relevant snippet... checking whether -lc should be explicitly linked in... no creating libtool checking for perl... perl checking for production mode... "production" checking for ceil in -lm... yes checking if should build only statically linked executables... no checking for large file support mode on Linux... enabled checking for f90... mpif90 checking whether the Fortran 9X compiler (mpif90 -Wl,-O1,--enable-new-dtags) works... no configure: error: installation or configuration problem: Fortran 9X compiler cannot create executables. configure: error: /bin/sh './configure' failed for fortran !!! ERROR: sci-libs/hdf5-1.6.5-r1 failed. Call stack: ebuild.sh, line 1545: Called dyn_compile ebuild.sh, line 940: Called src_compile hdf5-1.6.5-r1.ebuild, line 125: Called die !!! configure failed !!! If you need support, post the topmost build error, and the call stack if relevant.
Created attachment 92359 [details] emerge --info
Upgrading to gcc-4.1.1 with the USE="fortran" flag in place still does not fix this problem
Created attachment 97396 [details, diff] patch to use gfortran or g77 should fix for users with gcc-4
I've got some similar issues. My emerge info is attached. First, "out of the box" attempt: # emerge -v hdf5 2>&1 | tee ~znmeb/attempt1 Calculating dependencies ... done! >>> Emerging (1 of 1) sci-libs/hdf5-1.6.5-r1 to / * hdf5-1.6.5.tar.gz MD5 ;-) ... [ ok ] * hdf5-1.6.5.tar.gz RMD160 ;-) ... [ ok ] * hdf5-1.6.5.tar.gz SHA1 ;-) ... [ ok ] * hdf5-1.6.5.tar.gz SHA256 ;-) ... [ ok ] * hdf5-1.6.5.tar.gz size ;-) ... [ ok ] * checking ebuild checksums ;-) ... [ ok ] * checking auxfile checksums ;-) ... [ ok ] * checking miscfile checksums ;-) ... [ ok ] * checking hdf5-1.6.5.tar.gz ;-) ... [ ok ] * You need one of these Fortran Compilers: gfortran * Installed are: gfortran * Using gfortran * Configuring for GNU fortran... >>> Unpacking source... >>> Unpacking hdf5-1.6.5.tar.gz to /var/tmp/portage/sci-libs/hdf5-1.6.5-r1/work * Applying patches for selected FORTRAN compiler: gfortran >>> Source unpacked. >>> Compiling source in /var/tmp/portage/sci-libs/hdf5-1.6.5-r1/work/hdf5-1.6.5 ... * C++ support cannot be enabled without both enabling static * library support and disabling mpi. Fortran9X support needs * both static and mpi enabled to work with GNU fortran. * Suggested USE flags for fortran and mpi using gfortran: * USE="fortran mpi static -cxx" !!! ERROR: sci-libs/hdf5-1.6.5-r1 failed. Call stack: ebuild.sh, line 1568: Called dyn_compile ebuild.sh, line 937: Called src_compile hdf5-1.6.5-r1.ebuild, line 96: Called die !!! Please adjust your use flags. !!! If you need support, post the topmost build error, and the call stack if relevant. # ----------------------------------------------------------------------------- Well, they must know what they're talking about, right. So I changed the USE flags the way they asked for. It's a lot longer, but the news at the end is just as bad. See attached file "attempt2" for the details. I think there's some more magic that has to happen with the USE flags for "mpich", which got emerged the first time I set the "mpi" USE flag.
Created attachment 100466 [details] emerge info from znmeb
Created attachment 100467 [details] Log of first attempt with default USE flags
Created attachment 100468 [details] Log of second attempt, after applying recommended USE flags, which brought in "mpich"
It looks to me as if mpich doesn't provide a gfortran wrapper and the configure hence dies looking for one! I am not sure if that is mpich's expected behavior or not; I've cc'd the cluster folks who might be able to shed some light on this. Unfortunately, I don't know hdf5 very well, but it looks like the step from hdf5-1.6.5 to hdf5-1.6.5-r1 added a lot of complexity to the ebuild. In particular, I have no idea why enabling fortran requires pulling in mpi. Maybe somebody in the sci team who knows hdf5 better can comment on that. Thanks, Markus Thanks, Markus
*** Bug 205888 has been marked as a duplicate of this bug. ***
No longer in tree
