Created attachment 913431 [details] build.log … -- Found FFMPEG: /usr/include (found suitable version "6.1.2", minimum required is "2.3.3") found components: avformat avcodec avutil swscale swresample -- Found HDF5: /usr/lib64/libhdf5.so;/usr/lib64/libsz.so;/usr/lib64/libz.so;/usr/lib64/libdl.a;/usr/lib64/libm.so (found version "1.14.4-3") found components: C HL -- Found OGG: /usr/lib64/libogg.so -- Found THEORA: /usr/lib64/libtheora.so -- Found NetCDF: /usr/include (found version "4.9.2") -- Found nlohmann_json: /usr/share/cmake/nlohmann_json/nlohmann_jsonConfig.cmake (found version "3.11.3") -- Found SQLite3: /usr/include (found version "3.47.1") -- Found LibPROJ: /usr/lib64/libproj.so (found version "9.4.1") -- Searching for PEGTL -- Searching for PEGTL - found target taocpp::pegtl -- Found LibXml2: /usr/lib64/libxml2.so (found version "2.13.5") -- Found CGNS: /usr/include (found suitable version "4.40", minimum required is "4.10") -- Found LibHaru: /usr/lib64/libhpdf.so (found suitable version "2.4.2", minimum required is "2.4.0") -- Found GL2PS: /usr/lib64/libgl2ps.so (found suitable version "1.4.2", minimum required is "1.4.2") -- Could not use git to determine source version, using version 2.0.0 CMake Error at /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:233 (message): Could NOT find HDF5 (missing: HDF5_INCLUDE_DIRS) (found version "") Call Stack (most recent call first): /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:603 (_FPHSA_FAILURE_MESSAGE) CMake/patches/99/FindHDF5.cmake:1006 (find_package_handle_standard_args) ThirdParty/vtkm/vtkvtkm/vtk-m/CMakeLists.txt:128 (find_package)
it *does* build correctly when sci-libs/hdf5 has been built with USE=cxx. the ebuild of sci-libs/vtk-9.3.1 contains an unconditional RDEPEND to sci-libs/hdf5:=[mpi=] Additionally, there is a conditional dependency on sci-libs/cgnslib[hdf5] when USE=cgns As earlier <=vtk-9.3.0 had merged just fine with sci-libs/hdf5[-cxx], I guess it is the newly introduced USE=cgns that now would need USE=cxx enabled on hdf5. However, rather than extending the dependency here, it might be better suited to be built into sci-libs/cgnslib[hdf5]?
(In reply to Christian Bricart from comment #1) > > However, rather than extending the dependency here, it might be better > suited to be built into sci-libs/cgnslib[hdf5]? Let's get the obvious fix in for now and worry about that later.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=5aac98ba33e5dee3a4cb74d451fcfde895d14c2b commit 5aac98ba33e5dee3a4cb74d451fcfde895d14c2b Author: Sam James <sam@gentoo.org> AuthorDate: 2025-01-07 23:39:21 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2025-01-07 23:39:39 +0000 sci-libs/vtk: depend on sci-libs/hdf5[cxx] for USE=cgns Closes: https://bugs.gentoo.org/945964 Signed-off-by: Sam James <sam@gentoo.org> sci-libs/vtk/vtk-9.3.1.ebuild | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-)
After this change, for sci-libs/vtk to compile, sci-libs/hdf5, sci-libs/netcdf and sci-libs/vtk must have mpi USE flag disabled.
(In reply to Alex Belits from comment #4) > After this change, for sci-libs/vtk to compile, sci-libs/hdf5, > sci-libs/netcdf and sci-libs/vtk must have mpi USE flag disabled. It's not clear to me how this change could cause that. Please file a new bug with the full logs.
(In reply to Alex Belits from comment #4) > After this change, for sci-libs/vtk to compile, sci-libs/hdf5, > sci-libs/netcdf and sci-libs/vtk also sci-libs/cgnslib > must have mpi USE flag disabled.
(In reply to Sam James from comment #5) > (In reply to Alex Belits from comment #4) > > After this change, for sci-libs/vtk to compile, sci-libs/hdf5, > > sci-libs/netcdf and sci-libs/vtk must have mpi USE flag disabled. > > It's not clear to me how this change could cause that. Please file a new bug > with the full logs. sci-libs/hdf5-1.14.4_p3-r1 has > REQUIRED_USE=" > !unsupported? ( > cxx? ( !mpi ) mpi? ( !cxx ) > threads? ( !cxx !mpi !fortran !hl ) Disabling mpi for sci-libs/hdf5 requires disabling it for everything that uses hdf5, so when cxx is mandatory, mpi is broken.
Ah, I see.
> REQUIRED_USE=" > !unsupported? ( > cxx? ( !mpi ) mpi? ( !cxx ) > threads? ( !cxx !mpi !fortran !hl ) You can currently build it by setting USE=unsupported for sci-libs/hdf5. We could probably add that to the DEPEND of vtk if that makes the portage output easier to understand. BUT, fixing it properly requires dealing with the changes from https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=679682fb9fad0112c1f2684241db91814a66e7e0 and that will only happen in vtk-9.4.1. Upstream vtk doesn't seem to deal with passing --enable-unsupported when building hdf5, yet they build it with both enabled. They are also still on hdf-1.13 ( https://gitlab.kitware.com/vtk/vtk/-/issues/19527 ). Upstream hdf5 doesn't state explicitly why it's unsupported, only thing I found is https://github.com/HDFGroup/hdf5/blob/hdf5_1.14.4.3/configure.ac#L2112-L2114 . I also get no test failures for hdf5-1.14.5 with USE="cxx mpi unsupported". So if someone knows why it's unsupported, let me know.
(In reply to Paul Zander from comment #9) > So if someone knows why it's unsupported, let me know. As for the unsupported, there is a description which tells that basically all the high-level apis of the hdf5 are not thread-safe (and something similar for the parallelised version) --- thus unsupported (see https://github.com/HDFGroup/hdf5/blob/7bf340440909d468dbb3cf41f0ea0d87f5050cea/release_docs/INSTALL_CMake.txt#L925). Couldn't find anything else.
