Add function framework for job step node selection.
Remove reference to defunction job specification of procs_per_task.

Add function framework for job step node selection.

to cancel a job, issue the job_step_cancel RPC with a step_id
of NO_VAL.

and RPCs. This will all be managed directly by srun/slurmd.

each job table record.

Added prolog and epilog to job diagrams

Add cpucount by node for resource allocation

 of the required node specifications

changed unpack_strarray to null terminate the array of strings

removed gid from messages this will come from getpwuid call
added a needed include to scancel to compile
removed old test from job_mgr_test that wouldn't compile due to changes in the api

Added some new error codes for scancel API

Added pack/unpack functions and additional error codes.
Scontrol now parses and sends job update requests.
Update_job function now working with job_desc structure.

Added pack/unpack functions and additional error codes.
Scontrol now parses and sends job update requests.

Added pack/unpack functions and additional error codes.

must fork before dup in task initiation, cleaned it up and made it more straight forward.
test case runs, std in/out/err still need some work but they do connect back to a perl srun_emu.pl script

partition names exactly).

signal restoration now occurs in socket_impl

Start using slurm provided errno text from scontrol.
Better deal with partition update having invalid node names.

Start using slurm provided errno text from scontrol.