fix(lsp-mode.el): Make sure modules will be loaded

[jcs090218]

By looking at the macro require's description:

If the optional third argument NOERROR is non-nil, then, if
the file is not found, the function returns nil instead of signaling
an error.  Normally the return value is FEATURE.

It seems like it only check file not found condition; I think it's better if we ignore all conditions so one lsp packages break won't break all of them.

[yyoncho]

But that way if the package that you want to use is broken you won't be able to understand why it doesn't load.

[jcs090218]

But that way if the package that you want to use is broken you won't be able to understand why it doesn't load.

I have some thoughts on this issue,

1. That wouldn't cause too many hassles since this will only ignore errors during the requiring time. And loading a bunch of modules at a time doesn't help you isolate the issue (unless you have debug-on-error on), and it will cause confusion when you are trying to figure out what is broken in the chain. I prefer ignoring it here, and debugging it by evaluating (require 'xxx) directly. I don't think one extra lsp package would hamper one another is a good idea.
2. Add a debug flag, maybe a flag for developer/development mode?

;; just an example
(if lsp-debug
    (require package nil t)
  (ignore-errors (require package nil t)))

3. The issue detector tasks should hand it over to CI. Something like following is a good start,

(let ((lsp-debug t)
      (lsp-auto-configure t))  ; this is enabled by default
  (lsp))  ; start the requiring test

All of them are design decisions. WDYT?

[yyoncho]

Have you observed packages that fail when you so (require 'lsp-xxx)? Because this is an indication of a bug. I see that some packages are doing executable-find on top level for example - we should remove these instances.

[jcs090218]

lsp-metals is giving me several errors depends on your Emacs version,

Emacs 29+ snapshot (after emacs-mirror/emacs@80cf13a)

Eager macro-expansion failure: (wrong-number-of-arguments (1 . 1) 0)

Alexander-Miller/treemacs#982

Emacs 28.2 or earlier (before emacs-mirror/emacs@80cf13a)

load-with-code-conversion: Symbol’s function definition is void: treemacs-define-leaf-node

Alexander-Miller/treemacs#987

we should remove these instances.

You prefer to remove lsp-metals for now?

The worst-case scenario is me removing lsp-metals from my configuration which I don't want to do since I expect these bugs will be fixed one day. 😅 (mjhoy/dotfiles@0ee2c23)

[yyoncho]

We should fix metals.

[jcs090218]

Got it! I think there is no reason for this patch anymore! Closing this now!