Fix empty script interface crash on tscn load. #23495

[fire]

Add fail conditions to protect the visual script functions and also fix the initiating cause.

@akien-mga Do you prefer if I fix the initiating cause, add the ERR_FAILs or both?

Bugsquad edit: Fixes #23495.

[fire]

Noticed an error with visual script function creation adding all VS functions rather than just the node's virtuals.

I'll make an issue for it.

EDIT: Seems to be specific to the test case in the issue. Investigating.
EDIT: I can't reproduce the issue but I found another issue.

[akien-mga]

Do you prefer if I fix the initiating cause, add the ERR_FAILs or both?

Well the initiating cause should of course be fixed. As for the ERR_FAIL_CONDs, that's for you to judge whether those methods can be called with invalid input and/or if such invalid input could cause failures. If any of those is yes, then ERR_FAIL_COND checks are warranted.

For _update_members it makes sense to me. For has_function, is there any case where we allow creating (and then testing) for a function with an invalid identifier?

[fire]

I think you can name a function anything as a node, so it would be good if has_function didn't crash. However I can think of a case when a function is deprecated, not sure what happens then.

[akien-mga]

I think you can name a function anything as a node, so it would be good if has_function didn't crash.

Would it crash though? has_function only uses Map<StringName, ...>::has(), which can take any kind of StringName and should return false if it's not in the map.

If input validation needs to be done on function naming, it should be done further up when users actually define the name IMO.

[fire]

I think my problem is that functions becomes null but that's not solved by the fails check. Functions being null is when scripts is null.

[fire]

Removed has_function.

[akien-mga]

Thanks!