Ben Lerner

Hmm. To play devil's advocate for a moment -- ISTR that we'd _chosen_ to continue processing as far as we could, so that we could get as many errors up...

Not at my computer so can't check, but does it give bogus results in not-tc mode if you have a check block, and also define this particular variable? It might...

Does @MichaelMMacLeod 's fix above actually fix it? The fix seems to make an internal name, which (at least on a quick glance) seems independent of gensym numbering issues. Also...

Yes, I view that as a design feature, actually. * The outermost value you have on the penultimate line is an object, with a ref field that refers back to...

Err, no, isn't this working as intended? Because we _want_ to be able to write `examples` blocks _before_ the functions they're example'ing are defined, and they can use whatever helpers...

Ok, that second example is weird, but is unrelated to the original problem. The (implementation-level) reason here is because `desugar-scope-block` turns `s-check` statements into `add-letrec-binds`s of `s-underscore`s to the result...

The *original* error reported in this bug is complicated. Looking at the show-compilation output, the Resolved names pass shows ``` let var a = 1, b = 1: builtins.trace-value(srcloc("file:///home/blerner/pyret-lang/test1.arr", 4,...

We haven't done this, have we? Certainly not in the generality of extensibly hashable data...

_I_ want it, to improve parts of the compiler. We currently have no way to represent a key=>value mapping in a single object with O(1) lookup unless the key is...

Yep