Possible regression when calling evil-collection-*-setup

[gagbo]

It seems that de0b62b brought a regression

I triggered this by requiring 'profiler, which in turn triggered the evil-collection-setup call, which then triggered either the key definition hook or just the normal setup?

This error gets triggered by anything that's handled by evil-collection it seems.

Debugger entered--Lisp error: (cl-no-applicable-method seq-empty-p (normal))
  signal(cl-no-applicable-method (seq-empty-p (normal)))
  cl-no-applicable-method(#s(cl--generic :name seq-empty-p :dispatches nil :method-table nil :options nil) (normal))
  apply(cl-no-applicable-method #s(cl--generic :name seq-empty-p :dispatches nil :method-table nil :options nil) (normal))
  seq-empty-p((normal))
  seq-reduce(#f(compiled-function (acc elt) #<bytecode -0x542a0d6d15ae114>) (normal) nil)
  seq-difference((normal) nil)
  evil-collection--filter-states(normal)
  evil-collection-define-key(normal profiler-report-mode-map " " scroll-up-command [33554464] scroll-down-command [delete] scroll-down-command "j" profiler-report-next-entry "k" profiler-report-previous-entry [tab] profiler-report-toggle-entry "o" profiler-report-ascending-sort "O" profiler-report-descending-sort "c" profiler-report-render-calltree "C" profiler-report-render-reversed-calltree "i" profiler-report-describe-entry "=" profiler-report-compare-profile "\15" profiler-report-find-entry "gr" revert-buffer "q" quit-window "ZQ" evil-quit "ZZ" quit-windw)
  evil-collection-profiler-setup()
  #f(compiled-function () #<bytecode 0x88847b3ea7f1ffb>)()
  eval-after-load(profiler #f(compiled-function () #<bytecode 0x88847b3ea7f1ffb>))
  evil-collection-init((profiler))

In any case, I don't have the issue anymore when I check out 1ad283f so I assume it's a quoting issue in the change to evil-collection-translate-minor-mode-key ? Couldn't dig more into that, I'm just sticking to a commit that works for now.

[jojojames]

Was there a need for de0b62b @condy0919 ?

Might be worth reverting if it was just a code cleanup-esque change.

[condy0919]

Was there a need for de0b62b @condy0919 ?

Might be worth reverting if it was just a code cleanup-esque change.

Sorry, my computer is broken on Sept 15. Could you help fix that? The change is to make CI happy as evil-delay is marked obsolete and CI always complains.

[gagbo]

I suppose the issue is on the quoting of the condition somehow? I struggle a bit to see how this commit could be the issue, but so far, reverting to a commit before that seemed to fix my issues.

[condy0919]

Could you provide a reproducible recipe? I'm using Emacs 30, the latest evil and evil-collection without errors.

[gagbo]

Do you have a setup-hook for evil-collection? It's a bit hard to provide a 100% reproducible recipe because there are a lot of things that can trigger this regression, but maybe just having

(defun layout-foo-rotate-keymaps (_mode mode-keymaps &rest _rest)
  "A setup hook"
    (evil-collection-translate-key '(normal motion visual operator) mode-keymaps
      "p" "h"
      (kbd "C-p") (kbd "C-h")
      :destructive nil))

(add-hook 'evil-collection-setup-hook #'layout-foo-rotate-keymaps)

In your initialization code (+ maybe loading dired) could trigger a void-variable dired-filter-map-mode error – as long as you don't have dired-hacks installed.

It's pretty hard to find a good repro for this because I rely on this setup-hook and translate-key/translate-minor-key in a lot of code; and it seems that the error only triggers if some specific "which packages are already loaded?" conditions that I didn't really investigate too much. At least the PR seems to calm everything down, and removes the deprecation warning by backporting the backwards-compatible macro. Maybe the problem is that all the hook functions in evil-with-delay all share the same name, so that anytime one keymapp condition is true, all the hooks for every package trigger? I don't know, just played with macroexpand-step to understand that something's wrong with the evil-with-delay version, and sticking to evil-delay seems to fix my issues