Replace nameerror_after_exception by pytests's continue_on_failure

[Sheila-nk]

#87 (comment) please make sure that this test passes with the plugin

Originally posted by @ev-br in #87 (comment)

[ev-br]

This came first in scipy/scipy#13116 , might help cooking up a standalone test case.
Once there is a test, it'll be easier to decide if this can be merged with --continue_on_failure in the plugin layer.

[ev-br]

A test exists for the doctest layer in test_testmod.py::TestNameErrorAfterException. The CLI incantation is

$ pytest scpdt/tests/failure_cases_2.py --doctest-modules --doctest-continue-on-failure

and it indeed seems to match: with --doctest-continue-on-failure the traceback has four exceptions (as expected for name_error_after_exception=True); without it we only get one.

So need to golf sync the DTConfig setting and the pytest setting:

• if the pytest switch is on, it overwrites the DTConfig one (maybe emit a warning on mismatch)
• if the pytest switch is off, it is synced from DTConfig.

[ev-br]

Looked at it a while. This is low prio TBH.

A worthу aim would be to remove PytestDTRunner.report_unexpected_exception and .report_failure which were vendored from pytest's runner, https://github.com/pytest-dev/pytest/blob/8.0.2/src/_pytest/doctest.py#L184

However, this would require to multiple inherit PytestDTRunner from both DebugDTRunner (which it does now) and pytest's PytestDoctestRunner.

Does not seem worth it TBH.

Also, the PytestDTRunner inherits from DebugDTRunner, so it stops after the first exception anyway. Also, DebugDTRunner overrides all logic from nameerror_after_exeption.

[ev-br]

From the code health perspective, it's probably best to simply remove the nameerror_after_exception and always use the pytest naming, continue_on_failure throughout, both in pytest and "basic" doctest layer.