Bug#854905: nmu: petsc_3.7.5+dfsg1-4
user release.debian@packages.debian.org usertag 854905 nmu thanks On 11/02/17 21:41, Adrian Bunk wrote: > Control: reassign -1 release.debian.org > Control: retitle -1 nmu: petsc_3.7.5+dfsg1-4 > Control: severity -1 normal > Control: tags -1 - sid > Control: affects -1 libpetsc3.7.5-dev You forgot to usertag it... > > On Sat, Feb 11, 2017 at 11:32:39AM -0800, Dima Kogan wrote: >> Package: libpetsc3.7.5-dev >> Severity: grave >> >> Hi. Currently libpetsc3.7.5-dev is uninstallable. Sbuild resolver says: >> >> missing: >> pkg: >> package: libpetsc3.7.5-dev >> version: 3.7.5+dfsg1-4 >> architecture: amd64 >> unsat-dependency: libopenmpi-dev:amd64 (< 2.0.2~git.20161226) >> depchains: >> - >>depchain: >> - >> package: sbuild-build-depends-sundials-dummy >> version: 0.invalid.0 >> architecture: amd64 >> depends: libpetsc3.7-dev:amd64 >> - >> package: libpetsc3.7-dev >> version: 3.7.5+dfsg1-4 >> architecture: amd64 >> depends: libpetsc3.7.5-dev:amd64 >> >> >> Which is true, because libpetsc3.7.5-dev has >> >> Depends: libopenmpi-dev (>= 2.0.2~git.20161225), >> libopenmpi-dev (<< 2.0.2~git.20161226) >> >> But the only available libopenmpi-dev is 2.0.2-2 > > A binNMU seems to be sufficient here. > > nmu petsc_3.7.5+dfsg1-4 . ANY . unstable . -m "Rebuild with openmpi 2.0.2" We should probably wait until petsc migrates, and then maybe not do this unless we unblock openmpi. Emilio
Bug#854905: nmu: petsc_3.7.5+dfsg1-4
user release.debian@packages.debian.org usertag 854905 nmu thanks On Sun, Feb 12, 2017 at 08:16:06PM +0100, Emilio Pozuelo Monfort wrote: > user release.debian@packages.debian.org > usertag 854905 nmu > thanks > > On 11/02/17 21:41, Adrian Bunk wrote: > > Control: reassign -1 release.debian.org > > Control: retitle -1 nmu: petsc_3.7.5+dfsg1-4 > > Control: severity -1 normal > > Control: tags -1 - sid > > Control: affects -1 libpetsc3.7.5-dev > > You forgot to usertag it... You forgot to Cc control@bugs.d.o :) Regards, James
Bug#854905: nmu: petsc_3.7.5+dfsg1-4
On 14/02/17 23:36, James Clarke wrote: > You forgot to Cc control@bugs.d.o :) I Bcc'ed it :P Emilio
Bug#854905: nmu: petsc_3.7.5+dfsg1-4
On Sun, 12 Feb 2017 20:16:06 +0100 Emilio Pozuelo Monfort wrote: > > > > A binNMU seems to be sufficient here. > > > > nmu petsc_3.7.5+dfsg1-4 . ANY . unstable . -m "Rebuild with openmpi 2.0.2" > > We should probably wait until petsc migrates, and then maybe not do this unless > we unblock openmpi. > petsc 3.7.5+dfsg1-4 has now hit testing. Unblock for openmpi has been requested in #855217. Please binNMU petsc at your convenience. Drew p.s. I raised the question of the tight openmpi dependency with PETSc authors. They said the tight dependency was put in place so that PETSc would have consistent MPI handling. Otherwise configure, build and runtime of apps using PETSc could all be performed with different mpi versions, which can lead to error especially if the runtime version is older. Moreover mpich does not provide a simple major.minor.subminor decomposition. So the tight dependency on openmpi subminor version was put in place to match the handling used for the single version number available for mpich. But in principle later subminor versions of openmpi should work fine, and so the PETSc authors are putting in place a more relaxed test allowing use with later subminor versions (for the same major.minor). Once this patch is in place (in future versions of petsc), we won't be affected by the same version incompatibility reported here.
Bug#854905: nmu: petsc_3.7.5+dfsg1-4
On Thu, 16 Feb 2017 15:12:23 +0800 Drew Parsons wrote: > > petsc 3.7.5+dfsg1-4 has now hit testing. > Hi again. There's some discussion on what to do about openmpi in testing (#855217). Regardless of how we resolve that, can I affirmatively request we trigger the binNMU for petsc? This is for the sake of unstable users, who need the rebuild done anyway. And will make the transition faster if we do decide to let openmpi 2.0.2 into testing. nmu petsc_3.7.5+dfsg1-4 . ANY . unstable . -m "Rebuild with openmpi 2.0.2. Closes: #854905." Thanks, Drew