Clojurians Log v2

Just tripped over a change of behavior that I hope is unintentional:

(comment
  (def seo (component/start (build-system 9310 :http-kit)))
  (clojure.core.cache.wrapped/seed (-> seo :application :caches :i18n) {})
  (component/stop seo))

The REPL connection also seems really unstable in this latest version. I keep hitting errors inside unrepl code, weird nil pointers, and it seems to even be corrupting my main REPL at the command-line.

I'm going to turn off the Experimental Features and see if it's better...

OK, with Experimental Features turned off, it seems to be more stable -- a sequence of commands that blew something in unrepl worked just fine this time.

So what am I missing with EF turned off? (just Clojure -- I don't do cljs)

There were some instability issues with the block detection. I think I was able to find a way to keep the old behavior.

@seancorfield, did you upgrade your Clojure version? These last 3 versions did nothing (or almost nothing) on Clojure side...

I'll do a diff here on code between versions to see if there's something wrong...

(also, since 0.4.0, there's nothing on Clojure side in experimental features...)

Nothing's changed on my side. This is the same work setup I've been using for ages.

That's really strange. Do you have any examples on these unrepl errors?

If I see the weird behavior again, I'll try to repro more carefully (and I'll keep EF off for now -- good to know nothing there affects Clojure).

Execution error (IllegalStateException) at unrepl.core/non-eliding-write (REPL:31).
Attempting to call unbound fn: #'unrepl.core/write
user=> (pst)
IllegalStateException Attempting to call unbound fn: #'unrepl.core/write
	clojure.lang.Var$Unbound.throwArity (Var.java:45)
	unrepl.core/non-eliding-write (NO_SOURCE_FILE:31)
	unrepl.core/non-eliding-write (NO_SOURCE_FILE:27)
	unrepl.repl/3rcwMcc9rCLOH4gMQfLo21mMbCs/tagging-writer/fn--13607 (NO_SOURCE_FILE:503)
	unrepl.repl/3rcwMcc9rCLOH4gMQfLo21mMbCs/tagging-writer/fn--13603 (NO_SOURCE_FILE:498)
	unrepl.repl$3rcwMcc9rCLOH4gMQfLo21mMbCs.proxy$<http://java.io|java.io>.Writer$ff19274a.write (:-1)
	<http://java.io|java.io>.BufferedWriter.flushBuffer (BufferedWriter.java:120)
	<http://java.io|java.io>.BufferedWriter.flush (BufferedWriter.java:256)
	clojure.core/flush (core.clj:3711)
	clojure.core/prn (core.clj:3721)
	clojure.core/prn (core.clj:3714)
	clojure.core/apply (core.clj:665)

That was in my regular REPL after eval'ing some code via Chlorine into a Socket REPL started from that regular REPL.

I'll see if there's any change on the UNREPL repository, some fix or new version. In the meantime I'll try to debug here

I asked about Clojure version because I saw some strange behavior of the REPL on Clojure 1.10.1 (some functions disappearing from clojure.core namespace) but I wasn't able to reproduce most of the time...

Also, I just remembered that I changed the version of unrepl blob, I'll check if it can be an issue...

@seancorfield just one last question - these nil pointers are errors while evaluating Clojure core, or issues like you evaluate something and it just returns nil and not the result of the code?

I evaluated something that should return non-`nil` (and normally was doing so) and instead it displayed nil in red followed by an exception from somewhere inside the unrepl code -- that was inline in Atom (the above was in the command-line REPL).

As for Clojure 1.10.1, I've been using that ever since it appeared, with no issues until recently...

Ok, I'll look at it too :). I'll also see if I can get a version that doesn't use UNREPL (it will ease support for CLR and some other clojure flavors) so there are less "moving parts" that can fail

I would love a version that doesn't need to inject unrepl into my running process!