On Tue, 01 Dec 2009 14:42:43 strogdon wrote: > Now when I unpack the sage-4.2.1.spkg tarball and insert > > unset CFLAGS > unset CXXFLAGS > > near the top of the spkg-install script, repackage the tarball and > issue the above 'make' with custom CFLAGS; both sage and the > documentation build and 'sage -c quit' returns no errors. By altering > spkg-install in the above fashion every package in sage is exposed to > the custom CFLAGS with the exception of sage-4.2.1.spkg. A similar > hack to unset CFLAGS for the amd64 architecture > is utilized by Christopher Schwan and Francois Bissey (http:// > github.com/cschwan/sage-on-gentoo) to build a partially split sage. It > would appear that these custom amd64 flags cannot be used to > successfully build the sage-4.2.1.spkg tarball. Is there any known > reason why these custom CFLAGS cannot be used? Gentoo custom flags for > 32 bit architectures do not seem to have this build problem. On an > amd64 laptop I have, the CFLAGS setting > > CFLAGS="-march=k8 -O2 -pipe" > > also results in a sage that fails. > I am biting Steve. The current filtering in the ebuild is a terribly gross. We have compiler flag filtering functions. Although I think your question is interesting some flags cause a failure in the compilation of the sage spkg (doesn't spkg-install provide its own flags overriding yours?). We should try to find what exactly is causing the fault. Can you try without the -march=k8 please? If you are using >=gcc-4.3 could you also try -march=native?
Francois -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
