Handle negative timings

[kjh-giesbertz]

Describe the bug
For very short calculations timings can come out negative causing due to floating point errors. This causes a validation error getting the property results

To Reproduce
This was for a ccsd calculation on H2O in def2-svp, but it flaky, so not well reproducible and very system dependent.

Trace back

        # // Parse JSON
        self.property_json_data = self._process_json_file(self.property_json_file)
        # > Check in property JSON whether version fits:
        if self.do_version_check:
            self.check_version()
>       self.results_properties = PropertyResults(**self.property_json_data)
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
E       pydantic_core._pydantic_core.ValidationError: 2 validation errors for PropertyResults
E       calculation_timings.prop
E         Input should be greater than or equal to 0 [type=greater_than_equal, input_value=-0.6242519999999999, input_type=float]
E           For further information visit https://errors.pydantic.dev/2.13/v/greater_than_equal
E       calculation_timings.sum
E         Input should be greater than or equal to 0 [type=greater_than_equal, input_value=-0.1279849999999999, input_type=float]
E           For further information visit https://errors.pydantic.dev/2.13/v/greater_than_equal

/home/giesbertz/miniforge3/envs/livdft-fix/lib/python3.11/site-packages/opi/output/core.py:228: ValidationError

Expected behavior
I suppose the best solution is to clamp timings to be non-negative. An alternative would be to allow for negative timings, but since they are meaningless (though give an impression how imprecise timings are) I would go for clamping.

Version:

• ORCA 6.1.1
• OPI 2.0

Additional context
Add any other context about the problem here.

[haneug]

Thank you for reporting this! I was not aware that the timings can be negative. Clamping them sounds like a good solution assuming that negative timings only occur when the timings are close to 0.

[kjh-giesbertz]

Yeah, this was also a surprise to me, but Orca manages to produce this apparently.

[timmyte]

@kjh-giesbertz
Are you certain that this is due to numeric fluctuation?

I could also imagine the timings variable in ORCA is just not initialized properly. @haneug we should escalate that to the ORCA team.

[kjh-giesbertz]

@timmyte Good call. I have actually no clue what the cause is for negative timings in Orca. The numeric fluctuations are just a guess.

[haneug]

@kjh-giesbertz Since this issue is hard to reproduce could you supply us an ORCA output with a negative timing so that we can report this to the ORCA team?