TmpFS is really intended for user jobs and might not exist on submit
hosts, so /tmp is a better fallback if $TMPDIR is not set.

Bug 11932

It seems that setup.py is extra sensible about PYTHONPATH on newer
Python version (3.8).

Bug 11954

Signed-off-by: Nate Rini <nate@schedmd.com>

The check that the job is currently running has already happened.

A more aggressive refactoring could make this work, but skip for now.

Unused, and would likely need revision to handle HetJob/HetStep operation
properly even if a novel switch plugin desired to operate this way.

No switch plugins support this mode.

No-op, as no switch plugins support this mode of operation.

This is a huge performance win on it's own compared to the original
code since this is NO_VAL on all new step submissions.

This function, called at every step creation, leads to an excessive number
of lock/unlock operations on each list item scanning for a potential
duplicate.

(As was originally intended.)

Note that NULL is a valid argument - SELinux will simply reset the default
context in that case, which is effectively a no-op.

This sets the context up for all tasks launched, as well as the
--task-prolog and --task-epilog.

And job_record_t / struct job_details.

Add RPC plumbing as well.

Add required linking flags into common and slurmstepd builds.

Bug 11995.

This could cause several small issues related to cleaning up the build
directory. This includes the output of automake to regenerate the
Makefile.in from the modified Makefile.am.

Bug 11998

This is a tiny optimization to skip loop over all partitions in case
when it's not used - either priviledged user, or request doesn't have
SHOW_ALL flag.

Bug 11661

Bug 11661

validate_group() may be a heavy operation - repeated nsswitch calls for
users that don't have access to partition. We can avoid that just calling it
once for every partition and just comparing jobs with a prepared array of
allowed partitions.

Bug 11661