infrastructure is operational.

   if available at install time.

with FastSchedule==0 configuration.

   in an int instead of a short (truncation was causing strange results,
   especially when ORed with NODE_NO_RESPOND flag).
 o No longer bother processing the '*' at end of node state strings.
   sinfo -t now shows both responding and non-responding nodes by
   default --responding or --dead can be used to choose one or the other.

switch (was setting DRAINED nodes in state DRAINING).

option. Use SHOW_ALL as flag.

responding in a timely fashion) then send a registration RPC
to slurmctld.

are allocated. Replace polling to test for allocation with processing of
allocation response RPC from slurmctld at allocation time.

   to specified user.
 o New srun option --no-shell
 o Changed uid/gid_from_name () to uid/gid_from_string () and allow
   either user/group name or numerical user/group ID.
 o updated srun(1) documentation with new --no-shell option and
   changed semantics for --uid.

registration (in particular after changing state from UNKNOWN to DRAINING,
an inactive node was not registering to reset its state to DRAINED)

   of slurmd process.

    calls, improves performance and eliminates realloc failure on
    RH EL3 under extremely heavy workload apparently due to memory
    fragmentation

Requires setting counts in updated batch job launch request message.

    corresponding file descriptor for readability. Otherwise, srun
    and slurmd will spin when stdin file descriptors are full because
    pending data will always go unread. (gnats:434)

specifying an infinite time limit.

are specific to a particular plugin type.

commands modified. Info filtering added to slurmctld.

loadable as shared aliases. Give slurm functions a prefix of "slurm_" to
insure that no user functions with the same name is used instead.
[For some reason a this same commit a minute ago did not seem to take effect.]

right away and treat failure as a fatal error.