On 21/01/14 23:46, Sébastien Fabbro wrote:
> Some applications (I can see armadillo, scamp) actually use the
> atlas-specific routines clapack_* routines. 
> We could aim at keeping both the library soname and filename as close to
> upstream, and add an option to the alternative framework to create a
> ldscript with a generic soname and filename, turned on for blas and
> cblas at least.


I would really like to see that ldscript to be present. it would solve
many propblems including linking in distutils against mkl.



> 
> As a user, I don't particularly enjoy rebuilding octave, ROOT, R, or the
> scipy stack. @preserved-rebuild is more a hack than a feature.
> I don't know how many people actually switch providers, besides a first
> benchmark test. My guess is probably not many given the burden of
> re-compiling all the reverse dependencies and the previous fragility of
> the system.

But I still don't see the point of equalizing the soname. One advantage
I see from separate sonames is that you can link different applications
to different implementations. Of course this would require recompilation
of revdeps in case you remove one from your system. But I would argue
that switching/removing an implementation is a rare case for the average
user. Most will pick one or take the default aka reference
implementation and never touch it again.


Justin