Failed because I had JobRequeue=0 in my slurm.conf

bug 1370

as it is the only TRES that can be given to another job while suspended.

Added test to check that there are at least 2 nodes in
 configuration. Indeed this test seems to expect that there are at list 2
 machines in the hostlist returned by sinfo.

see bug 2050

This reverts commit 685eeae0.

bug 1742

Reformat of commit 9350c830
to replace spaces with tabs. No change in logic.

sbatch --ntasks option to take precedence over --ntasks-per-node plus node
    count, as documented. Set SLURM_NTASKS/SLURM_NPROCS environment variables
    accordingly.
bug 2015

Previous logic would report an error and stop the power_save thread.
bug 2042

Not sure when that went away...

If a suspended job is manually resumed and gang scheduling is
configured, but no time slices are available for the job being
resumed, then just resume it without adding it to a time slice.
The jobs previously running on those nodes will be replaced with
new jobs as resources become available and the resumed job will
basically be treated like a stray job.
bug 2031

Avoid reporting more allocated CPUs than exist on a node. This can be
    triggered by resuming a previosly suspended job, resulting in
    oversubscription of CPUs.
bug 2021

bug 2031

Add scancel -f/--full option to signal all steps including batch script and
    all of its child processes.
bug 2031

Bug 1888

out.

Remove unneeded code that commit 8274ea54 fixed.

This code would 0 out all GRES/TRES on a reconfig which isn't what we want.

8274ea54 does the right thing by itself.

bug 2032

bug 2039

bug 2037

backport of commit 4f2e2801 from v16.05

Needed to change a couple of variables from 32- to 64-bit.

Add new burst_buffer.conf parameters: ValidateTimeout and OtherTimeout.
See man page for details.