One of our testers created an illegal topology.conf file.

He has a config you probably wouldn't see in production, but can see in
testing when you are sometimes given a collection of miscellaneous
resources.

          |-- nodes
switch1 --|
          |-- switch2 -- nodes

He tried the topology.conf file below. Switch s1 is defined twice. Slurm
accepted this config, but wouldn't allocate nodes from both switches to
one job.

SwitchName=s1 Nodes=xna[14-26]
SwitchName=s2 Nodes=xna[41-43]
SwitchName=s1 Switches=s2

I believe slurm shouldn't allow the second definition of switch s1. The
attached patch checks for duplicate switch names.
Patch from Rod Schultz, Bull.

Bug in selec/cons_res using tasks-per-node and cpus-per-task

tasks_per_node and cpus_per_task not getting proper allocation and
failing in srun with "Requested node configuration is not	available"

window and not running in ncurses.

through that extra logic when starting up from smap.

into the block allocator system of bluegene.

This reverts commit e245c7bd.

This reverts commit bb8477b3.

This reverts commit 350ef5dc.

Modify salloc, sbatch, and srun node specification parsing to accept
a number followed by a suffix of "M" to multiply the numeric value
by 1,048,576 (1024 x 1024).

Large-scale update of bluegene.conf man page to incorporate BlueGene/Q
information plus some general updates.

Major update to BlueGene web page specifically to include BlueGene/Q information.

The squeue command was printing bad I/O nodes or c-nodes information
for pending jobs. There was on midplane name, just the I/O or c-node
specification.

test8.22   Bluegene/Q only: Stress test of running many job step allocations
           within the job's allocation
test8.23   Bluegene/Q only: Test that multple jobs allocations are properly
           packed within a midplane

These changes more thouroughly test Bluegene/Q job step placement
algorithms and validate several recent bug fixes in the SLURM code.

On Bluegene/Q systems, the job step allocation needs to be larger
than requested in some cases due to the job allocation geometry
(e.g. a 5 cnode allocation needs to be scaled up to at least
6 cnodes). This enhancement fixes that logic if multiple size
increases are needed.

next parallel step is ran on a sub block, SLURM won't over
subscribe cnodes.

This reverts commit c5d63854 from 8/11/2011. The memory
copy is not a leak, but is required to avoid memory corruption.

make sure that a job has a step_list before creating an interator for it