(I've re-orded the post a bit) Steve McMillan wrote: > > Jack Howarth wrote: > >> On Fri, Apr 17, 2009 at 10:09:23AM +0200, Martin Costabel wrote: >> >> >>> Steve McMillan wrote: >>> >>> >>>> I get the following error when I try to install openmpi v. 1.3.1-1000 >>>> from source on MacOS 10.5.6. The installation proceeds for ~15 >>>> minutes, then ends with: >>>> >>>> . >>>> Making all in doc >>>> make[7]: Nothing to be done for `all'. >>>> make[7]: Nothing to be done for `all-am'. >>>> Making all in vtfilter >>>> g++ -DHAVE_CONFIG_H -I. -I../.. -I../../extlib/otf/otflib >>>> -I../../extlib/otf/otflib -I../../vtlib/ -I../../vtlib -I/sw/include >>>> -fopenmp -DVT_OMP -O3 -DNDEBUG -finline-functions -c -o >>>> vtfilter-vt_filter.o `test -f 'vt_filter.cc' || echo './'`vt_filter.cc >>>> cc1plus: error: unrecognized command line option "-fopenmp" >>>> make[6]: *** [vtfilter-vt_filter.o] Error 1 >>>> >>>> >>> This is doubly weird: >>> >>> First, because the configure script explicitly tests for the >>> availability of the "-fopenmp" flag. In my case (as in Alexander's, >>> apparently), it finds that this flag does not work and then disables it >>> for the rest of the build procedure. In your case, configure seems to >>> think that the flag can be used, but then it turns out later that it >>> cannot. >>> >>> On a second degree, the package itself is weird, because it requires >>> gcc43 (which would understand the "-fopenmp" flag, contrary to Apple's >>> gcc-4.0.1), but then it doesn't use it for building. >>> >>> Maybe the package maintainer can explain. >>> >>> -- >>> Martin >>> >>> >> Martin, >> The openmpi package need gcc43 to build the fortran support in >> openmpi. If I had to guess, the most likely thing going on here is >> that a local symlink or alias has been made to gcc-4 so that it >> is being executed as gcc. A simple 'gcc -v' will reveal if that >> is the case. >> Jack >> >> > >
> OK. I updated my Xcode to version 3.1.2 and retried the installation, > and I get exactly the same error. I also thought that the version of > gcc might be the issue -- I have had fink packages fight one another in > the past, installing different versions of gcc over one another -- and > sometimes the version isn't what I thought. However, running gcc -v > says I have version 4.3.3. > > Steve > > They don't really fight. Most Fink packages use Xcode's compiler _regardless_ of whether a Fink gcc package is installed. Our gcc packages are specifically designed not to disrupt this. If in a batch build you happen to be building a Fink package that requires gcc42 and one that requires gcc43, say, things are set up to switch between the two cleanly. What you have going on appears to be exactly what the maintainer just mentioned. The openmpi build is assuming here that "gcc" is the system's version, and not Fink's, and that is _not_ 4.3.3. -- Alexander Hansen Fink User Liaison ------------------------------------------------------------------------------ Stay on top of everything new and different, both inside and around Java (TM) technology - register by April 22, and save $200 on the JavaOne (SM) conference, June 2-5, 2009, San Francisco. 300 plus technical and hands-on sessions. Register today. Use priority code J9JMT32. http://p.sf.net/sfu/p _______________________________________________ Fink-beginners mailing list [email protected] http://news.gmane.org/gmane.os.apple.fink.beginners
