Iain Buclaw

Could this be slightly ambiguous? Or at least, it would initially seem that there'll be a need to do a lookahead in order to determine whether `M`/`Nk` is for a...

> It's not ambiguous, and a lookahead isn't needed. You just read attributes until hitting Y or not. Not only attributes, but both attributes and parameter types.

Have I grok the proposed grammar change correctly? ``` TypeFunctionNoReturn: CallConvention FuncAttrsopt Parameters(opt) ParamClose Parameters: Parameter Parameter Parameters Parameter: Parameter2 M Parameter2 // scope Nk Parameter2 // return Parameter2: Type...

We have a bison implementation of the abi spec somewhere iirc, I'll see if I can find it.

> Not really, it won't match with Parameter because the "Type" will be a Y which is not a Type. But you don't know whether it is really a Type...

Not all addressed.

1. Would this be a precursor for diagnostics akin to: `error: func() throws but is not caught` 2. Possibility of future extensions to allow specifying which exception is thrown, e.g:...

> I assume I've broken one or part of the visitors The vtable layout has changed, so yes.

> @ljmf00 removing the union requires making my visitor much more complicated as I now have to walk everything myself versus just hooking where the identifiers appear Does it really...

> > Does it really though? Outside of unary and binary nodes, what else is there to have a visitor for? > > Just because something inherits from a UnaExp...