[BUG] run_matugen() regenerates the whole palette in light mode when gtk-application-prefer-dark-theme is 'true' instead of '1'

[juanjocop]

Describe the bug

run_matugen() in .config/ml4w/scripts/ml4w-wallpaper decides between light and dark mode with a numeric comparison against gtk-application-prefer-dark-theme:

dotfiles/dotfiles/.config/ml4w/scripts/ml4w-wallpaper

Lines 109 to 113 in 3de368e

	run_matugen() {
	local theme_pref
	theme_pref=$(grep -E '^gtk-application-prefer-dark-theme=' "$HOME/.config/gtk-3.0/settings.ini" | awk -F'=' '{print $2}')
	local mode="light"
	[ "$theme_pref" -eq 1 ] && mode="dark"

theme_pref=$(grep -E '^gtk-application-prefer-dark-theme=' "$HOME/.config/gtk-3.0/settings.ini" | awk -F'=' '{print $2}')
local mode="light"
[ "$theme_pref" -eq 1 ] && mode="dark"

gtk-application-prefer-dark-theme is a GTK boolean, so true/false is an equally valid spelling of 1/0. When the key holds true, [ -eq ] fails:

$ theme_pref="true"; [ "$theme_pref" -eq 1 ] && mode="dark"
bash: line 1: [: true: integer expected

The && chain short-circuits, mode keeps its default light, and matugen regenerates the entire palette in light mode — waybar, window borders, GTK, rofi, swaync, btop, kitty, nwg-dock and the quickshell components all go light on a dark setup. With wallpaper automation enabled it happens unattended, so it looks like an update silently changed the theme.

This is an internal inconsistency, not a missing feature. Two other consumers of the same key in this repo already accept both spellings:

• .config/ml4w/listeners/gtk-theme-switcher.sh — line 49 handles 1/true and 0/false explicitly, and even warns on unexpected values:

  if [[ "$THEME_PREF" == "1" || "$THEME_PREF" == "true" ]]; then

• .config/hypr/scripts/gtk.sh — line 28 treats only 0/false as light:

  if [[ $prefer_dark_theme == "0" || $prefer_dark_theme == "false" ]]; then

So of the three places that read this key, only run_matugen() was left with the numeric test.

The dotfiles ship gtk-application-prefer-dark-theme=1, which is why this doesn't show on a clean install. It surfaces once any third-party GTK settings tool normalizes settings.ini to the boolean spelling.

Steps to reproduce

On my machine the key was flipped to true by kde-gtk-config, which is part of the default CachyOS package set even in a Hyprland session. It rewrote ~/.config/gtk-3.0/settings.ini and ~/.config/gtk-4.0/settings.ini into KDE's format — both files and ~/.config/xsettingsd/xsettingsd.conf share the same mtime to the second, and the values it left behind are KDE's (breeze, breeze_cursors, ocean). I did not determine what invoked it, so I can't give exact steps for that part.

The bug itself is independent of that trigger and can be reproduced directly:

1. Confirm the current palette is dark: grep '"background"' ~/.config/ml4w/colors/colors.json
2. Set the key to its boolean spelling:
   sed -i 's/^gtk-application-prefer-dark-theme=1/gtk-application-prefer-dark-theme=true/' ~/.config/gtk-3.0/settings.ini
3. Re-run the wallpaper script:
   ~/.config/ml4w/scripts/ml4w-wallpaper "$(cat ~/.cache/ml4w/hyprland-dotfiles/current_wallpaper)"
4. It logs Running matugen in light mode, and colors.json comes back light ("background": "#faf8ff" on my source wallpaper, vs #121318 before).

Expected behavior

run_matugen() should treat true the same as 1, consistently with gtk-theme-switcher.sh and gtk.sh, and keep generating a dark palette.

A one-line change is enough:

case "$theme_pref" in 1 | true) mode="dark" ;; esac

I verified this against all four values: 1 → dark, true → dark, 0 → light, false → light, with no stderr output. After applying it, the script logs Running matugen in dark mode and the palette returns to its previous values.

Screenshots

No response

Distribution

Other (please specify below)

If Other, specify

CachyOS (Arch-based), ML4W OS 2.12.3, Hyprland 0.56.1

Additional context

There is a second-order effect worth flagging: while the key holds true, .config/ml4w/scripts/ml4w-toggle-theme cannot switch modes either. It tests grep -q "gtk-application-prefer-dark-theme=1", does not match true, falls through to the else branch, and runs sed 's/gtk-application-prefer-dark-theme=0/...=1/', which does not match either. It prints Switched to dark theme while leaving the GTK3 file unchanged — so the user cannot correct the state from the UI, which is what makes the palette change feel permanent.

I searched open and closed issues before reporting (prefer-dark-theme, matugen, light theme, run_matugen, settings.ini, kde-gtk-config) and did not find this. #1596 looked close but is the opposite direction and was fixed in 2.13.0.

I have no strong preference on the fix; the case above is just the smallest change that matches what the other two scripts already do. Happy to open a PR if that helps.

Code of Conduct

• I agree to follow this project's Code of Conduct

[juanjocop]

Correction and follow-up on my own report.

#1759 already fixes the run_matugen() half of this, and has been open since 22 July. I missed it when I filed this because I searched issues but not pull requests — apologies for the noise. Its change is the same one I suggested above, so please treat that PR as the fix for the main problem and this issue as background for it.

I verified #1759 independently on a different setup from the author's (CachyOS + Hyprland 0.56.1, where the trigger was kde-gtk-config rather than a manual change) and it resolves the light-palette regeneration.

One thing that came up while checking: there is a third consumer of this key, conf/hypr/scripts/gtk.sh (line 28), and it already handles both spellings too. So run_matugen() was the only one of the three left with the numeric test.

The ml4w-toggle-theme part of this issue is still unfixed and is not covered by #1759, which only touches ml4w-wallpaper. I opened #1766 for it. That one matters for recovery: with only #1759 merged, the palette regenerates correctly again, but a user already sitting on true still cannot toggle modes from the UI.