snixembed registered as the StatusNotifierWatcher but its
RegisterStatusNotifierItem implementation didn't accept SNI publishers
— Telegram, Vesktop, blueman-tray, and a libayatana-appindicator test
script all failed to register, leaving waybar's tray empty.
Removing snixembed lets waybar's tray module register itself as both
host AND watcher (when no other watcher exists). After this change,
all three SNI items registered immediately.
Trade-off: legacy XEmbed-based Wine tray icons no longer bridge to
SNI. The Wine System Tray host window rule is kept (still hides the
empty window off-screen) and the comment updated.
- niri/config.kdl: drop snixembed spawn-at-startup; update Wine rule
- packages.txt: drop snixembed (no longer used)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The niri 'environment' block only affects processes niri spawns
directly. Apps launched via dbus activation or systemd-user (most
.desktop launches go through one of these) get systemd-user's env,
which still has bare 'niri' — set by niri-session BEFORE niri itself
runs.
Adding a one-shot spawn-at-startup that re-imports XDG_CURRENT_DESKTOP
from niri's env into both systemd-user and dbus-activation, after
niri starts. Future dbus-activated apps will see niri:GNOME.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Many tray libraries (Electron's, Qt's StatusNotifier) gate icon
creation on a recognised XDG_CURRENT_DESKTOP value. Bare "niri"
makes them silently skip tray creation. Adding ":GNOME" as a
fallback identifier keeps niri primary while letting apps create
their tray icons.
Vesktop and Telegram are confirmed to want tray icons but were
not creating them; this fix unblocks both.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Both daemons were running but not registering SNI items on niri/Wayland,
leaving the waybar tray slot permanently empty. Network is already
accessible via the waybar network module (left-click → networkmanager_dmenu).
Bluetooth GUI (blueman-manager) is still available on demand; only the
applet is dropped.
Tray module stays configured so Electron apps (Discord/Vesktop, Steam,
etc.) still get a tray slot when they're running.
- niri/config.kdl: drop nm-applet and blueman-applet spawn-at-startup
- packages.txt: drop network-manager-applet (nothing else uses it);
keep blueman package for blueman-manager GUI
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Aligns the third sleep path (power menu) with the lid handler and
the swayidle 30-min timer — all three now suspend immediately and
hibernate after HibernateDelaySec (30 min).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Spec said `findmnt -no UUID /` and `resume=UUID=` but the script
correctly uses PARTUUID throughout (matches the system's existing
root=PARTUUID= convention).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
If `findmnt -no PARTUUID /` returns nothing (root on LVM, no GPT,
unusual mount state) the script would silently write a broken
resume=PARTUUID= line to /etc/kernel/cmdline. Bail with an error
message instead.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Flameshot's portal-based capture path on niri triggers an
xdg-desktop-portal access prompt on every invocation, blocking
the screenshot itself. grim talks to wlr-screencopy directly and
never touches the portal, so no prompt.
Pipeline: slurp (region) → grim (capture) → satty (annotate +
copy + save). satty replaces the wofi action menu by being the
post-capture surface itself.
- scripts/screenshot.sh added
- install.sh symlinks it; flameshot.ini generator removed
- niri Mod+Print → "screenshot"
- packages.txt drops flameshot
- docs tables updated
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Both plans were generated during their respective implementation
runs but never committed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The "Tooling" tables in README.md and ARCHITECTURE.md still listed
grim + slurp after the flameshot swap.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The progress window for copy/move/delete operations is small and
transient — floating it stops it from taking over a column.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
map_profile_to_strategy never writes to stderr — the suppression was
silencing nothing real and would mislead a future reader.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Trailing commas added throughout and modules-right expanded to one
key per line. Semantically identical — waybar's JSONC parser accepts
trailing commas.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
starship is required by fish/config.fish (starship init fish | source)
but was missing — fresh installs would have a broken shell prompt.
swww was listed but the actually-installed wallpaper daemon is awww
(matching the niri/config.kdl change in the previous commit).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- swww-daemon → awww-daemon (swww is not installed; awww 0.12.0-1 is)
- polkit-gnome-authentication-agent-1 → absolute path; the binary
lives in /usr/lib/polkit-gnome/ which is not on PATH
Both daemons were silently failing to start at every login.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
scripts/autostart.sh was never symlinked or referenced — niri's
spawn-at-startup directives cover everything it launched.
mako/config is gitignored (install.sh generates it from
theme/colors.json) but was still tracked from the initial commit.
The stale ~/.config/mako/config symlink that pointed at the repo
caused install.sh's writes to land back in the repo. Removed both
so install.sh now correctly produces a real generated file in
~/.config/mako/.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The tray drawer added complexity (drawer config, custom chevron module,
state-feedback CSS) for what amounts to a small visual refinement.
Removed in favour of the original always-visible tray. The mouse-battery
indicator was also removed from the bar.
Deletes:
- waybar/mouse-battery.sh + symlink + install.sh entry
- group/tray-drawer + custom/mouse-battery blocks in config.jsonc
- #custom-tray-arrow rules from style.css
- docs/superpowers/{specs,plans} for the drawer feature
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
scripts/waybar-restart.sh: kills waybar, waits for it to exit, then
relaunches detached via setsid. Symlinked into ~/.local/bin and bound
to Mod+Shift+B in niri.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Plan was originally written assuming `transform: rotate` would work
in GTK 3 CSS; it doesn't. Spec was already updated in 1646dfc; this
brings the plan in line so historical readers see the same approach
across spec, plan, and code.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
No CSS rule targets .tray-drawer-child — speculative key with no
consumer. YAGNI cleanup.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
GTK 3 CSS (used by waybar) treats `transform` as a fatal parse error,
so the rotation approach killed waybar entirely. Use a color shift
from @tn-fg-muted to @tn-fg as the expanded-state feedback instead.
Matches the existing style.css convention (battery/cpu/temperature
states all use color shifts, not transforms).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
waybar 0.15.0's clock module reads PST instead of PDT, showing time
an hour behind. Replace with custom/exec calling date directly.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
funman300