Bug#886421: python-escript: FTBFS on m68k: hwloc_set_cpubind returned "Error" for bitmap "0"

[Laurent Vivier]

Le 05/01/2018 à 20:35, Aaron M. Ucko a écrit :
> Source: python-escript
> Version: 5.1-5
> Severity: important
> Tags: upstream
> Justification: fails to build from source (but built successfully in the past)
> User: debian-68k at lists.debian.org
> Usertags: m68k
> 
> Builds of python-escript for m68k (admittedly not a release
> architecture) have been failing lately per the below excerpts from
> https://buildd.debian.org/status/fetch.php?pkg=python-escript&arch=m68k&ver=5.1-5&stamp=1514854858&raw=0.
> FWIW, automatic builds for this architecture run with nocheck in
> DEB_BUILD_OPTIONS because there's little overall incremental value to
> running tests on such a slow architecture.  However, it's perhaps just
> as well that that setting had no effect here, since this problem
> doesn't affect any other Debian architectures and may be worth
> investigating.
> 
> Could you please take a look?
> 
> Thanks!
> 
> --------------------------------------------------------------------------
> 
> /<<PKGBUILDDIR>>/debian/stage2M/bin/run-escript /<<PKGBUILDDIR>>/debian/tmp2M/scripts/release_sanity.py
> --------------------------------------------------------------------------
> WARNING: a request was made to bind a process. While the system
> supports binding the process itself, at least one node does NOT
> support binding memory to the process location.
> 
>   Node:  vs90
> 
> Open MPI uses the "hwloc" library to perform process and memory
> binding. This error message means that hwloc has indicated that
> processor binding support is not available on this machine.
> [...]
> This is a warning only; your job will continue, though performance may
> be degraded.
> --------------------------------------------------------------------------
> --------------------------------------------------------------------------
> Open MPI tried to bind a new process, but something went wrong.  The
> process was killed without launching the target application.  Your job
> will now abort.
> 
>   Local host:        vs90
>   Application name:  /<<PKGBUILDDIR>>/debian/stage2M/lib/pythonMPI
>   Error message:     hwloc_set_cpubind returned "Error" for bitmap "0"
>   Location:          rtc_hwloc.c:190
> --------------------------------------------------------------------------
> scons: *** [dummy] Error 213
> scons: building terminated because of errors.
> 
> *** Config Summary (see config.log and <prefix>/lib/buildvars for details) ***
> Escript revision 6608
>   Install prefix:  /<<PKGBUILDDIR>>/debian/stage2M
>           Python:  /usr/bin/python (Version 2.7.14)
>            boost:  /usr (Version 1.62.0)
>            numpy:  YES (with headers)
>              MPI:  OPENMPI (Version 2.1.1)
>   Solver library:  paso
>    Direct solver:  NONE
>          domains:  dudley, finley, ripley, speckley
>           netcdf:  YES (3)
>            weipa:  YES
>           openmp:  YES
> 
>   DISABLED features: boomeramg cppunit cuda debug gdal gmsh gzip lapack mkl papi parmetis pyproj scipy silo sympy trilinos umfpack visit
>   Treating warnings as errors
> 
> WARNING: Cannot import scipy. NetCDF sources will not be available for inversions.
> WARNING: Cannot import pyproj. Inversions may not work.
> WARNING: Cannot import gdal. Inversions will not honour WKT coordinate system information.
> WARNING: Cannot import sympy. Symbolic toolbox and nonlinear PDEs will not be available.
> WARNING: matplotlib not found, will skip some unit tests
> WARNING: gmsh not available. Skipping tests usersguide/trapezoid.py usersguide/quad.py usersguide/brick.py usersguide/refine.py cookbook/example04a.py cookbook/example04b.py cookbook/example05a.py cookbook/example05b.py cookbook/example05c.py cookbook/example06.py cookbook/example08c.py cookbook/example09m.py cookbook/example09a.py cookbook/example10m.py inversion/dc_forward.py!
> 
> ERROR: build stopped due to errors
> 
> debian/rules:57: recipe for target 'override_dh_auto_build' failed
> make[1]: *** [override_dh_auto_build] Error 2
> make[1]: Leaving directory '/<<PKGBUILDDIR>>'
> debian/rules:30: recipe for target 'build-arch' failed
> make: *** [build-arch] Error 2
> dpkg-buildpackage: error: debian/rules build-arch subprocess returned exit status 2
> 

It seems this buildd is running a m68k chroot on AMD64 using qemu-m68k.

Samuel Thibault has posted patches to qemu-devel mailing-list to fix CPU
affinity related functions:

  [PATCH] linux-user: Add getcpu() support
  https://www.mail-archive.com/qemu-devel@nongnu.org/msg502788.html

  [PATCH] linux-user: Fix sched_get/setaffinity conversion
  https://lists.gnu.org/archive/html/qemu-devel/2017-12/msg05242.html

Perhaps it could fix this problem.

Thanks,
Laurent