I'm not aware of a bug like that, and the project-root name should not affect anything, is it possible to create a simple repro where I can try?

It's also possible that it's a bug of my nvim. I'll try to make a repro.

Try this project: https://github.com/Invertisment/reprochildproj I added a readme It's very weird but the bug is different this time. This time it doesn't load one of two namespaces.

Also please disregard the test-bb task, it creates a JVM process and it's not correct.

From the readme, is that related to repl?

LSP has nothing related with REPL, it uses static analysis

I don't think so, but I always start a REPL.

Yes, I think REPL isn't related.

so, what's the exact lsp issue? the readme mentions loading in repl

anyway, I get the issue form readme on emacs + cider:

Could not find namespace: com.github.invertisment.repro-parentproj.core

I confused everything. I can't jump into the analyzer.clj using LSP. And REPL can't load it too.

the repl is another thing, probably related with the deps not being available in bb.edn, only on deps.edn

But I can jump into core.cljc

I can repro the find-definition issue for analyzer, let me take a closer look

I used bb.edn only to launch the repl task. So the process should exist but it should not do anything

the repl issue is that: your com.github.invertisment/repro-parentproj dep comes from deps.edn, when you start a bb repl, it won't check your deps.edn, but your bb.edn

and there is no com.github.invertisment/repro-parentproj dep there, so bb won't know about that

if you start a clojure repl, it should work though

the issue on the find definition of analyzer is that the code is invalid:

(ns com.github.invertisment.repro-parentproj.analyzer)

(defn hello []
  (println "hey") ;; missing a ) here

I'll try to fix it and try again. I was typing too fast then.

This time it works. Not sure what's the issue with my other project then.

yeah, I'd say 50% of fixing a issue is finding a way to repro 😅 , sometimes it's hard, but without reproing is not easy to understand how to fix it or if it's indeed a issue

In my other repository the issue is that with LSP I have several namespaces in one directory and when I jump into core.cljc then the file is blank. And then when I jump into function's source then it also can't find it. But REPL finds it because it loads it and runs it. But this happens only for one file in the same folder.

maybe some file with same name or something in the classpath?

(ns neil-proj.gitlab-dep
  (:require [io.gitlab.invertisment.____________.analyzer :as q]
            [io.gitlab.invertisment.____________.core :as qq] ;; blank file when navigated via LSP
            [io.gitlab.invertisment.____________.path :as qqq]
            [clojure.tools.build.api :as tools-build])) ;; this one is to test that some other source project works

(defn a []
  (qq/analyze-paths nil nil) ;; this code works via REPL but I can't LSP-jump there
  #_(tools-build/jar))

But when I jump into analyzer then in the bottom of the screen I see this path:

~/.gitlibs/libs/io.gitlab.invertisment/xxx-xxx/dd6aa966d5b8965400a7490c5c52c6d64d8fb232/src/io/gitlab/invertisment/xxx_xxx/analyzer.cljc

~/.gitlibs/libs/io.gitlab.invertisment/xxx-xxx/689d541bdbc59cc6c05c8cb7f099e2a53630cf58/src/clojure/io/gitlab/invertisment/xxx_xxx/core.cljc

are both namespaces (not the filename) equal?

All three namespaces are different, they contain code

I mean, the namespace name (ns foo.bar)

> ls | xargs -L 1 head -n 1

(ns io.gitlab.invertisment.------------.analyzer
(ns io.gitlab.invertisment.------------.core
(ns io.gitlab.invertisment.------------.path

Actually I can share the namespace names, there is nothing there:

(ns neil-proj.gitlab-dep
  (:require [io.gitlab.invertisment.query-scopes.analyzer :as q]
            [io.gitlab.invertisment.query-scopes.core :as qq] ;; blank file when navigated via LSP
            [io.gitlab.invertisment.query-scopes.path :as qqq]
            [clojure.tools.build.api :as tools-build]))

(defn a []
  (qq/analyze-paths nil nil) ;; this code works via REPL
  #_(tools-build/jar))

[martin@qwe query_scopes HEAD]$ ls | xargs -L 1 head -n 1
(ns io.gitlab.invertisment.query-scopes.analyzer
(ns io.gitlab.invertisment.query-scopes.core
(ns io.gitlab.invertisment.query-scopes.path
[martin@qwe query_scopes HEAD]$ ls
analyzer.cljc  core.cljc  path.cljc
[martin@qwe query_scopes HEAD]$

I deleted project's .clj-kondo directory and it jumps now.

Unfortunately I deleted it and I can't send it to you 😕

I think it should be something with previous commits of the project. Maybe I tried to load it while the project wasn't fully working yet. I was testing things out and moving files around.

that sounds like git branch switchs and file watch was not properly working. LSP file watch is a feature that the editor keeps listening for the project files, and when they change outside the editor the editor tells the lsp server that those files were created/changed/deleted, so the server can update analysis of the project correctly. This is mostly a feature to avoid the need of restarting your editor or lsp server to re-analyze the project, only for consistency.

But I have been restarting my editor (and with it LSP) and REPL multiple times :thinking_face: And it persisted until I deleted that directory.

Hum, weird, we don't have cache issues for a long time after we fixed some known bugs, again, if you can find a way to repro LMK, but usually is hard to repro cache issues 😔

I think for now I won't be doing a repro 😄 There is one and it could be played around to remove and add remote project's sources. And that may be enough.

But I don't want to look for it more for now

I tried the nightly release using the pinned comment, same issue

@UG00LE5TN Try the static release @UKFSJSM38 I've had the same issue with clj-kondo. The issue was that the build environment used a too new version of glibc. I downgraded the build image again and this "fixed it"

Also I install the static binary in the clj-kondo install script by default now

@UG00LE5TN You can install the static image with --static

thanks @U04V15CAJ! @UKFSJSM38 let me know if you'd like to open an issue to track this

please do

@U04V15CAJ Do you have the commit so I can do the same to clojure-lsp?

what's funny is that this started happening when building on github actions too. I've had the similar issue on circleci

@UKFSJSM38 I guess your issue is that you're using ubuntu-latest

which you should not do, use a fixed version of ubuntu and not the newest

Ok

I made --static the default now in the install script too for linux amd64

Just see latest commits in the clj-kondo repo

thanks

@UKFSJSM38 https://github.com/clojure-lsp/clojure-lsp/issues/1415

The script should be fixed @UG00LE5TN LMK otherwise

@UKFSJSM38 I think you forgot that you also have linux aarch64 which doesn't have a static binary

oh yeah, forgot about that

also you should probably update the brew formula to use the static binary on amd64 (only that again)

will do thanks

I just did another release for this in the end since some github actions also downloaded the amd64 dynamic binary and started failing

Yeah, I think I'll do a new one just to avoid that kind of problem

you might want to double-check what ldd --version returns in the 20.04 runners

if it says 2.31 it's good

👍

@UG00LE5TN What does ldd --version return on your system?

ubuntu-20.04 https://github.com/clojure-lsp/clojure-lsp/actions/runs/3670681448/jobs/6205385271 2.31, so it's safe

@U04V15CAJ

robertfw@orion:~$ ldd --version
ldd (Ubuntu GLIBC 2.31-0ubuntu9.9) 2.31
Copyright (C) 2020 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
Written by Roland McGrath and Ulrich Drepper.