How to Debug the AutoPot Workflow?
==================================

As with every new package, AutoPot will not be bug-free. Presently, our
experience is that most workflow failures happen because execution of a task is
not configured properly, for example because of wrong filenames.

In such cases the workflow will not abort but starts hanging, because the
AutoPot orchestrator is written so that new events will only be triggered when
all tasks on which the event depends do not have the state ``FAILED``.

The state of runs can be checked by running the following command in the
``workflow`` directory:

.. code-block:: bash

   motoko info --verbose

The output may look like this:

.. code-block:: text

   ******************************
   TaskManager: md_select
   ******************************
   Loaded 2 jobs: OK
   Loaded job infos: OK
   Table create: OK

   | cfg_fname                        |   id |
   |----------------------------------|------|
   | md_cfgs.xyz,0                    |    1 |
   | md_cfgs.xyz,1                    |    2 |
   Loaded 2 runs: OK
   Loaded run infos: OK
   Table create: OK

   | run_name   |   id |   job_id | state     |   nproc | machine_name   | start_time     | last step   | last update   | Time/step   | Total Time   |
   |------------|------|----------|-----------|---------|----------------|----------------|-------------|---------------|-------------|--------------|
   | test       |    1 |        1 | FINISHED  |       1 | localhost      | 10:49 04/11/25 | None        | None          | None        | None         |
   | test       |    2 |        2 | FAILED    |       1 | localhost      | 10:52 04/11/25 | None        | None          | None        | None         |

Here, the run with ``id = 2`` has failed. To fix it, go to
``md_select/BD-md_select-runs/run-2`` and inspect ``test.e2`` and ``test.o2``
to check what caused the error.

If the error is caused by ``launch.sh`` or ``doIt.py``, it is probably
necessary to also fix them in the corresponding task directory so that future
runs are not impacted by the same error.