within the job step credential.
svn merge -r15520:15530 https://eris.llnl.gov/svn/slurm/branches/multi-core

failures in scancel or squeue commands.

    killed even if the node where srun executes goes down (unless the
    --no-kill option is used, previous termination logic would fail if 
    srun was not responding).

leave it pointed to by the job record (eventually resulting in a second 
free call and invalid memory references (assuming the memory gets re-used).

    specific subsystems. This is just the infrastructure, flag is unused
    right now (work on tomorrow).

    (cons_res_preempt.patch from Chris Holmes, HP).

previously only printed the last argv value.

    the TaskPluginParam configuration parameter rather than slurmd's
    SLURM_ENFORCED_CPU_BIND environment variable.

  (also added output of SacctDefaultCommand from "scontrol show config")

  and Moe Jette. There are still a couple of end-cases that need to be fixed. We also
  need to push the core identification out to the slurmd for task binding. sched/gang
  needs revision to support this too.

  enabled. Only effects sched/backfill, but make sure the thread is 
  finished before purging job data structures