gh-106368: Argument Clinic: Add tests for cloned functions with custom C base names

[erlend-aasland]

• Issue: Increase Argument Clinic test coverage #106368

[AlexWaygood]

There is still one case in this block that is untested:

cpython/Tools/clinic/clinic.py

Lines 4888 to 4890 in e28b0dc

	if (is_legal_py_identifier(full_name) and
	(notc_basenameoris_legal_c_identifier(c_basename)) and
	is_legal_py_identifier(existing)):

What if is_legal_py_identifier(full_name) evaluates to True, not c_basename evaluates to False, is_legal_c_identifier(c_basename) evaluates to False and is_legal_py_identifier(existing) evaluates to True?

I.e., if I apply this diff to your PR branch, the assertion is never triggered:

- if (is_legal_py_identifier(full_name) and- (not c_basename or is_legal_c_identifier(c_basename)) and- is_legal_py_identifier(existing)):+ if is_legal_py_identifier(full_name) and is_legal_py_identifier(existing):+ if c_basename:+ assert is_legal_c_identifier(c_basename)

...Could you also add a test with a cloned function with a custom C base name where the custom C base name is not a legal C identifier?

[AlexWaygood]

I'm getting warnings about the execution environment when I'm running python -m test test_clinic -v with this PR branch locally, btw:

Ran 243 tests in 0.620s
OK
Warning -- files was modified by test_clinic
Warning -- Before: []
Warning -- After: ['clinic/']
Warning -- files was modified by test_clinic
Warning -- Before: []
Warning -- After: ['clinic/']
test_clinic failed (env changed)
== Tests result: SUCCESS ==
1 test altered the execution environment:
test_clinic
Total duration: 1.2 sec
Tests result: SUCCESS

[AlexWaygood]

I'm getting warnings about the execution environment when I'm running python -m test test_clinic -v with this PR branch locally, btw:

Ah, and it looks like the CI is complaining about the same thing: https://github.com/python/cpython/actions/runs/5868181574/job/15910445072

[erlend-aasland]

I'm getting warnings about the execution environment when I'm running python -m test test_clinic -v with this PR branch locally, btw:

Ah, and it looks like the CI is complaining about the same thing: https://github.com/python/cpython/actions/runs/5868181574/job/15910445072

Ah, yeah, I noticed earlier today, but got sidetracked. Since we're running a "expect success" test, we will actually generate clinic output in the added test (hence the sudden clinic/ directory). A workaround is to register a cleanUp() method, another is to just suppress all output, and yet another, is to run a destination file clear at the end of the clinic input.

[erlend-aasland]

I wonder why we're not seeing complaints about clinic/ directories for the other self.clinic.parse tests...

[erlend-aasland]

I wonder why we're not seeing complaints about clinic/ directories for the other self.clinic.parse tests...

... because they don't create output; either they're dumping to block or buffer, or they're testing some weird directive, or they're run with precomputed checksums in the input (hence no regenerated output).

[erlend-aasland]

...Could you also add a test with a cloned function with a custom C base name where the custom C base name is not a legal C identifier?

Ah, good call. On it!

[AlexWaygood]

Looks great, thank you!

[erlend-aasland]

thank you!

Likewise!