Re: [gentoo-user] Problem with cblas libraries

[denis cohen <denis.cohen@gmail.com>]

Following Michael Mol's advice I masked these in /etc/portage/package.mask:

>=virtual/blas-2.0
>=virtual/cblas-2.0
>=virtual/lapack-3.1
>=sci-libs/cblas-reference-20030223-r5
>=sci-libs/lapack-reference-3.2.1
>=sci-libs/gsl-1.14

and did an "emerge -uD world".

All blocks were gone and after the emerge I got gsl working again.
I am also not sure what use is eselect-cblas since eselect blas works
without it.

gmsh still won't emerge with these types of errors:
...
[100%] Building CXX object CMakeFiles/gmsh.dir/contrib/bamg/bamglib/Meshio.cpp.o
[100%] Building CXX object
CMakeFiles/gmsh.dir/contrib/bamg/bamglib/QuadTree.cpp.o
[100%] Building CXX object
CMakeFiles/gmsh.dir/contrib/bamg/bamglib/SetOfE4.cpp.o
Linking CXX executable gmsh
CMakeFiles/gmsh.dir/Numeric/fullMatrix.cpp.o: In function
`fullMatrix<double>::invertInPlace()':
fullMatrix.cpp:(.text+0xcc): undefined reference to `dgesv_'
CMakeFiles/gmsh.dir/Numeric/fullMatrix.cpp.o: In function
`fullMatrix<double>::gemm(fullMatrix<double> const&,
fullMatrix<double> const&, double, double)':
fullMatrix.cpp:(.text+0x1e9): undefined reference to `dgemm_'
CMakeFiles/gmsh.dir/Numeric/fullMatrix.cpp.o: In function
`fullVector<double>::axpy(fullVector<double> const&, double)':
fullMatrix.cpp:(.text+0x260): undefined reference to `daxpy_'
...

So making some progress.

Denis



On Wed, Aug 31, 2011 at 01:02, Alex Schuster <wonko@wonkology.org> wrote:
> denis cohen writes:
>
>> I've unmerged cblas* blas* gsl, synced, and "emerge -uDN world "also
>> without success (also tried cblas-external USE flag but could not see
>> it with eix).
>
> What's your problem with the world update? Adding --tree to the emerge
> command might show what pulls in what. I had to unmerge clapack and lapack-
> reference yesterday, but I cannot remember the details - it all started with
> digikam not building because some blas stuff was missing. But finally the
> @world update was possible. And both got re-installed.
>
>> eselect cblas also gave no results.
>
> Here it is set to the only option gsl.
>
>> One problem I may have is with eselect. eselect-cblas has been
>> unmerged but trying to re-emerge it gives this block.
>
> I get the very same blocker as you. eselect cblas is working, though. So is
> eselect-cblas being needed at all?
>
>        Wonko
>
>