Add one-step Japanese input setup

[komagata]

Summary

• add Setup > Input Method > Mozc (Japanese) to the Omarchy menu
• install fcitx5-mozc, restart Fcitx 5, and append Mozc to the current input-method group
• document the setup and add shell plus desktop acceptance coverage

Why

Japanese input is not optional for Japanese-speaking users. Windows and macOS finish setup with a working Japanese IME, while Omarchy currently requires users to discover the package and configure Fcitx 5 manually.

This keeps the first change deliberately small: it adds a one-step setup action rather than changing the main installer.

Scope

This PR intentionally treats input-method setup separately from locale and UI internationalization. These concerns may be connected during initial setup, but users may need Japanese input regardless of their system locale or UI language.

Keeping this change focused also makes it possible to improve the input experience incrementally without first deciding the broader locale and i18n architecture.

Naming the menu item after the input engine rather than equating Japanese input with Mozc also leaves room to add options such as Hazkey (Japanese) or Rime (Chinese) alongside it in the future.

Testing

• bash test/shell.d/input-japanese-test.sh
• bash test/shell.d/menu-test.sh
• bash test/shell.d/menu-guards-test.sh
• ./test/cli
• QEMU/KVM desktop acceptance: installed Mozc, registered it with Fcitx 5, and converted nihongo to 日本語 in a real terminal

The broader VM suite still has unrelated existing environment/OCR failures; the new Japanese input acceptance test passes end to end.

[akitaonrails]

Ran this on Omarchy 4.0.4 (fcitx5 5.1.22, Hyprland 0.56.2) against an install that already had Mozc registered: idempotent, exit 0, no group rewrite. The D-Bus registration is the right call — editing profile directly loses the change, because fcitx5 rewrites that file from live state when it exits.

Two small things from using it, plus one gap that bites right after it succeeds.

fcitx5-remote --check prints the input state to stdout. It writes a digit (1 here), so the readiness loop leaks it into the presentation terminal before the success message. >/dev/null covers it:

if fcitx5-remote --check >/dev/null 2>&1; then

#11719 hit the same thing in the Korean sibling.

fcitx5-configtool is hidden from the launcher. The manual edit still points other-language users at it:

install an input engine like fcitx5-chinese-addons with omarchy pkg add, plus fcitx5-configtool to add the engine to your input methods

default/omarchy/launcher.hides:11 lists fcitx5-configtool, so it never appears in the launcher after installing. Worth saying "run fcitx5-configtool from a terminal" so people do not go looking for it — that was a few confused minutes here.

Chromium ignores the IME unless --enable-wayland-ime is set. This is outside your scope, but it lands immediately after your script succeeds: the engine is installed and registered, every GTK/Qt app and terminal accepts Japanese, and the default browser silently does not. I opened #12139 to add the flag to config/chromium-flags.conf with a migration for existing installs, since the per-browser flags files are only copied at install time.

Unrelated to the setup path, for anyone who finds this thread: once input works in Chromium, the candidate window is misplaced there (#7559), and the popup is invisible in Hyprland true fullscreen (#11303).

For what it is worth, the shape of this PR — one menu action, engine named rather than language — reads right. A sibling omarchy-theme-set-fcitx5 in #12142 makes the candidate window follow the active Omarchy theme, which pairs with this.

[komagata]

@akitaonrails Thank you so much for testing this on your setup and taking the time to write such detailed feedback! It’s especially helpful to have confirmation that running it with Mozc already registered leaves the group unchanged.

I’ve addressed both points in f903b4e:

• Suppressed the output from fcitx5-remote --check.
• Clarified that fcitx5-configtool should be run from a terminal, since it’s hidden from the launcher.

I also updated the test to catch the stray output. The related shell tests and CLI suite pass.

Thanks as well for opening #12139 and pointing out the related issues and theme integration. I really appreciate you looking beyond setup to what people encounter when they start using Japanese input.