Fork me on GitHub
#tools-deps
<
2019-10-04
>
dominicm07:10:33

The alternative in this case is creating impl classes, and I'm not sure if that's better or worse. It's unclear whether that's a stable api to me, because the artefact is named impl too

dominicm07:10:03

So the repro is this:

;; deps.edn
{:paths ["src"]
 :deps
 {org.clojure/tools.deps.alpha {:mvn/version "0.7.541"}
  metasoarous/oz {:mvn/version "1.6.0-alpha5"}}}
;; src/io/dominic/tda_npe/core.clj
(ns io.dominic.tda-npe.core
  (:require
    [clojure.tools.deps.alpha :as tools.deps]))

(def deps-map
  '{:paths ["src"]
    :deps {org.clojure/clojure #:mvn{:version "1.9.0"} }
    :mvn/repos {"central" {:url ""}
                "clojars" {:url ""}}})

(defn -main
  [& _]
  (tools.deps/resolve-deps deps-map {}))

dominicm08:10:31

clojure -Spath | while IFS= read -rd: line; do
    if [ -f "${line}" ]; then
        jar tf "${line}" | while read -r file; do echo "${line}: ${file}"; done
    fi
done | sort -k2 | uniq -f1 --all-repeated=separate
Here's a script for finding duplicates on the classpath 🙂

dominicm08:10:34

metasoarous/oz {:mvn/version "1.6.0-alpha5"
                  :exclusions [org.apache.maven/maven-aether-provider]}
This fixes it (haven't tested oz). Looks like there's 2 sources for the default impl of VersionResolver!

dominicm08:10:43

It's coming transitively from pomegranate, so I think you can't have pomegranate and t.d.a on the classpath together without this exception.

delaguardo11:10:11

https://github.com/marketplace/actions/setup-clojure I wrote github action to setup hosted runner with clojure cli.

👍 8
parrot 9
vlaaad12:10:15

This is amazing, thank you very much!

delaguardo12:10:03

you are welcome to open an issue if something isn’t working as expected)

delaguardo16:10:35

also published action for provision graalvm instead of java - https://github.com/marketplace/actions/setup-graalvm-environment

seancorfield18:10:34

Adding this to next.jdbc to matrix test on Java 8 and Java 11. Thank you!

seancorfield18:10:52

Wow, it takes a long time for those jobs to get started! 21 minutes for the first one. Still waiting for the second one in the matrix.

seancorfield19:10:23

29 minutes end-to-end. It'll be interesting to see how they tune that during the beta. CircleCI started almost immediately (and completed in 30 seconds, compared to an actual run time of about a minute on GitHub).

vlaaad19:10:09

For me it usually starts immediately...

seancorfield19:10:43

Maybe it was just because this was the first time I've used it? I'll report back on subsequent runs.

seancorfield20:10:54

Just pushed another commit and, yeah, it started up nearly immediately.