-l accelerator=true|false	(GPU use)
-l mpiprocs=#		(processors per node)
-l naccelerators=#	(GPU count)
-l select=#		(node count)
-l ncpus=#		(task count)
-v key=value		(environment variable)
-W umask=#		(set job's umask)

If a user job tries to load his environment remotely and the slurmd
is not running as root, the previous logic would generate a fatal
error, now it just logs the error and proceeds with the passed user
environment.

that ended in the OverTimeLimit interval.

The DEFAULT node/frontend/partition values will add to previous
default values and not reinitialize.
bug 397

someone.  They can still add or remove a user of any admin level.

Previously nodes in FUTURE state would have no name field sent,
resulting in siview core dumping

bug 403

bug 403

bug 403

bug 403

bug 403

bug 403

Try to avoid problems when the job record is changed.

Caused by allocating single adapter per node of specific adapter type.

Does optimization of resource allocation for GRES with topology
Similar to changes already added to select/cons_res.
bug 410

Partitial reversal of commit 16143b0f
bug 403

Additional work for 6f50943c
This fixes some more GRES allocation test cases