Fixed AnimationTrackEditor's TransformTrack can't edit

[TokageItLab]

2021-05-07.5.13.14-1.mov

AnimationTrackEditor's TransformTrack was failing to get/set value because the type of key when getting dictionary's value was wrong in 4.0. This PR fixed it.

However, TransformTrack is working correctly in 3.2, so it is possible that there is a fundamental casting problem going on somewhere in 4.0.

[akien-mga]

AnimationTrackEditor's TransformTrack was failing to get/set value because the type of key when getting dictionary's value was wrong in 4.0. This PR fixed it.

However, TransformTrack is working correctly in 3.2, so it is possible that there is a fundamental casting problem going on somewhere in 4.0.

Your fix basically changes the argument from StringName to String. There was a number of changes to the core around StringNames are possibly that creates an issue here. But I don't know if the proper fix is to pass a String as key, or fix whatever makes it fail with StringNames when it used to work.

Maybe the problem is in Dictionary.has()?

[TokageItLab]

@akien-mga StringName in the Variant, StringName was cast to String in 3.2, but in 4.0, StringName is still kept StringName. This makes it appear that the Dictionary operator [] doesn't work as desired, so it may be better to cast StringName to String in the Dictionary operator [].

Variant.cpp
3.2

Variant::Variant(const StringName &p_string) {

	type = STRING;
	memnew_placement(_data._mem, String(p_string.operator String()));
}

Variant::operator StringName() const {

	if (type == NODE_PATH) {
		return reinterpret_cast<const NodePath *>(_data._mem)->get_sname();
	}
	return StringName(operator String());
}

4.0

Variant::Variant(const StringName &p_string) {
	type = STRING_NAME;
	memnew_placement(_data._mem, StringName(p_string));
}

Variant::operator StringName() const {
	if (type == STRING_NAME) {
		return *reinterpret_cast<const StringName *>(_data._mem);
	} else if (type == STRING) {
		return *reinterpret_cast<const String *>(_data._mem);
	}

	return StringName();
}

[TokageItLab]

Close this and move to #48523

[TokageItLab]

#43371, #44310, #48139, #48523
As with some PRs, it is still undecided where the cast from StringName to String should be done in c++. If we don't want internal casting, this PR is expected to make the appropriate changes, so I reopen it.

[TokageItLab]

Fixed by #50057.