-
|
Hi, I'm facing a strange issue that only global
Minimal configuration (mandatory);;; emacs -Q -l ~/.config/emacs/test.el
(setq package-archives '(("melpa" . "https://melpa.org/packages/")
("org" . "https://orgmode.org/elpa/")
("gnu" . "https://elpa.gnu.org/packages/")))
(package-initialize)
(package-install 'eglot)
(require 'eglot)
(require 'tramp)
(add-to-list 'tramp-remote-path 'tramp-own-remote-path)
(add-hook 'python-mode-hook
(lambda ()
(setq-local eglot-workspace-configuration
'((:python . (:venvPath "/data/wangtianshu/.local/share/conda/envs"))))
(eglot-ensure)))manually run If I use global eglot-workspace-configuration, it takes effect. ;;; emacs -Q -l ~/.config/emacs/test.el
(setq package-archives '(("melpa" . "https://melpa.org/packages/")
("org" . "https://orgmode.org/elpa/")
("gnu" . "https://elpa.gnu.org/packages/")))
(package-initialize)
(package-install 'eglot)
(require 'eglot)
(require 'tramp)
(add-to-list 'tramp-remote-path 'tramp-own-remote-path)
(add-hook 'python-mode-hook
(lambda ()
(setq-default eglot-workspace-configuration
'((:python . (:venvPath "/data/wangtianshu/.local/share/conda/envs"))))
(eglot-ensure)))LSP transcript - M-x eglot-events-buffer (mandatory unless Emacs inoperable)important diff part I think (remove the different of id and time) of two events buffer (global vs local) setq-local:setq-globalBacktrace (mandatory, unless no error message seen or heard):No. |
Beta Was this translation helpful? Give feedback.
Replies: 8 comments 14 replies
-
|
I can reproduce this, and what's more, I can reproduce this by setting |
Beta Was this translation helpful? Give feedback.
-
|
Stepping through the E.g., if I'm visiting the file (defun eglot--uri-to-path (uri)
"Convert URI to file path, helped by `eglot--current-server'."
(when (keywordp uri) (setq uri (substring (symbol-name uri) 1)))
(let* ((server (eglot-current-server))
(remote-prefix (when (and server
(file-remote-p
(project-root (eglot--project server))))
(project-root (eglot--project server))))
(retval (url-filename (url-generic-parse-url (url-unhex-string uri))))
;; Remove the leading "/" for local MS Windows-style paths.
(normalized (if (and (not remote-prefix)
(eq system-type 'windows-nt)
(cl-plusp (length retval)))
(substring retval 1)
retval)))
(concat remote-prefix normalized)))I filed #879 to change this behavior, but I'm not sure there isn't a deeper bug here |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your analysis, Andreas!
I'll have a look at your PR
…On Sun, Mar 13, 2022, 17:06 Andreas Fuchs ***@***.***> wrote:
Stepping through the eglot-handle-request method for
workspace/configuration, I think I see why this doesn't set local
variables from the .dir-locals.el file, at least: The form (eglot--uri-to-path
scopeUri) on a remote server doesn't return the path to the project - it
only returns the path to the TRAMP prefix.
E.g., if I'm visiting the file
/scpx:ferdl.hack:/home/asf/Hacks/cloud/infra/backend.py, that form
returns /scpx:ferdl.hack: instead of (what I'd expect, given the
description and behavior in local files)
/scpx:ferdl.hack:/home/asf/Hacks/cloud. Indeed, manually patching the
function to return the full project prefix instead of just the remote part
does apply local variables set via .dir-locals.el correctly:
(defun eglot--uri-to-path (uri)
"Convert URI to file path, helped by `eglot--current-server'."
(when (keywordp uri) (setq uri (substring (symbol-name uri) 1)))
(let* ((server (eglot-current-server))
(remote-prefix (when (and server
(file-remote-p
(project-root (eglot--project server))))
(project-root (eglot--project server))))
(retval (url-filename (url-generic-parse-url (url-unhex-string uri))))
;; Remove the leading "/" for local MS Windows-style paths.
(normalized (if (and (not remote-prefix)
(eq system-type 'windows-nt)
(cl-plusp (length retval)))
(substring retval 1)
retval)))
(concat remote-prefix normalized)))
—
Reply to this email directly, view it on GitHub
<#876 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAC6PQ25PT3WVPVJLA7ZSYTU7YVCBANCNFSM5QSYHKVQ>
.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
|
@joaotavora Should this move to issue as it have been confirmed by @antifuchs |
Beta Was this translation helpful? Give feedback.
-
|
It's good that the issue has been reproduced by someone else, but that doesn't mean that I have enough information to reproduce it reliably and easily. It doesn't mean your problem is i being ignored, only that it's not something that should be in the "issues" basket. |
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
This is being handled in #879. I don't use Tramp, so fixing someone else's problem is done on a "best effort" basis. In particular I can't do testing because noone has provided a short from-scratch recipe small enough so that I can witness and fix the problem myself. Until someone does, there are some commits in the aforementioned PR that seem to solve the problem, so that you don't need, presumably, to use |
Beta Was this translation helpful? Give feedback.
-
|
Since you don't use tramp, I don't quite understand how I can provide a short from-scratch recipe. |
Beta Was this translation helpful? Give feedback.
-
|
What I mean is that I don't use tramp for my daily work, I have no need for it. I'm not ideologically opposed to using it. But if you give me that short recipe, I have no problem in following it. Then, during the time that I will be following your recipe, I will be using tramp. In other words, you give the recipe 📄 , I bake the cake 🧑🍳 , I fix the cake 🧑🔧 so it's more delicious 😋 and then you eat the cake! 🍰 🍰 🍰 All for free, I think it's a fair bargain 😉 Anyway, I wasn't talking about you in particular. There are many situations in this bug tracker where people facing problems tend to forget that to fix a problem i usually have to reproduce it locally. Otherwise my brain is usually not powerful enough to confidently guess where the problem is and propose a fix that I'm sure won't create new problems. I do try, as I have done here, but it's not very effective. I spend a lot of time reading people's logs, scrolling up and down in discussion threads, trying to guess what this person means when writing this or that. Running experiments with that person. It's fun to a certain point, but when you're very busy with other stuff, not so much. Above all, it's not very effective. If the bug reporter does the work of putting together a recipe that will almost 100% guarantee that if I follow it from start to finish I will see the same problem that the bug reporter is seeing, things are made much easier. It helps if the recipe doesn't force me to install Windows, or enormous Java toolchains. But I guess a python recipe, or a clangd recipe and some kind of local ssh server (or a shared playground ssh server) would do nicely. I think tramp even has a sudo method where you can access files of different users, maybe that is enough to reproduce it. |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your reply, in fact often we are not sure what is needed in this recipe. Can you see if this is enough (now I can reproduce it without tramp). |
Beta Was this translation helpful? Give feedback.
-
|
Reproduce steps without tramp:
mkdir test_project
cd test_project
git init
touch test.py
;;; emacs -Q -l ~/.config/emacs/test.el
(setq package-archives '(("melpa" . "https://melpa.org/packages/")
("gnu" . "https://elpa.gnu.org/packages/")
("nongnu" . "https://elpa.nongnu.org/nongnu/")))
(package-initialize)
(package-install 'eglot)
(require 'eglot)
(require 'tramp)
(setq native-comp-async-report-warnings-errors nil)
(add-to-list 'tramp-remote-path 'tramp-own-remote-path)
(add-hook 'python-mode-hook
(lambda ()
(setq-local eglot-workspace-configuration
'((:python . (:venvPath "/Users/wangtianshu/.local/share/conda/envs"))))
(eglot-ensure)))
;;; emacs -Q -l ~/.config/emacs/test.el
(setq package-archives '(("melpa" . "https://melpa.org/packages/")
("gnu" . "https://elpa.gnu.org/packages/")
("nongnu" . "https://elpa.nongnu.org/nongnu/")))
(package-initialize)
(package-install 'eglot)
(require 'eglot)
(require 'tramp)
(setq native-comp-async-report-warnings-errors nil)
(add-to-list 'tramp-remote-path 'tramp-own-remote-path)
(add-hook 'python-mode-hook
(lambda ()
(setq-default eglot-workspace-configuration
'((:python . (:venvPath "/Users/wangtianshu/.local/share/conda/envs"))))
(eglot-ensure)))
As we can see: local |
Beta Was this translation helpful? Give feedback.
-
|
See also the very closely related #967. The behavior there occurs locally, no tramp involvement needed. I believe it is this (cl-defmethod eglot-handle-request
(server (_method (eql workspace/configuration)) &key items)
"Handle server request workspace/configuration."
(apply #'vector
(mapcar
(eglot--lambda ((ConfigurationItem) scopeUri section)
(with-temp-buffer |
Beta Was this translation helpful? Give feedback.
-
|
This is by design. You're supposed to set eglot-workspace-confuguration
project-locally, because that's what it is, a project-local setting
(project ~= workspace).
And the only current way to do that is either globally or through
dir-locals.el.
The real problem of that other issue is that tramp directories aren't --
apparently -- being correctly calculated.
…On Fri, Jul 15, 2022, 18:09 JD Smith ***@***.***> wrote:
See also the very closely related #967
<#967>. The behavior there
occurs locally, no tramp involvement needed. I believe it is this
with-temp-buffer that causes the "falling back on the default value"
issue:
(cl-defmethod eglot-handle-request
(server (_method (eql workspace/configuration)) &key items)
"Handle server request workspace/configuration."
(apply #'vector
(mapcar
(eglot--lambda ((ConfigurationItem) scopeUri section)
(with-temp-buffer
—
Reply to this email directly, view it on GitHub
<#876 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAC6PQY5L3ZNUSOHUJ33H2DVUGLLDANCNFSM5QSYHKVQ>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
Sometimes this isn't possible or desirable, for example with other packages that alter the virtual environment dynamically, as in #967. In those cases, it would be very helpful to be able to alter the workspace config "just in time" as eglot is spinning up a server from some particular buffer. Is the |
Beta Was this translation helpful? Give feedback.
-
Alright but that doesn't change the fact that using a buffer-local value of But setting a buffer-locally is still wrong. When the server asks the client "what is your configuration for this project you declared" it doesn't say "...and by the way, I'm asking for this file specifically". And before you say "oh, just pick any file and grab that value", remember that maybe all the files are closed, i.e. no active buffers for a given active Eglot management session. What would you do then? It doesn't make sense to have this buffer-locally. |
Beta Was this translation helpful? Give feedback.
-
|
@tshu-w that is a very good reproduction recipe that I'm willing to try if you adapt it slightly for the TRAMP/.dir-locals case. Let me just confirm some things
|
Beta Was this translation helpful? Give feedback.
-
|
Great, I guess. What works exactly? Tramp + python workspace things? |
Beta Was this translation helpful? Give feedback.
-
|
@joaotavora Yes, tramp + python workspace things. |
Beta Was this translation helpful? Give feedback.
-
I claim you do want some variability based on the singular & special buffer from which the lsp server is started. For example, I have a small package which updates the path to the python executable, setting up a virtual environment, for each new python buffer opened. This runs before eglot is started. There are many tools which do something like this. I want any LSP server started from there to have the correct venv configured. A function call e-w-c, which could examine default directory, could also work well for this.
Doesn't eglot consider the current buffer when starting the server, examining the buffer's mode for example to pick a server command to run? So it already does consider "one file from the project". The config functionality people are after is in that same vein: on first start of a server, setup not just the server command, but alter some of its initial config as well. Or do you mean, having started a server from a buffer, you later close all buffers from the associated workspace? And then for some reason the already running server (sans buffers) asks again for the configuration? I can certainly see how that would be a problem if the correct information is kept in (now lost) buffer-local state.
I can see the argument here. And I now understand why eglot processes requests in a temp buffer with the .dir-locals applied. A function call version of BTW, eglot declaring the variable with |
Beta Was this translation helpful? Give feedback.
-
Good.
You can't have directory-local variables without buffer-local varlues, so you're looking at an implementation detail. I'll work on some more documentation and overhaul this variable to be a function to address #967. Someone also complained that the alist/plist syntax is confusing, which i agree, so maybe I'll create a new variable that addresses both problems and obsolete this one. |
Beta Was this translation helpful? Give feedback.
-
|
Great, thanks. FYI, here's how I'm currently setting it (in python-mode-hook, before eglot startup): (if-let ((venv (simple-venv-find))
((not (eq venv 'none))))
(setq-default eglot-workspace-configuration
(list (cons :python
(list :venvPath venv
:pythonPath (simple-venv-interpreter venv))))))Some of the confusion I had:
Most of this is not eglot's issue per se, but comes from ambiguity/server-specificity. Once the new workspace-config variable settles, perhaps soliciting some wiki contributions of "common" workspace/configuration settings for different servers would be good. |
Beta Was this translation helpful? Give feedback.
-
|
After trying it out in-depth today I found that it now works fine without the hack. |
Beta Was this translation helpful? Give feedback.
After trying it out in-depth today I found that it now works fine without the hack.