Re: [gentoo-user] Kicad and complications from hdf5 and vtk USE flags. Not package specific tho.

[Dale]

Miles Malone wrote:
> Well I just installed kicad-meta (6.0.1), and neither vtk nor hdf5 are
> pulled into the dep tree as a result...  Given opencascade[vtk] is
> what's pulling in vtk in your emerge above, disabling the vtk
> useflag's probably going to help.
>
> In addition, I cant imagine why in a million years MPI should be
> getting pulled in unless you've explicitly enabled it somewhere else,
> MPI's hardly something you'd be using if you werent USING, i.e. on a
> distributed memory system.  Looking at the dependency tree, the most
> likely place for an MPI use flag to be causing all kinds of mpi
> related shenanigans with other packages is Boost.  If you've enabled
> MPI for Boost without actually needing it, that'll cause all kinds of
> tree problems.  In my experience if you're using MPI you need to be
> really really careful about enabling it globally, it's incredibly hard
> for the devs to test given test systems are rarely clusters, so it's
> often got significant breakage. Of course if you're developing for a
> cluster, which I do from time to time, this can be a pain.
>
> So assuming you've got *absolutely no libraries* in your world file
> (there really shouldnt be), make sure you've got mpi and vtk disabled
> globally, and not per-package enabled for boost and opencascade.  Get
> rid of vtk and hdf5, and try again? (Assuming you dont need hdf5 for
> something other than kicad that is, of course. It's the main format I
> use for data storage as it plays nicely with Matlab and okay with c++
> and python)
>
> That's my best guess as to what's causing your blockers, anyways.
> Give it a go and see where the error moves to
>
>

I'm fine with disabling it globally so I added it to make.conf.  I also
made sure any entries in package.use were commented out.  Now I get this:


root@fireball / # emerge -auDN world

These are the packages that would be merged, in order:

Calculating dependencies... done!

!!! The ebuild selected to satisfy "<sci-libs/hdf5-1.12:=[mpi=]" has
unmet requirements.
- sci-libs/hdf5-1.10.5-r1::gentoo USE="cxx hl threads zlib -debug
-examples -fortran -mpi -szip -unsupported" ABI_X86="(64)"

  The following REQUIRED_USE flag constraints are unsatisfied:
    !unsupported? ( threads? ( !cxx !hl ) )

  The above constraints are a subset of the following complete expression:
    !unsupported? ( at-most-one-of ( cxx mpi ) threads? ( !cxx !mpi
!fortran !hl ) )

(dependency required by "sci-libs/vtk-9.0.3-r4::gentoo" [ebuild])
(dependency required by "sci-libs/opencascade-7.5.2-r5::gentoo" [installed])
(dependency required by "sci-electronics/kicad-5.1.12-r2::gentoo[occ]"
[installed])
(dependency required by "sci-electronics/kicad-symbols-5.1.12::gentoo"
[installed])
(dependency required by "sci-electronics/kicad-meta-5.1.12::gentoo"
[installed])
(dependency required by "@selected" [set])
(dependency required by "@world" [argument])
root@fireball / #


Also, keep in mind I'm still on V5 of Kicad.  I'm wanting some more time
for bug fixes and such before I update to V6.  It's sort of a one way
street.  I might add tho, there was a update to V5 a few weeks ago that
started this whole mess.  Ever since that new version of V5 hit the
tree, I've had issues getting clean paths to upgrade with Kicad
installed.  If I could, I'd go back to the previous version but when
they did the update, they removed the previous version that was
working.  I've considered putting off my projects and just leaving Kicad
uninstalled for a while. 

Thanks for the help.  My decoder ring just can't figure out what emerge
is trying to say, especially when it conflicts with itself.

Dale

:-)  :-)