limitation in terms of scalability, I did some tests with cluster sizes
varying from 64 to 64k nodes. I saw the same kind of performance issues
that you saw with large sizes (more than a few thousands nodes).

Looking at the code, I noticed that the hash table of the node records
was not built before the construction of the topology information in
slurmctld ! Making sure that the hash table is present really reduces
the built time.

This patch ensures that and enable to get sub-second built time
for topology information in scenarios that were used to require a
minute before.

I think that using this patch, the load time of slurm confs is less
problematic.

implemented.

using switch topology information.

Job requeue test assumed the job would not be started immediately
after requeue, but that could happend and result in a test
reporting FAILURE, when the job really was requeued.

Conflicts:
	testsuite/expect/test21.24

(commit 4cd63575)
with sacctmgr load when Parent has "'" around it

page.

in theory that is a possible combo returned.

configured then srun will use its dynamic ports only from the configured
range.

Also make wait_for_job in the global tests file more robust by
checking for a zero job ID

Conflicts:
	META
	NEWS

Conflicts:
	testsuite/expect/test17.34

special characters in their names like ':'

It seems to have been broken for some time due to logic blocking
SIGWINCH being removed from srun, which could result in the
signal being sent to any thread, not the thread designed to handle
window resizing.

There was an error reported by slurmstepd trying to read the window
size from srun when srun was exiting. This eliminates that spurious
error message.

Added support for job email triggers: TIME_LIMIT, TIME_LIMIT_90 (reached
90% of time limit), TIME_LIMIT_80 (reached 80% of time limit). Applies to
salloc, sbatch and srun commands.