Packaging question on MPI requires

[Dave Love]

Zbigniew Jędrzejewski-Szmek <zbyszek at in.waw.pl> writes:

>> Surely this isn't scalable.  What would happen with a common API for N
>> different Scheme implementations, for instance?  I'm also interested in
>> Haskell; others might want the bindings to Common Lisp, ruby, Tcl...
>> Currently CLASSPATH isn't set for the Java bindings.  
> If by "scalable" you mean that the rpm-mpi-hooks package cannot scale,
> I disagree. There's maybe a dozen script languages that might care
> about MPI, I'm sure that the rules can be added. Nevertheless,
> extensibility hook sounds useful.

"Scalable" is probably the wrong word, but I assume there will be policy
virtual paperwork, and it means you have to persuade MPI maintainers to
support it before you do anything new, or you need to fork the MPI
packages.

>> This is the sort of reason I think the rules for MPI packaging are
>> unfortunate.
> I see that this causes additional problem for you, but in the previous
> situation automatic requirements were completely broken, and now they
> seem to work fine (for the supported languages). For a user of software
> packaged natively for Fedora this is a definite win.

Of course I'm happy with automation and generally having things just
work.  I was only meaning to use what I've done as an example.  It's the
general setup I don't like, but this doubtless isn't the place...
However, I do have problems even with Fortran packaging, given the
nightmare of module file incompatibilities and needing to use devtoolset
compilers.