Add missing test test_fixed_foot_detector.py to Python tests

[traversaro]

There was an unexpected test failure in conda-forge/bipedal-locomotion-framework-feedstock#25, and it turned out a test was not run in the CI.

[traversaro]

I also tried to fix the test, with this modification:

+++ b/bindings/python/Contacts/tests/test_fixed_foot_detector.py
@@ -5,13 +5,12 @@ pytestmark = pytest.mark.contact_detectors
 import bipedal_locomotion_framework.bindings as blf
 import manifpy as manif
 import numpy as np
-
+from datetime import timedelta

 def test_fixed_foot_detector():
-    dt = 0.01
-
     parameters_handler = blf.parameters_handler.StdParametersHandler()
-    parameters_handler.set_parameter_float("sampling_time", dt)
+    dt = timedelta(seconds=0.01)
+    parameters_handler.set_parameter_datetime("sampling_time", dt)

     detector = blf.contacts.FixedFootDetector()
     assert (detector.initialize(parameters_handler))
@@ -65,7 +64,7 @@ def test_fixed_foot_detector():
     phase_list = blf.contacts.ContactPhaseList()
     phase_list.set_lists(contact_lists=contact_list_map)

-    assert (detector.set_contact_phase_list(phase_list))
+    detector.set_contact_phase_list(phase_list)

     # Retrieve initial fixed foot
     fixed_foot = detector.get_fixed_foot()

but now I am hitting the problem:

        phase_list = blf.contacts.ContactPhaseList()
        phase_list.set_lists(contact_lists=contact_list_map)

        detector.set_contact_phase_list(phase_list)

        # Retrieve initial fixed foot
        fixed_foot = detector.get_fixed_foot()

>       assert (fixed_foot.name == "l_sole")
E       AssertionError: assert 'Contact' == 'l_sole'
E         - l_sole
E         + Contact

python/Contacts/tests/test_fixed_foot_detector.py:72: AssertionError
---------------------------------------------------------------------- Captured stderr call -----------------------------------------------------------------------
[ERROR] [2023-04-27 23:01:53.522] [thread: 7556] [blf] [FixedFootDetector::getFixedFoot] Unable to find the fixed foot. This should never happen.

[GiulioRomualdi]

Let me try to give it a look

[traversaro]

I guess it is a matter of adding timedelta(seconds=<>) in the right places.

[traversaro]

Small OT: there are also a lot of warnings like:

PytestUnknownMarkWarning: Unknown pytest.mark.parameters_handler - is this a typo?  You can register custom marks to avoid this warning - for details, see https://docs.pytest.org/en/stable/how-to/mark.html
    pytestmark = pytest.mark.parameters_handler

not sure if this is intentional or not.

[traversaro]

There are also a lot of Ubuntu failure like:

TypeError: Unregistered type : manif::SE3<double>

this is because the compilers meta package installs a gcc version different from the one that is used in building the manifpy recipe, I will try to fix that at the conda-forge level.

[GiulioRomualdi]

Hi tried to replicate the failure but everything seems working fine on my PC.
All the dependencies are installed with apt and superbuild compiled from src

I've tested this commit 3628840 (The last of this PR)

[traversaro]

There are also a lot of Ubuntu failure like:

TypeError: Unregistered type : manif::SE3<double>

this is because the compilers meta package installs a gcc version different from the one that is used in building the manifpy recipe, I will try to fix that at the conda-forge level.

This will be fixed by conda-forge/compilers-feedstock#57 .

[GiulioRomualdi]

Hi tried to replicate the failure but everything seems working fine on my PC. All the dependencies are installed with apt and superbuild compiled from src

I've tested this commit 3628840 (The last of this PR)

Ok that's strange the test is not called via ctest probably because I've a dirty build folder 🤔

[GiulioRomualdi]

The test #658 (comment) was failing because of a regression caused by 95a29b4.
I fixed it in b91fb10

[traversaro]

Thanks, great! If the Windows Conda CI passes I think we can merge, the other CIs will work as soon as conda-forge/compilers-feedstock#57 is merged, but I guess we need anyhow to wait at least for the US morning.

[GiulioRomualdi]

Shall we release a new version?

[traversaro]

Shall we release a new version?

For conda-forge it is not necessary, I just skipped the broken test there, but there was no actual bug in the non-test code.

[GiulioRomualdi]

Since windows conda ci is passing I'm going to merge the PR as @traversaro suggested in #658 (comment)