just run both builds in one instance? npx shadow-cljs watch dev test?

it checks automatically whether an instance is running

just may not detect it in time they are started too fast concurrently

> just run both builds in one instance? npx shadow-cljs watch dev test? I’ll try doing this but I think this would cause a bit of a mismatch in our tooling. I have the project setup with kaocha + kaocha-cljs2 + funnel, all of which are bundled together using kaocha hooks. Then to enable proper hot reloading with kaocha-cljs2 I manually hook into the shadow-api/watch-compile!. By running both shadow dev and test builds together, I’ll also have to run Kaocha and funnel everytime I run the dev repl which may not be required. Also won’t get terminal logs separately. Also when the dev repl is running we can no longer run kaochas --focus tests or any other runs from the CLI. Everything related to koacha will need a repl interface -or- dev repl will need to be closed. Any thoughts here? Also would a PR be accepted adding an extra option in config and making this check overridable?

run npx shadow-cljs server first

then you can run all other commands separately

We’re running the shadow server from clj (embedded). Any way for the kaocha process running in a separate process to connect to that embedded shadow server?

the check will not be overridable, the check was added for good reasons. basically shadow-cljs is not intended to run two instances per project. they WILL interfere with each other.

if you want to create a repo with that setup I can take a look. so far I don't quite understand what you are doing

Thanks I can do that.

Would super appreciate it 🙌 🙏

embedded server is fine. the npx command will also connect to that one

Process A: cider-jack-in-clj&cljs or clj -A:dev launches a server using shadow-server/start!and then does shadow-api/watch :dev Process B: Then I have another process for tests clj -A:tests -m kaocha.runner which also is trying to shadow-api/watch :test but it can’t find the shadow server running in process A

the CLJ API never looks for a new server, only the npx shadow-cljs command does

why are these using different aliases?

The clj commands? The :dev repl has some dev specific paths and debugging libraries. The :test repl has kaocha and other libraries + test paths. > the CLJ API never looks for a new server, only the npx shadow-cljs command does This is the issue because we need to use the shadow-api/watch-compile! trigger on kaocha pre-load. So we have to use the clj api.

can you explain the last bit? I have never used kaocha, why does it need to trigger this?

I guess I can add an option to remove this check in case of two embedded servers. the reason it exists is that a server writes several "state" files (nrepl.port, http.port, etc) to .shadow-cljs. if you start two the second will overwrite the files from the first

so in essence the first becomes a "ghost" that nothing external (ie. your editor) can find

so in case of your setup I guess thats fine since probably both test and dev are self contained with no "network access" required from external processes?

maybe I could add a special for that, that doesn't even attempt to write those files

> the CLJ API never looks for a new server, only the npx shadow-cljs command does 💡 Maybe that ^ could be changed instead of disabling the check?

I am currently working in a separate profile that just has;

:cljs {:extra-deps {org.clojure/clojurescript {:mvn/version "1.11.60"}
                                   thheller/shadow-cljs {:mvn/version "2.22.0"}
                                   binaryage/devtools {:mvn/version "1.0.6"}
                                   com.google.javascript/closure-compiler-unshaded {:mvn/version "v20230228"}}}

Why do you have shadow-cljs dep commented out in the :dev profile? #_#_thheller/shadow-cljs {:mvn/version "2.22.0"}

yeah, the shadow-cljs dependency needs to be active

I was troubleshooting. It was giving me the same error with and without ignoring the shadow-cljs in the deps profile. I was a bit confused with a recommendation to run shadow standalone in the docs. The whole thing gives me a bit of confusion for a moment every time. Imagine it’s a me thing. At any rate, I’m quite pleased with putting all my cljs deps in it’s own profile and my clj dev deps separate. Everything is working as I expect and workflow is quite enjoyable. Good times. CHEERS!

yes, and no. This version of shadow-cljs bumps the underlying version of the google-closure-lib, which provides http://goog.net.cookies. goog.net.cookies got deprecated (and then removed) in favour of http://goog.net.Cookies.

Ohhh, I have both there, I'll remove the lower case

I hit this the other day if you are interested https://clojurians.slack.com/archives/C0620C0C8/p1678379726699909

it is a bit more involved than just changing the cookies to Cookies... you need to use http://goog.net.Cookies/getInstance now

example code?

the gist is if you get a warning, add a hint. if you get no warning you should be safe.

No warning. Here is an example:

(ns demo
  (:require ["js-code" :refer [SomeObject]]))

(defn example []
  (let [obj (SomeObject.)]
    (set! (.-property obj) val)))

and it gets renamed?

I haven't checked that. The code seemingly works under both development and production builds

as I said .. no warning means you should be safe.

I was just wondering if what I'm doing is a bad idea and whether I should prefer goog.object/set here

Ok good to know

set! is correct there

noted! Yeah I had an issue sometime where i got into a crazy state, and wanted to rule out the shadow cache as the issue, but will avoid going forward