Ancient versions of OpenMPI and their derivatives (i.e. Cray MPI) are
dependent upon communication ports being assigned to them by Slurm. Such MPI
jobs will experience step launch failure if any component of a
heterogeneous job step is unable to acquire the allocated ports.
Non-heterogeneous job steps will retry step launch using a new set of
communication ports (no change in Slurm behavior).

NOTE: Correcting this would necessitate assigning the same set of ports
to all components of the heterogeneous job (not possible today) plus changes to
srun in order to better synchronize the step startup and error handling.

Same logic as done in commit fb296c70 done for energy.

Bug 4336

If heterogeneous job step is unable to acquire MPI reserved ports then
    avoid referencing NULL pointer.
bug 4333

Bug 3647.

%{_libdir} is /usr/lib64. Adding this can cause ordering
problems with the library search path, and is redundant.

And %{_libdir}/slurm is not needed by Slurm itself; it relies on
rpath to resolve the install location instead.

So, this is only useful on a non-standard install, in which case
it's best left to the site to decide how to handle this instead.

Bug 4344.

Bug 4141

When updating NumNodes, cpus, tasks, gres, memory, etc. need to adjusted
so that counts are correct.

Co-authored with Danny Auble

This requires changing the resv_id and then also updating that on all
the jobs.  Since this requires having the job write lock we spawn
a new thread instead of altering all the paths into
_advance_resv_time().

Bug 4246

Issue could be triggered when updating a partition node(s) with node(s)
that were already in the partition, incorrectly increasing the
node_record->part_cnt (number of associated partitions) and thus
incorrectly extending the array of pointers to partitions associated
with this node, leading to an array with repeated associated
partitions pointers.

Bug 4318.

On CLE 6.0 mungedir is /usr; a 'module unload' call then removes /usr/bin
from PATH which is rather inconvenient.

Bug 4334.

A slurmctld lock needs to at least have a node and partition write lock set
before set_cluster_tres() is called. The slurmctld "cold-restart" path
was not covered.

Bug 4318.

Memory TRES was getting the pn_min_memroy value when updating the job.
But the TRES memory value is the total memory of the job
(e.g pn_min_memory * cpus || pn_min_memory * nodes).

Bug 4177

Bug 4264
Bug 3837

CVE-2017-15566.

Bug 4228.

Added SchedulerParameters configuration option "disable_hetero_steps" to
    disable job steps that span multiple components of a heterogeneous job.
    Disabled by default except with mpi/none plugin. This limitation to be
    removed in Slurm version 18.08.
bug 4322

node_features/knl_generic plugin: Do not clear a node's non-KNL features
    specified in slurm.conf.
bug 4294

Avoid rebooting a node if a job's requested feature is not under the
control of the node_features plugin and is not currently active.

bug 4294

Bug 4294

Added more validation logic for updates to node features.
Added node_features_p_node_update_valid() function to node_features plugin.
bug 4294

Starting in MariaDB 10.2 many of the api commands started
setting errno erroneously.

Backport of 5b934425

This reverts commit 7e5d3d15.

Turns out the spank_task_privileged needs to execute inside the child
process instead of the slurmstepd.

Bug 4298

If configured with NodeFeatures=knl_cray and there are non-KNL
nodes which include no features the slurmctld will abort without
this patch when attempting strtok_r(NULL).

bug 4294

instead of sometimes calling functions directly.

instead of sometimes calling functions directly.

instead of sometimes calling functions directly.

As well as jobcomp functions.

This also renames some to match the calls of other things.

Bug 4292

No other options use underscores, so make these consistent with
the rest. Both old and new syntax is supported for now.

Change all 'verbose' option output to use dashes, and adjust
the testsuite to match.

Fix a few related documentation errors while here.

If configured with NodeFeatures=knl_cray and there are non-KNL
  nodes which include no features the slurmctld will abort without
  this patch when attemping xstrdup(NULL).
bug 4294

Bug 4279

Before it would allow all sorts of things like
adding/subtracting/multiplying/etc.  It would cause warnings such as

/home/bart/slurm-tmp5/src/common/layouts_mgr.c: In function ‘_layouts_load_automerge’:
/home/bart/slurm-tmp5/src/common/layouts_mgr.c:363:21: error: ‘*’ in boolean context, suggest ‘&&’ instead [-Werror=int-in-bool-context]
   *lvalue = *lvalue * *rvalue;     \
             ~~~~~~~~^~~~
/home/bart/slurm-tmp5/src/common/layouts_mgr.c:1034:4: note: in expansion of macro ‘_entity_update_kv_helper’
    _entity_update_kv_helper(type_t, operator); \
    ^~~~~~~~~~~~~~~~~~~~~~~~
/home/bart/slurm-tmp5/src/common/layouts_mgr.c:1086:4: note: in expansion of macro ‘_layouts_load_merge’
    _layouts_load_merge(bool, s_p_get_boolean);
[tag] [reply] [−] Private Comment 16

Bug 4062

Setting -Werror tricks the test into failing on the

"error: In the GNU C Library, "major" is defined by <sys/sysmacros.h>.
 For historical compatibility, it is  currently defined by <sys/types.h>
 as well, but we plan to remove this soon. To use "major", include
 <sys/sysmacros.h> directly. If you did not intend to use a system-defined
 macro "major", you should undefine it after including <sys/types.h>."

error. Since the normal Slurm build uses -Werror, this warning on including
both headers will then cause the build itself to fail.

Bug 3982.