Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 384159 - [science-overlay] sys-cluster/openmpi conflicts with sys-cluster/nullmpi
Summary: [science-overlay] sys-cluster/openmpi conflicts with sys-cluster/nullmpi
Status: RESOLVED INVALID
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: Current packages (show other bugs)
Hardware: AMD64 Linux
: Normal normal (vote)
Assignee: Gentoo Cluster Team
URL:
Whiteboard: [science overlay]
Keywords:
Depends on:
Blocks:
 
Reported: 2011-09-23 05:19 UTC by Micah Shennum
Modified: 2011-10-23 07:59 UTC (History)
1 user (show)

See Also:
Package list:
Runtime testing required: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Micah Shennum 2011-09-23 05:19:37 UTC
My normal upgrade fails with sys-cluster/nullmpi colliding with sys-cluster/openmpi on 

/usr/bin/mpiCC
/usr/bin/mpicc
/usr/bin/mpicxx
/usr/bin/mpiexec
/usr/bin/mpirun
/usr/include/mpi.h
Comment 1 Micah Shennum 2011-09-23 06:17:04 UTC
I "fixed" this by masking nullmpi
Comment 2 Kacper Kowalik (Xarthisius) (RETIRED) gentoo-dev 2011-09-27 09:42:14 UTC
That's expected, you need to install virtual/mpi-2.0 from science overlay to avoid other packages merging openmpi if you've already have one implementation on your system.
Comment 3 Micah Shennum 2011-09-28 04:29:49 UTC
Thank you.
Comment 4 Mathieu Z 2011-10-22 19:43:28 UTC
(In reply to comment #2)
> That's expected, you need to install virtual/mpi-2.0 from science overlay to
> avoid other packages merging openmpi if you've already have one implementation
> on your system.

Then some ebuild is having a wrong DEPEND issue. I removed virtual/mpi, sys-cluster/openmpi, then ran and update world and it wants to pull in BOTH:

# emerge -auDN world

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

Calculating dependencies... done!
[ebuild  N     ] sys-cluster/openmpi-1.5.4  USE="cxx fortran ipv6 threads -heterogeneous -infiniband -mpi-threads -pbs -romio -sctp -slurm -vt" 
[ebuild     U  ] www-client/google-chrome-15.0.874.100_beta106180 [15.0.874.92_beta104978]
[ebuild  N     ] sys-cluster/nullmpi-0.7  USE="-static-libs" 
[ebuild   R    ] dev-libs/protobuf-2.4.1  USE="-python*" 
[ebuild  N     ] virtual/mpi-2.0-r2  USE="cxx fortran romio" 
[ebuild  N     ] sci-libs/hdf5-1.8.7  USE="cxx fortran mpi threads zlib -debug -examples -szip" 
[ebuild  N     ] media-libs/vigra-1.7.1-r1  USE="fftw hdf5 jpeg png python tiff -doc -test" 
[ebuild     U  ] app-office/libreoffice-bin-3.4.3.2-r1 [3.4.3-r2] USE="(-aqua) -kde% (-kdeenablefinal)" 
[ebuild     U  ] dev-java/icedtea-7.2.0 [7.2.0_pre]
[ebuild     U  ] dev-tex/latex-beamer-3.12 [3.11]
Comment 5 Mathieu Z 2011-10-22 19:50:16 UTC
Actually virtual/mpi is pulling in nullmpi if it has USE=romio enable but openmpi does not. So either enable USE=romio globally or turn it on for sys-cluster/openmpi as well.
Comment 6 Kacper Kowalik (Xarthisius) (RETIRED) gentoo-dev 2011-10-23 07:59:07 UTC
(In reply to comment #5)
> Actually virtual/mpi is pulling in nullmpi if it has USE=romio enable but
> openmpi does not. So either enable USE=romio globally or turn it on for
> sys-cluster/openmpi as well.

Yeah use flags need to be consistent here or those deps that have [romio?] will be skipped, so either set it globally or add specific flavour of mpi to package.use that matches virtual/mpi.

We're working on solution to such kind of problems, sorry for the inconvience.