-- BLUEGENE - This fixes a race condition with dynamic mode to make it so we copy the booted and job list of blocks before trying to create.

    so the highest job priority becomes TOP_PRIORITY.

with multiple front-end nodes. Use these new fields for Cray
information. Use information in front-end table for step 
credentials and communications.

slurmdb_setup_cluster_dims(void)

    at FirstJobId.

task/affinity configured instead of when NOT configured. Patch from
Martin Perry

06_node-reordering-first-run-test-always.diff

24_Cray-deprecate-HAVE_CRAY.diff and

various files: deprecate HAVE_CRAY

This 
 * replaces HAVE_CRAY (which is an untested configure option) 
 * with HAVE_NATIVE_CRAY which tests for Cray-specific files and
 * deprecates the apbasil location (which has been /usr/bin/apbasil since XT3).

This distinction is used to disable code on systems that do not support Cray
ALPS interaction.

25_Cray-test-for-native-cray-system.diff
Cray: test for native system

This changes the autoconf logic so that
 * HAVE_CRAY is defined to enable basic Cray XT/XE support;
 * HAVE_NATIVE_CRAY is defined on native Cray XT/XE systems only.

The distinction is necessary due to the user of multi-cluster systems, which
may need to load the select/cray plugin, but which are not native Cray systems
(such as e.g. the slurmdbd host).

In particular, when building with support for remote execution of Cray jobs,
it would be wrong to set FRONT_END mode locally on the clients. Hence only
set this when building on a native Cray system.

select/cray: disable node state alteration via slurmd

This disables unconditionally setting up DOWN nodes on Cray systems, since this
interferes with the periodic node-state query via Basil/ALPS inventory requests.

Cray: SlurmUser needs to be 'root'

This disallows a SlurmUser other than root. The reason is that slurmctld forks
processes talking to ALPS/BASIL. Doing this requires root permissions.

The following rules apply:
 * the RESERVE method requires that 
   - either the effective UID of the calling process resolves to that of the
     'user_name' attribute of the ReserveParamArray or
   - the effective UID is 0 (super-user), i.e.
   - only root can make reservations for other users;
 * the RELEASE method requires root permissions in all cases (slurmctld uses
   this method on each scheduling cycle to clean up vestigial ALPS reservations).

To avoid entanglements with sudo, this patch simply throws a fatal() message
if the SlurmUser is not root. (An alternative would be to issue an error 
message, but to continue processing.)     

Based upon patch from Gerrit
21_job-info-debuglevel.diff