I have to ask...how does this affect you?

It's mostly an inconsistency that I accidentally caught when a test started failing.

It's a test for a piece of code that tries to figure out how many arguments it can pass to a listener (which can be a block, a lambda, a method, or anything callable). When the arity is -2 it can't really know how many are ok to pass so it just tries to call the method. In 9K the listener is called, but in previous versions of JRuby and in MRI there's an ArgumentError.

If I could choose I'd like all Rubies to work like 9K, but I thought I'd report it since it's an inconsistency with MRI.

The reason why I'm doing the arity checking at all is for backwards compatibility. A previous version called the listeners with other arguments.

I would like to see this solved before we entertain call protocol changes in IR for blocks which will not happen for 9k.0 but still we should fully nail all semantics of blocks down before trying to tackle it.

Ok...this is likely a (reasonably) simple problem with arity-checking in lambdas. We'll get it fixed up, because there's other logic that depends on lambdas enforcing arity strictly.

cc @enebo @subbuss

My concern with this is that lambdas have always raised arity errors, and if they aren't now we've got a significant behavioral difference. It's a difference most people won't see, but I don't like it anyway.

Fixed in commits today including b6d996d and 7e82e91.