Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 20 Oct 2022 21:19:08 +0200
From:      Christoph Moench-Tegeder <cmt@burggraben.net>
To:        Yuri Victorovich <yuri@freebsd.org>
Cc:        ports-committers@freebsd.org, dev-commits-ports-all@freebsd.org, dev-commits-ports-main@freebsd.org
Subject:   Re: git: b0c2eb73e1e0 - main - math/vtk9: Add net/py-mpi4py dependency
Message-ID:  <Y1GfLNk6c3hLQDhs@elch.exwg.net>
In-Reply-To: <202210190121.29J1LZuE036720@gitrepo.freebsd.org>
References:  <202210190121.29J1LZuE036720@gitrepo.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
## Yuri Victorovich (yuri@FreeBSD.org):

>     math/vtk9: Add net/py-mpi4py dependency

That doesn't work like this (or any of the follow-ups): net/py-mpi4py
pulls in net/openmpi by default ("the only non-failing MPI implementation"
py-mpi4py's Makefile says in a comment), and net/openmpi brings, well
the whole MPI stack, including ${LOCALBASE}/lib/libmpi.so, which
vtk9's MPI option uses as the indicator file when trying to depend
on net/mpich. With that file present (although from the wrong package)
the build system does not install mpich, vtk9 picks up openmpi,
links against that and gets that dependency registered:
  $ pkg info --dependencies vtk9 | grep mpi
        py39-mpi4py-3.1.3
        openmpi-4.1.4
On the next run, poudriere detects that the package should depend on
mpich, deletes the vtk9 package and everything which depends on it
and rebuilds vtk9, which again picks up openmpi... circle.

Regards,
Christoph

-- 
Spare Space



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Y1GfLNk6c3hLQDhs>