So I'm looking at the codebase to try and add an exclude config to use blocks. It looks like the :use lint is different from the other ns lints in that it's registered immediately rather than just associng in something new to register a finding for later. For this would it be better to try and refactor the use finding to work the same way as e.g. refer-all, or would it be better to just add an additional requirement in the and expression that gates registering the finding?

The latter is most certainly easier, but it's inconsistent with the rest.

What is the problem you're trying to solve?

you're trying to unify the code for :use and :refer :all in clj-kondo's codebase?

More or less. I just want the :use linter to have an :exclude config

you could also migrate to :refer :all

but that's only a workaround ;)

but I'm open to improvements, as long as the tests pass and there are no perf regressions

Yeah.

Right. I was more or less asking if you'd prefer if I just add a one-liner that just doesn't register the use finding if the config excludes the ns, or if you'd prefer that I try to make use work like refer does

I would have to look into that, don't remember the exact code. I think :refer :all has some logic that prefetches some vars from the cache to improve linting, does :use also have that?

yeah, looks like there's a call to reg-var-usage!

I actually meant this part: https://github.com/clj-kondo/clj-kondo/blob/2bfc8bfe50d143f5806f489a074ff601ab834a16/src/clj_kondo/impl/analyzer/namespace.clj#L221-L226

What it does is look at the cache in case of :refer :all so it can actually resolve names better. clj-kondo has a hard time with use and refer all because it has nothing to go by statically

it just assoces stuff into the :referred-all segment and passes that on so that it would get used here

I don't know if this is a problem with Cursive, the IJ file watcher plugin, or clj-kondo, but I'm seeing the following, where the indicator in the IDE is highlighting the fn inside the sexpr instead of either the whole sexpr or the int fn itself (what I'd expect). The output below shows the correct columns, which correspond to the open paren for the subs sexprs.

can you make a minimal repro in text? then I'd be happy to try out what it does in emacs. you could also just disable clj-kondo to see if that is or is not the problem?

https://github.com/clj-kondo/clj-kondo/pull/1290

Does the lint-as clojure.core/defn have issues with multiple fn bodies?

it should not

Hmm. I'm trying to mess with specter again and multi-path isn't linting correctly. I've set the lint-as for defdynamicnav to lint-as defn, but it's saying it can't resolve path1 and path2

(defdynamicnav multi-path
  "A path that branches on multiple paths. For updates,
   applies updates to the paths in order."
  ([] STAY)
  ([path] path)
  ([path1 path2]
   (late-bound-richnav [late1 (late-path path1)
                        late2 (late-path path2)]
     (select* [this vals structure next-fn]
       (let [res1 (i/exec-select* late1 vals structure next-fn)]
         (if (reduced? res1)
           res1
...

can you make an example I can actually try locally, so a complete form? this will make it easier for me to help you

Sure, sorry about that

(defdynamicnav something
  "blah"
  ([] nil)
  ([path] path)
  ([item1 item2]
   [item1 item2]))

This recrates the error, all defdynamicnav has to be is some random macro that has lint-as set to defn

Hmm. It seems like it might be something specific to this config. I want to do a little more testing.

This works:

(ns foo
  {:clj-kondo/config '{:lint-as {bar/defdynamicnav clojure.core/defn}}}
  (:require [bar :refer [defdynamicnav]]))

(defdynamicnav something
  "blah"
  ([] nil)
  ([path] path)
  ([item1 item2]
   [item1 item2]))

Yeah, it worked fine for me just now too when I defined my own macro. Maybe it's that defdynamicnav is being created wrong or something.

I need to go through this a little more slowly.