On Wed, Feb 03, 2016 at 12:09:18AM +0100, Michael Banck wrote: > On Tue, Feb 02, 2016 at 12:19:10AM +0100, Matthias Klose wrote: > > On 01.02.2016 15:38, Emilio Pozuelo Monfort wrote: > > >On 01/02/16 12:41, Alastair McKinstry wrote: > > >> > > >>The upload of openmpi1.10 triggered an auto-openmpi transition, as > > >>expected and wanted. > > >>The rest, is premature. > > > > > >The problem is that you uploaded the package directly to unstable, without > > >asking for a transition slot and without fixing those problems you mention > > >beforehand. It'd have been better to upload it to experimental until > > >things were > > >ready. > > > > > >Since that has already happened, it'd be good to know what packages build > > >fine > > >against the new openmpi and which ones don't. > > > > fyi, I was surprised as well, tried to go forward, and ended up with > > http://people.canonical.com/~ubuntu-archive/transitions/html/openmpi.html > > (only tried dependency level 1 to build). > > Thanks! > > I look at some of mine and they seem to FTBFS due to scalapack on longer > being installable, which in turn is due to blacs-mpi FTBFSing: > > /usr/bin/ld: cannot find -lmpi_f77 > > That library has vanished from libopenmpi-dev, Alastair, was that > intentional, and if so, how is Fortran linking supposed to work now? > > I removed that library from blacs-mpi link line and it built fine, but I > am not sure this is the right fix.
Nevermind, I saw Alastair has filed a patch (#813490). As blacs-mpi has been NMU'd before and looks rather central, I suggest to zero-day NMU this, unless somebody objects. Michael
