On cray systems with step NHC, the step launches are delayed and
  produce a pair of messages (below) that caused the test to fail:
  srun: Job step creation temporarily disabled, retrying
  srun: Job step created

The v17.02 updates appear to have only gone into master rather than
  the slurm-17.02 branch

CRAY systems only: TaskPlugins must list task/cgroup before task/cray in
    order for the cgroup files to be created before task/cray runs. Without
    this change, the task/cray plugin frequently produces errors about the
    "mems" file being missing. The errors don't seem consistent, so this
    probably involves a race condition. Note that NERSC uses this order
    today and I changed read_config.c to produce a fatal error if the order
    is reversed.

Failure to do so results in a bunch of task/cray errors about
  not finding the cgroup set up.

Avoid a test failing of all nodes in a partition are not usable
  (down, drained, reserved, or otherwise unusable).

Disable test if underlying select/linear use

Modify qsub test to explicitly create and destroy the error files
  to avoid leaving around a bunch of error files (even if they are
  normally empty).

Provide better insight into the failure

Turns out this never worked, ever.  What used to happen is if the protocol_version that was
read in didn't match the rpc_version given to unpack things was just 0.  What this does
now is set the rpc_version to what was stored making it all good.

Most of these tests could be modified to work with ALPS, but since
  ALPS support is going way soon, just disable the tests as appropriate.

correctly.

Revert logic originally added for bug 3166.
  Revisit as time permits.
bug 3166

job_queue_rec wasn't being free'd before continues and breaks.

Add new-lines in some log messages to better reflect the actual I/O
  to apbasil.

unbuffered stdout.  After that commit 4ce0be75 was added which appeared to bandaid the
fallout making the only place using the unbuffered option was introduced in the former commit and
only really have unbuffered output on a PTY.  The latter commit would potentially send back
unfinished output without a newline causing jumbled output when expecting buffered output thus
resulting in potentially jumbled/unbuffered output.

I was able to reproduce this regularly with test1.31.

In any case making it so the code is clearer to read is a win.

This reverts commit a6148002.