lsp--document-highlight fails if textDocument/documentHighlight is not supported #2378
Labels
Comments
|
Willing to open a PR? In general, we are not supposed to call that method at all when |
|
Yes, I would send a PR. I've dug a little bit deeper and I think this is the code path that adds the function: Don't care about the error. Is caused by my debugging code. This is the return value of Does the detection of a not available feature happen after the first request? Must the function get unbound from the hook when this happens? |
This means that highlights are supported.
Can you elaborate? I am interested in the value of |
|
This case is very difficult, but after a long research I've got all pieces together. The core of the problem is desktop-save-mode. In desktop-restore-file-buffer it does: (let* ((auto-insert nil) ; Disable auto insertion
(coding-system-for-read
(or coding-system-for-read
(cdr (assq 'buffer-file-coding-system
desktop-buffer-locals))))
(buf (find-file-noselect buffer-filename :nowarn)))
(condition-case nil
(switch-to-buffer buf)
(error (pop-to-buffer buf)))
(and (not (eq major-mode desktop-buffer-major-mode))
(functionp desktop-buffer-major-mode)
(funcall desktop-buffer-major-mode))It opens the file and if the major-mode do not match the saved major mode, it switches. A few weeks ago, I've installed rustic-mode and lsp-mode via package-list-packages. Rustic sets itself as the major mode for rs files and starts lsp on startup. Some of my open buffers are still using rust-mode. Hence, desktop-mode reopens a file that triggers rustic and lsp. Lsp starts the rls and sets the buffer local variable lsp--buffer-workspaces. But because the major mode mismatch desktop-mode switches to rust-mode and this calls (prog-mode → fundamental-mode →) kill-all-local-variables which removes lsp--buffer-workspaces, while the timer stay active. And the rest of the story is obvious: the timer gets triggered, pulls lsp--buffer-workspaces and ends with error in lsp--send-request-async. At least, I would suggest the following change to detect such missettings, but I'm not really familiar with Emacs and would hope there's a hook where lsp can get a message to unset it's timer or bind the timer to a local variable that becomes killed, too. diff --git i/lsp-mode.el w/lsp-mode.el
index 8d0d5e99..b7b9c2cd 100644
--- i/lsp-mode.el
+++ w/lsp-mode.el
@@ -4284,6 +4284,7 @@ Added to `after-change-functions'."
(equal buffer (current-buffer))
(not lsp-inhibit-lsp-hooks)
lsp-managed-mode)
+ (lsp--cur-workspace-check)
(run-hooks 'lsp-on-idle-hook)))
(defun lsp--on-change-debounce (buffer) |
|
Thank you, this is was a really tricky thing to find! It seems to me that we should use change-major-mode-hook which is called in kill-all-local-variables to call lsp-disconnect. |
|
What do you think about this change? Is it right? diff --git i/lsp-mode.el w/lsp-mode.el
index 63fafe02..dbc63036 100644
--- i/lsp-mode.el
+++ w/lsp-mode.el
@@ -4284,6 +4284,7 @@ Added to `after-change-functions'."
(equal buffer (current-buffer))
(not lsp-inhibit-lsp-hooks)
lsp-managed-mode)
+ (lsp--cur-workspace-check)
(run-hooks 'lsp-on-idle-hook)))
(defun lsp--on-change-debounce (buffer)
@@ -7820,6 +7821,7 @@ argument ask the user to select which language server to start."
(lsp--try-project-root-workspaces (equal arg '(4))
(and arg (not (equal arg 1))))))
(lsp-mode 1)
+ (add-hook 'change-major-mode-hook 'lsp-disconnect t t)
(when lsp-auto-configure (lsp--auto-configure))
(setq lsp-buffer-uri (lsp--buffer-uri))
(lsp--info "Connected to %s." |
|
Looks good to me. I would put it in lsp-managed-mode and add also remove-hook when lsp-managed mode is turned off. |
|
Pushed a fix. Please let me know if it works. Thank you for your analysis once again, it was really helpful. |
|
I'm getting this error with Haskell. It works fine on the first .hs file I open, but every other .hs file in the workspace after that first one, I get this error. I don't use desktop-save-mode, btw. |
are you on the latest melpa version? |
|
Yep. |
|
@spacekitteh I've written up my research for the cause of this bug. https://jo-so.de/2020-12/Emacs-CPU-Verbrauch.html#verbrauchersucheimemacs Maybe you can use a translator to read it. In the end, I used this code in my init.el to analyze what's happening. This creates large logs, so maybe only use lsp-mode in one buffer. (require 'lsp-mode)
(with-eval-after-load 'lsp-mode
(let ((trace-buf "*trace*")
(ctx
(lambda ()
(format
" <+<%s|+|%s|+|%s|+|%s|+|%s>+>"
(buffer-name)
(local-variable-p 'lsp--cur-workspace)
(local-variable-p 'lsp--buffer-workspaces)
lsp--cur-workspace lsp--buffer-workspaces)))
(watcher
(lambda (symbol newval operation where)
(let ((buf (buffer-name)))
(with-current-buffer "*trace*"
(goto-char (point-max))
(insert (format "==>> %s|-|%s|-|%s|-|%s|-|%s|-|%s\n"
buf where symbol
(local-variable-p symbol)
operation (if newval "*val*")))
;; (when (and (not (string-match-p "\.rs$" buf))
;; )
;; (error "open back trace 1"))
))
;; (when (string= operation "makunbound")
;; (error "open back trace 2"))
;; (when (and (eq symbol 'lsp--buffer-workspaces)
;; (string= operation "set")
;; (if newval t))
;; (error "open back trace 3"))
))
)
(get-buffer-create trace-buf)
(trace-function-background 'lsp-feature? trace-buf ctx)
(trace-function-background 'lsp--find-workspaces-for trace-buf ctx)
(add-variable-watcher 'lsp--cur-workspace watcher)
(add-variable-watcher 'lsp--buffer-workspaces watcher)
)) |
|
@spacekitteh do you use |
|
I don't, but I tried anyway, and that didn't fix it (also, the files I'm opening aren't empty). |
|
@spacekitteh are you able to reproduce with |
|
@spacekitteh try removing the haskell workspace and adding it again: |
That didn't work. I did notice this though: and |
I'll try in the next few days. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Describe the bug
I get this error message for a Rust file that uses rls:
This is the backtrace of the error:
Which Language Server did you use
Rust
OS
Linux/Debian, GNU Emacs 27.1, lsp-mode 20201129.1832 (from melpa)
Suggested fix
The text was updated successfully, but these errors were encountered: