then send tasks SIGTERM. smgr handles SIGXCPU by printing a message
   that job exceeded timelimit.

 o src/common/slurm_cred.c : assert non-null credential on get_signature
   operation

setting flag).

This would occur for example if we killed a job that never started any
steps on a node.
slurmd/req.c responds with INVALID_JOBID when an attempt is made to
effect a job that is not found.

This would occur for example if we killed a job that never started any
steps on a node.

slurmctld expects no response.

improved performance in production.

 o abolish references to libcred

   slurm_cred_t job credential.

Major enhancements to jsspp.tex.

and project.bib. Continued development of jsspp.tex.

   needed there)

 o src/common/slurm_cred.[ch] : add function slurm_cred_insert_jobid()
   for updating the internal job state on the verifier side without
   verifying a credential. (Used when running batch jobs)

slurm_errno.h: added new error code for invalid job credential

   the old slurm_job_credential_t, which was one cred per job, to a
   one cred per *job step* scheme. Creds are now limited lifetime and
   cannot be replayed (cred is invalidated once used).

   removed old implementation signature/credential_utils.[ch] and
   replaced with slurm_cred.[ch].

calls _shm_reopen() calls _shm_lock_and_initialize() etc.

if the requested node count or time limit are outside of the partition's
configured limits.

of the partition limits.

A job time limit of NO_VAL is valid while the job is queued, set its
value to the partition's limit only when the job is allocated resources.
Srun reports when a queued job is finally allocated resources.

A job time limit of NO_VAL is valid while the job is queued, set its
value to the partition's limit only when the job is allocated resources.