An over-engineering audit of the whole tree, applied in five reviewed
batches. Behaviour stays the same except where UPGRADE.md says otherwise.
PHP: the showcase and error-page stylesheets are prebuilt into
resources/dist by bin/stylesheets.mjs, through Vite's own postcss-import
(first occurrence kept, the order an application's build gives), instead
of Stylesheets::bundle() inlining imports on every request; only the
import walk DesignGuard needs stays. SchemeStylesheet::withProfiles()
replaces three copies of the scheme-plus-profiles loop, material:scheme
leaves spec and contrast checks to the node script that already made
them, and the error page's scheme cache, the hashed view namespace, the
translations path with no lang/ folder and DesignGuard's 1.x-name hints
are gone.
JS: the androidx shape port progress.js and both bin scripts each carried
lives once in resources/js/shapes.js (the generated SVGs are unchanged);
util.js holds ringIndex(), ms(), reopenGuard() and remember(), which
were written out several times; listeners are released through
AbortController; tooltip.js's hoverPopover() serves the rich tooltip too.
CSS: every rule for an element inside the navigation rail queries
`--md-navigation-rail-value` instead of repeating the seven collapsed
conditions under five media branches; badge, alert, progress, slider and
button read one non-inheriting colour-role table (components/color.css);
the dialog chrome, the submenu's popover chrome, the chip's state layer
and touch target, and the visually-hidden inputs use the shared rules
they copied; foundation/tokens.css is folded into foundation.css.
Views: Support\Field and Support\Link replace the error-key, bound-value
and link-attribute blocks copied into the fields and link components;
the timepicker period group, the menu filter and the showcase head are
partials; the datepicker's steppers and entry fields are loops; component
docblocks no longer restate SKILL.md.
Tests and tooling: one dataset-driven ComponentStylesheetsTest replaces
four per-group files, DesignGuardTest and the layout-component tests use
datasets, browser tests share one ready() helper, CSS parsing lives in
ComponentStylesheet alone. docs/audits and the finding IDs citing it are
removed, as are pestphp/pest-plugin-laravel, the unused composer scripts
and check:font; the lint job runs in the feature job, which now installs
node packages so the prebuilt-stylesheet staleness test runs in CI.
Feature suite 1177 passed, Chrome browser suite 299 passed.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Google publishes 67 symbols per cut (auto_awesome, insights,
tips_and_updates, battery_50 ...) with width and height but no viewBox,
their paths in 24 units, or 20 in the 20 cut, where every other symbol
draws in `0 -960 960 960`. bin/fetch-symbols dropped the size so CSS
decides it, which left these with no coordinate system: `<x-icon>`
drew them at 1:1, whole only at exactly 24px and cut off or misplaced
at any other size, the 20 cut every small button uses included.
normalise() now turns a file's width and height into its viewBox when
it has none, before dropping them. The 268 files are regenerated from
Google's originals at the recorded SOURCE commit with the fixed step
(each shipped file matched the old step's output first; the other
16,272 do not change), and every path lies inside its new viewBox.
IconTest checks that each shipped symbol has a viewBox and that
`<x-icon name="auto_awesome">` renders in 20 units at size 16 and in 24
at size 32; both fail without the change.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The three state colours are DynamicColors on their own tonal palette now, built
exactly as Google builds error/on-error/error-container/on-error-container in both
specs, so the scheme's contrast level, dark tones, spec version and platform reach
them as they reach every other role; --harmonize (and a profile's harmonize) pulls
each source towards the seed, off by default. material:scheme also writes M3's
medium (0.5) and high (1.0) levels for both themes and every profile, keyed on
data-contrast and never on a media query, and --contrast now moves the standard
block alone. Scheme::load() takes the level; the mail theme stays on standard.
Plan: docs/plans/material-3-alignment.md, steps 6 and 7 (core C1, C2, C15).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
bin/springs.mjs is the sampler behind motion.css: damping ratio and stiffness in,
the 49-point linear() and the settle time out. Run with --settle it reproduces
the six curves the file carried byte for byte, so only the window changed.
Each spring is now sampled over the duration M3 publishes for the web (spatial
350/500/650 ms, effects 150/200/300 ms) instead of stopping at its own settle
time, so the *-duration tokens read Google's numbers. Sampling stays in real
time — point i is the spring at duration × i / 48 — so the bounce lands at the
same millisecond as before; a longer duration just holds the tail flat at 1, and
a shorter one pins the last point, at most 0.3% of the travel in one frame. Each
token's comment records damping, stiffness, settle time and sampled duration.
M3's second motion scheme is here too: [data-motion="standard"] swaps the three
spatial springs (damping 0.9, 0.2% overshoot instead of 9%) and their durations;
the effects springs sample identically in both schemes, so they are shared. The
reduced-motion block names both selectors, so it still zeroes every duration.
Plan: docs/plans/material-3-alignment.md step 5, findings core C8, C9.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
Tracking now comes from androidx's TypeScaleTokens.kt as sp/16 rem, so Display
Large is -0.0125rem, Title Medium and Body Medium 0.0125rem where material-web's
pre-Expressive numbers stood. The emphasized set gets its own
--md-sys-typescale-emphasized-<style>-tracking, which the emphasized utilities
read: M3 widens four of them and takes Display Large back to zero. Every regular
utility sets `font-variation-settings: normal`, because the property inherits and
body copy inside an emphasized heading is meant to read as itself.
bin/check-font.mjs (fontkit) prints the packaged woff2's fvar axes, so a
re-subset cannot silently drop the ROND axis the emphasized styles depend on;
`npm run check:font` runs it and tests/Feature/FontTest.php runs it through the
configured node binary, skipping when node cannot run (PHP cannot open a woff2
without Brotli).
Plan: docs/plans/material-3-alignment.md step 4, findings core C5, C6, C7.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
M3 draws icons at 20 dp from the optical size 20 cut, whose strokes stay as
heavy as the 24 dp design's instead of thinning with the glyph (styles.md
§ Icons, the optical size axis). The package only shipped the 24 cut, so every
20 px icon was a scaled-down 24 (core audit C16). bin/fetch-symbols now checks
out both cuts from the same commit of google/material-design-icons and writes
the new one to resources/svg/symbols/{outlined-20,filled-20}: 4,135 symbols
each, the same names and the same normalisation as the 24 cut.
Plan step 10; finding C16.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Qwx5USif3wFFmxtHg5U1g9
The scheme script always generated with material-color-utilities' 2025 spec.
An application whose palette was generated with M3's original 2021 colour
(ReStride) could not reproduce it. The script now takes `spec` ('2025' by
default, or '2021') and refuses anything else, and the command takes
`--spec`, validated before Node runs.
Colour profiles may set their own `spec`, `success`, `warning` and `info`;
without them the command's options apply. The stylesheet header writes out
every option that differs from its default, so the recorded command
regenerates the file, and a profile whose spec differs from the header's
names it. Default output is byte-for-byte unchanged.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RHoXZSHc8gGpZjFmA5fPc2
<x-button> (label buttons, icon buttons and toggles in five sizes, with
filled, tonal, outlined, elevated and text variants in any colour role),
<x-tooltip>, <x-menu> with items, groups and separators, <x-button-group>,
<x-group> as a connected button group, <x-split-button>, <x-fab>,
<x-fab-menu> and <x-loading>. Sizes, colours and shapes come from
androidx Compose Material 3's tokens; the loading indicator ports its
Morph into SVG + SMIL.
Menus follow WAI-ARIA's menu button pattern on popovers placed by CSS
anchor positioning. Browser tests run in Chromium, Firefox and WebKit.
The showcase fetches the icon names on demand: inlined, they tripped
Pest's test server into HTTP 431s under Firefox.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01V9NnLxnPp8vaaurb3Z1MFy
Colour, shape, type, elevation and motion as tokens and Tailwind
utilities; `php artisan material:scheme`, which generates an app's colour
roles with Google's material-color-utilities (spec 2025); the theme head
script with light, dark and system and its Alpine store; Google Sans Flex;
every Material Symbol (4,135, outlined and filled) drawn by <x-icon>
without blade-icons; all 35 M3 Expressive shapes, ported from androidx, as
<x-shape>; the x-figure directive; the Toasts concern; DesignGuard for
applications' tests; and a showcase with every token, both themes side by
side and an icon search.
Colour utilities are `@theme inline`, so a section with its own
data-theme repaints; without it they resolve once on :root.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01V9NnLxnPp8vaaurb3Z1MFy
