#38065: openmpi @1.6.3_0+gcc47+valgrind -- header locations not self-consistent ----------------------------+------------------- Reporter: beany_kelly@… | Owner: mww@… Type: defect | Status: new Priority: Normal | Milestone: Component: ports | Version: 2.1.3 Resolution: | Keywords: Port: openmpi | ----------------------------+------------------- Changes (by larryv@…): * owner: macports-tickets@… => mww@… * keywords: headers => Old description:

I have a Mountain Lion system (10.8.2) with an up-to-date MacPorts installation. I've installed OpenMPI as above (+gcc47 +valgrind), but I'm having trouble compiling parallel code with it, because the headers aren't where the system expects. Here's a sample error from the compiler:

/opt/local/include/mpi.h:2087:41: fatal error: openmpi/ompi/mpi/cxx/mpicxx.h: No such file or directory

The problem is that apart from mpi.h itself, all the headers are buried deep down, one layer feeper than mpi.h indicates. For instance, mpicxx.h is *actually* in (relative to /opt/local/include):

openmpi/openmpi/ompi/mpi/cxx/mpicxx.h

I don't know why so much nesting is needed to begin with, but it looks like the path on line 2087 of mpi.h is out by one "openmpi/".

I can fix this path by hand, of course, but then the same error crops up again with another path ...

I've tried a clean port uninstall and new installation, but no change.

New description: I have a Mountain Lion system (10.8.2) with an up-to-date MacPorts installation. I've installed OpenMPI as above (+gcc47 +valgrind), but I'm having trouble compiling parallel code with it, because the headers aren't where the system expects. Here's a sample error from the compiler: {{{ /opt/local/include/mpi.h:2087:41: fatal error: openmpi/ompi/mpi/cxx/mpicxx.h: No such file or directory }}} The problem is that apart from mpi.h itself, all the headers are buried deep down, one layer feeper than mpi.h indicates. For instance, mpicxx.h is *actually* in (relative to /opt/local/include): {{{ openmpi/openmpi/ompi/mpi/cxx/mpicxx.h }}} I don't know why so much nesting is needed to begin with, but it looks like the path on line 2087 of mpi.h is out by one "openmpi/". I can fix this path by hand, of course, but then the same error crops up again with another path ... I've tried a clean port uninstall and new installation, but no change. -- Comment: Thanks for the ticket. In the future, please Cc relevant port maintainers and use [[WikiFormatting]] to format the description. -- Ticket URL: <https://trac.macports.org/ticket/38065#comment:1> MacPorts <http://www.macports.org/> Ports system for Mac OS

#38065: openmpi @1.6.3_0+gcc47+valgrind -- header locations not self-consistent ----------------------------+------------------- Reporter: beany_kelly@… | Owner: mww@… Type: defect | Status: new Priority: Normal | Milestone: Component: ports | Version: 2.1.3 Resolution: | Keywords: Port: openmpi | ----------------------------+------------------- Comment (by beany_kelly@…): Hi & thanks. Actually, I ''wasn't'' using the wrappers --- though I had thought I was. Now that I try to use them, I'm not getting the header issues any more, but things still aren't compiling. The wrappers aren't reporting what I expect, either: {{{ openmpicc --version }}} yields the following: {{{ Apple clang version 4.0 (tags/Apple/clang-421.0.60) (based on LLVM 3.1svn) Target: x86_64-apple-darwin12.2.0 Thread model: posix }}} This seems strange to me, because I requested openmpi +gcc47, as noted above. Why would it be reporting Clang? The which and locate commands don't show any other instances of openmpicc on my system. -- Ticket URL: <https://trac.macports.org/ticket/38065#comment:4> MacPorts <http://www.macports.org/> Ports system for Mac OS

#38065: openmpi @1.6.3_0+gcc47+valgrind -- header locations not self-consistent ----------------------------+------------------- Reporter: beany_kelly@… | Owner: mww@… Type: defect | Status: new Priority: Normal | Milestone: Component: ports | Version: 2.1.3 Resolution: | Keywords: Port: openmpi | ----------------------------+------------------- Comment (by beany_kelly@…): Hi eborisch, and thanks for the information. Wow; I didn't realise I had no control over the C/C++ compilers with the MacPorts OpenMPI installation. This mixed-compiler approach seems bizarre to me. I'd appreciate your linking to this discussion, as I can't find it using the Ticket Search box. In the mean time, I'll try MPICH instead, as per your suggestion. -- Ticket URL: <https://trac.macports.org/ticket/38065#comment:6> MacPorts <http://www.macports.org/> Ports system for Mac OS