Bug 4105

Add mapping from task to originating pack index to allow the stepd to identify which tasks are part of which packs

Bug 4105

Not totally working, but not ever used.

Bug 4105

Bug 4105

Bug 4105

Bug 4105

The task/cgroup plugin sets the UID patch based upon the pack_jobid
   (if a pack job). This is needed for Cray systems.

Bug 4105

Without this change, a step that did not use heterogeneous job
  component zero would not run

For a pack step launch with one job component, always set pack_jobid
   field and all of the other "pack_" fields. This will be needed
   for Cray support.

# Conflicts:
#	testsuite/expect/README

Previously scontrol update PartitionName=* MaxNodes=* would
print an error and exit of the MaxNodes value was invalid (e.g.
not a number). Now it prints an error and can continue, say
processing multiple requests in a script

Coverity CID 197669

Keep log message string on single line

Bug 6873

When specifying -j and not -s then non-eligible jobs will be shown
by sacct.
Time windows can also be used with -S and -E.
If --state is also used then non-eligible jobs won't be shown,
because non-eligible are not actually PD.

Bug 6873

# Conflicts:
#	NEWS

i.e. Centos7

Bug 6805

Bug 6898

Continuation of commit cc153e03

Cancelled jobs before start have time_start=0 in the DB, but their
time_end!=0 (the cancel time).
Query fixed to handle properly these cases for -s PD.

Bug 6894

Bug 6878

Test was always leaving around a file named "slurm.conf.orig".
Now delete the file on test success.

bug 6878

Previous logic would wait up to 5 seconds. New logic will extend
the time limit up to half of MessageTimeout (minimum value still
5 seconds).

Previous logic lacked locks on global data structure.
Also initialized some variables to NULL before use.

Previous logic placement would result in sending NULL for
topology information. Moved buffer build later in the logic.

bug 5667

# Conflicts:
#	src/slurmctld/job_mgr.c
#	src/slurmctld/proc_req.c
#	src/slurmd/slurmd/req.c
#	testsuite/expect/test21.39

We will store the pid of running scripts on a list in order to kill them
when slurmctld is shutting down. This will make the thread running the script
to end, so we will join this thread too, but only during a specified timeout,
because if the pid forked by the caller thread is stuck in IO, slurmctld will
be stuck too.

We also need to diferentiate when a sigkill is send by a human or OOM, from
the one send by slurmctld when killing it, so the one way wil be to set
ctld_script_rec_t jobid to -1 when we are killing the process.

Work to do includes adding this logic to the burst_buffer functions which use
the run_command() function.

Work to do also may need to add locks for accessing ctld_script_rec_t records.

Bug 5913

Now that the pid is no longer an argument, the name doesn't make sense.
Change to container_g_join() to reflect the underlying functionality.

Bug 6567.

Removal of this argument forces the use of this function call only for
adding the calling process - and not any other process - to the job
container. This is needed for an in-development job container
plugin that adds jobs to namespaces. As it uses setns() system call -
which can only work on the calling process - the plugin thus relies
on being called from the calling process.

Bug 6567.

Every once in a while I want to remove an element from a list but not
delete it.  This function does that.