CIDER hanging with newer babashka.nrepl

[giuliano108]

I'm having issues getting CIDER to talk to babashka 0.6.1 nREPL . CIDER hangs afer evaluating the first form. When using babashka 0.4.1 everything works fine. I tried multiple versions of CIDER / babashka.

In Emacs, I've set nrepl-log-messages to t, causing CIDER to log all the nREPL messages sent/received.

With bb 0.4.1 , I see CIDER sending 2 eldoc ops (id=6 and id=7), folloed by an eval (id=8). bb immediately replies no-eldoc to id=6 and id=7 . id=8, as expected gets the result of the eval.

With bb 0.6.1 , bb doesn't reply to the 2 eldoc ops or the eval. Ops 6, 7 and 8 go unanswered. I double checked with tcpdump to confirm that the responses are missing from the "wire" as well.

You should be able to see the full side-by-side logs here:

https://gist.github.com/giuliano108/47edecde5f64d964a596ffda1d2ba907/revisions?diff=split#diff-eda0f6fe7e39e7b95b3db19a55a8f05c2f3e3c84d72039dc945aa9a17fb582cbL92

[borkdude]

@giuliano108 It would be useful if you can detect the most recent bb that still worked before 0.6.1:

https://github.com/babashka/babashka/releases

[borkdude]

It's still working for me with CIDER but I haven't upgraded CIDER for a while. Could you also specify your CIDER version?

[borkdude]

For the record, I'm using cider 1.2.0-snapshot (from July).

[borkdude]

Now upgraded to the newest 1.2.0-snapshot released on 14th of September, still works. Do you have some code you are running for reproduction?

[giuliano108]

Now upgraded to the newest 1.2.0-snapshot released on 14th of September, still works. Do you have some code you are running for reproduction?

I tried both 1.2.0-snapshot and now I'm on 1.1.1. With a clj file open in Emacs, I'm running bb nrepl-server, then cider-connect. At that point, evaluating (+ 1 1) in the cider-repl buffer, without doing anything else, was causing CIDER to hang.

I'm saying "was" because as I was trying every bb release in turn, I stopped being able to reproduce the issue... 😞

The same operation that was getting no response before (eldoc for the -> symbol) is now being regularly answered.

I'm sorry for wasting your time, I'll close the issue and will come back with better data if I come across it again in the future.

Thanks for babashka!

(-->
  id         "6"
  op         "eldoc"
  session    "e7fe485f-e0cc-4b6a-8514-a3293af8f99f"
  time-stamp "2021-09-17 21:59:17.901560000"
  ns         #("workday" 0 7 (fontified nil help-echo cider--help-echo cider-locals nil cider-block-dynamic-font-lock t face font-lock-type-face))
  sym        "->"
)
(<--
  id         "6"
  session    "e7fe485f-e0cc-4b6a-8514-a3293af8f99f"
  time-stamp "2021-09-17 21:59:18.074796000"
  eldoc      nil
  name       nil
  ns         "clojure.core"
  status     ("done")
  type       "function"
)

[giuliano108]

@borkdude, I've managed to reproduce the issue outside of CIDER, by using the "official" nrepl/nrepl.

This is the repro script:

(require '[clojure.pprint :refer [pprint]])
(require '[nrepl.core :as nrepl])

(with-open [conn (nrepl/connect :port 1667)]
  (let [client (nrepl/client conn 1000)]
    (println "### 1 - Sending")
    (time
      (-> client
          ;; setting a :ns other than "user" triggers a different behaviour
          ;; depending on the version of babashka
          ;; bb 0.5.0 -> all OK
          ;; bb 0.5.1 and newer -> both nrepl/message calls time out
          (nrepl/message {:op "eldoc" :sym "->>" :ns "somens"})
          doall
          pprint))
    (println "### 1 - Received\n")

    (println "### 2 - Sending")
    (time
      (-> client
          (nrepl/message {:op "eval" :code "(+ 1 1)" :ns "user"})
          doall
          pprint))
    (println "### 2 - Received\n")))

(System/exit 0)

I run it with:

clj -Sdeps '{:deps {nrepl/nrepl {:mvn/version "LATEST"}}}' -M nrepl_test.clj

bb version 0.5.0 causes the script to generate the output below (nothing unexpected):

### 1 - Sending
({:id "f129d10f-7b3f-488f-bb41-86383f7112dc",
  :session "none",
  :status ["no-eldoc" "done"]})
"Elapsed time: 41.194489 msecs"
### 1 - Received

### 2 - Sending
({:id "01baf091-397c-4484-b0ab-e893a8d168d4",
  :ns "user",
  :session "none",
  :value "2"}
 {:id "01baf091-397c-4484-b0ab-e893a8d168d4",
  :session "none",
  :status ["done"]})
"Elapsed time: 18.036323 msecs"
### 2 - Received

Newer babashkas, including 0.6.1 cause the script to hang/timeout waiting on bb's nREPL server. That's the same behaviour I was seeing in CIDER.

### 1 - Sending
()
"Elapsed time: 1024.387576 msecs"
### 1 - Received

### 2 - Sending
()
"Elapsed time: 1005.941852 msecs"
### 2 - Received

[borkdude]

Thank you!

[borkdude]

I can reproduce this with executing your script against v0.5.0 vs v0.6.0:

docker run --rm -p 1667:1667 babashka/babashka:0.6.0 bb nrepl-server 0.0.0.0:1667

### 1 - Sending
()
"Elapsed time: 1016.894966 msecs"
### 1 - Received

### 2 - Sending
()
"Elapsed time: 1003.025573 msecs"
### 2 - Received

[borkdude]

Found the issue. What OS are you using with bb? I can provide a binary for testing later today.

[borkdude]

Here is a linux binary:

https://22909-201467090-gh.circle-artifacts.com/0/release/babashka-0.6.2-SNAPSHOT-linux-amd64.tar.gz

In case you need a different OS, let me know.