Control: tag -1 confirmed On Tue, Sep 2, 2014 at 16:29:45 +0200, Salvatore Bonaccorso wrote: > I guess this request is quite unfortunate as it affects a rebuild of a > ppackage in stable. Note I'm not the maintainer but we runned into > this problem. > > It is tracked in BTS as #748130. The binary packages produced by > src:scotch in wheezy were not (re-)builded at the time picking the > correct default mpi-implementation. > > On Wed, May 14, 2014 at 05:42:48PM +0200, Pierre Saramito wrote: > > Hi all, > > > > The stable (wheezy) debian package "libscotch-5.1" depends on MPI lib as : > > > > dep: libmpich2-3 [amd64, mips, mipsel, s390, s390x] > > dep: libopenmpi1.3 [not amd64, mips, mipsel, s390, s390x] > > > > while the "mpi-default-bin" package has the following dependencies : > > > > dep: mpich2 [mips, mipsel, s390, s390x] > > dep: openmpi-bin [not mips, mipsel, s390, s390x] > > > > The same bug occurs for all MPI-dependent packages from the "scotch" source package. > > Thus, on amd64, the scotch library is not usable: > > package depending on scotch will compile but segfault at run time. > > > > A possible solution should be to recompile the wheezy/amd64 scotch package. > > Can the scotch package be binnmu'ed on stable for amd64? > > Is > > nmu scotch_5.1.12b.dfsg-1.2 . amd64 . -d wheezy . -m "Rebuild to pick the correct mpi-default implementation (Closes: #748130)" > > correct? > The -d (or the . after it) is wrong, and binNMU changelogs can't close bugs, but other than that, yes. I've scheduled the rebuild. Cheers, Julien
Attachment:
signature.asc
Description: Digital signature