Source: openmpi Version: 4.1.2~rc1-2 Severity: serious Justification: FTBFS
The new version of openmpi seems to be segfaulting in mca_io_romio321.so The segfault is found in nearly all arches in mpi4py tests, https://buildd.debian.org/status/package.php?p=mpi4py (it causes mpi4py to FTBFS) The crash is bad enough that most arches are not able to provide a backtrace, although some arches reference libpthread.so.0 (which could suggest instability from the libc6 2.32 transition) ppc64el and ia64 however provide a full backtrace, both pointing at mca_io_romio321.so (MPIOI_File_iwrite_all) >From ppc64el https://buildd.debian.org/status/fetch.php?pkg=mpi4py&arch=ppc64el&ver=3.1.1-8&stamp=1632692000&raw=0 testIReadIWrite (test_io.TestIOSelf) ... ok testIReadIWriteAll (test_io.TestIOSelf) ... [ppc64el-osuosl-01:1381766] *** Process received signal *** [ppc64el-osuosl-01:1381766] Signal: Segmentation fault (11) [ppc64el-osuosl-01:1381766] Signal code: Address not mapped (1) [ppc64el-osuosl-01:1381766] Failing at address: (nil) [ppc64el-osuosl-01:1381766] [ 0] linux-vdso64.so.1(__kernel_sigtramp_rt64+0x0)[0x7fff92ec0514] [ppc64el-osuosl-01:1381766] [ 1] [0x0] [ppc64el-osuosl-01:1381766] [ 2] /usr/lib/powerpc64le-linux-gnu/openmpi/lib/openmpi3/mca_io_romio321.so(mca_io_romio_dist_MPI_File_iwrite_all+0x38)[0x7fff831135e8] [ppc64el-osuosl-01:1381766] [ 3] /usr/lib/powerpc64le-linux-gnu/openmpi/lib/openmpi3/mca_io_romio321.so(mca_io_romio321_file_iwrite_all+0xa8)[0x7fff831100f8] [ppc64el-osuosl-01:1381766] [ 4] /usr/lib/powerpc64le-linux-gnu/libmpi.so.40(PMPI_File_iwrite_all+0x100)[0x7fff9221e890] [ppc64el-osuosl-01:1381766] [ 5] /<<PKGBUILDDIR>>/.pybuild/cpython3_3.9/build/mpi4py/MPI.cpython-39-powerpc64le-linux-gnu.so(+0x12b654)[0x7fff9245b654] [ppc64el-osuosl-01:1381766] [ 6] /usr/bin/python3.9[0x100f9e84] [ppc64el-osuosl-01:1381766] [ 7] /usr/bin/python3.9(_PyEval_EvalFrameDefault+0xa9c)[0x100d2a6c] [ppc64el-osuosl-01:1381766] [ 8] /usr/bin/python3.9(_PyFunction_Vectorcall+0x20c)[0x100f631c]
