(gdb) where
#0 0x0000155553245775 in __strncmp_avx2 () from /lib64/libc.so.6
#1 0x000015554043b392 in dohash () from /usr/lib64/pmix/mca_gds_hash.so
#2 0x000015554043c2bd in hash_fetch () from /usr/lib64/pmix/mca_gds_hash.so
#3 0x000015554064b286 in _store_job_info (ds_ctx=ds_ctx@entry=0x7484f0,
ns_map=ns_map@entry=0x7486e8, proc=proc@entry=0x15553f6115f0)
at dstore_base.c:2720
#4 0x0000155540654476 in pmix_common_dstor_register_job_info (
ds_ctx=0x7484f0, pr=0x1555380169e0, reply=<optimized out>)
at dstore_base.c:2892
#5 0x00001555438ffb31 in server_switchyard () from /lib64/libpmix.so.2
#6 0x0000155543900c26 in pmix_server_message_handler ()
from /lib64/libpmix.so.2
#7 0x00001555439834b5 in pmix_ptl_base_process_msg () from /lib64/libpmix.so.2
#8 0x000015554364cd85 in event_process_active_single_queue ()
from /lib64/libevent-2.1.so.6
#9 0x000015554364d787 in event_base_loop () from /lib64/libevent-2.1.so.6
#10 0x000015554391e6a6 in progress_engine () from /lib64/libpmix.so.2
#11 0x000015555465f1ca in start_thread () from /lib64/libpthread.so.0
#12 0x00001555531b7e73 in clone () from /lib64/libc.so.6
This is run in a flux alloc -N16 allocation in the "rabbit" queue.
Note that when running something else like /bin/true under -opmi=pmix there is no segfault.
Problem: openmpi hello world causes the flux shell to segfault when run with
-opmi=pmix
on el capEven with one task.
Launch script:
Software versions are
And the failure:
@grondo got this stack trace
This is run in a
flux alloc -N16
allocation in the "rabbit" queue.Note that when running something else like /bin/true under -opmi=pmix there is no segfault.