bpo-38818: PyInterpreterState.eval_frame now pass tstate

[vstinner]

PyInterpreterState.eval_frame (PEP 523) now requires a new mandatory
tstate parameter (PyThreadState*).

https://bugs.python.org/issue38818

[vstinner]

This change is backward incompatible and changes the PEP 523. I would prefer to have it carefully reviewed ;-)

[shihai1991]

This change is backward incompatible and changes the PEP 523. I would prefer to have it carefully reviewed ;-)

In order to keep backward compatibility, can we use a new function to do this action?

[vstinner]

In order to keep backward compatibility, can we use a new function to do this action?

PyInterpreterState.eval_frame is a structure field. What do you mean?

[corona10]

@shihai1991

It can happen from the contributors who are not familiar with the codebase.
@shihai1991 I think that the snippet I attach links below will help your future contribution.

cpython/Include/pystate.h

Line 25 in 2bc43cd

typedef struct _is PyInterpreterState;

cpython/Include/internal/pycore_pystate.h

Line 109 in 9def81a

_PyFrameEvalFunction eval_frame;

@vstinner
Looks good to me.

[shihai1991]

In order to keep backward compatibility, can we use a new function to do this action?

PyInterpreterState.eval_frame is a structure field. What do you mean?

Sorry, i didn't express cleary. I just worried about whether we have external user use _PyEval_EvalFrameDefault(struct _frame *f, int exc)(although it's a inner function).

[vstinner]

frame_eval is specified by https://www.python.org/dev/peps/pep-0523/

The question if it should be public or not is discussed at https://bugs.python.org/issue38500

At least, the API (frame_eval) is used by PyCharm to implement their debugger ;-)

[brettcannon]

I don't think this should go in until bpo-38500 is resolved.

[vstinner]

I included this change into PR #17340. I close this PR.