The output of "scontrol show node" changed with respect to GPU bound
sockets. There was no error reported by test20.4, but the output was
wrong without this commit.

# Conflicts:
#	NEWS

support.

Bug 4871

is removed from the system.

Bug 5665

burst buffer.

Bug 5665

block distributions are failing with some configurations
ompi 4.0.x (head as of 10 sept 2018)
slurm 18.08
PMIx 3.0
The test with a block distribution is hanging. This will let the
remaining 3 MPI steps in test1.18 run to completion while we wait
for a fix from ompi.
Note that Alex and Danny are not seeing this failure, but they
might not be working off the latest ompi code.

test39.12 was failing with some configurations by double counting
available cores. Moving some code fixes that problem and speeds
up the logic as well.

Specifically log Cores/CPUs related errors using proper name,
not logging bad CPUs information as bad Core information.

Log SLURMD_HOSTNAME on remote host rather than "hostname", which
is more meaningful on a multi-slurmd configuration

or a database admin.

Bug 5679

Make tests work properly if GPU count exceeds CPU count on a node

The GRES information now contains socket binding information,
necessitating this change to the parsing function

Bug 5359

Bug 5679

regression from combination of 037bb3cb and 5a6a26f6

Bug 4479

node_registration_status_msg.

Bug 5672

Bug 5672

Coverity CID 164845

"Argument with 'nonnull' attribute passed null"

Bug 5529

Seconded by Moe.

If the Cores count in gres.conf is bad, slurmstepd was generating
a fatal error at time of step launch. Rather than quitting then,
this change logs the error and ignores the Cores value.

Coverity CID 187778

Coverity CID 187927

Coverity CID 187926

This reverts some logic added in commit 01336077 for debugging

"no_send_gids to go back to the original behavior.

Bug 5620

LaunchParameters=send_gids.

Bug 5620

sizeof(char) is defined to be 1, and Slurm is already heavily
reliant on that throughout. Stop pretending Slurm would work
on a hypothetical system in which this was not true.

(C99 Standard - 6.5.3.4.3.)

Strip off unnecessary casts while here as well.

Replace with xstrfmtcat(), and resolve this overrun warning.

Cannot return NULL, it will abort() before then.
Remove the unnecessary cast while here.

Add configuration parameter "GpuFreqDef" to control a job's default GPU
    frequency.
Also Coverity CID 187908

Bug 5659