Get rid of any race conditions and call anything that was in
_pre_task_privileged from the parent instead of the child.

NOTE: This should be safe as we don't execute the task until after
_exec_wait_child_wait_for_parent is signaled which happens after all this is
long over.

Bug 3865

Fix regression in commit e5c05549 that would put the stepd pid into the memory cgroup instead of the task's pid.

Beforehand this would put the result of getpid() into the cgroup.  Before
e5c05549 this was done in the child of the fork which would get you
the task's pid, but moving it to run in the parent broke this logic.

What this patch does is adds pid to the input parameters of
task_g_pre_launch_priv making it so we could use the correct pid.

step wasn't always gathered correctly.

Bug 3531

jobid and stepid to be used if needing to kill job because it hadn't started yet.

The old logic would result in test16.4 failing some of the time.
  The failure was caused by the sattach command attaching to a
  job step before the original srun command received a
  RESPONSE_LAUNCH_TASKS message. That messsage  would then be sent
  to the salloc command. Since srun never got the message, it
  would hang. This change does not mark the job step as RUNNING
  until after the original srun gets sent the RESPONSE_LAUNCH_TASKS
  message and sattach requests are blocked until that time.

Identify the function where an error is generated.

Note that this does not protect against all possible problems here.
The setgroups() call in Linux at least is willing to set any gid_t
value except -1 on a group, so calls will not always fail on corrupted
group lists.

Bug 3320.

Add "flag" field to launch_tasks_request_msg. Remove the following fields
    (moved into flags): multi_prog, task_flags, user_managed_io, pty,
    buffered_stdio, and labelio. More flags to be added later.

Previous cap was 2 sec (default TCP timeout) times the node count
  and divided by 1000. A 9000 node job would have the messages
  spread out over 18 seconds. This change caps the spread at
  5 seconds and assumes the normal TCP logic can handle the rest
bug 3044

slurmstepd modified to pre-load all relevant plugins at startup to avoid
    the possibility of modified plugins later resulting in inconsistent API
    or data structures and a failure of slurmstepd.
bug 2334

This patch builds upon commit 0c27392b,
adding calls to the new function so the slurmctld daemon will log
reasons for jobs failing to start.
bug 1828

or failed based on the signal that would always be killing it.

What this does is set the state earlier to match a normal set.

Remove the unneeded _send_pending_exit_msgs.  There is only one task and
we have the message for it, so don't worry about that one.

Most important, wait for the other slurmstepd's to send their message,
otherwise they could be lost on the other end.

This will keep from referencing the task array that might not be set up
correctly in src/common/plugstack.c _spank_handle_init().

sleep is killed.

we have the pids added to the system correctly.

Most likely related to bug 2874

back to the slurmctld.

unsetenv is POSIX1.2001, and should always be available.

Coverity 44992.

Add slurm_cond_* macros to wrap pthread_cond_* functions and log
  error on failures.

make it to the slurmctld when using message aggregation.

POSIX, Linux, FreeBSD all agree setpgrp(0, 0) is equivalent to setpgid(0, 0),
and setpgrp() is deprecated in POSIX.

Continue sorting system includes and un-#ifdef'ing a few additional headers.
(HAVE_STRING, HAVE_UNISTD, HAVE_LIMITS, HAVE_PTHREAD).

Unconditionally define _GNU_SOURCE when used. Should audit use of this further
and possibly define this in config.h directly along with POSIX macros.

Remove last vestige of src/common/malloc.[ch].

This might possibly be related to bug 2334, but it's a long shot.

With slurm-15.08.10, we're seeing occasional segfaults in slurmstepd. The logs point to the following line: slurm-15.08.10/src/slurmd/slurmstepd/mgr.c:2612

On that line, _get_primary_group() is accessing the results of getpwnam_r():
    *gid = pwd0->pw_gid;

If getpwnam_r() cannot find a matching password record, it will set the result (pwd0) to NULL, but still return 0. When the pointer is accessed, it will cause a segfault.

Checking the result variable (pwd0) to determine success should fix the issue.

RHEL6 requires resetting the processes "dumpable" flag after all
seteuid calls complete in order to generate a core file.
bug 2334

make the wait to return data only hit after 500 nodes and configurable
based on the TcpTimeout value.

This may be helpful for timing purposes. Added by Cray request.

This fixes a bug introduced in commit 52fe3de1
in the event the fork() call fails in slurmstepd.

Insure that a job is completely launched before trying to suspend it.
Previous logic would start suspend logic early in the life of the
slurmstepd process, after it's listening socket was open but before
the tasks were launched. This defers the suspend logic until after
all prologs and setup completes and the tasks are launched. This is
important in the case of gang scheduling, in which newly launched
jobs can be immediately suspended.
bug 2494