Bug#586149: Please consider the usage of update-alternatives for hdf5 libraries
Hi, There's a better way to do this by just having different shared library file names and sonames for each of the shared library packages. Then the -dev packages can conflict, while one can install multiple packages using the different shared libraries simultaneously, not one-at-a-time as update-alternatives would do. Note this would also allow simultaneous install of hdf5-tools and libhdf5-mpi-dev which is required to build several HDF5 reverse-depends, such as med-fichier. This represents a major regression vs. 1.8.6 and should be fixed as soon as possible. I'll work on a patch tomorrow and will post it to this bug. -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-grass-devel
Bug#586149: Please consider the usage of update-alternatives for hdf5 libraries
severity 586149 serious retitle 586149 Conflict between hdf5-tools and libhdf5-mpi-dev prevents several packages from building block 661301 by 586149 thanks Hi again, Because this causes other packages to FTBFS, it really needs to be an RC bug. -Adam On Sun, 2012-02-26 at 00:12 -0500, Adam C Powell IV wrote: Hi, There's a better way to do this by just having different shared library file names and sonames for each of the shared library packages. Then the -dev packages can conflict, while one can install multiple packages using the different shared libraries simultaneously, not one-at-a-time as update-alternatives would do. Note this would also allow simultaneous install of hdf5-tools and libhdf5-mpi-dev which is required to build several HDF5 reverse-depends, such as med-fichier. This represents a major regression vs. 1.8.6 and should be fixed as soon as possible. I'll work on a patch tomorrow and will post it to this bug. -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-grass-devel
Re: paraview and gmsh are not installable together, hdf5-issue.
Hi Anton, [Just addressing d-s-m since I think we're all subscribed, and adding the HDF5 maintainers.] On Wed, 2012-02-15 at 21:00 +0100, Anton Gladky wrote: Hi all, there is a problem with installing paraview and gmsh together in sid now. gmsh requires libpetsc3.2, which is dependent on libhdf5-openmpi-7. paraview is dependent on hdf5-7, which is conflicting with libhdf5-openmpi-7. There's a similar problem with the MED library: it build-depends on libhdf5-mpi-dev, which depends on libhdf5-openmpi-7, and hdf5-tools, which depends on libhdf5-7. Any idea how to solve it? Given that Paraview and PETSc are basically parallel (can run on 1 CPU but are really built for more; MED is in the same category), I think it makes sense for them both to link with libhdf5-openmpi-7. So if Paraview builds with libhdf5-mpi-dev that should solve this problem -- unless it requires hdf5-tools to build like MED. Here are two other ways to solve the problem: * Change the sonames of the -openmpi, -mpich[2] and -lam HDF5 libraries so libhdf5-7 does not conflict with libhdf5-openmpi-7 etc. * Make a new hdf5-mpi-tools package, or a set of four (one for each MPI implementation) with a meta-package depending on the arch-specific MPI implementation. HDF5 team, what do you think? I'd be happy to propose a patch for either of the above remedies. -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#604714: libhdf5-mpi-dev: Wrong MPI dependency on LAM platforms
Package: libhdf5-mpi-dev Version: 1.8.4-patch1-2 Severity: important Tags: patch Greetings, The libhdf5-mpi-dev package depends on libopenmpi-dev whether the platform defaults to openmpi or lam. This is because it checks the /etc/alternatives symlink to mpi, which will always point to openmpi for hdf5, because it build-depends on all of the MPI implementations, and OpenMPI has the highest priority. (This was my mistake in copying the old PETSc MPI rule; PETSc's build process installs only the arch's default MPI so there's only one /etc/alternatives candidate.) The attached patch vs. current SVN checks the dependency of the mpi-default-dev package instead, so it is always consistent with mpi-defaults. Thanks, Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ Index: trunk/debian/changelog === --- trunk/debian/changelog (revision 2951) +++ trunk/debian/changelog (working copy) @@ -1,3 +1,11 @@ +hdf5 (1.8.4-patch1-3) unstable; urgency=low + + [ Adam C. Powell, IV ] + * Correct the libhdf5-mpi-dev dependency on the default MPI version +(closes: #6x). + + -- + hdf5 (1.8.4-patch1-2) unstable; urgency=low [ Adam C. Powell, IV ] Index: trunk/debian/rules === --- trunk/debian/rules (revision 2951) +++ trunk/debian/rules (working copy) @@ -22,7 +22,7 @@ DEB_BUILD_GNU_TYPE ?= $(shell dpkg-architecture -qDEB_BUILD_GNU_TYPE) DEB_BUILD_ARCH ?= $(shell dpkg-architecture -qDEB_BUILD_ARCH) -DEFAULT_MPI=$(shell readlink /etc/alternatives/mpi | sed s/usr//g | sed s/include//g | sed s/lib//g | sed s/\\///g) +DEFAULT_MPI=$(shell dpkg -s mpi-default-dev | grep Depends | sed s/Depends: lib// | sed s/-dev// | sed s/lam4/lam/) patch: patch-stamp patch-stamp: signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#604714: libhdf5-mpi-dev: Wrong MPI dependency on LAM platforms
Hello again, Please discard the last patch, a better one is attached. -Adam On Tue, 2010-11-23 at 13:57 -0500, Adam C Powell IV wrote: Package: libhdf5-mpi-dev Version: 1.8.4-patch1-2 Severity: important Tags: patch Greetings, The libhdf5-mpi-dev package depends on libopenmpi-dev whether the platform defaults to openmpi or lam. This is because it checks the /etc/alternatives symlink to mpi, which will always point to openmpi for hdf5, because it build-depends on all of the MPI implementations, and OpenMPI has the highest priority. (This was my mistake in copying the old PETSc MPI rule; PETSc's build process installs only the arch's default MPI so there's only one /etc/alternatives candidate.) The attached patch vs. current SVN checks the dependency of the mpi-default-dev package instead, so it is always consistent with mpi-defaults. Thanks, Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ Index: trunk/debian/changelog === --- trunk/debian/changelog (revision 2951) +++ trunk/debian/changelog (working copy) @@ -1,3 +1,11 @@ +hdf5 (1.8.4-patch1-3) unstable; urgency=low + + [ Adam C. Powell, IV ] + * Correct the libhdf5-mpi-dev dependency on the default MPI version +(closes: #604714). + + -- + hdf5 (1.8.4-patch1-2) unstable; urgency=low [ Adam C. Powell, IV ] Index: trunk/debian/rules === --- trunk/debian/rules (revision 2951) +++ trunk/debian/rules (working copy) @@ -22,7 +22,7 @@ DEB_BUILD_GNU_TYPE ?= $(shell dpkg-architecture -qDEB_BUILD_GNU_TYPE) DEB_BUILD_ARCH ?= $(shell dpkg-architecture -qDEB_BUILD_ARCH) -DEFAULT_MPI=$(shell readlink /etc/alternatives/mpi | sed s/usr//g | sed s/include//g | sed s/lib//g | sed s/\\///g) +DEFAULT_MPI=$(shell dpkg -s mpi-default-dev | grep Depends | sed s/Depends: lib// | sed s/Depends: // | sed s/-dev// | sed s/lam4/lam/) patch: patch-stamp patch-stamp: signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#510057: Bug#510057: Consider creating default MPI version
Thank you Francesco for taking care of this! -Adam On Thu, 2010-04-22 at 19:00 -0400, Adam C Powell IV wrote: Dear Francesco, This patch has been available since early February now, and without it a number of packages are blocked. Those include new packages for Salomé and Code-Aster, a mpi-defaults transition for Octave, and a working version of med-fichier. With the freeze coming up soon, it would help a lot of people to get this done, as these engineering simulation packages could enter Debian in time for the squeeze release. Also, mpi-defaults in unstable still uses lam, so the patch is still current, though it won't be for long. What more do you need to make this happen? How can I help? -Adam On Wed, 2010-02-03 at 09:29 -0500, Adam C Powell IV wrote: tags 510057 +patch thanks Hello Francesco, On Wed, 2010-01-27 at 14:28 -0500, Adam C Powell IV wrote: On Wed, 2010-01-27 at 20:10 +0100, Francesco P. Lovergine wrote: That said, a patch for add mpi-defaults to current support is of course welcome. Okay, I will put some time into this approach (adding mpi-defaults, leaving the three in place) when I get some time, probably next week. I'm attaching a patch vs. current SVN which adds an extra package libhdf5-mpi-dev, that package is empty and depends on the default MPI version of libhdf5-*-dev for each platform. It has to be Arch: any because it will depend on a different package for different platforms. This will provide a good migration path for squeeze: people who have been using a specific MPI implementation can still do so, while those who want to make their package depend on the default MPI version can have something in squeeze which should be compatible with squeeze+1. There's one problem though: mpi-defaults is about to transition from lam to mpich2 on platforms which don't support openmpi. This will require a new mpich2 version of HDF5 to make it work. I'd be glad to make a patch for that if you like. -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#571453: hdf5: Please drop build-dependency on LAM/MPI and/or MPICH
Hi Manuel and GIS devs, Just FYI, Manuel, Francesco has indicated that he would rather continue to build all flavors of hdf5 for squeeze in bug 510057, so it's not likely that this bug will be closed by the squeeze release. That said, if 510057 is closed, it will give the reverse-depends an mpi-defaults transition path for squeeze, which would help with the removal of MPICH and LAM after the release... -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#510057: Bug#510057: Consider creating default MPI version
tags 510057 +patch thanks Hello Francesco, On Wed, 2010-01-27 at 14:28 -0500, Adam C Powell IV wrote: On Wed, 2010-01-27 at 20:10 +0100, Francesco P. Lovergine wrote: That said, a patch for add mpi-defaults to current support is of course welcome. Okay, I will put some time into this approach (adding mpi-defaults, leaving the three in place) when I get some time, probably next week. I'm attaching a patch vs. current SVN which adds an extra package libhdf5-mpi-dev, that package is empty and depends on the default MPI version of libhdf5-*-dev for each platform. It has to be Arch: any because it will depend on a different package for different platforms. This will provide a good migration path for squeeze: people who have been using a specific MPI implementation can still do so, while those who want to make their package depend on the default MPI version can have something in squeeze which should be compatible with squeeze+1. There's one problem though: mpi-defaults is about to transition from lam to mpich2 on platforms which don't support openmpi. This will require a new mpich2 version of HDF5 to make it work. I'd be glad to make a patch for that if you like. -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ diff -urN debian.old/changelog debian/changelog --- debian.old/changelog 2010-02-03 08:04:03.0 -0500 +++ debian/changelog 2010-02-03 09:27:24.0 -0500 @@ -1,3 +1,10 @@ +hdf5 (1.8.4-7) unstable; urgency=low + + * Added libhdf5-mpi-dev package which simply depends on default MPI version +of HDF5 for each platform. + + -- Adam C. Powell, IV hazel...@debian.org Wed, 03 Feb 2010 09:02:05 -0500 + hdf5 (1.8.4-6) unstable; urgency=low * Fixed typo in 1.8.4-4 changelog. diff -urN debian.old/control debian/control --- debian.old/control 2010-02-03 08:04:03.0 -0500 +++ debian/control 2010-02-03 09:27:24.0 -0500 @@ -6,7 +6,7 @@ Build-Depends: libmpich1.0-dev (= 1.2.7-1), zlib1g-dev, lam4-dev (= 7.1.1-3.2), quilt, libopenmpi-dev [!arm !armel !hppa !mips !mipsel !s390 !sh4 !m68k], libjpeg62-dev | libjpeg-dev, debhelper ( 7), sed (=4.1.5), gfortran, libibverbs-dev [!arm !armel !hppa !mips !mipsel !s390 !kfreebsd-i386 !kfreebsd-amd64 !hurd-i386 !sh4], - sharutils + sharutils, mpi-default-dev Standards-Version: 3.8.3 Homepage: http://hdfgroup.org/HDF5/ Vcs-Browser: http://svn.debian.org/viewsvn/pkg-grass/packages/hdf5/trunk/ @@ -140,6 +140,19 @@ This package contains development files for use with MPICH. Warning: the C++ interface is not provided for this version. +Package: libhdf5-mpi-dev +Section: libdevel +Priority: extra +Architecture: any +Depends: ${hdf5-mpi-dev}, mpi-default-dev +Description: Hierarchical Data Format 5 (HDF5) - development files - MPICH version + HDF5 is a file format and library for storing scientific data. + HDF5 was designed and implemented to address the deficiencies of + HDF4.x. It has a more powerful and flexible data model, supports + files larger than 2 GB, and supports parallel I/O. + . + This package depends on the default MPI version of HDF5 for each platform. + Package: libhdf5-doc Section: doc Architecture: all diff -urN debian.old/control.in debian/control.in --- debian.old/control.in 2010-02-03 08:04:03.0 -0500 +++ debian/control.in 2010-02-03 09:27:24.0 -0500 @@ -6,7 +6,7 @@ Build-Depends: libmpich1.0-dev (= 1.2.7-1), zlib1g-dev, lam4-dev (= 7.1.1-3.2), quilt, libopenmpi-dev [!arm !armel !hppa !mips !mipsel !s390 !sh4 !m68k], libjpeg62-dev | libjpeg-dev, debhelper ( 7), sed (=4.1.5), gfortran, libibverbs-dev [!arm !armel !hppa !mips !mipsel !s390 !kfreebsd-i386 !kfreebsd-amd64 !hurd-i386 !sh4], - sharutils + sharutils, mpi-default-dev Standards-Version: 3.8.3 Homepage: http://hdfgroup.org/HDF5/ Vcs-Browser: http://svn.debian.org/viewsvn/pkg-grass/packages/hdf5/trunk/ @@ -140,6 +140,19 @@ This package contains development files for use with MPICH. Warning: the C++ interface is not provided for this version. +Package: libhdf5-mpi-dev +Section: libdevel +Priority: extra +Architecture: any +Depends: ${hdf5-mpi-dev}, mpi-default-dev +Description: Hierarchical Data Format 5 (HDF5) - development files - MPICH version + HDF5 is a file format and library for storing scientific data. + HDF5 was designed and implemented to address the deficiencies of + HDF4.x. It has a more powerful and flexible data model, supports + files larger than 2 GB, and supports parallel I/O. + . + This package depends on the default MPI version of HDF5 for each platform. + Package: libhdf5-doc Section: doc Architecture: all diff -urN debian.old/rules debian/rules --- debian.old/rules 2010-02-03 08:04:03.0 -0500 +++ debian/rules 2010-02-03 09:27:24.0 -0500 @@ -22,6 +22,8 @@ DEB_BUILD_GNU_TYPE ?= $(shell dpkg-architecture -qDEB_BUILD_GNU_TYPE) DEB_BUILD_ARCH ?= $(shell dpkg
[DebianGIS-dev] Bug#510057: Consider creating default MPI version
Greetings, Because MPICH1 and LAM are long since end-of-life, this is taking on a new urgency. The Debian Science team is planning to deprecate the LAM and MPICH1 packages for squeeze, and remove them entirely for squeeze+1. [1] For this reason, a package built against mpi-defaults should be the only MPI version of most packages for squeeze and moving forward. [1] http://lists.debian.org/debian-science/2009/11/msg00010.html I would be happy to provide a patch to make this happen, as I did to create the openmpi version. If you prefer to have the current three plus an mpi-defaults version, I would be happy to make that happen as well, just let me know either way. Regards, Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
Re: [DebianGIS-dev] [Pkg-scicomp-devel] Open MPI 1.3 related rebuilds
On Sun, 2009-02-01 at 17:05 -0600, Steve M. Robbins wrote: On Sun, Feb 01, 2009 at 10:52:18AM -0600, Dirk Eddelbuettel wrote: On 1 February 2009 at 00:37, Steve M. Robbins wrote: | Hello Dirk, | | On Mon, Jan 26, 2009 at 09:10:04PM -0600, Dirk Eddelbuettel wrote: | | Open MPI 1.3, released a few days ago, recommends a rebuild of its | dependencies. | | Given the fairly small number of affected packages, we are hoping do drive | this rebuild 'informally' rather than with the full force of libopenmpi-1.2 | and libopenmpi-1.3 libraries (which would still require rebuilds, but also | delays via NEW etc pp). Details and logs of this initiative are on a new | wiki page at http://wiki.debian.org/OpenMPI13Transition | | The wiki page states that Open MPI 1.3 is not binary compatible with | Open MPI 1.2. Presumably, therefore, the SONAME has changed. Not quite. It's more complicated because of the fact that at least three different implementations (LAM/MPI, MPICH and MPICH2, OpenMPI) provide /usr/lib/libmpi.so and there is no soname encoded. That's not quite the case; the SONAME for the OpenMPI-supplied library is libmpi.so.0: $ objdump -p /usr/lib/libmpi.so|grep SON SONAME libmpi.so.0 If a program linked against libmpi.so.0 fails to run with the new libmpi.so.0, then either there is a bug in OpenMPI 1.3 or the SONAME should change. This is not confined to relinking dependent libraries in Debian packages: you will also break user-built code. Also note that it isn't strictly incompatible. Some users of Open MPI that were built agains 1.2.* continue to work under 1.3. Sure, but the test for re-using an existing SONAME is Is the ABI the same, or not? i.e. Does *every* program continue to run, or not? My understanding is that the answer is no. I have to agree with Steve. We have versioned shlib package names for a reason, though the second of these conditions (do old programs still run) is more important than the first (has the ABI changed at all). If you keep the same lib package name but can't provide continuity, it's a bug. It's not just a matter of best practices. Debian policy is pretty clear about this. (It might even be an RC bug, though I'm not sure and don't have the time to look into policy just now...) If upstream doesn't change their soname, the new package should use something like -1.3 in the shlib package name. IIRC it's not a requirement that the old and new versions be simultaneously installable or that the soname change (e.g. compiler ABI changes don't require this), so feel free to leave the soname as upstream has it. -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ signature.asc Description: This is a digitally signed message part ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#473871: hdf5: Doesn't actually build openmpi libs
Package: hdf5 Version: 1.6.6-3 Greetings, I'm afraid your rules file doesn't build openmpi libraries on the architectures where it's supposed to, so the openmpi packages are empty. Looking through the build logs on every architecture where it's supposed to work (amd64, ia64, sparc, powerpc), if I search the log for openmpi, I get the package installation, a couple of symlinks at the top, then dh_makeshlibs at the very end, with no references to openmpi in between. I don't understand why this is; your system seems sound and logical. I realize that I'm the cause of a lot of your openmpi troubles, having suggested this in the first place. So I think I've come up with an elegant solution to this problem, found in the patch. Instead of testing for the architecture in rules, each openmpi target just tests for the file /usr/lib/openmpi/include/mpi.h, and if it finds it, configures/builds/installs. This patch works on amd64, haven't tested it elsewhere. The advantage of this approach is that when more arches support openmpi, you can just adjust the arch lists in control, and rules will automatically build the new openmpi packages. I hope this patch meets with your approval, and solves this problem of openmpi support once and for all. Regards, -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ --- hdf5-1.6.6/debian/rules~ 2008-04-02 02:41:06.0 + +++ hdf5-1.6.6/debian/rules 2008-04-02 02:40:36.0 + @@ -18,38 +18,11 @@ DEB_HOST_GNU_TYPE ?= $(shell dpkg-architecture -qDEB_HOST_GNU_TYPE) DEB_BUILD_GNU_TYPE ?= $(shell dpkg-architecture -qDEB_BUILD_GNU_TYPE) -DEB_BUILD_ARCH ?= $(shell dpkg-architecture -qDEB_BUILD_ARCH) # To be used if binary packages were buildable on any arch #ARCH_FLAG=-a ARCH_FLAG=-s -ifeq ($(DEB_BUILD_ARCH),arm) -build_openmpi = no -else ifeq ($(DEB_BUILD_ARCH),armel) -build_openmpi = no -else ifeq ($(DEB_BUILD_ARCH),hppa) -build_openmpi = no -else ifeq ($(DEB_BUILD_ARCH),mips) -build_openmpi = no -else ifeq ($(DEB_BUILD_ARCH),mipsel) -build_openmpi = no -else ifeq ($(DEB_BUILD_ARCH),s390) -build_openmpi = no -else -build_openmpi = yes -endif - -ifeq ($(build_openmpi),yes) -configure_stamp_openmpi = configure-stamp-openmpi -build_stamp_openmpi = build-stamp-openmpi -install_openmpi = install-openmpi -else -configure_stamp_openmpi = -build_stamp_openmpi = -install_openmpi = -endif - ifeq (,$(findstring noopt,$(DEB_BUILD_OPTIONS))) USE_PROD = yes else @@ -80,7 +53,7 @@ --enable-shared --enable-production=$(USE_PROD) -configure: configure-stamp-debian configure-stamp-serial configure-stamp-lam $(configure-stamp-openmpi) configure-stamp-mpich +configure: configure-stamp-debian configure-stamp-serial configure-stamp-lam configure-stamp-openmpi configure-stamp-mpich configure-stamp-debian: debian/control.in cd debian for i in *_devlib; do j=`basename $$i _devlib`; \ @@ -119,12 +92,14 @@ configure-stamp-openmpi: configure-stamp-debian dh_testdir - -mkdir debian/build-openmpi -# configure version with lam +# configure version with openmpi + if [ -e /usr/lib/openmpi/include/mpi.h ]; then \ + mkdir debian/build-openmpi; \ cd debian/build-openmpi CPPFLAGS=-I/usr/lib/openmpi/include \ CC=mpicc.openmpi CXX=mpic++.openmpi RUNPARALLEL=/usr/bin/mpirun.openmpi \ ../../configure $(CONFIGURE_FLAGS) \ - --enable-parallel=yes + --enable-parallel=yes; \ + fi touch configure-stamp-openmpi configure-stamp-mpich: configure-stamp-debian @@ -138,7 +113,7 @@ --enable-parallel=yes touch configure-stamp-mpich -build: build-stamp-serial build-stamp-lam $(build-stamp-openmpi) build-stamp-mpich +build: build-stamp-serial build-stamp-lam build-stamp-openmpi build-stamp-mpich build-stamp-serial: configure-stamp-serial dh_testdir @@ -152,7 +127,9 @@ build-stamp-openmpi: configure-stamp-openmpi dh_testdir - $(MAKE) -C debian/build-openmpi/ + if [ -e /usr/lib/openmpi/include/mpi.h ]; then \ + $(MAKE) -C debian/build-openmpi/; \ + fi touch build-stamp-openmpi build-stamp-mpich: configure-stamp-mpich @@ -192,10 +169,12 @@ install-openmpi: build-stamp-openmpi dh_testdir dh_testroot - -mkdir debian/build-openmpi/tmpinst - $(MAKE) -C debian/build-openmpi/ install prefix=$(CURDIR)/debian/build-openmpi/tmpinst/usr + if [ -e /usr/lib/openmpi/include/mpi.h ]; then \ + mkdir debian/build-openmpi/tmpinst; \ + $(MAKE) -C debian/build-openmpi/ install prefix=$(CURDIR)/debian/build-openmpi/tmpinst/usr; \ dh_install -p$(openmpipack) -p$(package)-openmpi-dev \ - --sourcedir=debian/build-openmpi/tmpinst + --sourcedir=debian/build-openmpi/tmpinst; \ + fi install-mpich: build-stamp-mpich dh_testdir @@ -222,7 +201,7 @@ dh_md5sums -i dh_builddeb -i -binary-arch: build install-serial install-lam $(install-openmpi) install-mpich +binary-arch: build install-serial install-lam install-openmpi install-mpich dh_testdir
[DebianGIS-dev] Bug#457080: Bug#457080: hdf5: Please add OpenMPI build
On Sun, 2008-01-13 at 15:07 -0500, Adam C Powell IV wrote: Hi again, On Wed, 2008-01-09 at 13:14 +0100, Francesco P. Lovergine wrote: On Wed, Jan 09, 2008 at 06:56:11AM -0500, Adam C Powell IV wrote: On Wed, 2008-01-02 at 12:34 +0100, Francesco P. Lovergine wrote: On Mon, Dec 31, 2007 at 09:26:10PM -0500, Adam C Powell IV wrote: On Wed, 2007-12-19 at 11:04 -0500, Adam C Powell IV wrote: As OpenMPI is the successor to LAM (the LAM developers now work on OpenMPI), please add an OpenMPI build of hdf5. Will you accept a patch if I provide one? Cheers, -Adam Sure. I put up the patch about a week ago (before this reply). Does it look acceptable? [BTW, what is DebianGIS-dev?] Yes, it is now integrated ini the svn repository for the next upstream release. It will integrate a version transition, so it is still not uploaded. See http://wiki.debian.org/DebianGis Thanks again. I'm afraid there's a problem with my patch. Please remove the file hdf5-H5public-openmpi.patch and the line in the debian/rules install-openmpi target which applies it. It could cause breakage and should not be there. Sorry about that! Just curious, any progress or timeline hints for the next upload? (Are you packaging 1.8.0?) I have a some packages I'd like to upload which use this, and would like to start to port them sooner rather than later. That is, unless you're planning to just stick with 1.6.x for the lenny release... Thanks, -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#457080: Bug#457080: hdf5: Please add OpenMPI build
On Thu, 2008-03-20 at 15:31 +0100, Francesco P. Lovergine wrote: On Thu, Mar 20, 2008 at 09:51:53AM -0400, Adam C Powell IV wrote: On Sun, 2008-01-13 at 15:07 -0500, Adam C Powell IV wrote: Hi again, On Wed, 2008-01-09 at 13:14 +0100, Francesco P. Lovergine wrote: On Wed, Jan 09, 2008 at 06:56:11AM -0500, Adam C Powell IV wrote: On Wed, 2008-01-02 at 12:34 +0100, Francesco P. Lovergine wrote: On Mon, Dec 31, 2007 at 09:26:10PM -0500, Adam C Powell IV wrote: On Wed, 2007-12-19 at 11:04 -0500, Adam C Powell IV wrote: As OpenMPI is the successor to LAM (the LAM developers now work on OpenMPI), please add an OpenMPI build of hdf5. Will you accept a patch if I provide one? Cheers, -Adam Sure. I put up the patch about a week ago (before this reply). Does it look acceptable? [BTW, what is DebianGIS-dev?] Yes, it is now integrated ini the svn repository for the next upstream release. It will integrate a version transition, so it is still not uploaded. See http://wiki.debian.org/DebianGis Thanks again. I'm afraid there's a problem with my patch. Please remove the file hdf5-H5public-openmpi.patch and the line in the debian/rules install-openmpi target which applies it. It could cause breakage and should not be there. Sorry about that! Just curious, any progress or timeline hints for the next upload? (Are you packaging 1.8.0?) I have a some packages I'd like to upload which use this, and would like to start to port them sooner rather than later. That is, unless you're planning to just stick with 1.6.x for the lenny release... Done a few hours ago, package is pending in the NEW queue. Ah, terrific! Thanks, hadn't looked there; good timing. :-) -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#463393: hdf5: FTBFS: missing library libh5test-1.6.5.so.0
Package: hdf5 Version: 1.6.5-5 Severity: serious Tags: patch Greetings, The binary h5perf in the hdf5-tools package requires the library libh5test-1.6.5.so.0 which is not in any binary package. With hdf5-tools installed, this results in: # /usr/bin/h5perf h5perf: error while loading shared libraries: libh5test-1.6.5.so.0: cannot open shared object file: No such file or directory Because of this, dh_shlibdeps is failing: dh_shlibdeps -phdf5-tools -Llibhdf5-serial-1.6.5-0 -ldebian/libhdf5-serial-1.6.5-0/usr/lib dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/gif2h5 shouldn't be linked with libpthread.so.0 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/gif2h5 shouldn't be linked with libz.so.1 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/gif2h5 shouldn't be linked with libm.so.6 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5dump shouldn't be linked with libpthread.so.0 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5dump shouldn't be linked with libz.so.1 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5dump shouldn't be linked with libm.so.6 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h52gif shouldn't be linked with libpthread.so.0 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h52gif shouldn't be linked with libz.so.1 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h52gif shouldn't be linked with libm.so.6 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5jam shouldn't be linked with libpthread.so.0 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5jam shouldn't be linked with libz.so.1 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5jam shouldn't be linked with libm.so.6 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5import shouldn't be linked with libpthread.so.0 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5import shouldn't be linked with libz.so.1 (it uses none of its symbols). dpkg-shlibdeps: warning: debian/hdf5-tools/usr/bin/h5import shouldn't be linked with libm.so.6 (it uses none of its symbols). dpkg-shlibdeps: failure: couldn't find library libh5test-1.6.5.so.0 needed by debian/hdf5-tools/usr/bin/h5perf (its RPATH is ''). Note: libraries are not searched in other binary packages that do not have any shlibs or symbols file. To help dpkg-shlibdeps find private libraries, you might need to set LD_LIBRARY_PATH. dh_shlibdeps: command returned error code 512 make: *** [binary-arch] Error 1 It seems dh_shlibdeps has become stricter, so this now causes FTBFS, which is RC. :-( Note the library is sitting right in the test directory. Aha, it defines PUB_LIB= so PUB_LIB=$(LIB) is ignored. The attached patch fixes this, and makes the install_shlib and install_devlib files more general so it gets installed in the package. Now the package builds, and the result is: # /usr/bin/h5perf No parallel IO performance because parallel is not configured Cheers, -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ --- hdf5-1.6.5/test/Makefile.in~ 2005-07-21 14:49:03.0 + +++ hdf5-1.6.5/test/Makefile.in 2008-01-30 23:01:19.0 + @@ -39,7 +39,6 @@ LIB=libh5test.la LIB_SRC=h5test.c testframe.c LIB_OBJ=$(LIB_SRC:.c=.lo) -PUB_LIB= ## Temporary files. These files are the ones created by setting the ## HDF5_NOCLEANUP environment variable and running `make test' without --- hdf5-1.6.5/debian/install_shlib~ 2008-01-30 23:58:30.0 + +++ hdf5-1.6.5/debian/install_shlib 2008-01-30 23:59:03.0 + @@ -1 +1 @@ -usr/lib/libhdf5*.so.* +usr/lib/libh*.so.* --- hdf5-1.6.5/debian/install_devlib~ 2008-01-30 23:58:30.0 + +++ hdf5-1.6.5/debian/install_devlib 2008-01-30 23:59:54.0 + @@ -1,5 +1,5 @@ usr/include -usr/lib/libhdf5*.so -usr/lib/libhdf5*.settings -usr/lib/libhdf5*.la -usr/lib/libhdf5*.a +usr/lib/libh*.so +usr/lib/libh*.settings +usr/lib/libh*.la +usr/lib/libh*.a ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#457080: Bug#457080: hdf5: Please add OpenMPI build
Hi again, On Wed, 2008-01-09 at 13:14 +0100, Francesco P. Lovergine wrote: On Wed, Jan 09, 2008 at 06:56:11AM -0500, Adam C Powell IV wrote: On Wed, 2008-01-02 at 12:34 +0100, Francesco P. Lovergine wrote: On Mon, Dec 31, 2007 at 09:26:10PM -0500, Adam C Powell IV wrote: On Wed, 2007-12-19 at 11:04 -0500, Adam C Powell IV wrote: As OpenMPI is the successor to LAM (the LAM developers now work on OpenMPI), please add an OpenMPI build of hdf5. Will you accept a patch if I provide one? Cheers, -Adam Sure. I put up the patch about a week ago (before this reply). Does it look acceptable? [BTW, what is DebianGIS-dev?] Yes, it is now integrated ini the svn repository for the next upstream release. It will integrate a version transition, so it is still not uploaded. See http://wiki.debian.org/DebianGis Thanks again. I'm afraid there's a problem with my patch. Please remove the file hdf5-H5public-openmpi.patch and the line in the debian/rules install-openmpi target which applies it. It could cause breakage and should not be there. Sorry about that! -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#457080: Bug#457080: hdf5: Please add OpenMPI build
On Wed, 2008-01-02 at 12:34 +0100, Francesco P. Lovergine wrote: On Mon, Dec 31, 2007 at 09:26:10PM -0500, Adam C Powell IV wrote: On Wed, 2007-12-19 at 11:04 -0500, Adam C Powell IV wrote: As OpenMPI is the successor to LAM (the LAM developers now work on OpenMPI), please add an OpenMPI build of hdf5. Will you accept a patch if I provide one? Cheers, -Adam Sure. I put up the patch about a week ago (before this reply). Does it look acceptable? [BTW, what is DebianGIS-dev?] Thanks, -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel
[DebianGIS-dev] Bug#457080: hdf5: Please add OpenMPI build
On Wed, 2007-12-19 at 11:04 -0500, Adam C Powell IV wrote: As OpenMPI is the successor to LAM (the LAM developers now work on OpenMPI), please add an OpenMPI build of hdf5. Will you accept a patch if I provide one? Cheers, -Adam -- GPG fingerprint: D54D 1AEE B11C CE9B A02B C5DD 526F 01E8 564E E4B6 Engineering consulting with open source tools http://www.opennovation.com/ ___ Pkg-grass-devel mailing list Pkg-grass-devel@lists.alioth.debian.org http://lists.alioth.debian.org/mailman/listinfo/pkg-grass-devel