I think this is the essentially same issue as in `afl` in the thread
where I just re-replied.
It looks like `curly-fn` wasn't introducing a mark the way that `afl`
did, and so it wasn't defending against a local `let` binding with the
old expander. For example,
#lang curly-fn racket
(let ([let 5])
#{values})
reports an error with the old expander, instead of avoiding the local
binding of `let` for the identifier introduced by `#{....}`.
For the new expander, the idea is to add a scope to the entire module
read by `#lang curly-fn`, but arrange for that scope to be removed in
the identifiers introduced by `#{....}` so they won't be bound by
anything in the enclosing module.
> --
> You received this message because you are subscribed to the Google Groups
> "Racket Developers" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to
racket-dev+...@googlegroups.com.
> To post to this group, send email to
racke...@googlegroups.com.
> To view this discussion on the web visit
>
https://groups.google.com/d/msgid/racket-dev/B5D45B8F-6059-4104-AC35-CB0B683EFF
> BA%
40gmail.com.
> For more options, visit
https://groups.google.com/d/optout.