octave-suitesparse 3.2.3-1
http://www.octave.org
A high-level language, primarily intended for numerical computations. Complete build against SuiteSparse
unsupported :: science
Maintainer: Demind
Votes: 4
License: GPL
Last Updated: Mon, 05 Oct 2009 17:08:13 +0000
First Submitted: Wed, 17 Sep 2008 06:49:40 +0000
Dependencies arpack curl fftw fltk ftgl gcc-libs>=4.3.0 glpk gnuplot gperf graphicsmagick hdf5 pcre qhull qrupdate readline suitesparse>=3.2.0 zlib
Updated, new version and no more "problematic" configure options
I'm trying to investigate this
From Jaroslav Hajek while diagnosing an error I was seeing
Also, you should be aware that --enable-64 is experimental. To
successfully use that, your BLAS and LAPACK *must* be compiled with
64-bit INTEGERs (a lot of ops will work if they're not, but some will
crash silently). Are you sure of that? If you really mean to use the
64-bit indexing (required for matrices with more than 2e9 elements), I
suggest you try the development sources, which includes some
improvements and checks whether everything is all right.
Rebuilding without this option fixed my problem.
Also, the discussion at
# http://www.nabble.com/Random-rounding-errors-td16010966.html
is not suggesting that octave should be built with -ffloat-store, it is saying that you could try compiling ATLAS with that flag. Note that SSE registers do not have extra bits like x87, so this option is not relevant to SSE. I suggest removing this option unless you have a different reason for having it.
Updated to new version
I think you need to add the version to the provide entry, i.e.
provides=('octave=3.2.0' 'octave3')
which is needed by octave-forge. Otherwise pacman will complain when installing something with this warning
missing dependency for octave-forge : octave>=3.2.0
missing dependency for octave-forge : octave<3.3
Updated
I'm ready to update the package providing the whole functionality. I'll upload it in some hours thanks
3.2.0 is out.
release 3.0.5 f octave is out. Changing $pkgrel and md5sum is sufficient to build.
Can you add this to the PKGBUILD:
# stricter compilers
patch -Np1 <<EOF
--- orig/octave-3.0.3/liboctave/sparse-util.h 2008-09-23 23:13:47.000000000 -0800
+++ octave-3.0.3/liboctave/sparse-util.h 2009-01-28 09:47:47.000000000 -0900
@@ -24,7 +24,7 @@
#if !defined (octave_sparse_util_h)
#define octave_sparse_util_h 1
-extern OCTAVE_API void SparseCholError (int status, char *file, int line, char *message);
+extern OCTAVE_API void SparseCholError (int status, const char *file, int line, const char *message);
extern OCTAVE_API int SparseCholPrint (const char *fmt, ...);
#endif
--- orig/octave-3.0.3/liboctave/sparse-util.cc 2008-09-23 23:13:47.000000000 -0800
+++ octave-3.0.3/liboctave/sparse-util.cc 2009-01-28 09:47:59.000000000 -0900
@@ -31,7 +31,7 @@
#include "sparse-util.h"
void
-SparseCholError (int status, char *file, int line, char *message)
+SparseCholError (int status, const char *file, int line, const char *message)
{
(*current_liboctave_warning_handler)("warning %i, at line %i in file %s",
status, line, file);
EOF
Thank you, I will delete it.
Hi Stefan, my suitesparse package depends on intel_tbb now. If I could I would remove the tbb package but I can only orphan it
There is a package intel-tbb in AUR which servers the same purpose than your package tbb. Can you pleas let octave-suitesparse depend against it? tbb should be removed from AUR.
Updated
Builds fine for Octave 3.0.3 with only changing $pkgver and md5sum on i686.
Updated. Thanks to Stefan
In this period i have very little time, if you want help i'll be pleased to add your contribution. If you prefer waiting for a new build i apologize for the delay you'll have in an update
Please add an install file with info file handling. See /usr/share/pacman/proto-info.install
Thank you for your contribution. The provides line does not work as expected for me. If I try to install qtoctave, pacman wants to deinstall octave-suitesparse and to install octave instead.
v1.6.0