allocation with respect to network topology.
svn merge -r16770:16795 https://eris.llnl.gov/svn/slurm/branches/topology

It's too difficult to keep in sync, just check the first two lines.

Add web page stub for checkpoint/blcr.

a separate file is written per node or per task.

    re-launch a job step if the step fails with a reserved port conflict.

to busy reserved ports, move on to new set of ports for the next job step
create, also change formatting of the port hostlist to avoid port expressions
like "1[000-003]", we push entries on the list like this "[1000]" to fix.

modifications to allow people to be able to see what task layout a step had and also correct node counts for jobs/steps