Local manager does not limit the number of running tasks

[RaphaelRobidas]

When running a flow (for example, this one), the manager configuration seems to be read correctly, but the hardware limits are completely disregarded.

I use the following manager.yml file:

qadapters:

- priority: 1
  queue:
      qtype: shell 
      qname: localhost

  job:
      mpi_runner: "mpirun"
      mpi_runner_options: "--bind-to none"
      omp_env: {"OMP_NUM_THREADS": 1}

  limits:
      timelimit: 1:00:00
      max_cores: 1

  hardware:
      num_nodes: 1
      sockets_per_node: 1
      cores_per_socket: 1
      mem_per_node: 1 Gb

When launching a flow with abirun.py FLOW_DIR scheduler, the initial printout contains this, confirming that the specifications have been ready correctly:

AbiPy Manager:
[Qadapter 0]
ShellAdapter:localhost
Hardware:
   num_nodes: 1, sockets_per_node: 1, cores_per_socket: 1, mem_per_node 1024,
{'OMP_NUM_THREADS': 1}
Qadapter selected: 0

However, more than one task will run at once, although only one core should be available. This is very problematic for larger flows where the number of tasks is much greater than the number of cores, causing the machine to crash.

Abipy version: git+https://github.com/abinit/abipy@59e3816168ea11ab90622531948d0dac4cd31df9 OpenMPI version: 4.1.6 OS: Debian Trixie