Apply BatchStartTimeout configuration to task launch and avoid aborting
srun commands due to long running Prolog scripts.
bug 978

Provide better description of the slurm.conf configuration parameter
BatchStartTimeout.
bug 979

Disable a partition test if JobSubmitPlugins=all_partitions

When nodes scheduled for reboot, set state to DOWN rather than FUTURE so
they are still visible to sinfo. State set to IDLE after reboot completes.
bug 1007

Srun executable names beginning with "." will be resolved based upon the
working directory and path on the compute node rather than the submit node.

suggest parentheses around assignment used as truth value.

The code tries to load libslurm.so even if precedent dlopen calls
succeeded. The code is structured so that we have to "return;" as
soon as a dlopen succeeds.

This corrects logic introduced yesterday in commit
6f89dc9d which introduced a double
free of step records, at least on job requeue.
bug 1012

Describe restrictions on specific job and step record purging functions
with respect to "cleaning" flag used for Node Health Check on Cray systems.

Always call select_g_step_finish() when terminating a job step,
even if the job is also being terminated. This is needed for Cray
systems.
bug 1012

When a job is requeued, call deallocate_nodes() with a job state
if COMPLETING. Previously it was called with a state of JOB_REQUEUE,
which could be problematic for step complete function calls (which
I am working on fixing now).

Remove some duplicate code. No change in functionality.

When a job record is purged, simply purge the step list rather than possibly invoking a node health check on Cray systems.

No checking or other operations are performed on this list, just a purge.

If an attempt is made to submit a job explicitly using a job ID that already exists, then do not try to purge and re-use it, but return an error. The slow clean-up of job steps on Cray systems due to node health check makes me wary of preserving the existing code. Returning an error seems a safer option.

Call delete_step_records() before clearing the job's JOB_COMPLETING
state flag. This would make a difference in the case of jobs automatically
requeued based upon their exit code, but probably not in other cases.
Also in the select plugins, check not only for a job state of JOB_COMPLETING,
but also FINISHED states. In either case, we are not in a position to
gracefully clean up the step.

Fix race condition in CPU frequency set with job preemption.
When the preemptor job completed, it would notify the srun, which
would notify the slurmctld, which could resume a preempted job.
That preempted job could reset the CPU frequency before the
preemptor. This change has the slurmstepd resetting a job's
CPU frequency prior to notifying srun of completion, which
eliminates the race condition.
bug 1011

This corrects logic added in commit
738913fa
for BGQ systems only

"job_comp/mysql" setting an incorrect default database.

This helps reduce a race condition reported in test1.64. Log termination
message right away rather than trying to terminate the job and then
log the event before the srun program exits

Previous logic did not work properly to allocate specific GRES
model types to job steps from the matching job model types.

Broken in the course of adding support for GRES type field
bug 633

Conflicts:
	src/common/gres.c

Add logic to confirm that gres bitmaps exist before trying to use
them. We haven't seen this failure mode in v14.03, but in v14.11.

Conflicts:
	src/slurmctld/job_mgr.c

database index for the array elements avoiding duplicate database values.

Issue single RPC to update a job array and return separate
error codes as needed.