Clojurians Log v2

Does it work outside of nREPL?

Yes, it works if I start bb repl

Have you tried version 225?

Yes, I think that's the version where I first noticed the problem, then I did a binary search to try to identify where it started.

I see Babashka v1.13.221 above though?

Yes, that was the first version where I saw the problem. Let me double-check with v1.13.225...

ok, I think I see the problem, it has to do with (.getContextClassLoader (Thread/currentThread)) (clojure.lang.RT/baseLoader). I siwtched nREPL to use nREPL's own source and this needs to be fixed. I think a workaround is to do this in the nREPL first:

(.setContextClassLoader (Thread/currentThread) (.getClassLoader clojure.lang.RT))

If I do that, I get a different error:

;; Connected to nREPL server - 
;; CIDER 2.1.0-snapshot (package: 2.1.0-snapshot), nREPL 1.7.0
;; Clojure 1.12.6-SCI, Java 25.0.4
user> (.setContextClassLoader (Thread/currentThread) (.getClassLoader clojure.lang.RT))
;; => nil
user> (require '[cognitect.aws.client.api :as aws])
;; => nil
user> (aws/client {:api :dynamodb})
clojure.lang.ExceptionInfo: Cannot find resource cognitect/aws/dynamodb/service.edn. [at cognitect/aws/service.clj:32:5]
user>

and if you do this?

(.setContextClassLoader (Thread/currentThread)
  (some #(when (= "main" (.getName ^Thread %)) (.getContextClassLoader ^Thread %))
        (.keySet (Thread/getAllStackTraces))))

do it before anything else

That gives a NPE again

Thank you, much appreciate the help.

The workaround above: you need to restart the nREPL server first if you haven't done so and execute this before loading any aws namespace

I think I did, let me double-check...

OK, so if I start bb repl and start an nrepl server from the REPL, then connect with cider and try your work-around, it works. But if I start the nrepl with this Babashka task and try the work-around, I get the NPE I was seeing before:

{:doc "Start an nREPL server"
 :depends [gen-version]
 :extra-paths ["test"]
 :extra-deps {org.clojure/test.check
              {:mvn/version "0.9.0"}}
  :requires ([babashka.nrepl.server :as srv])
  task (srv/start-server! {:host "localhost" :port 1339})}

this is a workaround that should cover all scenarios:

export BABASHKA_PRELOADS="(require 'nrepl.util.classloader) (alter-var-root (resolve 'nrepl.util.classloader/dynamic-classloader) (constantly (fn ([] (.getContextClassLoader (Thread/currentThread))) ([cl] cl))))"

Meanwhile I've merged the fix. Once master CI finishes you can install the dev build and remove your workaround:

bash <(curl ) --dev-build --dir /tmp

please test so I won't release a version that didn't fix it properly. should be good in 15 minutes

Thank you, confirmed it works with the BABASHKA_PRELOADS work-around. I will try the dev build in 15 mins or so.

👍 Make sure to unset BABASHKA_PRELOADS with the fixed version :)

Will do 🙂

Should be there now

That seems to be working!

Thank you so much for your help and the quick fix - I was at a complete loss.

thanks for reporting!

Reporting it was the easy bit, you did all the hard work!