Of course you're right. I bumped shadow and that bumped cljs too, plus tests hang in CI for an hour - off to debug!

Though it is a problem I guess in this case, since I am explicitly overriding + and don't need the validation anymore. Maybe if I exclude I can kill it

Vs just replacing the var

a quick look suggests that you are missing a :require-macros in this ns

https://github.com/sicmutils/sicmutils/blob/main/src/sicmutils/env.cljc

so macros are only regcognized if this ns is loaded by other means, which may be unpredicatable

I'd assume the infix thing is a macro so for CLJS this needs a self-require (:require-macros [sicmutils.env])

Nope, just a function

The example is the same without the ->infix, that was just for printing a string

No macros at all in the example, just a function + from my namespace

ah ok. then I guess I don't get what the point of that is

Oh, it's an extensible + that is adding symbols in this case, but can do numbers, vectors, functions, and anything you extend it to

Imagine from clojure.algo.generic

all this potemkin import-vars stuff probably doesn't work for cljs

It totally does

It works great in the REPL usually and in all tests, etc and has for a year

And works in the example where I bind the + to f as an example

This is the reader making an incorrect assumption about what the symbol + means

last time I checked it was purely CLJ so I can't really see how it works?

The sicmutils library? All cljc

It seems like it is just a problem with the REPL launched like in my example , not shadow's test runner or the cider repl

For cljs

I can prove it when I'm at a computer, I'll put this together for the cljs folks and see if this changed with the recent version or it's been this way

oh I didn't see the :refer. that was just about regular +

yeah that doesn't seem right

but fwiw potemkin import-def is absolutely not ideal for CLJS

Yup, I wrote my own lighter weight one for cljs vs using Potemkin… it doesn't pull any metadata over etc which is not ideal

https://github.com/sicmutils/sicmutils/blob/45e251d70b7199a33de8774eec3dde0a1639c3f1/src/sicmutils/util/def.cljc#L91

Not ideal but works for now

did you try this with regular CLJS? as far as I can tell shadow-cljs is doing everything it is supposed to

hmm yeah I can reproduce this with just cljs.main

not sure what is happening

guessing something with the inference stuff around those functions

repro reported in #cljs-dev maybe someone has an idea. not sure if this is a known issue.

Thank you for the repro!