[Debian-med-packaging] Bug#721931:

[tom fogal]

Frank Loeffler writes:
> Is it known (ideally by upstream) whether the OpenMP pragmas are
> correct, i.e. that there are no bugs when OpenMP is enabled?

Well, there ae no *known* bugs, but absence of proof is not proof of
absence, of course :-)

> I agree with upstream that disabling OpenMP by not giving the
> relevant compiler flags is not a problem, but it is still strange to
> have pragmas but not to use them when compiling.

OpenMP support isn't universal; last we looked, it was such a nightmare
on Mac that we do not bother there, for example.

> Usually this points to either code in development or some bug in the
> OpenMP imlementation. In both cases disabling OpenMP via compiler
> flags is a quick way to avoid problems on user side, but it would be
> nice to know what upstreams reason to do this really is.

Is it guaranteed that OpenMP is available in Debian?  If so, then I
think unconditionally adding it in is a fine solution for Debian.  We
would need to add gomp (I think package "libgomp1") to the depends,
then.

I guess it's really a question as to how much Debian wants packages to
depend on libgomp.  In ImageVis3D's case, it is an optional dependency
with a very minor performance impact.  Thus, we feel the distributor
should be at liberty to choose.