Could you share a print of what you are seeing?

Sure.

If I start to type a defn like this, with my cursor as shown:

@UKFSJSM38 This is what goes into the minibuffer:

does that shows on minibuffer?

I see, I didn't even know that was possible 😂 That probably happens because you don't have flycheck or flymake configured properly?

It's not WRONG but it's also not helping me.

Ah hm, that's interesting.

Isn't this driven by flycheck:

even so, I suggest you install lsp-ui to have those warnings on your buffer like:

Yeah, ok - I had an inkling that if I installed lsp-UI, at least the errors would go somewhere else. I'll go ahead and do that, thank you.

I'll give it a try!

👍 lsp-ui is cool, but some features are a little bit too much, so I suggest add:

(setq lsp-ui-doc-enable nil
      lsp-ui-peek-enable nil)

all the others are already disabled by default IIRC

Ok thank you - I'm adding that to my use-package declaration and then I'll probably just have to customize the colors!

I think the colors should work fine OOTB, they use emacs default faces for success, warning, error etc

@UKFSJSM38 It works as you suggested - the errors appear in the buffer itself as overlays (or whatever). Interestingly, though, and possibly not a bug, but it still stops the normal (quite helpful) display from CIDER in the minibuffer in this case.

So if I type this:

Sorry - nevermind. It is working, but not always.

If I manage to determine when it works and when it doesn't, I'll come back with a more useful (precise) question.

I found pretty weird to show lint in the minibuffer, the default behavior without lsp-mode is not that AFAIK, I suggest you ask in the Discord server of emacs-lsp: https://discord.com/channels/789885435026604033/789890622424219658

Ok, thanks for the heads up - I had very limited luck asking for help there with lsp-treemacs (which I'm convinced is just broken on a number of platforms) but I'm not quite good enough yet at elisp to debug it myself.

I will try to post there later just in case they are interested.

Thanks for all the help -it's very appreciated.

Good corner case, ATM clojure-lsp caches the sha of the project.clj, and since that would not change, it will use the cached classpath, one thing you can try to confirm that is to change anything in your project.clj, like add a space or newline and after a server restart it should work

not sure how we would fix that , pretty hard to know, would need to check the sha in the .m2 or something

Thanks for the quick reply. The edit of the deps/project.clj + LspStop clojure_lsp + LspStart clojure_lsp works. I’m not sure how this would be done through the lsp interface but perhaps adding a manual call to make the cache stale ( reset the shah value ) would at least give a manual approach ?

yeah, we fixed the last cache issue I was aware in the previous release, so this is the only one that we know that we would like to reset that cache, even so, not sure worth it. At least a issue makes sense for that if you don't mind opening it, there we could discuss fixing the root cause or other alternatives

sure - I’ll open one

https://github.com/clojure-lsp/clojure-lsp/issues/1488

hum, not sure what if the metadata is written in this way means even checking the print, maybe sharing the draft PR would be easier :p ?

AFAICS the exception is happening https://github.com/clojure-lsp/clojure-lsp/blob/c04df1c50afa5738e3e84f6367b1875ad84c3c8b/lib/src/clojure_lsp/feature/hover.clj#L30

Yeah, I've got a simple fix and a complex fix lol

maybe doc is not only string? maybe a str would fix it already?

https://github.com/clojure-lsp/clojure-lsp/pull/1490

The bug is that doc-line checks (when (seq doc), which doesn't actually verify the docs are a string. If markdown is enabled, docstring->formatted-markdown throws an Exception as you noted.

the simple fix is change the check to (when (and (string? doc) (seq do)

yeah, it never occurred that doc could be other than string, that's new to me hehe

maybe we should confirm is doc that comes from clj-kondo should really be other things than string

the more complex and fun fix is to check if the metadata is referencing another var's metadata and then use q/find-definition to get information about it, and check it for a doc string lol

as you can see in the image I posted, base-var has a docstring, and then existing-var is referencing it and target-var is referencing existing-var . My PR implements tracing those references

oh got it

sounds cool that find-definition, so it seems to be 2 different issues: • fix the exception, show the hover as string • new feature: follow metadata docs

Yes

No problem fixing in the same PR, added 2 comments there, LMK what you think

of course, I'll fix those right up

that's a nice feature, we should probably mention in the docs somewhere in features.clj I guess

I can write some up tomorrow if you'd like

Would be nice! I think a simple mention in the hover section in the features would be enough

One thing would be nice to improve is the follow arity as well, at Nubank we have hundreds of functions like:

(defn ^:private bar-impl [{:keys [a b c d]}])

(defn bar [options] ,,, (bar options))

hmm! that's interesting

I've seen folks hardcode the arglists before

yeah, that should fix that as well, not sure if using the same :doc field though