Induced by bf_continue option and deleting a partition.

to reflect only the latest supported format.

if bf_continue option is configured and slurm is reconfigured during
one of the sleep cycles, then the backfill scheduler will reference
an invalid partition pointer.

Previous logic would place more tasks on each node than specified
by --ntasks-per-node, using fewer nodes than desired. This only
happens with exclusive node allocations (e.g. in partition
configuration Shared=Exclusive).

EpilogSlurmctld pthread is passed required arguments rather than a pointer
to the job record, which under some conditions could be purged and result
in an invalid memory reference.

segfault.

is not parsed correctly.

gres/gpu and gres/mic - Do not treat the existence of an empty gres.conf
file as a fatal error. There may be no gres devices on that node so we
do not require the file. Assume gres counts of zero if no file.

bug 436

job CPU count not loaded correctly
Partition time limit format wrong (minutes rather than hhmmss format).

isn't doing the launching.

updated the srun man page.

sched/backfill - Prevent possible memory corruption due to use of
bf_continue option and long running scheduling cycle (pending jobs could
have been cancelled and purged).

Change max message length from 100MB to 1GB before generating "Insane
message length" error.

can only be a TORUS (1).

This specifies the maximum depth the backfill scheduler should go
in any single partition.

srun is pending.  The step creation would use the old values and possibly
hang srun since the step wouldn't be able to be created in the modified
allocation.

into the future.

bug 428

for setdebugflags command, avoid parsing "-flagname" as
an scontrol command line option.

for CR_Socket.

Add support for "on" and "before*" options

bugfixes and the new item extraction feature.

Set PBS_ACCOUNT, PBS_ENVIRONMENT, and PBS_QUEUE only for batch
jobs and only if the user submission sets the account and partition.