Re: F23 Self Contained Change: RPM MPI Requires Provides
Sandro Mani manisan...@gmail.com writes: Where are we with this? I see some F23 builds but no updates yet. Formally, this Change has been removed from the scope of F23 as it was not ready for Alpha, which is a requirement to have a Change in a release. However Zbyszek told me yesterday that he is going to work on it and would like to push this F23. So, lets see when it will be ready and then decide whether it can still meet the F23. Regards, Jan So the issues I pointed out in my previous mail (conerning MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR) above have been resolved. A batch of F24 rebuilts has been done by Zbyszek, hitting some build failures along the way. Most of these are now fixed, still needing a fix is nwchem which fails [1] due to what appears to be a fortran use before decleared issue. Once a full set of working F23 package is built, the plan is to submit one big mass-update. Should MPI packages be OK to push in f23 currently? I've just had failures from auto-qa like this, though the build actually ran an MPI test successfully: not ok - depcheck for Koji build scalasca-2.2.2-2.fc23# FAIL --- arch: x86_64 details: output: |- conflicting requests nothing provides libmpi.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 nothing provides libmpicxx.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 conflicting requests nothing provides libmpi.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 nothing provides libmpicxx.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 conflicting requests nothing provides libmpi.so.1()(64bit)(openmpi-x86_64) needed by scalasca-openmpi-2.2.2-2.fc23.x86_64 nothing provides libmpi_cxx.so.1()(64bit)(openmpi-x86_64) needed by scalasca-openmpi-2.2.2-2.fc23.x86_64 If anyone knowledgable of fortran has an idea how to fix the nwchem issue, that should resolve the last pending issue. Appears to have been triggered by a recent change of some other component, since it rebuilt fine just a few weeks back when testing in copr. Sandro [1] http://koji.fedoraproject.org/koji/buildinfo?buildID=676763 I can take a look if no-one else is going to. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 08/19/2015 06:36 AM, Dave Love wrote: Should MPI packages be OK to push in f23 currently? I've just had failures from auto-qa like this, though the build actually ran an MPI test successfully: not ok - depcheck for Koji build scalasca-2.2.2-2.fc23 # FAIL --- arch: x86_64 details: output: |- conflicting requests nothing provides libmpi.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 nothing provides libmpicxx.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 conflicting requests nothing provides libmpi.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 nothing provides libmpicxx.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 conflicting requests nothing provides libmpi.so.1()(64bit)(openmpi-x86_64) needed by scalasca-openmpi-2.2.2-2.fc23.x86_64 nothing provides libmpi_cxx.so.1()(64bit)(openmpi-x86_64) needed by scalasca-openmpi-2.2.2-2.fc23.x86_64 You're going to need to wait for the big MPI update: https://admin.fedoraproject.org/updates/rpm-mpi-hooks-3-1.fc23,mathgl-2.3-10.fc23,elpa-2015.05.001-2.fc23,freefem++-3.31-7.3.fc23,netcdf-4.3.3.1-5.fc23,MUMPS-5.0.1-2.fc23,gpaw-0.11.0.13004-15.fc23,ga-5.3b-18.fc23,valgrind-3.10.1-19.fc23,boost-1.58.0-6.fc23,dl_poly-1.9.20140324-14.fc23,hpl-2.1-13.fc23,netgen-mesher-5.3.1-8.fc23,espresso-3.3.0-7.fc23,scotch-6.0.4-5.fc23,hdf5-1.8.15-6.patch1.fc23,scalapack-2.0.2-10.fc23,gromacs-5.0.6-6.fc23,scorep-1.4.2-2.fc23,sundials-2.6.2-4.fc23,Ray-2.3.1-11.fc23,orsa-0.7.0-33.fc23,elk-3.1.12-14.fc23,mpich-3.1.4-5.fc23,openmpi-1.8.8-3.fc23,mpi4py-1.3.1-14.fc23,pypar-2.1.5_108-11.fc23,paraview-4.3.1-11.fc23 -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On Wed, Aug 19, 2015 at 10:11:04AM -0600, Orion Poplawski wrote: On 08/19/2015 06:36 AM, Dave Love wrote: Should MPI packages be OK to push in f23 currently? I've just had failures from auto-qa like this, though the build actually ran an MPI test successfully: not ok - depcheck for Koji build scalasca-2.2.2-2.fc23# FAIL --- arch: x86_64 details: output: |- conflicting requests nothing provides libmpi.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 nothing provides libmpicxx.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 conflicting requests nothing provides libmpi.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 nothing provides libmpicxx.so.12()(64bit)(mpich-x86_64) needed by scalasca-mpich-2.2.2-2.fc23.x86_64 conflicting requests nothing provides libmpi.so.1()(64bit)(openmpi-x86_64) needed by scalasca-openmpi-2.2.2-2.fc23.x86_64 nothing provides libmpi_cxx.so.1()(64bit)(openmpi-x86_64) needed by scalasca-openmpi-2.2.2-2.fc23.x86_64 You're going to need to wait for the big MPI update: https://admin.fedoraproject.org/updates/rpm-mpi-hooks-3-1.fc23,mathgl-2.3-10.fc23,elpa-2015.05.001-2.fc23,freefem++-3.31-7.3.fc23,netcdf-4.3.3.1-5.fc23,MUMPS-5.0.1-2.fc23,gpaw-0.11.0.13004-15.fc23,ga-5.3b-18.fc23,valgrind-3.10.1-19.fc23,boost-1.58.0-6.fc23,dl_poly-1.9.20140324-14.fc23,hpl-2.1-13.fc23,netgen-mesher-5.3.1-8.fc23,espresso-3.3.0-7.fc23,scotch-6.0.4-5.fc23,hdf5-1.8.15-6.patch1.fc23,scalapack-2.0.2-10.fc23,gromacs-5.0.6-6.fc23,scorep-1.4.2-2.fc23,sundials-2.6.2-4.fc23,Ray-2.3.1-11.fc23,orsa-0.7.0-33.fc23,elk-3.1.12-14.fc23,mpich-3.1.4-5.fc23,openmpi-1.8.8-3.fc23,mpi4py-1.3.1-14.fc23,pypar-2.1.5_108-11.fc23,paraview-4.3.1-11.fc23 The update is now https://admin.fedoraproject.org/updates/rpm-mpi-hooks-3-2.fc23 If you made an update with scalasca for F23, please cancel it and I'll add scalasca to the big update. (I cannot check the status because admin.fedoraproject.org is overloaded.) Zbyszek -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On Sat, Aug 15, 2015 at 2:49 PM, Sandro Mani manisan...@gmail.com wrote: So the issues I pointed out in my previous mail (conerning MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR) above have been resolved. A batch of F24 rebuilts has been done by Zbyszek, hitting some build failures along the way. Most of these are now fixed, still needing a fix is nwchem which fails [1] due to what appears to be a fortran use before decleared issue. Once a full set of working F23 package is built, the plan is to submit one big mass-update. If anyone knowledgable of fortran has an idea how to fix the nwchem issue, that should resolve the last pending issue. Appears to have been triggered by a recent change of some other component, since it rebuilt fine just a few weeks back when testing in copr. Sandro [1] http://koji.fedoraproject.org/koji/buildinfo?buildID=676763 I haven't touched fortran in quite some time, so take this with with a grain of salt (preferably more). The issue likely has to do with some transliterations (?) from 32-bit to 64-bit and vice versa when using particular libraries. Upstream has released quite a few patches for build 26243, which might address this: http://www.nwchem-sw.org/index.php/Download#Patches_for_the_26243_revision_of_NWChem_6.5 my money would be on Tddft_grad.patch.gz but I don't bet as I don't have any luck... -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 15.08.2015 13:31, Jan Kurik wrote: On Sat, Aug 15, 2015 at 5:45 AM, Orion Poplawski or...@cora.nwra.com mailto:or...@cora.nwra.com wrote: On 07/27/2015 04:05 PM, Sandro Mani wrote: On 27.07.2015 23 tel:27.07.2015%2023:56, Sandro Mani wrote: Ok I've now got one full build of all MPI packages [1]. Investigating the output, things are looking good, except for the fact that I realized that I'll also need to handle binaries MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR - these directories are outside MPI_HOME and hence currently don't get the enhanced requires/provides strings. Since change deadline is tomorrow, I'd like to see the rebuilds starting to land in rawhide+F23 despite of this remaining issue, and then I'll look at cleanly adapting the scripts to also handle the (few) remaining cases. I've prepared a set of git am-able patches to apply to the various packages here [2]. The rebuild sequence needed is posted below. Know FTBFS are paraview (protobuf incompatibility) and netgen-mesher (orrection: that would be gmsh (someone had the great idea of doing something like #define Status int in a public header and I haven't yet got around to finding out who). Affected by the missing handling of MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR are elpa, sundials, pypar and mpi4py. So if any proven packager could fire the rebuilds, that would be much appreciated. Where are we with this? I see some F23 builds but no updates yet. Formally, this Change has been removed from the scope of F23 as it was not ready for Alpha, which is a requirement to have a Change in a release. However Zbyszek told me yesterday that he is going to work on it and would like to push this F23. So, lets see when it will be ready and then decide whether it can still meet the F23. Regards, Jan So the issues I pointed out in my previous mail (conerning MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR) above have been resolved. A batch of F24 rebuilts has been done by Zbyszek, hitting some build failures along the way. Most of these are now fixed, still needing a fix is nwchem which fails [1] due to what appears to be a fortran use before decleared issue. Once a full set of working F23 package is built, the plan is to submit one big mass-update. If anyone knowledgable of fortran has an idea how to fix the nwchem issue, that should resolve the last pending issue. Appears to have been triggered by a recent change of some other component, since it rebuilt fine just a few weeks back when testing in copr. Sandro [1] http://koji.fedoraproject.org/koji/buildinfo?buildID=676763 -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On Sat, Aug 15, 2015 at 5:45 AM, Orion Poplawski or...@cora.nwra.com wrote: On 07/27/2015 04:05 PM, Sandro Mani wrote: On 27.07.2015 23:56, Sandro Mani wrote: Ok I've now got one full build of all MPI packages [1]. Investigating the output, things are looking good, except for the fact that I realized that I'll also need to handle binaries MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR - these directories are outside MPI_HOME and hence currently don't get the enhanced requires/provides strings. Since change deadline is tomorrow, I'd like to see the rebuilds starting to land in rawhide+F23 despite of this remaining issue, and then I'll look at cleanly adapting the scripts to also handle the (few) remaining cases. I've prepared a set of git am-able patches to apply to the various packages here [2]. The rebuild sequence needed is posted below. Know FTBFS are paraview (protobuf incompatibility) and netgen-mesher (orrection: that would be gmsh (someone had the great idea of doing something like #define Status int in a public header and I haven't yet got around to finding out who). Affected by the missing handling of MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR are elpa, sundials, pypar and mpi4py. So if any proven packager could fire the rebuilds, that would be much appreciated. Where are we with this? I see some F23 builds but no updates yet. Formally, this Change has been removed from the scope of F23 as it was not ready for Alpha, which is a requirement to have a Change in a release. However Zbyszek told me yesterday that he is going to work on it and would like to push this F23. So, lets see when it will be ready and then decide whether it can still meet the F23. Regards, Jan Thanks Sandro [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing2 [2] https://smani.fedorapeople.org/rpm-mpi-hooks/patches/ Rebuild sequence (these are four groups, items within a group can be built together): mpich openmpi mpi4py valgrind pypar boost dl_poly hpl netgen-mesher espresso scotch hdf5 scalapack gromacs scorep sundials Ray orsa mathgl gmsh elk elpa freefem++ netcdf MUMPS gpaw ga nwchem cp2k netcdf-cxx4 netcdf-fortran coin-or-Ipopt -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA/CoRA DivisionFAX: 303-415-9702 3380 Mitchell Lane or...@cora.nwra.com Boulder, CO 80301 http://www.cora.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/27/2015 04:05 PM, Sandro Mani wrote: On 27.07.2015 23:56, Sandro Mani wrote: Ok I've now got one full build of all MPI packages [1]. Investigating the output, things are looking good, except for the fact that I realized that I'll also need to handle binaries MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR - these directories are outside MPI_HOME and hence currently don't get the enhanced requires/provides strings. Since change deadline is tomorrow, I'd like to see the rebuilds starting to land in rawhide+F23 despite of this remaining issue, and then I'll look at cleanly adapting the scripts to also handle the (few) remaining cases. I've prepared a set of git am-able patches to apply to the various packages here [2]. The rebuild sequence needed is posted below. Know FTBFS are paraview (protobuf incompatibility) and netgen-mesher (orrection: that would be gmsh (someone had the great idea of doing something like #define Status int in a public header and I haven't yet got around to finding out who). Affected by the missing handling of MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR are elpa, sundials, pypar and mpi4py. So if any proven packager could fire the rebuilds, that would be much appreciated. Where are we with this? I see some F23 builds but no updates yet. Thanks Sandro [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing2 [2] https://smani.fedorapeople.org/rpm-mpi-hooks/patches/ Rebuild sequence (these are four groups, items within a group can be built together): mpich openmpi mpi4py valgrind pypar boost dl_poly hpl netgen-mesher espresso scotch hdf5 scalapack gromacs scorep sundials Ray orsa mathgl gmsh elk elpa freefem++ netcdf MUMPS gpaw ga nwchem cp2k netcdf-cxx4 netcdf-fortran coin-or-Ipopt -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA/CoRA DivisionFAX: 303-415-9702 3380 Mitchell Lane or...@cora.nwra.com Boulder, CO 80301 http://www.cora.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
Ok I've now got one full build of all MPI packages [1]. Investigating the output, things are looking good, except for the fact that I realized that I'll also need to handle binaries MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR - these directories are outside MPI_HOME and hence currently don't get the enhanced requires/provides strings. Since change deadline is tomorrow, I'd like to see the rebuilds starting to land in rawhide+F23 despite of this remaining issue, and then I'll look at cleanly adapting the scripts to also handle the (few) remaining cases. I've prepared a set of git am-able patches to apply to the various packages here [2]. The rebuild sequence needed is posted below. Know FTBFS are paraview (protobuf incompatibility) and netgen-mesher (someone had the great idea of doing something like #define Status int in a public header and I haven't yet got around to finding out who). Affected by the missing handling of MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR are elpa, sundials, pypar and mpi4py. So if any proven packager could fire the rebuilds, that would be much appreciated. Thanks Sandro [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing2 [2] https://smani.fedorapeople.org/rpm-mpi-hooks/patches/ Rebuild sequence (these are four groups, items within a group can be built together): mpich openmpi mpi4py valgrind pypar boost dl_poly hpl netgen-mesher espresso scotch hdf5 scalapack gromacs scorep sundials Ray orsa mathgl gmsh elk elpa freefem++ netcdf MUMPS gpaw ga nwchem cp2k netcdf-cxx4 netcdf-fortran coin-or-Ipopt -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 27.07.2015 23:56, Sandro Mani wrote: Ok I've now got one full build of all MPI packages [1]. Investigating the output, things are looking good, except for the fact that I realized that I'll also need to handle binaries MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR - these directories are outside MPI_HOME and hence currently don't get the enhanced requires/provides strings. Since change deadline is tomorrow, I'd like to see the rebuilds starting to land in rawhide+F23 despite of this remaining issue, and then I'll look at cleanly adapting the scripts to also handle the (few) remaining cases. I've prepared a set of git am-able patches to apply to the various packages here [2]. The rebuild sequence needed is posted below. Know FTBFS are paraview (protobuf incompatibility) and netgen-mesher (orrection: that would be gmsh (someone had the great idea of doing something like #define Status int in a public header and I haven't yet got around to finding out who). Affected by the missing handling of MPI_PYTHON_SITEARCH and MPI_FORTRAN_MOD_DIR are elpa, sundials, pypar and mpi4py. So if any proven packager could fire the rebuilds, that would be much appreciated. Thanks Sandro [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing2 [2] https://smani.fedorapeople.org/rpm-mpi-hooks/patches/ Rebuild sequence (these are four groups, items within a group can be built together): mpich openmpi mpi4py valgrind pypar boost dl_poly hpl netgen-mesher espresso scotch hdf5 scalapack gromacs scorep sundials Ray orsa mathgl gmsh elk elpa freefem++ netcdf MUMPS gpaw ga nwchem cp2k netcdf-cxx4 netcdf-fortran coin-or-Ipopt -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/21/2015 09:45 AM, Sandro Mani wrote: On 21.07.2015 17:41, Orion Poplawski wrote: On 07/17/2015 09:50 AM, Sandro Mani wrote: Yep - I'm now building things in copr [1]. [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing/ Great. Looks like you need to build a newer openmpi package in your copr since I updating it to 1.8.7 in rawhide. Yes - I'm having some problems with yum in copr not resolving the dependencies, i.e. DEBUG util.py:378: Error: Package: blacs-openmpi-2.0.2-8.1.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: Requires: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-1.8.6-1.2.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-devel-1.8.6-1.2.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-1.8.7-1.fc24.i686 (fedora) DEBUG util.py:378: Not found DEBUG util.py:378: Available: openmpi-devel-1.8.7-1.fc24.i686 (fedora) DEBUG util.py:378: Not found but cannot reproduce this locally (neither with dnf nor with yum-deprecated). Haven't had time to investigate properly yet though (oh and the -devel packages definitely should not be providing the libraries). Again, you need to build a new openmpi in your copr that's newer than the 1.8.7-1 in Fedora now. For the -devel issue, we need: %__mpi_magic^(setuid )?(setgid )?(sticky )?ELF (32|64)-bit.*$ %__mpi_flagsexeonly to match the %__elf_* options. -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 21.07.2015 18:03, Orion Poplawski wrote: On 07/21/2015 09:45 AM, Sandro Mani wrote: On 21.07.2015 17:41, Orion Poplawski wrote: On 07/17/2015 09:50 AM, Sandro Mani wrote: Yep - I'm now building things in copr [1]. [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing/ Great. Looks like you need to build a newer openmpi package in your copr since I updating it to 1.8.7 in rawhide. Yes - I'm having some problems with yum in copr not resolving the dependencies, i.e. DEBUG util.py:378: Error: Package: blacs-openmpi-2.0.2-8.1.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: Requires: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-1.8.6-1.2.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-devel-1.8.6-1.2.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-1.8.7-1.fc24.i686 (fedora) DEBUG util.py:378: Not found DEBUG util.py:378: Available: openmpi-devel-1.8.7-1.fc24.i686 (fedora) DEBUG util.py:378: Not found but cannot reproduce this locally (neither with dnf nor with yum-deprecated). Haven't had time to investigate properly yet though (oh and the -devel packages definitely should not be providing the libraries). Again, you need to build a new openmpi in your copr that's newer than the 1.8.7-1 in Fedora now. For the -devel issue, we need: %__mpi_magic^(setuid )?(setgid )?(sticky )?ELF (32|64)-bit.*$ %__mpi_flagsexeonly to match the %__elf_* options. Ahh sorry didn't realize that this was related to the version mismatch. Ok thanks! -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/17/2015 09:50 AM, Sandro Mani wrote: Yep - I'm now building things in copr [1]. [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing/ Great. Looks like you need to build a newer openmpi package in your copr since I updating it to 1.8.7 in rawhide. -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 21.07.2015 17:41, Orion Poplawski wrote: On 07/17/2015 09:50 AM, Sandro Mani wrote: Yep - I'm now building things in copr [1]. [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing/ Great. Looks like you need to build a newer openmpi package in your copr since I updating it to 1.8.7 in rawhide. Yes - I'm having some problems with yum in copr not resolving the dependencies, i.e. DEBUG util.py:378: Error: Package: blacs-openmpi-2.0.2-8.1.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: Requires: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-1.8.6-1.2.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-devel-1.8.6-1.2.fc24.x86_64 (coprbecloudfedoraprojectorg_results_smani_rpmmpihookstesting_fedorarawhidex86_64_) DEBUG util.py:378: libmpi_mpifh.so.2()(64bit)(openmpi-x86_64) DEBUG util.py:378: Available: openmpi-1.8.7-1.fc24.i686 (fedora) DEBUG util.py:378: Not found DEBUG util.py:378: Available: openmpi-devel-1.8.7-1.fc24.i686 (fedora) DEBUG util.py:378: Not found but cannot reproduce this locally (neither with dnf nor with yum-deprecated). Haven't had time to investigate properly yet though (oh and the -devel packages definitely should not be providing the libraries). -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/14/2015 08:09 AM, Sandro Mani wrote: On 09.07.2015 23:17, Orion Poplawski wrote: On 07/09/2015 03:06 PM, Sandro Mani wrote: Ah yes sorry didn't read the contents of fileattrs/libsymlink.attr properly. But couldn't that be handled with a %__libsymlink_path ^.*\.so$ %__libsymlink_flags magic_and_path instead of the %__libsymlink_exclude_path in libsymlink.attr? Could be. File a bug against redhat-rpm-config for that then. FWIW, filed as #1241737, waiting for a reply. And quickly applied. Excellent, thanks. -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 17.07.2015 17:40, Orion Poplawski wrote: On 07/14/2015 08:09 AM, Sandro Mani wrote: On 09.07.2015 23:17, Orion Poplawski wrote: On 07/09/2015 03:06 PM, Sandro Mani wrote: Ah yes sorry didn't read the contents of fileattrs/libsymlink.attr properly. But couldn't that be handled with a %__libsymlink_path ^.*\.so$ %__libsymlink_flags magic_and_path instead of the %__libsymlink_exclude_path in libsymlink.attr? Could be. File a bug against redhat-rpm-config for that then. FWIW, filed as #1241737, waiting for a reply. And quickly applied. Excellent, thanks. Yep - I'm now building things in copr [1]. [1] https://copr.fedoraproject.org/coprs/smani/rpm-mpi-hooks-testing/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 09.07.2015 23:17, Orion Poplawski wrote:
On 07/09/2015 03:06 PM, Sandro Mani wrote:
On 09.07.2015 21:42, Orion Poplawski wrote:
On 07/09/2015 01:14 PM, Sandro Mani wrote:
On 09.07.2015 20:18, Orion Poplawski wrote:
Also, it doesn't seem to get all of the requires quite right. For
scorep-openmpi I have:
Provides: libscorep_adapter_compiler_event.so.2()(64bit)(openmpi-x86_64)
but:
Requires: libscorep_adapter_compiler_event.so.2()(64bit)
is being emitted. This appears to be coming from:
./fileattrs/libsymlink.attr:%__libsymlink_requires
%{_rpmconfigdir}/elfdeps --provides --soname-only
So it looks like we need to contend with that as well.
# cat fileattrs/libsymlink.attr
# Make libfoo.so symlinks require the soname-provide of the target library
%__libsymlink_requires %{_rpmconfigdir}/elfdeps --provides
--soname-only
%__libsymlink_magic ^symbolic link to `.*lib.*\.so\..*'$
%__libsymlink_exclude_path ^.*[[:digit:]]$
Perhaps with:
%global __libsymlink_exclude_path
^%{_prefix}/lib(64)?/(openmpi|mpich)/.*$
in mpi packages.
Or directly in mpi.attrs?
No, because there is a default %__libsymlink_exclude_path that we don't want
to override in general (think people building rpms locally with
openmpi/mpich-devel installed).
Ah yes sorry didn't read the contents of fileattrs/libsymlink.attr properly.
But couldn't that be handled with a
%__libsymlink_path ^.*\.so$
%__libsymlink_flags magic_and_path
instead of the %__libsymlink_exclude_path in libsymlink.attr?
Could be. File a bug against redhat-rpm-config for that then.
FWIW, filed as #1241737, waiting for a reply.
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/07/2015 03:12 AM, Sandro Mani wrote:
Hello
I've got an initial implementation of this using the rpm dependency generator
hooks, as suggested in the other thread [1].
The resulting scripts are here: https://smani.fedorapeople.org/rpm-mpi-hooks/
There is just one problem: an elf binary in an $MPI_HOME subfolder will now
trigger both the elf as well as the mpi dependency generator, resulting in
both
libfoo.so()(64bit)
libfoo.so()(64bit)(openmpi-x86_64)
Also, isn't the -x86_64 redundant? Also though I guess we don't have an mpi
variable MPI_NAME.
Also, your trick of using:
for module in $(module avail 21 | grep ^mpi/); do
to find the available mpi modules needs a -t option for Lmod. Fortunately
this also works with environment-modules:
for module in $(module -t avail 21 | grep ^mpi/); do
Also, it doesn't seem to get all of the requires quite right. For
scorep-openmpi I have:
Provides: libscorep_adapter_compiler_event.so.2()(64bit)(openmpi-x86_64)
but:
Requires: libscorep_adapter_compiler_event.so.2()(64bit)
is being emitted. This appears to be coming from:
./fileattrs/libsymlink.attr:%__libsymlink_requires
%{_rpmconfigdir}/elfdeps --provides --soname-only
So it looks like we need to contend with that as well.
# cat fileattrs/libsymlink.attr
# Make libfoo.so symlinks require the soname-provide of the target library
%__libsymlink_requires %{_rpmconfigdir}/elfdeps --provides
--soname-only
%__libsymlink_magic ^symbolic link to `.*lib.*\.so\..*'$
%__libsymlink_exclude_path ^.*[[:digit:]]$
Perhaps with:
%global __libsymlink_exclude_path ^%{_prefix}/lib(64)?/(openmpi|mpich)/.*$
in mpi packages.
--
Orion Poplawski
Technical Manager 303-415-9701 x222
NWRA, Boulder/CoRA Office FAX: 303-415-9702
3380 Mitchell Lane or...@nwra.com
Boulder, CO 80301 http://www.nwra.com
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides - Review request
On 07/08/2015 05:03 PM, Sandro Mani wrote: On 08.07.2015 22:28, Orion Poplawski wrote: It appears to be sufficient to define this macro anywhere, not just in elf.attr. So I think it could be added in to a rpm macros file in openmpi/mpich-devel. Ah cool, actually it seems to also work if I add it directly to mpi.attrs, meaning that all pieces are nicely confined in one place. Agreed. Also I think you need (64)? or (|64). Correct, fixed. I'm seeing this also in cmake: $ grep '(64)' /usr/lib/rpm/fileattrs/cmake.attr %__cmake_path ^/usr/lib(64)/cmake/.*/.*(Config\.cmake|-config\.cmake)$ suppose it should be fixed there too. Thanks. Fixed in rawhide now. -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 09.07.2015 20:18, Orion Poplawski wrote:
Also, isn't the -x86_64 redundant? Also though I guess we don't have an mpi
variable MPI_NAME.
Yes it is redundant, but it is the prettiest variable I could find,
given the lack of MPI_NAME.
Also, your trick of using:
for module in $(module avail 21 | grep ^mpi/); do
to find the available mpi modules needs a -t option for Lmod. Fortunately
this also works with environment-modules:
for module in $(module -t avail 21 | grep ^mpi/); do
Ok
Also, it doesn't seem to get all of the requires quite right. For
scorep-openmpi I have:
Provides: libscorep_adapter_compiler_event.so.2()(64bit)(openmpi-x86_64)
but:
Requires: libscorep_adapter_compiler_event.so.2()(64bit)
is being emitted. This appears to be coming from:
./fileattrs/libsymlink.attr:%__libsymlink_requires
%{_rpmconfigdir}/elfdeps --provides --soname-only
So it looks like we need to contend with that as well.
# cat fileattrs/libsymlink.attr
# Make libfoo.so symlinks require the soname-provide of the target library
%__libsymlink_requires %{_rpmconfigdir}/elfdeps --provides
--soname-only
%__libsymlink_magic ^symbolic link to `.*lib.*\.so\..*'$
%__libsymlink_exclude_path ^.*[[:digit:]]$
Perhaps with:
%global __libsymlink_exclude_path ^%{_prefix}/lib(64)?/(openmpi|mpich)/.*$
in mpi packages.
Or directly in mpi.attrs?
Thanks,
Sandro
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides - Review request
On 09.07.2015 21:39, Orion Poplawski wrote: On 07/08/2015 05:03 PM, Sandro Mani wrote: Once the package is approved, openmpi and mpich will need to BuildRequire: rpm-mpi-hooks, and openmpi-devel and mpich-devel will need to Require: rpm-mpi-hooks. And finally, once that is done too, all *-openmpi and *-mpich packages will need to be rebuilt. We (i.e. you :) ) should build this up in a copr first for testing. Yes good idea, I'll start firing builds as soon as the pending issues with rpm-mpi-hooks are resolved. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/09/2015 03:06 PM, Sandro Mani wrote:
On 09.07.2015 21:42, Orion Poplawski wrote:
On 07/09/2015 01:14 PM, Sandro Mani wrote:
On 09.07.2015 20:18, Orion Poplawski wrote:
Also, it doesn't seem to get all of the requires quite right. For
scorep-openmpi I have:
Provides: libscorep_adapter_compiler_event.so.2()(64bit)(openmpi-x86_64)
but:
Requires: libscorep_adapter_compiler_event.so.2()(64bit)
is being emitted. This appears to be coming from:
./fileattrs/libsymlink.attr:%__libsymlink_requires
%{_rpmconfigdir}/elfdeps --provides --soname-only
So it looks like we need to contend with that as well.
# cat fileattrs/libsymlink.attr
# Make libfoo.so symlinks require the soname-provide of the target library
%__libsymlink_requires %{_rpmconfigdir}/elfdeps --provides
--soname-only
%__libsymlink_magic ^symbolic link to `.*lib.*\.so\..*'$
%__libsymlink_exclude_path ^.*[[:digit:]]$
Perhaps with:
%global __libsymlink_exclude_path
^%{_prefix}/lib(64)?/(openmpi|mpich)/.*$
in mpi packages.
Or directly in mpi.attrs?
No, because there is a default %__libsymlink_exclude_path that we don't want
to override in general (think people building rpms locally with
openmpi/mpich-devel installed).
Ah yes sorry didn't read the contents of fileattrs/libsymlink.attr properly.
But couldn't that be handled with a
%__libsymlink_path ^.*\.so$
%__libsymlink_flags magic_and_path
instead of the %__libsymlink_exclude_path in libsymlink.attr?
Could be. File a bug against redhat-rpm-config for that then.
--
Orion Poplawski
Technical Manager 303-415-9701 x222
NWRA, Boulder/CoRA Office FAX: 303-415-9702
3380 Mitchell Lane or...@nwra.com
Boulder, CO 80301 http://www.nwra.com
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 09.07.2015 21:42, Orion Poplawski wrote:
On 07/09/2015 01:14 PM, Sandro Mani wrote:
On 09.07.2015 20:18, Orion Poplawski wrote:
Also, it doesn't seem to get all of the requires quite right. For
scorep-openmpi I have:
Provides: libscorep_adapter_compiler_event.so.2()(64bit)(openmpi-x86_64)
but:
Requires: libscorep_adapter_compiler_event.so.2()(64bit)
is being emitted. This appears to be coming from:
./fileattrs/libsymlink.attr:%__libsymlink_requires
%{_rpmconfigdir}/elfdeps --provides --soname-only
So it looks like we need to contend with that as well.
# cat fileattrs/libsymlink.attr
# Make libfoo.so symlinks require the soname-provide of the target library
%__libsymlink_requires %{_rpmconfigdir}/elfdeps --provides
--soname-only
%__libsymlink_magic ^symbolic link to `.*lib.*\.so\..*'$
%__libsymlink_exclude_path ^.*[[:digit:]]$
Perhaps with:
%global __libsymlink_exclude_path ^%{_prefix}/lib(64)?/(openmpi|mpich)/.*$
in mpi packages.
Or directly in mpi.attrs?
No, because there is a default %__libsymlink_exclude_path that we don't want
to override in general (think people building rpms locally with
openmpi/mpich-devel installed).
Ah yes sorry didn't read the contents of fileattrs/libsymlink.attr
properly. But couldn't that be handled with a
%__libsymlink_path ^.*\.so$
%__libsymlink_flags magic_and_path
instead of the %__libsymlink_exclude_path in libsymlink.attr?
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides - Review request
On 07/08/2015 05:03 PM, Sandro Mani wrote: Once the package is approved, openmpi and mpich will need to BuildRequire: rpm-mpi-hooks, and openmpi-devel and mpich-devel will need to Require: rpm-mpi-hooks. And finally, once that is done too, all *-openmpi and *-mpich packages will need to be rebuilt. We (i.e. you :) ) should build this up in a copr first for testing. -- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane or...@nwra.com Boulder, CO 80301 http://www.nwra.com -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/09/2015 01:14 PM, Sandro Mani wrote:
On 09.07.2015 20:18, Orion Poplawski wrote:
Also, it doesn't seem to get all of the requires quite right. For
scorep-openmpi I have:
Provides: libscorep_adapter_compiler_event.so.2()(64bit)(openmpi-x86_64)
but:
Requires: libscorep_adapter_compiler_event.so.2()(64bit)
is being emitted. This appears to be coming from:
./fileattrs/libsymlink.attr:%__libsymlink_requires
%{_rpmconfigdir}/elfdeps --provides --soname-only
So it looks like we need to contend with that as well.
# cat fileattrs/libsymlink.attr
# Make libfoo.so symlinks require the soname-provide of the target library
%__libsymlink_requires %{_rpmconfigdir}/elfdeps --provides
--soname-only
%__libsymlink_magic ^symbolic link to `.*lib.*\.so\..*'$
%__libsymlink_exclude_path ^.*[[:digit:]]$
Perhaps with:
%global __libsymlink_exclude_path
^%{_prefix}/lib(64)?/(openmpi|mpich)/.*$
in mpi packages.
Or directly in mpi.attrs?
No, because there is a default %__libsymlink_exclude_path that we don't want
to override in general (think people building rpms locally with
openmpi/mpich-devel installed).
--
Orion Poplawski
Technical Manager 303-415-9701 x222
NWRA, Boulder/CoRA Office FAX: 303-415-9702
3380 Mitchell Lane or...@nwra.com
Boulder, CO 80301 http://www.nwra.com
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
On 07/07/2015 03:12 AM, Sandro Mani wrote:
Hello
I've got an initial implementation of this using the rpm dependency generator
hooks, as suggested in the other thread [1].
The resulting scripts are here: https://smani.fedorapeople.org/rpm-mpi-hooks/
There is just one problem: an elf binary in an $MPI_HOME subfolder will now
trigger both the elf as well as the mpi dependency generator, resulting in
both
libfoo.so()(64bit)
libfoo.so()(64bit)(openmpi-x86_64)
being generated for /usr/lib64/openmpi/lib/libfoo.so. However, to achieve the
goal of disambiguating the provides of libfoo and libfoo-openmpi,
libfoo-openmpi should only provide libfoo.so()(64bit)(openmpi-x86_64).
The only solution which comes to mind is adding
%__elf_exclude_path ^%{_prefix}/lib(64)/(openmpi|mpich)/.*$
to elf.attr. Is this acceptable?
It appears to be sufficient to define this macro anywhere, not just in
elf.attr. So I think it could be added in to a rpm macros file in
openmpi/mpich-devel. Also I think you need (64)? or (|64).
Cool stuff, thanks for working on this.
Thanks,
Sandro
[1] https://lists.fedoraproject.org/pipermail/devel/2015-June/211570.html
--
Orion Poplawski
Technical Manager 303-415-9701 x222
NWRA, Boulder/CoRA Office FAX: 303-415-9702
3380 Mitchell Lane or...@nwra.com
Boulder, CO 80301 http://www.nwra.com
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides - Review request
On 08.07.2015 22:28, Orion Poplawski wrote: It appears to be sufficient to define this macro anywhere, not just in elf.attr. So I think it could be added in to a rpm macros file in openmpi/mpich-devel. Ah cool, actually it seems to also work if I add it directly to mpi.attrs, meaning that all pieces are nicely confined in one place. Also I think you need (64)? or (|64). Correct, fixed. I'm seeing this also in cmake: $ grep '(64)' /usr/lib/rpm/fileattrs/cmake.attr %__cmake_path ^/usr/lib(64)/cmake/.*/.*(Config\.cmake|-config\.cmake)$ suppose it should be fixed there too. I've now posted the review request for the rpm-mpi-hooks package at https://bugzilla.redhat.com/show_bug.cgi?id=1241282 A quick double check of the mpi.req and mpi.prov as part of the review would also be appreciated. Once the package is approved, openmpi and mpich will need to BuildRequire: rpm-mpi-hooks, and openmpi-devel and mpich-devel will need to Require: rpm-mpi-hooks. And finally, once that is done too, all *-openmpi and *-mpich packages will need to be rebuilt. Thanks, Sandro -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: F23 Self Contained Change: RPM MPI Requires Provides
Hello
I've got an initial implementation of this using the rpm dependency
generator hooks, as suggested in the other thread [1].
The resulting scripts are here:
https://smani.fedorapeople.org/rpm-mpi-hooks/
There is just one problem: an elf binary in an $MPI_HOME subfolder will
now trigger both the elf as well as the mpi dependency generator,
resulting in both
libfoo.so()(64bit)
libfoo.so()(64bit)(openmpi-x86_64)
being generated for /usr/lib64/openmpi/lib/libfoo.so. However, to
achieve the goal of disambiguating the provides of libfoo and
libfoo-openmpi, libfoo-openmpi should only provide
libfoo.so()(64bit)(openmpi-x86_64).
The only solution which comes to mind is adding
%__elf_exclude_path ^%{_prefix}/lib(64)/(openmpi|mpich)/.*$
to elf.attr. Is this acceptable?
Thanks,
Sandro
[1] https://lists.fedoraproject.org/pipermail/devel/2015-June/211570.html
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
F23 Self Contained Change: RPM MPI Requires Provides
= Proposed Self Contained Change: RPM MPI Requires Provides = https://fedoraproject.org/wiki/Changes/RpmMPIReqProv Change owner(s): Sandro Mani manisandro at gmail dot com Have the rpm-build find-provides and find-requires scripts encode the MPI compiler name in the provides string of a binary to distinguish otherwise identical provides between packages $foo, $foo-openmpi and $foo-mpich. == Detailed Description == Currently, the packages libfoo, libfoo-openmpi and libfoo-mpich providing the library libfoo.so all have a provides string of i.e. libfoo.so()(64bit) While yum used a shortest-package-name rule to choose which package to pick, dnf does not have any rules, and seems to just pick the first match it comes across. Currently the only solution would be to filter the provides from the -openmpi, -mpich packages and add explicit Requires: where needed. I'd like to propose to extend the provides string in such way that it also encodes the MPI implementation, i.e.: $ rpm -qp --provides libfoo libfoo.so()(64bit) $ rpm -qp --provides libfoo-mpich libfoo.so()(64bit)(mpich-x86_64) $ rpm -qp --provides libfoo-openmpi libfoo.so()(64bit)(openmpi-x86_64) To this end, I'm proposing to adapt the find-requires and find-provides scripts as (or similarly to): find-provides find-requires Discussion of these changes are tracked in bug #1232504. This change is intended to coordinate the rebuild of all MPI related packages to ensure all such packages consistently use the new provides format. == Scope == * Proposal owners: - Work on find-provides and find-requires based on feedback. - As soon as updated find-provides and find-requires shipped with rpm-build, do a mass-rebuild of all MPI packages. * Other developers: N/A (not a System Wide Change) * Release engineering: N/A (not a System Wide Change) * Policies and guidelines: N/A (not a System Wide Change) * Trademark approval: N/A (not needed for this Change) -- Jan Kuřík ___ devel-announce mailing list devel-announce@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel-announce
F23 Self Contained Change: RPM MPI Requires Provides
= Proposed Self Contained Change: RPM MPI Requires Provides = https://fedoraproject.org/wiki/Changes/RpmMPIReqProv Change owner(s): Sandro Mani manisandro at gmail dot com Have the rpm-build find-provides and find-requires scripts encode the MPI compiler name in the provides string of a binary to distinguish otherwise identical provides between packages $foo, $foo-openmpi and $foo-mpich. == Detailed Description == Currently, the packages libfoo, libfoo-openmpi and libfoo-mpich providing the library libfoo.so all have a provides string of i.e. libfoo.so()(64bit) While yum used a shortest-package-name rule to choose which package to pick, dnf does not have any rules, and seems to just pick the first match it comes across. Currently the only solution would be to filter the provides from the -openmpi, -mpich packages and add explicit Requires: where needed. I'd like to propose to extend the provides string in such way that it also encodes the MPI implementation, i.e.: $ rpm -qp --provides libfoo libfoo.so()(64bit) $ rpm -qp --provides libfoo-mpich libfoo.so()(64bit)(mpich-x86_64) $ rpm -qp --provides libfoo-openmpi libfoo.so()(64bit)(openmpi-x86_64) To this end, I'm proposing to adapt the find-requires and find-provides scripts as (or similarly to): find-provides find-requires Discussion of these changes are tracked in bug #1232504. This change is intended to coordinate the rebuild of all MPI related packages to ensure all such packages consistently use the new provides format. == Scope == * Proposal owners: - Work on find-provides and find-requires based on feedback. - As soon as updated find-provides and find-requires shipped with rpm-build, do a mass-rebuild of all MPI packages. * Other developers: N/A (not a System Wide Change) * Release engineering: N/A (not a System Wide Change) * Policies and guidelines: N/A (not a System Wide Change) * Trademark approval: N/A (not needed for this Change) -- Jan Kuřík ___ devel-announce mailing list devel-annou...@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel-announce -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
