fail statement

[gavinking]

Should we introduce a fail statement, that means, approximately, assert (false)? I hate adding a whole new keyword just for this, but it's conceptually useful.

At the very least we should make definite assignment / definite return analysis treat assert (false) as a special case.

[RossTate]

This sounds more like unreachable in that it's a declaration that this location can't be reached and not a command that if reached it should throw an error.

[FroMage]

I'm very doubtful of adding a new keyword just for this. Why not make it a toplevel function of ceylon.language and treat it specially? After all, we already have some special support for Tuple and stuff, why not consider that if we can prove that x() is or is an alias for ceylon.language::fail() then it throws.

Also it reminds me a lot of the throw; we used to have at some point. Not sure how to unify both.

[gavinking]

Definite return checking now special-cases assert (false).

[gavinking]

Also relevant for definite assignment. Done now.