Viv Kendon wrote: > On Wed, 1 Jul 2009, Alexander Hansen wrote: > >>> /sw/fink/dists/unstable/main/binary-darwin-i386/devel/openmpi_1.3.2-1002_darwin-i386.deb >>> >>> (--install): >>> trying to overwrite `/sw/bin/mpicc', which is also in >>> package openmpi-dev > >> The issue here is that openmpi and openmpi-dev have files that install >> in the same directory but only differ by case, e.g. 'mpiCC' vs. >> 'mpicc'. On a case-insensitive filesystem these look the same. > > This would explain everything, except it is happening on a case > sensitive file system: > > Volumes: > hachi: > Capacity: 232.57 GB > Available: 170.74 GB > Writable: Yes > File System: Case-sensitive Journaled HFS+ > BSD Name: disk0s2 > Mount Point: / > > (I had to go check!) Is fink itself doing a case-insensitive comparison? Oh, right; the dpkg package manager was modified to do a case-insensitive comparison, so that we were consistent between case-sensitive and case-insensitive (pick the more restrictive option). > >> I _think_ you may be able to work around this by using >> >> sudo dpkg -r openmpi openmpi-dev >> fink install openmpi >> sudo dpkg -i --force-overwrite >> /sw/fink/dists/unstable/main/binary-darwin-i386/devel/openmpi-dev_1.3.2-1002_darwin-i386.deb >> > > I followed the above, it worked, with a warning about mpiCC this time > (upper case instead of lower case). And here is what I now have > installed: > > ls -l /sw/bin/*mpicc /sw/bin/*mpiCC > 8 lrwxr-xr-x 1 root admin 16 Jul 1 17:20 /sw/bin/mpiCC@ -> > /sw/bin/om-mpicc > 8 lrwxr-xr-x 1 root admin 12 Jul 1 17:19 /sw/bin/om-mpicc@ -> > opal_wrapper > > And to prove it is case sensitive, so the file system isn't the cause > of the problem: > > sudo touch /sw/bin/mpicc; ls -l /sw/bin/*mpicc /sw/bin/*mpiCC > 8 lrwxr-xr-x 1 root admin 16 Jul 1 17:20 /sw/bin/mpiCC@ -> > /sw/bin/om-mpicc > 0 -rw-r--r-- 1 root admin 0 Jul 1 17:25 /sw/bin/mpicc > 8 lrwxr-xr-x 1 root admin 12 Jul 1 17:19 /sw/bin/om-mpicc@ -> > opal_wrapper > >> I'm not sure why the upstream developers insist on this behavior, but >> it's something that needs to be properly addressed within the Fink >> packaging. > > Maybe someone tried to do this and it isn't working as intended? Not as of yet; the maintainer is working on it. > > Meanwhile, my rebuild grace is proceeding nicely, fingers crossed for > it solving my original problem. > > many thanks, > -- Viv > ________________________________________________ > Dr Viv Kendon http://quantum.leeds.ac.uk/~viv > tel: +44 113 343 4864 Physics and Astronomy > Quantum Information Group University of Leeds >
-- Alexander Hansen Fink User Liaison ------------------------------------------------------------------------------ _______________________________________________ Fink-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fink-users
