# Conflicts:
#	src/common/slurm_acct_gather.c
#	src/plugins/sched/backfill/backfill.c
#	src/slurmctld/node_scheduler.c

done in that commit, but only in the main scheduler instead of backfill.

'scontrol update reservationname=foo duration=unlimited' sets INFINITE
as the duration, which needs to be translated to a year as is done
elsewhere. Otherwise it'll convert to 49710 days, which is definitely
wrong.

Bug 3836.

And handle memory leak.
Continuation of d3ee28fa

build_all_frontend_info() could be called twice unintentionally.

Bug 3443.

_get_begin_next_month() may be called twice in succession as
a side-effect of the MIN macro. Store in a local variable first.

MAX() will re-evaluate the higher value argument; if this is a function
is may be called twice over, leading to unintended side effects or a
crash.

Bug 3443.

fed.weight was removed in ae0c71d0.

fed.weight was removed in ae0c71d0 but the packing of it wasn't
removed in the case of a null slurmdb_cluster_rec_t.

Otherwise the recently newly added '#include <lua.h>' condition will
break the build on systems without any Lua installation.

Follow on to commit 7cc4d0d8. Bug 3681.

This also fixes the fed_mgr on the backup slurmctld to start backup
correctly if the backup takes control more than once.

Bug 3827

Continuation of a6e19307
CID:169658

Since the cluster records are merged into one cluster record in a
federation, the total time needs to account for each cluster's time.

It will result in the job being held if it is requeued at a later time.

Bug 3751.

Previously, incorrect partitions and duplicated partition names
could be shown.

The array needs to be incremented by two, not one, as each
element is a start-end pair.

Bug 3793.

when configured with memory-leak-debug

Made obsolete by moving the federation to an async model.

Don't need it anymore. Was used to hosw communications between sibling
clusters.

Previously it was only opening the connections at startup and then if
had something to send to a sibling that it didn't have a connection to.
It's better to have them establish the connection when the cluster is
added to the federation so that it can detect if the cluster is down
later and the clusters can sync jobs when joing the federation.

Since moving to a asynchronous model for submitting jobs in the
federation, LLC doesn't apply anymore. And since LLC was the only flag,
don't display the federation flags in outputs by default anymore.
Leaving the variable in the struct in case it is needed later.

Doesn't apply anymore now that federation communication is asynchronous
and jobs are submitted to all siblings without regards to a cluster's
weight.

I cannot fathom why this was added - this would only have
influenced something if listen() was implemented as an inline
function or a macro, which I don't believe has ever been the
case.

Regardless, on modern Linux systems that limit is set dynamically
through /proc/sys/net/core/somaxconn.

While not POSIX compliant, BSD and Linux platforms have shipped this
function for a long time, and slurmctld/slurmdbd/slurmd have actually
been using the version from libc instead of this local redefinition.

Zero cleanup functions exist within the source, remove all this
dead infrastructure.