Four separate per-monitor theming bugs in the window-bind module, plus
the first tests the module has ever had.
ENTER_HOOKED never cleaned up on destroy (#2). It's keyed by GdkSurface
pointer and inserted once per surface, but ensure_destroy_cleanup only
removed the widget-pointer-keyed sets (DESTROY_HOOKED / AUTO_HOOKED). A
closed window left its ENTER_HOOKED entry forever, so a later window
whose surface was allocated at the same address hit the `already` early-
return in attach_enter_monitor and silently never re-themed when moved
between monitors. Record widget->surface in ENTER_SURFACE at hook time so
the single connect_destroy path (now forget_bind, also folding in the
previously separate MAP_HOOKED cleanup) can clear the enter hook by its
real key.
Subtrees grown deeper than one level after bind never got the sheet (#3).
watch_root_children only observed the root's *direct* children, and a
style-context provider isn't inherited by children in GTK4 -- so a
popover's contents, menu items, or rows appended into an existing box
after bind/map rendered with the display-wide shared sheet (the wrong
monitor's colours). Replace it with hook_subtree: recursively observe
every current container and, from inside items-changed, re-walk the tree
and hook any newly-spliced-in containers too. All the child-models are
parked in the WidgetBind so they drop together on destroy.
One-shot watch arming (#5). ensure_themes_watch / the apply_shared /
apply_app_css watchers stored None on a monitor_directory failure; the
watch was only ever re-armed as a side effect of another bind_window*
call, so a consumer that binds exactly once (bread-polkit) lost per-
monitor reload permanently after a single transient failure. Add
arm_or_retry: a bounded lazy retry (5 attempts, 2s apart) guarded so a
burst of binds doesn't each spawn a timer.
Palette-only writes didn't reach binds (#6). Reloads keyed strictly on
.css events under themes/; writing only palettes/<out>.json left bound
windows stale (bites third-party tooling -- the CLI writes both). Also
watch palettes/ for .json; the reload path re-reads the JSON via
load_palette_for regardless, so reacting to either file is correct (a
paired write just reloads twice, harmlessly).
Tests: the module was feature-gated behind `gtk` and had zero coverage
(the audit noted #2/#3/#5 are "exactly what tests would have caught").
Adds pure-logic tests for forget_bind's surface-keyed cleanup, the
css/json reload routing, and watch arming; plus one #[ignore]d
display+GTK test for the deep-subtree re-attach and bind lifecycle
(gtk4::init() binds GTK to one thread for the process and pumps the glib
main context, which deadlocks the shell::hotreload tests -- so it can't
run in the shared binary; it still compiles every build).
The old sanitiser mapped '/', ':' and space -- and every other unsafe
byte -- all to '_', so "a/b" and "a:b" both became "a_b" and silently
shared one palettes/<name>.json + themes/<name>.css, last writer wins.
No collision detection at all.
Percent-encode every byte outside [A-Za-z0-9._-] as %XX instead. That is
injective, so two connectors can never land on the same file. '%' itself
isn't in the keep-set so it encodes to %25 and the mapping stays
reversible in principle (nothing decodes today -- reload_binds_for_
sanitized only compares sanitize_output(name) against the on-disk stem).
Real Hyprland connectors (eDP-1, HDMI-A-1, DP-2) are already only
keep-set bytes, so their filenames are unchanged; the files are
session-scoped under XDG_RUNTIME_DIR and regenerated at login anyway.
stylesheet_resolved() is loaded *only* by the per-monitor bind path
(gtk::bind_window* / reload_binds_for_sanitized), which attaches it as a
widget-tree CssProvider carrying one output's palette. It already hex-
inlines every rule body, but it still emitted the leading @define-color
accent/bg/on-* block verbatim.
@define-color in GTK4 is stylesheet-global -- not provider- or subtree-
scoped (confirmed by GTK's CSS maintainers). So every bind and every
per-output reload redefined the display-global @accent/@bg/@on-* to that
monitor's palette, and any CSS parsed afterwards in another window --
apply_shared in every other bread app, apply_app_css, apply_user_css --
resolved its named colours against whichever monitor bound last. That is
exactly the "wrong monitor's accent leaks" failure the hex-inlining was
added to prevent, just displaced out of the bound tree.
Drop every @define-color line from the _resolved output: the rules are
fully self-contained hex, so the block was pure liability there. The
display-global sheet (stylesheet() / render(), loaded at APPLICATION
priority) keeps its block -- that is the one provider meant to own those
names. Tightens the existing regression test to assert the block is now
actually absent, and adds one guarding the display-global sheet still
carries it.
The gap that let the shell-theme breadbar ship half-verified: a builtin
theme.toml can parse in isolation but fail `resolve_theme` at runtime (a
slot name the manifest no longer knows, a field rename, a broken
`extends`), and `load()` silently falls back to liquid-motion. breadbar
has no tracing subscriber so that fallback's `warn!` goes nowhere — the
shell just quietly ignores `active = "spotlight"`.
Three tests over `builtin::ALL` (not a hand-maintained id list):
- every_builtin_theme_resolves: load_named() succeeds, id/name match, has
window anchors
- active_theme_id_env_selects_each_builtin_and_never_falls_back: BREAD_SHELL_THEME
set to each id → load() returns that id, not the fallback
- unknown_theme_id_falls_back_to_liquid_motion: the fallback still works
Adding a fifth builtin now needs its theme.toml to be loadable, not just
parseable.
- `XDG_CONFIG_HOME` / `XDG_RUNTIME_DIR` are honoured only when absolute
(XDG spec; matches `bread_utils::xdg`), in layerrules and output.
- `output::atomic_write` pid-suffixes its temp name (`.<name>.tmp.<pid>`)
so two concurrent `bread-theme generate-output` runs writing the same
`themes/<output>.css` can't race on one shared `.tmp`.
- `hypr::socket_path` reconstructs `/run/user/<uid>` from the process's
real uid when `XDG_RUNTIME_DIR` is unset, instead of assuming 1000.
- `bread-theme --help` prints to stdout (pipeable); usage on an error
path (missing args) goes to stderr.
Toolchain drift (clippy 0.1.97): pre-existing on main, all mechanical
and behaviour-preserving.
- bread-theme/gtk.rs: `type AppCssBuilder` alias for the repeated
`Rc<dyn Fn(&Palette) -> String>` (type_complexity ×4).
- bread-theme/lib.rs: `sort_by_key(|..| Reverse(len))` (unnecessary_sort_by).
- bread-theme/output.rs: `io::Error::other`; struct-init in a test.
- bread-utils/singleton.rs: explicit `.truncate(false)` on the lock file
open — we only clear it after winning the lock (suspicious_open_options).
- bread-polkit/identity.rs: elide `pick_user` lifetimes.
Fourth compiled-in shell theme (bread-theme/src/shell/builtin.rs). Stresses
four axes no existing theme touched:
- light surfaces (new tokens.light flag, swaps which of the fixed @bg/
@on-bg pair plays paper-surface vs ink — see Tokens::light's doc comment)
- bottom anchoring (bar.window.anchors = ["bottom", ...])
- segmented bar chrome (new bar_border = "segmented" value: the bar window
itself draws no fill/border, three slot-group pills draw their own)
- blur disabled per-theme ([compositor.*] blur = false everywhere)
Also: a new bottom_right surface anchor (manifest.rs + types.rs), since the
three existing anchor shapes all assume a top-anchored bar's satellites
belong in the top-right corner. tokens.accent2() gives the equaliser its
own accent independent of accent_from/accent_to.
tokens.bg_alpha governs the bar, which stays readable at 0.72 because it
covers a thin strip of wallpaper. A full launcher panel at that alpha washes
out badly over a bright wallpaper and its text becomes hard to read — which
is what breadbox looked like, since it hardcoded 0.60 and read no theme value
at all.
Adds [launcher].panel_alpha, set to the approved reference's own values: 0.95
for glass-workbench, 0.93 for liquid-motion.
breadbar had to hardcode past these: the manifests carried pre-review values
(chip_height 32/20/36, dot_widths [6,10,14,18]) that disagreed with what the
bar actually draws, so a chrome pass overrode them locally rather than let the
manifest dictate wrong geometry.
Updated to what the user approved on the interactive reference in
bos-ui-demos/proposed/: one chip highlight height per bar (26 island, 22 flush,
22 capsule) and Option-B dot widths [8,13,17,22]. Two tests pinned the old
numbers and are updated with the reason.
A manifest that states one thing while the app draws another is the defect
class this project has hit repeatedly; the local hardcodes in breadbar can now
be removed in favour of these.
The two built-in launcher themes previously shared the same row radius,
padding, icon treatment, and selection fill (only mode/width/top/radius/
icon_px were themed), so breadbox's overlay rendered as one launcher
recoloured rather than two different instruments.
Adds row_radius, row_inset, row_padding_v/h, icon_radius, search_font_size,
search_padding_v/h, and selection_alpha to [launcher], all consumed by
breadbox's build_css in the paired breadbox commit. Sets liquid-motion to
the soft/roomy demo numbers (row radius 12, 28px icons, 22% selection alpha,
Recent/Apps section headers) and glass-workbench to the dense/technical demo
numbers (row radius 6, 22px icons, 28% selection alpha, flat list).
Also wires up three previously-declared-but-unconsumed tokens
(font_family, font_fallback, font_size_base) for breadbox's launcher panel
specifically — scoped consumption, not the full ecosystem-wide font system
replacement (breadbar/stylesheet() are unaffected). Updated the stale
doc comments in types.rs/mod.rs and both theme.toml files to match, and
updated the two existing launcher-geometry tests plus added a glass-workbench
equivalent.
Defaults for all new keys reproduce breadbox's pre-redesign hardcoded CSS
values, so themes that don't set them (spotlight's capsule doesn't read any
of this) are unaffected.
audit-schema-consumption.md's recurring defect: a key that validates
but does nothing, five separate instances (css/extra.css overlay,
tokens.font_family, tokens.font_fallback, tokens.font_size_base,
tokens.accent_to). All five are consumed only by breadbar/breadbox,
which this pass doesn't touch, so per the audit's own (a)/(b)/(c)
choice this is (b): document the gap at both the schema accessor and
every theme.toml site that sets the key, so the manifest states its
design intent honestly instead of implying the value has an effect it
doesn't.
Also documents modules.clock.format/show_date (consumed only by
ClockStyle::Plain, silently ignored by Flip/None) and
bar.window.margin.bottom (parsed, never applied by breadbar) the same
way.
Neither doc site previously said which of the two theme-selection
mechanisms is meant to coordinate multiple already-running processes.
The env var is read once per process and cannot do that; shell.toml's
active key is the one every host reads, so it's the only one that can.
watch() panicked (.expect) if monitor_directory failed (e.g.
fs.inotify.max_user_watches exhaustion), contradicting the crate's own
'the shell must never fail to start because a theme file is malformed'
stance. Now logs and continues without a watch, same posture as
bread_theme::gtk::watch_theme_file's existing .ok()? handling.
watch() also resolved the active theme's directory once and pinned a
monitor there forever: switching shell.toml's active key while a host
was running left edits to the newly-active theme unobserved until
restart. Now also watches shell.toml's own directory and re-arms the
theme-directory monitor when active changes.
Public API change: shell::watch now returns ThemeWatch instead of a
bare gio::FileMonitor (still opaque, still just needs to be kept
alive). breadbar's theme.rs:525-531 stores the return value in a
RefCell<Option<gio::FileMonitor>> and wraps it in Some(monitor) - both
need updating to RefCell<Option<bread_theme:🐚:ThemeWatch>> and
*cell.borrow_mut() = monitor (no Some(..)), left to the breadbar-side
agent since this repo doesn't touch breadbar's files.
's role is now documented explicitly on watch() as a
single-process testing override, not the coordination mechanism -
shell.toml's active key is.
anchor was the one enum-shaped schema field taken via
unwrap_or_default() with no manifest-time validation, while width and
layer both bail! on an unrecognized value (as does bar.window.anchors).
A typo'd anchor silently fell through to breadbar's own runtime
eprintln + unanchored-window fallback instead of the load-time hard
error every other typo'd key gets. Validate against the three shapes
breadbar/src/surface.rs actually implements.
spring_ease's backOut overshoot (peaks ~1.065) drove a shrinking
animation's interpolated value below its target, which could go
negative for a size request and trip GTK's height >= -1 assertion.
Previously only patched at breadbar's call sites; clamp inside
spring_to itself so the safety is automatic for every caller.
[launcher].modes now lists calc/cmd/url alongside apps, sections is
enabled, and two new keys (search_width/search_radius) let an embedded
launcher theme declare its search-state capsule geometry, defaulting to
the idle value when a theme omits them. Also gives spotlight's [bar.slots]
a widget:left_of_stats entry so a Lua widget requesting that placement
(e.g. git-branch-widget.lua) has a home instead of being silently dropped.
- bread-theme:🐚 WorkspacesModule.dot_widths and ClockModule.placeholder_clock,
resolved/validated in manifest.rs and modeled in types.rs, so a theme can
actually configure dot-pill widths and an entry-as-clock placeholder
instead of those staying schema-only.
- New compiled-in builtin "spotlight" (assets/shell/spotlight/), registered
in builtin.rs so list() returns three themes: centred capsule window spec
(anchors=[top] only, width=480, exclusive=none, keyboard=on_demand),
workspaces.style=dots, clock.style=none+placeholder_clock, launcher.mode=
embedded, drawer=[launcher_results]. Colours are palette token names
(pink), never hex.
- bread-theme::anim::spring_to: a reusable TickCallback size-interpolation
helper (GTK4 has no CSS width/height transition on a widget or layer-shell
surface), generalized from WorkspaceTrail's existing ease/tick pattern, for
breadbar's capsule-drawer expand/collapse.
- bread_launcher::LAUNCHER_APP: the launcher's one shared on-disk identity
(cache/config/history), so breadbar's embedded capsule and breadbox's
overlay window read and write the SAME cache and launch history rather
than forking into two rankings of a user's apps.
Tests: +6 spotlight builtin tests (loads/in list/capsule window shape/
embedded mode/dots widths/flat-pink-not-hex), bread-theme --lib 63->69,
bread-launcher --lib unchanged at 18.
Phase 5 of the shell theme system (THEME_SYSTEM_PLAN.md §11): a second
builtin, demo 02's flush edge-to-edge bar with pill workspaces, a plain
date+time clock, and cpu/ram chips instead of the media widget.
- bread-theme/assets/shell/glass-workbench/: theme.toml + CSS template,
faithful to bos-ui-demos/02-glass-workbench.html. Accent maps to the
`green` palette token (flat, not a gradient) rather than a hex literal,
so pywal theming still works.
- builtin.rs: generalized from a single hardcoded liquid-motion constant
pair to a small BuiltinTheme registry (builtin::ALL / builtin::find),
so mod.rs's discovery/list()/resolve_builtin no longer special-case one
id. liquid-motion stays the pinned fallback in resolve_builtin().
- manifest.rs: KNOWN_MODULES gains "cpu"/"ram".
- types.rs: new Tokens::bar_border() ("full" default vs "bottom") so a
flush bar can ask for a single hairline instead of an island's full
border.
- Tests: builtin loads, appears in list() alongside liquid-motion, and its
window spec is the flush/edge shape (36px, zero margin, radius 0).
cargo test -p bread-theme --lib: 63 passing (59 prior + 4 new).
radius (20 -> 8) and icon_px (36 -> 32) were demo-derived aspirations, not
what breadbox implements today (.launcher-bg's actual border-radius and
make_icon's actual set_pixel_size). row_anim/rule/footer/sections/modes are
kept but marked declared-but-not-yet-consumed: breadbox implements none of
row animation, a rule/divider, a footer, sections, or query modes today, so
the manifest should say so rather than imply they're live. Adds a builtin
launcher test mirroring the existing window/tokens fidelity tests.
Adds `bread-theme layerrules`, which writes the active shell theme's
[compositor] table to ~/.config/hypr/layerrules.json (atomic write). This
lets ~/.config/hypr/scripts/ui/rules.lua read theme-driven blur/transparency/
animation for the breadbar/breadbox layer-shell namespaces instead of having
them hardcoded, following THEME_SYSTEM_PLAN.md §9. rules.lua keeps its
previous hardcoded rules as a pcall-guarded fallback for when the JSON is
missing or malformed (lives outside this repo, so not part of this commit).
Also fixes a pre-existing test race: shell::tests and the new
layerrules::tests both mutate XDG_CONFIG_HOME in parallel `cargo test`
threads but previously used separate, unrelated locks (or none), so they
could observe each other's env var changes mid-test. Both now share
bread_theme::test_support::XDG_CONFIG_HOME_LOCK.
Reproduces breadbar's now-removed fixed Lua-widget interleave
(right-of-workspaces, left/right-of-clock, left-of-stats) as explicit
widget: slot entries, so the builtin theme still renders pixel-identical
to today's bar under breadbar's new theme-driven widget placement.
`tray` deliberately has no slot entry anywhere — it stays in the
control-panel popover regardless of [bar.slots].
Updates the one test asserting the builtin's slot contents.
Implements the shell theme manifest layer from THEME_SYSTEM_PLAN.md
§4-5: ShellTheme/WindowSpec/Slots/Tokens/LayerRule types, TOML
discovery (user -> system -> compiled-in builtin), one level of
`extends` deep-merge, deny_unknown_fields validation naming the
offending key, slot module-name validation, and css() token
substitution with an extra.css overlay. load() never fails, falling
back to the compiled-in builtin and logging once.
Ships exactly one builtin manifest, liquid-motion, describing
breadbar/breadbox as they exist today (not the design-doc demo, which
disagrees with the code on bar side margin, launcher geometry, and the
easing curves). Compositor rules and surface specs are keyed by
layer-shell namespace and cover all five breadbar namespaces plus
breadbox/breadbar-panel/breadbar-dismiss.
watch() is gated behind the existing `gtk` feature (gio::FileMonitor
is a gtk4 dependency); the rest of the module is gtk-free so bread and
breadcrumbs can validate a theme without linking GTK. No consumer
changes — breadbar/breadbox still use their own hardcoded values.
FONT_FAMILY is "Varela Round, sans-serif" but emission wrapped the whole
string in quotes, so CSS looked up one family named that string. Emit
'Varela Round', sans-serif instead and lock that in the tests.
Each Hyprland/GDK connector can have its own palette and stylesheet
under $XDG_RUNTIME_DIR/bread/{palettes,themes}/. GTK apps bind a
widget-level provider so two windows in one process can follow
different wallpapers. Bump workspace version to 0.7.4 for the tag.
pywal derives @red from the wallpaper and can hand it any hue - on a
blue-toned wallpaper the "red" slot is itself blue, making destructive
buttons indistinguishable from normal accent/confirm buttons. GNOME's
own destructive-action style is a fixed red for the same reason; this
is now the one button in the shared stylesheet that intentionally
ignores the palette.
Two bare class selectors (.title, .subtitle) were colliding with
libadwaita's own internal row/window-title label classes of the same
name, causing every AdwActionRow/AdwSwitchRow/AdwSpinRow title to
inherit the 1.4em heading size meant for app view-titles - the root
cause of breadman settings' ~24px row-title bug found in design review.
Renamed to .page-title/.page-subtitle (breadhelp, the only caller,
updated separately).
Also scoped a .boxed-list override so AdwPreferencesGroup's boxed-list
GtkListBox gets its surface fill + radius back - the shared
`list, listbox { background-color: transparent }` rule (needed for
plain GTK4 sidebars) was stripping it with equal specificity.
New `adw` feature (gated separately from `gtk`, since AdwApplicationWindow
isn't compatible with gtk4-layer-shell — the five panel/launcher apps stay
on plain `gtk`, only breadman/breadhelp-style plain-window apps want this):
preferences_group/toggle_row/spin_row/action_row/preferences_page, wrapping
libadwaita's PreferencesGroup/SwitchRow/SpinRow/ActionRow/PreferencesPage.
adw::init() also forces dark color-scheme, since bread-theme's whole design
is a fixed dark base regardless of system GTK preference.
These directly target defects a design critique found: hand-rolled
switch+label rows with no intrinsic width (breadman/settings' ~1400px
stretched toggles) and spinners stranded far from their label — both just
don't happen when the row is a real AdwSwitchRow/AdwSpinRow instead of a
box assembled from scratch.
Also, two shared-stylesheet fixes usable by every app immediately, gtk
feature only:
- `scale` (slider) had no rule at all, so every volume/brightness slider
showed GTK's own default blue instead of the palette accent — the same
critique flagged breadbar's control-panel sliders contradicting its own
on-brand OSD fill two clicks away.
- A new `chip()`/`set_chip_active()` helper in gtk.rs uses the existing
(already-tokenized, already-defined) `.chip`/`.pill` stylesheet rule
instead of each app hand-rolling its own filter-chip CSS — which is how
breadclip/breadpad/breadman ended up with three different, mutually
disagreeing pill fills for what's supposed to be one shared component.
- Add minisign-based signing/verification for the bakery index:
scripts/gen-index.sh signs index.json (MINISIGN_SEC_KEY env var, dormant
no-op with a loud warning until a key is provisioned); bakery/src/manifest.rs
fetches index.json.minisig and verifies it with minisign-verify against a
hardcoded PUBKEY before parsing/caching, and re-verifies the cached copy
on every load (falls back to one re-fetch if the cache predates signing
or fails verification; a fresh fetch that fails verification is a hard
error).
- Close the previously-unchecksummed config-example and systemd-unit
downloads in bakery/src/install.rs (scaffold_config, install_service):
index.json now carries `sha256`/`example_sha256` for these artifacts
(computed in gen-index.sh), verified via the same download::verify_sha256
used for binaries. Downloads without a matching sha256 in the index are
refused rather than installed unverified.
- scripts/get.sh now verifies the bakery release binary itself against a
pinned minisign public key before installing it (falls back to the
existing sha256-only check with a loud warning if no .minisig is
published yet or minisign isn't installed; a present-but-invalid
signature is a hard failure).
- Add dormant "sign release binary" steps to the bakery and bread-theme
release workflows (.github/workflows/release.yml,
.forgejo/workflows/release-bread-theme.yml), gated on secrets that are
not yet configured — binaries ship unsigned exactly as before until the
owner wires up the secret.
- .gitignore: add *.minisign-sec / minisign.key so the signing key can
never be committed by accident.
- bread-theme: fix stale docs describing a "Catppuccin Mocha fallback"
(BREAD_DESIGN_SYSTEM.md, README.md, Cargo.toml/bakery.toml/registry
descriptions) — the actual implementation (palette.rs) uses a fixed BOS
dark base with only accent colors from pywal.
- bread-theme: fix the legacy css_vars() path, which had its own
hand-written @define-color block that predated the `accent` and computed
`on-*` ink colors used by the rest of the stylesheet — any caller whose
CSS referenced those names against css_vars()'s output would hit
undefined colors (the illegible-text bug). css_vars() now delegates to
the same define_colors() the full stylesheet uses, so the two can't
drift apart again.
Was previously an unmanaged binary baked into /etc/skel — never updated by
bakery or pacman, which is why fresh BOS installs (and existing ones) kept
regenerating the shared theme.css with the pre-fix pywal-driven background
even after bos-settings itself shipped the fix.
A light or muddy-toned wallpaper (e.g. an all-beige photo) makes pywal hand
back a light or off-hue background/surface slot, and every bread GUI's
panels inherited it directly — the app stopped looking like a dark BOS tool
and started looking like whatever colour the wallpaper happened to be.
background/foreground/color0(surface)/color7(overlay) are now fixed BOS
dark-theme constants, never read from pywal's special/colors blocks. Only
color1-6 (the actual accent slots) still track the wallpaper via pywal, which
is the feature breadpaper is for.
The stylesheet is written with write-tmp-then-rename (atomic), which replaces the
inode. A monitor on the file itself caught the first replace then went deaf
(inotify reports DELETE_SELF and never re-arms), so `bread-theme reload` updated
the file but no running GUI ever recoloured. Monitor the parent directory and
filter for the stylesheet filename instead — that fires on every reload. Verified
against a real atomic-rename write (event arrives as Renamed with the new name in
other_file, so match both file and other_file).
Readability: pywal can emit a light value in any palette slot, and the shared
sheet assumed dark backgrounds (white text), so text vanished on light surfaces/
accents. Add ink_on() — a WCAG-luminance pick of near-black/near-white per
background — exposed as @on-bg/@on-surface/@on-accent/@on-red/@on-overlay. The
component sheet now sets colour on containers and lets labels inherit (de-emphasis
via opacity), dropping the blanket `label { color }` rule that overrode
coloured-background text. pywal hues are untouched.
Hot reload: add gtk::apply_app_css(closure) — applies an app's own CSS now and
re-runs the closure whenever the shared theme file is rewritten, so apps recolour
in place. New `bread-theme reload` verb rewrites the file (atomic rename trips
every running GUI's monitor) — the command to run after changing pywal colours.
Adds the single source of truth for bread GUI styling so the apps stop
each re-implementing (and drifting on) component CSS:
- stylesheet(&Palette): full component sheet (buttons, entries, switches,
dropdowns, lists/rows/sidebars, cards, chips, scrollbars, headings) built
from the design tokens + a canonical @define-color block (surface=color0,
overlay=color7, accent=color4).
- render() / shared_css_path() / write_shared_css(): render for the current
pywal palette and write to $XDG_RUNTIME_DIR/bread/theme.css.
- gtk::apply_shared(): load that file (or a rendered fallback) at APPLICATION
priority and watch it, so every app recolours live with no rebuild.
- new `bread-theme` CLI (generate|path|print) — gtk-free, light. Run at
session start and on palette change; apps pick it up via the file watch.
The contract is a CSS *file*, so apps stay decoupled from this crate's gtk4
version. Tests cover the stylesheet, path, and render helpers.
