Orion Poplawski <or...@nwra.com> writes: > With the help of the very useful libSegFault I was able to generate a > backtrace in COPR with: > > export LD_PRELOAD=/usr/lib64/libSegFault.so > export SEGFAULT_SIGNALS=ill > > which pointed to the libpsm2 library. I've filed > https://bugzilla.redhat.com/show_bug.cgi?id=1659852
Well done. (Arguably the different HPC transports should be optional, as Infinipath was.) > Any Koji/COPR debug tips pages out there that could benefit from > mentioning libSegFault/catchsegv? I can't remember -- does that depend on gdb being available? For what it's worth, openmpi can be configured to provide backtraces directly. I've always turned it on, and forget it's not the default: opal_abort_print_stack=1, at least in older versions. [I don't know if it's been fixed recently, but last I knew, BFD-based backtraces had an off-by-one error in the line number, which you had to take care of.] I've seen a different problem with MPI-based packages that's started fairly recently, with tests just hanging non-reproducibly in koji and copr (most often on arm and i686), so I've turned them off in two or three packages pending time to try to get a stack trace. Any suggestions on such problems gratefully received if they're not just library-specific deadlocks. This stuff is rather painful to do in batch/asynchronously, when all the output you can get has to be in build.log. _______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://getfedora.org/code-of-conduct.html List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org
