Hello, I've just finished installing R 2.0.1 for our UNIX systems, which include Tru64, Solaris and IRIX, plus Linux on x86 and PPC. I am doing this on AFS, which allows me to minimize the number of files in the installation by keeping one copy only of everything that's generic and separate copies of platform-specific files only, with run-time linking to the correct files for each platform provided through the AFS "@sys" magic, which expands to a particular system name at run-time. I then proceeded to install Bioconductor. It created some files which are specific to a particular platform. Installing for another platform would overwrite these files, so I renamed them temporarily and proceeded to install Bioconductor on another platform. All the generic files are now there, only the .so's for a different platform are missing. getBioC() on the other platform says that all the packages are up to date, but obviously R would barf trying to load Linux x86 .so's on any other platform, so... How can I force "getBioC()" to recreate the .so's for every platform? If I can avoid installing R and Bioconductor separately for each platform, I would very much like to do that. -- Atro Tossavainen (Mr.) / The Institute of Biotechnology at Systems Analyst, Techno-Amish & / the University of Helsinki, Finland, +358-9-19158939 UNIX Dinosaur / employs me, but my opinions are my own. < URL : http : / / www . helsinki . fi / %7E atossava / > NO FILE ATTACHMENTS

Hi Atro, On Feb 9, 2005, at 11:14 PM, Atro Tossavainen wrote: > How can I force "getBioC()" to recreate the .so's for every platform? > One possibility is to not use getBioC(). Instead download the package tarballs and install them with R CMD INSTALL. This should force a compile even if the package is already installed. You might also try specifying the destdir parameter. Given an empty destdir, I'm hoping you'll get the behavior you want. > If I can avoid installing R and Bioconductor separately for each > platform, > I would very much like to do that. Eh? Haven't you already had to do this? Given that the platforms are not all binary compatible, I don't think there's any way to avoid this. If it were me, I would go for the "disk space is cheap" notion and spend time automating install scripts to make admin'ing all those installs easier. Good luck, + seth

Hello Seth, > > How can I force "getBioC()" to recreate the .so's for every platform? > > One possibility is to not use getBioC(). Instead download the package > tarballs and install them with R CMD INSTALL. This should force a > compile even if the package is already installed. Will try, thanks. > You might also try specifying the destdir parameter. Given an empty > destdir, I'm hoping you'll get the behavior you want. I'm sorry, but I can't see where there would be an empty destdir I could use. > > If I can avoid installing R and Bioconductor separately for each > > platform, I would very much like to do that. > > Eh? Haven't you already had to do this? Given that the platforms are > not all binary compatible, I don't think there's any way to avoid this. In a way, yes. I've compiled everything separately on all the platforms we use, then installed everything in the same place with just one copy of the files that are identical across platforms and separate copies for those that aren't, with AFS magic taking care of picking the right copy of non-identical files on each platform. (On AFS, you can use the @sys macro that the client will expand to its own platform name at runtime in file names. Therefore, if "file" is a symlink to "file.@sys", the client will look for, say, "file.i386_linux24", or "file.alpha_dux40", or... you get the picture.) > If it were me, I would go for the "disk space is cheap" notion and > spend time automating install scripts to make admin'ing all those > installs easier. Disk space is cheap, but making sure that all of the separate installs are up to the same level with add-on packages could potentially become a major pain, which is why I try to do the same with anything I install on AFS for multiple platforms. -- Atro Tossavainen (Mr.) / The Institute of Biotechnology at Systems Analyst, Techno-Amish & / the University of Helsinki, Finland, +358-9-19158939 UNIX Dinosaur / employs me, but my opinions are my own. < URL : http : / / www . helsinki . fi / %7E atossava / > NO FILE ATTACHMENTS