Arch Linux User Repository

Doesn't compile for me:

/tmp/elrick-temp/makepkg/build/mingw-w64-hdf5/src/hdf5-1.12.2/fortran/src/H5_buildiface.F90:55:53:

   55 |   H5_H5CONFIG_F_RKIND
      |                                                     1
Error: Syntax error in array constructor at (1)
/tmp/elrick-temp/makepkg/build/mingw-w64-hdf5/src/hdf5-1.12.2/fortran/src/H5_buildiface.F90:57:41:

   57 |   H5_H5CONFIG_F_NUM_IKIND
      |                                         1
Error: Invalid character in name at (1)
/tmp/elrick-temp/makepkg/build/mingw-w64-hdf5/src/hdf5-1.12.2/fortran/src/H5_buildiface.F90:58:44:

   58 |   H5_H5CONFIG_F_IKIND
      |                                            1
Error: Explicit array shape at (1) must be constant of INTEGER type and not UNKNOWN type
...

Any idea?

yes, mingw-w64-wine should be updated first to silence wine warnings

I'm getting a CMake error with the latest version:

CMake Warning at config/cmake/HDF5UseFortran.cmake:337 (message):


            Fortran REAL(KIND=0050:err:winediag:nodrv_CreateWindow L"The explorer process failed to start.") is $max_real_fortran_sizeof Bytes, but no corresponding C float type exists of that size
                                             !!! Fortran interfaces will not be generated for REAL(KIND=0050:err:winediag:nodrv_CreateWindow L"The explorer process failed to start.") !!!
Call Stack (most recent call first):
  CMakeLists.txt:1069 (include)


CMake Error at config/cmake/HDF5UseFortran.cmake:342 (math):
  math cannot parse the expression: "4,8,10,16 - 1": syntax error, unexpected
  exp_NUMBER, expecting end of file (4).
Call Stack (most recent call first):
  CMakeLists.txt:1069 (include

Can you please add: -DHDF5_BUILD_FORTRAN=ON \ ?

From 1.12.0-2 to 1.12.1-1 support for C++ has been removed. Could you please add

--- PKGBUILD?h=mingw-w64-hdf5   2021-10-11 12:50:49.000000000 +0200
+++ PKGBUILD    2021-10-11 12:16:46.139345645 +0200
@@ -33,6 +33,7 @@
     mkdir -p build-${_arch} && pushd build-${_arch}
     ${_arch}-cmake \
       -DHDF5_ENABLE_Z_LIB_SUPPORT=ON \
+      -DHDF5_BUILD_CPP_LIB=ON \
       -DBUILD_TESTING=OFF \
       -DHDF5_BUILD_TOOLS=OFF \
       -DHDF5_BUILD_EXAMPLES=OFF \

Line 59 of PKGBUILD is incorrect. Here is a patch to rectify that issue:

--- PKGBUILD.old        2018-06-16 09:52:43.000000000 +0000
+++ PKGBUILD            2018-06-16 09:23:50.263953815 +0000
@@ -56,7 +56,7 @@

 package() {
   for _arch in $_architectures; do
-    cd "$srcdir/hdf5-${_arch}/build-${_arch}"
+    cd "$srcdir/hdf5-${pkgver/_/-}/build-${_arch}"
     make DESTDIR="${pkgdir}" install
     rm "$pkgdir"/usr/${_arch}/share/{COPYING,*.txt}
     ${_arch}-strip --strip-unneeded "$pkgdir"/usr/${_arch}/bin/*.dll

Thanks for the further comments. I agree that it is simpler to do so, specially if you mention that there is further hassle in maintenance. I tried without wine and setting -DHDF5_BUILD_TOOLS=ON. This attempts (and fails) to generate H5lib_settings.c, but at first glance the problem seems to be that the mingw-cmake is used when building the H5make_libsettings file, which results in a windows binary (therefore needing the workaround of using wine). This happens in file src/hdf5-1.10.1/build-x86_64-w64-mingw32/src/CMakeFiles/hdf5-static.dir/build.make. If we, on the other hand, were to specify that these few files should be built with the regular Linux cmake, then wine would not be needed. I do not know enough of the cmake system to say if this is easy to do, and I have the feeling that it is more hassle than worth as you suggested earlier. Let us leave it for now :) Thanks again!

Its easier to just require wine for the try compile checks instead of setting each cmake results var because it's a pain to maintain as they may change between releases or 32/64 arch. HDF5_BUILD_TOOLS is already off, and I doubt it has an impact on runtime checks.

Thanks again, and sorry for taking so long to come back with some feedback. The szip package compiled without problems but I did not attempt to build/test if it works. The new hdf5 PKGBUILD, on the other hand, works and the zlib compression is also effective when used. One comment on requiring wine to build this package, though. It occurred to me that perhaps we can compile the configuration tools (such as src/H5make_libsettings.c, which is responsible to generate H5lib_settings.c in the source before the actual compilation). Have you ever tried using "DHDF5_BUILD_TOOLS" to avoid the calls to wine?

updated, could you check ezacaria? also added szip package if you want to enable it.