Should setting the line number be allowed for PY_RESUME events?

[markshannon]

Currently, setting the line number of a frame in a PY_RESUME event is allowed, but was accidentally broken in d27acd4

However, setting a line number is not allowed for PY_START events.
Historically both these events were treated as CALL events by sys.settrace, and setting the line number was not allowed.

I think, that for consistency, setting the line number of a frame in a PY_RESUME event should not be allowed.

@gaogaotiantian thoughts?

Linked PRs

• gh-113102: Fix typo in INSTRUMENTED_RESUME #114349

[gaogaotiantian]

I guess the question is - do we benefit from being able to jump in a PY_RESUME event.

For example, a dummy code:

def f():
    yield a; b = 3;
    yield b;
for i in f():
    print(i)

Being able to jump between yield a and b = 3 actually gives the user something - might not be that useful in this case. There might be actually useful cases in asyncio where PY_RESUME happens much frequently.

I don't think we ever documented this behavior, at least not in pdb. We say not all jumps are allowed, but never mention CALL event.

However, honestly, I don't think a lot of people are using this feature and I doubt anyone will even notice this. I'm totally fine with simply make the behavior consistent and disallow jump for PY_RESUME - especially if that makes the interpreter code simpler.

[hugovk]

Triage: can this be closed or is there more to do?

[neonene]

Ah, this issue is still reconsidering the current situation restored by #114349.