Logs for jobs which are explicitly requeued will say so rather than saying
that a node in their allocation failed.

This is a minor change to commit 4b8cdd4c
from yesterday and is needed to work with launch/poe (which was
broken by the commit).

--export=NONE is not supported, but --export=name=value is supported.

Conflicts:
	src/slurmctld/reservation.c

a reservation.

which they have access to (rather then preventing them from seeing ANY
reservation).  Backport from 14.11 commit 77c2bd25.

arbitrary layouts (test1.59).

option (since it isn't).

If job or step request a gres count of zero, do not test that the
gres name is valid (e.g. "--gres=foo:0" is fine even if "goo" is
not a valid gres).

Replace putenv() with setenv() to eliminate last CLANG error reported.
setenv makes its own copies of the name/value.

Do not cause it to be rebooted (powered up).

This permits a sys admin to power down a node that should already
be powered down, but avoids setting the NO_RESPOND bit in the
node state. Doing so under some conditions prevented the node from
being scheduled. The downside is that the node could possibly be
allocated when it really isn't ready for use.

When a node's state is set to power_down, then execute SuspendProgram
even if previously executed for that node.

which protects against race conditions with the reservations.

Conflicts:
	src/common/assoc_mgr.c
	src/slurmctld/job_mgr.c

Fix logic determining when job configuration (i.e. running node power up
logic) is complete. (Will look at better solution for v14.11).

When a node's state is set to power_up, then execute ResumeProgram even if
previously executed for that node.

different times when reservations are using the associations that are
being deleted.

Conflicts:
	doc/html/faq.shtml