How to propagate `TypeError`?

[rigorpls]

Following my attempts at #122, I'm trying to get the right output for the following snippet:

(condition-case thecond
    (length t)
  (error thecond))
;;=> (wrong-type-argument sequencep t)

Emacs has the wrong-type-argument symbol and an 'error-conditions property for it, which is used by ert.el.

However, rune gives me (error . "expected Sequence, found Symbol: t") here, and I don't see the TypeError originally created in length surviving the macros up until the condition_case implementation.

I've managed to get a similar output by replacing it with a LispError in 2d6d766, which is probably the wrong thing to do.

I've gathered the changes I needed to see basic should and should-error tests succeed in #123, in case it might be useful.

[CeleritasCelery]

This is something that needs to be fundamentally fixed. Initially, I was just using anyhow for the errors because it was easy and could handle anything. But that is is not how Emacs errors work. They have specific types of errors that can be thrown. So this led to me adding EvalError, LispError, and other types. But it is really a mess. We need to (at some point) get rid of anyhow and use a error type that correctly holds and propagates the type of the error without the need for downcasting. Your workaround in #123 is fine for now