Pressing "rerun workflow" still results in the same error

can you merge with master and push that?

It has to be a force-push, is that alright?

why does it have to be a force push?

oh sorry, I rebased on master instead of merging

hang on

I think CircleCI gets confused when you force push a commit when another is still running, or something along those lines

Pushing a totally new commit should fix it

Maybe I'm missing something, but I could only get my branch to run on Circleci by editing the yml config file and pushing a temporary commit

the force push was to remove that commit after making the PR, sorry about the trouble it caused

same error.. should I make a new branch and try opening another PR?

yeah, let's just do that

btw, the spec for get is obviously wrong if you look at https://github.com/clojure/clojure/blob/653b8465845a78ef7543e0a250078eea2d56b659/src/jvm/clojure/lang/RT.java#L758 anything may implement the ILookup interface, that's why it's spec'ed as ::any in speculative: https://github.com/borkdude/speculative/blob/master/src/speculative/core.cljc#L202

ah, my mistake

should I change it to :any or remove the commit altogether?

maybe make a note there, so it will be left untouched

keys nilable associative is also wrong since (keys (filter ... {:a 1})) works

that's why it's seqable

so only the x -> nilable/x things are good to add I think

hmm, alright

yeah, it's complicated. the specs in speculative already dealt with a lot of examples from the wild, so its good to look at those first

okay

so like the type system isn't strong enough to capture those functions

should get be left unspecced then? since it's essentially an any->any function

yes

the type of keys is Seqable[MapEntry] if you would have higher kinded types

but we don't have that, it's a pretty weak typesystem that only screams at you if something is clearly wrong

alright, pushed with comments

Did you do any changes to .circleci/config.yml? I think that might be the reason why the JVM tests aren't passing

not on the PR branch

normally the script looks at CIRCLE_PR_USERNAME, but somehow that's not set

for example, if you look at another PR: https://github.com/borkdude/clj-kondo/pull/651 it just works

maybe you should try to unfollow the project on circleci again, because it seems the project isn't running the PR branch under my name

confusing, this

yeah sorry, it's my first time on circleci and I might have messed things up without knowing how

@U0K592YDP it seems like his PR isn't running on my account somehow: https://circleci.com/gh/borkdude/clj-kondo

also it seems his macOS step isn't running

I think that was because I'm on the free plan and macOS jobs are only for paid plans

I'm also on the free plan

@qythium can you delete clj-kondo from your circleci account and then do another commit?

(dummy commit to re-trigger circleci)

I've already "unfollowed" the project, is deleting a separate thing?

not sure

it seems like it's now correctly building

so I probably gave the wrong tip of following the project

Interesting, but that would make it complicated for contributors to run and test things like the performance script on their end without opening a PR

or creating a local Docker container and running Circleci there, from what I gathered

Actually my main motivation for looking into clj-kondo's type checker was to have it warn me about using get on vectors that had accidentally been converted to lazy seqs - I found myself making that error a few times resulting in hard-to-trace bugs because it would propagate nils silently

It's good to know that I can choose to override it at a local config level, even if :associative is technically too narrow

yeah

as a last step (and I hope you're not getting annoyed now): it would be good to add a couple of unit tests for the things you changed, so they will lock down the changes you made

pushed a test myself now

merged

oops, only just saw this

thanks for your time! I wasn't annoyed at all :)

and thank you for your contribution 🙂

Updating on Nix 😃 : https://github.com/NixOS/nixpkgs/pull/75733

Thank you!