hdf5 @1.8.15-patch1_0+gcc5+mpich test fails

[dstrubbe (David Strubbe)]

The test phase fails for hdf5 +gcc5 +mpich -cxx. This is not one of the "experimental" variants that the port warns about. The failure seems to lead to test failures in the dependents netcdf, netcdf-fortran, and octopus. I found that +gcc5 and +gcc5 +cxx both were successful, but +gcc5 +mpich +cxx also failed. Unfortunately the test log does not provide much information about the problem; I'm not sure if there are more details to be found anywhere else. The file hdf5/work/hdf5-1.8.15-patch1/test/testhdf5.chklog just has the part below, which is from the main.log. I am using XCode 6.4 on OSX 10.10.5 Yosemite.

/Applications/Xcode.app/Contents/Developer/usr/bin/make  check-TESTS
===Serial tests in test begin Wed Oct  7 12:19:44 EDT 2015===
============================
Testing  testhdf5 
============================
 testhdf5  Test Log
============================
Command terminated abnormally.
       11.57 real         0.03 user         0.03 sys
make[4]: *** [testhdf5.chkexe_] Error 1
make[3]: *** [build-check-s] Error 2
make[2]: *** [test] Error 2
make[1]: *** [check-am] Error 2
make: *** [check-recursive] Error 1
Command failed:  cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_tarballs_ports_science_hdf5/hdf5/work/hdf5-1.8.15-patch1" && /usr/bin/make check 
Exit code: 2
Error: org.macports.test for port hdf5 returned: command execution failed

[mascguy (Christopher Nielsen)]

This still looks like an issue, even with a newer gcc, like gcc 7.

Interestingly enough, though, I'm not seeing that failure when utilizing openmpi as the MPI provider.

[mascguy (Christopher Nielsen)]

There's not much we can do to fix, without someone spending significant time digging into this.

If runtime issues are occurring, then we can take another look. Otherwise let's close.