This is an internal inconsistency. I have not identified any surface level bugs, but it is a trap for future compiler work.

This is a duplicate of issue42725.

No this is not a duplicate.
This is about making the symbol table correct for current semantics.
#42725 is about changing the behavior

Does not it change the current behavior? I do not see a difference with issue42725.

>>> def f():
...     x: (yield)
... 
>>> f()
<generator object f at 0x7f971b45d950>

The aim is to change the behavior of the symbol table to match the compiler.
The behavior has already changed at module scope.

Python 3.9.0+
>>> x:(yield None)
  File "<stdin>", line 1
SyntaxError: 'yield' outside function
>>> 


Python 3.10.0a4+
>>> x:(yield None)
>>>

Then why not fix the compiler?

As I understand, it is not clear yet what behavior should be, and discussing it is the matter of issue42725. Any changes before the decision be made are preliminary.

It is very likely that the decision will be in favor of ignoring annotations. In that case your PR or even more radical changes can be merged.

OK, I won't merge anything yet.

Well even though I still don't understand 'how this issue is different than issue42725', I don't think you actually fix the symbol table;
see this one: https://bugs.python.org/msg383659

Annotations still have side effects on the symbol table (even though they shouldn't have). and I already had a fix but decided not to go for it until issue42737 is resolved since it would make the language inconsistent. Just FYI.

I guess this will be resolved when the great PEP 563/649 debate is concluded. No need for another issue.

AFAIK this is already resolved on the main (https://bugs.python.org/issue42725)