clang::Cursor::specialized should return an Option<clang::Cursor>

[fitzgen]

Right now, clang::Cursor::specialized returns a clang::Cursor regardless if that resulting cursor is_valid() or not. This is a footgun for callers, which have to remember to check resulting_cursor.is_valid(). Instead, specialized should check for them and return None if it is not a valid cursor, and Some if it is.

I can mentor someone who would like to pick up this bug.

[highfive]

Please make a comment here if you intend to work on this issue. Thank you!

[glasserc]

This looks fun and I will give it a shot today!

[fitzgen]

Great! Don't hesitate to reach out if you run into any roadblocks :)

[glasserc]

This might be a silly question, but why is it possible for a cursor to be invalid? Shouldn't the presence of a Cursor mean that it is valid and that you can use it? Shouldn't invalid cursors themselves be represented using None or something? I guess that's a larger issue than just changing this one method though..

[fitzgen]

Agreed! But we're using FFI to Clang and we don't get to decide how Clang represents its data structures :)

We do want to make it so that our wrapper Cursor type never contains a null or invalid Clang cursor, but this is a bit more work (or at least requires some more investigation and experimentation). See #119.

Fixing issues like this one will certainly help us get there, though!