Following Roberts huge effort to get a lot of .spkg updates into
sage-4.5.0.alpha0, a 64-bit build of Sage on both Solaris 10 SPARC and
OpenSolaris x64 does not look to be a distant dream any more. I think it will be
close race between a Cygwin port and 64-bit Solaris or OpenSolaris port.
There are still however 5 or 6 problems with a Solaris build, so any help would
be appreciated.
One or two problems might be specific to Solaris 10 SPARC or specific to
OpenSolaris x64, but most of the problems appear on all 64-bit versions of Solaris.
There are two ways to attack this port.
1) Install OpenSolaris, which is a free download, as a virtual machine. One
problem with this can be that if your BIOS on your PC is crippled, you can only
install 32-bit operating systems as virtual machines. This is the case with my
Sony Vaio.
2) Use 't2' and build Sage on there in 64-bit mode. Now the parallel building of
packages is working, building on 't2' is nowhere near the pain it used to be,
though it is still not exactly fast.
The only changes for those used to building Sage on t2 are
$ export SAGE64=yes
$ export
SAGE_FORTRAN_LIB=/usr/local/gcc-4.4.1-sun-linker/lib/sparcv9/libgfortran.so
There's a list of the issues on a 'metaticket'
http://trac.sagemath.org/sage_trac/ticket/9026
Most of the issues are resolved in sage-4.5.0.alpha0. The few remaining are:
1) Pynac - http://trac.sagemath.org/sage_trac/ticket/9026
Awaiting review.
http://boxen.math.washington.edu/home/kirkby/patches/pynac-0.2.0.p4.spkg
2) Libfpll
http://trac.sagemath.org/sage_trac/ticket/7864
awaiting review
http://boxen.math.washington.edu/home/kirkby/patches/libfplll-3.0.12.p1.spkg
3) Numpy
http://trac.sagemath.org/sage_trac/ticket/8086
awaiting review
http://boxen.math.washington.edu/home/kirkby/patches/numpy-1.3.0.p4.spkg
4) R.
http://trac.sagemath.org/sage_trac/ticket/9040
This is potentially problematic, as R does not apparently build with gcc on x86
versions of Solaris, though it does with Sun Studio
I'm hoping to get the help of Dennis Clarke of Blastwave on this one. We don't
really want to make have two compilers a prerequisite for building Sage.
But since R is only called via the command line, and not a library, we can just
forget R for a minute.
5) Maxima
http://trac.sagemath.org/sage_trac/ticket/9099
6) Sage library.
This patch (awaiting review) allows c_lib to build
http://trac.sagemath.org/sage_trac/ticket/9097
but then things go wrong and I can't work out where the compiler flags are
coming from. Can someone take a look at this, and tell me what's up
http://groups.google.co.uk/group/sage-devel/browse_thread/thread/14155a7f42d4df7a#
7) Scipy installs, but I'm very suspicious it is not a full 64-bit install.
http://trac.sagemath.org/sage_trac/ticket/9100
This is because some of the object files created are 32-bit, though the final
build is fully 64-bit. Something is wrong there.
8) Linbox
http://trac.sagemath.org/sage_trac/ticket/9101
This is problematic on Solaris 10 SPARC, but works fine on OpenSolaris. As such,
that can only be solved with the aid of 't2' or similar SPARC based system.
9) Singular
http://trac.sagemath.org/sage_trac/ticket/9033
I find the libsingular build will not complete, though the other bits in the
Singular package do. One can just comment out the line that builds libsingular,
though clearly that is not ideal.
For more detailed discussions, we should take this to sage-solaris, or drop me
private emails, But I'd like to get a bit of feedback here. See who is willing
to do what. I think any one of many people probably here probably know how to
solve the issue with the Sage library.
http://groups.google.co.uk/group/sage-devel/browse_thread/thread/14155a7f42d4df7a#
Somehow it needs to be told to add the -m64 flag, but I don't know how.
Dave
--
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
