix=%p --with-netlib-lapack= >>> %b/../LAPACK/lapack_LINUX.a -b 64 -F ic "$iflags" -F if "$iflags" >>> $cflags >>> + else ../ATLAS/configure -v 2 --prefix=%p --with-netlib-lapack= >>> %b/../LAPACK/lapack_LINUX.a -b 32 -F ic "$iflags" -F if "$iflags" >>> $cflags >>> + fi >>> make >>> cd lib >>> # Apparently on some machines (=monoprocesors ?) the threaded libs >>> are not made; >>> @@ -88,7 +91,7 @@ >>> # We link 'manually', with ld, to avoid having unnecessary libs like >>> lgcc_s among the load commands. >>> # This way the libs can be safely used in linking with any compiler: >>> they will not bring themselves >>> # the wrong lgcc_s in the search list. >>> - ld="ld -dynamic -dylib -single_module -dead_strip -x -all_load -L. >>> -L%p/lib/gcc4.3/lib -ldylib1.o -dylib_install_name" >>> + ld="ld -dynamic -dylib -single_module -dead_strip -x -all_load -L. >>> -L%p/lib/gcc4.4/lib -ldylib1.o -dylib_install_name" >>> $ld %p/lib/libatlas.dylib libatlas.a -o libatlas.dylib -lSystem >>> $ld %p/lib/libcblas.dylib libcblas.a -o libcblas.dylib -latlas - >>> lSystem >>> $ld %p/lib/libf77blas.dylib libf77blas.a -o libf77blas.dylib - >>> lgfortran -latlas -lSystem >>> @@ -144,7 +147,7 @@ >>> << >>> Splitoff: << >>> Package: %N-shlibs >>> - Depends: gcc43-shlibs >>> + Depends: gcc44-shlibs >>> Files: lib/*.dylib >>> InstallScript: mkdir -p %i/share/doc; ln -s %N %i/share/doc/%n >>> Shlibs: << >>> >>> The packaging passes 'fink -m rebuild' on x86_64 fink as well as 'make >>> check'. >>> Jack >>> >>> ------------------------------------------------------------------------------ >>> _______________________________________________ >>> Fink-devel mailing list >>> Fink-devel@lists.sourceforge.net >>> http://news.gmane.org/gmane.os.apple.fink.devel >>> Subscription management: >>> https://lists.sourceforge.net/lists/listinfo/fink-devel >>> -- Alexander Hansen Fink User Liaison ------------------------------------------------------------------------------ _______________________________________________ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel Subscription management: https://lists.sourceforge.net/lists/listinfo/fink-devel
