@gvanrossum, do we know why we were special-casingBEFORE_WITH like this?

I ask because it looks like this PR madeBEFORE_ASYNC_WITH (which also uses_PyObject_CallNoArgs) a viable uop. Itseems fine, but maybe I'm missing something.

@gvanrossum, do we know why we were special-casingBEFORE_WITH like this?

I ask because it looks like this PR madeBEFORE_ASYNC_WITH (which also uses_PyObject_CallNoArgs) a viable uop. Itseems fine, but maybe I'm missing something.

Urgh, I have no idea. :-( Agreed that both seem fine. Either both should have theTIER_ONE_ONLY marker, or neither should. Let me dig through some history:

It's been like this since the cases_generator code first landed (#105924). And even then, commenting out that line makesBEFORE_*WITH appear in the output and it compiles just fine. In the PR I found this commit though:
e3471da

The commit message gives a clue:

Fix test_memoryview by excluding BEFORE_WITH

Unclear what's wrong with that opcode."

Commenting out the line, regenerating, and then running./python.exe -Xuops -m test test_memoryview gives this error:

test test_memoryview failed -- Traceback (most recent call last):  File "/Users/guido/cpython/Lib/test/test_memoryview.py", line 288, in test_contextmanager    self._check_released(m, tp)  File "/Users/guido/cpython/Lib/test/test_memoryview.py", line 273, in _check_released    with m:         ^ValueError: operation forbidden on released memoryview object

Doing the same on main, however, causes no crashes. Quite possibly whatever was defective in the original code generator and Tier 2 interpreter has since been fixed, and the TIER_ONE_ONLY marker can safely be removed from BEFORE_WITH. I don't feel like doing a bisection to figure out the exact commit that made the test_memoryview failure go away.