There were a few minor problems and enhancements to the original code
* There were some missing header files includes
* derived_exitcode needs to be printed as %lu rather than %hu.
* the fsize variable was set and never read
* I added some "static" definitions on functions
* Some compilers report warnings for variables definitions NOT at the start of a block (I needed to use xmalloc for the pop_marks variable)
* Some compilers report warnings for the xrealloc return value not used
* In the log functions you are testing for empty strings, so testing if the first character is '\0' is faster than using strlen()
* The xstrfmtcat is faster than using a temporary value and copying data
* slurmdb_association_rec_t was changed to slurmdb_assoc_rec_t

Also see:
https://github.com/asanchez1987/jobcomp-elasticsearch

Conflicts:
	doc/man/man5/slurm.conf.5

job.  This is already handled on the stepd when the script finishes.

often an inventory request is handled.

Conflicts:
	src/common/layouts_mgr.c

free afterwards would not have zeroed out memory on the variables that
didn't get unpacked.

This fixes the setting of PMI_CRAY_NO_SMP_ENV for v15.08.
The previous commit (25b03b8c)
worked fine with v14.11, but changes were required to function
with v15.08 and commit 54e53857,
which gives user control over packing on a job-by-job basis in
order to over-ride system with packing configuration.
See bugs 819 and 1585

sbatch to stop parsing script for "#SBATCH" directives after first command.
It keeps parsing so long as lines contain only white space or comments
(first non-white space character is '#').

bug 1608

This initializes some variables used with the srun --no-alloc option
that can result in test1.26 failing sporatically.

Bug 1384

Add SchedulerParameters option of "sched_max_job_start="  to limit the
number of jobs that can be started in any single execution of the main
scheduling logic.

Bug 1601