Stefan Behnel wrote: > Dag Sverre Seljebotn, 18.05.2010 16:13: >> Stefan Behnel wrote: >>> Lisandro Dalcin, 18.05.2010 15:38: >>>> building any of these MPI from sources takes time, it will load the >>>> machine paying very little to us. >>> Hudson only has to do it once, so that the packaged binary becomes >>> available. Artifacts created in one build job can be reused in another, and >>> Hudson keeps a trace back to the original build job that the artifact >>> originated from. >> If Sage is built in such a job (the Sage distro, not the library, but >> the latter cannot build without the former) > > I didn't know that. Does this mean we have to build basically-all-of Sage > before we can build the sagelib to test Cython with it? Or can we cut down > the dependencies to something smaller and more interesting? (I guess > 'interesting' is pretty relative here, though...) > > >> then NumPy is available in Sage by default.
This might not be the best idea -- Sage only works for one Python version at the time (currently 2.6), while we want NumPy to be present on all Python versions, in particular we want to test NumPy+Cython+Python 3.x way before Sage takes the leap to 3.x. >> >> OpenMPI is also (if this is not outdated information) available as an >> SPKG, and so easily installable in Sage through >> >> ./sage -i openmpi >> >> or something like that. > > It would also need the -devel packages (header files & friends), no idea > how those are installed. I'm pretty sure SPKGs include that stuff by default (they end up in the-sage/local/include). -- Dag Sverre _______________________________________________ Cython-dev mailing list [email protected] http://codespeak.net/mailman/listinfo/cython-dev
