Really strange: If I run the bit of fireplace that gets the classpath and path separator, the bdecoder fails. No idea why, or how this runs normally in fireplace, or even if it does? Maybe it just gets swallowed on Lein!

However, I have found the real culprit. For whatever reason, when evaluating code in Figwheel with an id, the output goes to the ORIGINAL id, not the new id.

So the '1' is in the output, but it's not picked up as belonging to the current operation.

For some reason, it also locks up figwheel on subsequent requests when I do it. Probably because we're looking in the wrong id again

YEP. That's it! It goes to the original "id" not the new one, so it gets lost. Fireplace filters differently to other editors I would guess. I think CIDER uses a subscribe system, where it listens to everything to build it's CLI.

It only locked up once, I fiddled with the code a little. It's mildly worrying, but not hugely concerning.

But yes, it keeps printing output to the id I used to create the session for figwheel.

I don't know how to push further. I guess look through figwheel source / report to @bhauman

https://clojurians.slack.com/archives/clojurescript/p1469104022000508 in case someone wants to jump in before me

@dominicm: I'm still having problems with Figwheel snapshot... hmm

@juhoteperi: hmm indeed. I'll dig in tomorrow. I was hopeful it would just work