And from underlying slurm_msg_sendto_timeout call as well.

Add function to clear total_gres at start of scheduling cycle
Modify logic to avoid overflow on gpu counter

bug 4373, comment #24

Removed the sentence which incorrectly stated that when not using the
gres flag enforce-binding option, cpus other than the ones defined in gres.conf
could be used for a gpu.

Bug 5189

bug 5189

Add salloc/sbatch/srun option of --gres-flags=disable-binding to disable
    filtering of CPUs with respect to generic resource locality. This option is
    currently required to use more CPUs than are bound to a GRES (i.e. if a GPU
    is bound to the CPUs on one socket, but resources on more than one socket
    are required to run the job). This option may permit a job to be allocated
    resources sooner than otherwise possible, but may result in lower job
    performance.
bug 5189

'have_innodb' is deprecated.

Bug 5317.

Cleanup printf formaters and ensure they match the types:
%zu for size_t
%zd for ssize_t

Bug 5417.

When printing arrays in squeue and setting the SLURM_BITSTR_LEN variable to 0
or to NULL, the length of the output defaulted to 64, when the documentation
says it will default to "unlimited". This patch fixes this situation.

Bug 5440

Because of a bug in the some versions of the Linux kernel, disable
constraining kernel memory space with cgroups by default.

Bug 5223.

Coverity CID 186991

Logic was switched around in 17.11, enable_user_top is now the
correct option.

Bug 5165.

This is not working reliably even when setting
SchedulerParameters=enable_hetero_steps and/or using OpenMPI with Slurm's
mpi/pmi2, as it was previously documented.

Bug 5309.

Bug 5426.

Bug 5426.

Bug 5426.

Bug 5426.

Still needs some work here. Ideally the length of the
script should be read out of the msg struct, rather
than recomputed. That would also move us closer to allowing
binary data in the script.

As we already copy this on the normal path out of here,
just use a heap allocation to begin with.

Sync error messages between _log_init and _sched_log_init.

Add _GNU_SOURCE define to expose O_CLOEXEC on older platforms.

Return a Buf instead. Adjust _pack_batch_job_launch_msg()
and _pack_job_desc_msg() to work with the Buf instead
of a char *.

Note one minor functional change in _open_part_state_file() -
no error will the thrown in the state file is < 10 bytes. I do not
believe this to be a big deal - the only such valid file that would
not immediate throw an error is 4-bytes containing just the
SLURM_PROTOCOL_VERSION and nothing else.

Given a filename, open() and mmap() it as the data field of a Buf.

Tag the Buf with 'mmaped' boolean to make sure free_buf knows
to munmap rather than xfree the data structure, and so we can
call fatal_abort if xfer_buf_data or grow_buf are called.

Documented, and code reads as needing, the node lock. But these
were incorrectly set as the job locks.

Bug 5394.

Needs the job write lock, as it may change job status not just node
status. Especially after commit 33e352a6.

Bug 5406.

harden some gres related tests

An is_*() named function should return true/false, but not modify things.
It has been renamed to a more adequate name: remove_path_slashes(char *).

Bug 4859