/**
 * core/button rules, each scoped to the unit that owns it.
 *
 * This file is shared by construction: every unit that places a core/button lands here.
 * Nothing in it may be written unscoped -- a rule with no unit hook class repaints every
 * button on the site, which is a theme decision and belongs to wp-theme-scaffold.
 *
 * The one rule below is that theme decision, written by wp-theme-scaffold and deliberately
 * NOT unit-scoped: it is keyed to core's own is-style-outline class, so every unfilled CTA
 * on the site gets it. A unit-scoped rule belongs under its own hook class, as above.
 */

/*
 * The unfilled CTA is an inline box, not an inline-block one.
 *
 * Measured on the source's two outline CTAs at all four viewports
 * (ai/reference/units/section-content@www.washingtonmills-com.l5.wfdev.net.json):
 * `display: inline`, against core's `display: inline-block` on .wp-block-button__link. That
 * is what makes the source's box 25px tall on a 21px line instead of boxing to its padding.
 *
 * Everything else about this variation -- colour, background, border, radius and the
 * asymmetric 0/20px/0/25px padding -- is theme.json's
 * styles.blocks.core/button.variations.outline, and is NOT repeated here. `display` is the
 * only one of the measured properties theme.json has no key for, which is the whole reason
 * this file still has a rule in it.
 */
.wp-block-button.is-style-outline > .wp-block-button__link {
	display: inline;
}

/*
 * THE FILLED CTA'S OFFSET FILL-PLATE -- PROMOTED FROM A PER-UNIT LIST TO A THEME-WIDE RULE,
 * 2026-09-13 (fix/global-link-decoration). Same shape and same reason as the unfilled rule
 * above it: `%cta-button`'s own `:after` (scss/base/_common.scss:37-47) is core's own default
 * (`:not(.is-style-outline)`) button state on the SOURCE side too, not a look one particular
 * unit chose, so it belongs unscoped, owned by wp-theme-scaffold, exactly as the outline rule
 * already was.
 *
 * WHAT IT IS. A `content: ''` box at `top: 0; left: 30px; width: 100%; height: 100%` with
 * `background: $highlight-1` (this theme's `primary`) and `z-index: -1`, BEHIND an anchor that
 * carries the same background itself -- so the union of the two paints 30px wider than the
 * anchor's own box, on the right. `core/button` already emits the anchor's own fill
 * (`__experimentalSkipSerialization` on its colour support puts `has-primary-background-color`
 * straight on the anchor from the block's own attributes); only the plate was missing.
 *
 * WHY IT WAS FIVE SELECTOR LISTS UNTIL NOW. `.wm-section-content`, `.wm-centered-text`,
 * `.wm-cta-band`, `.wm-inner-section-content` and `.wm-testimonials` each reached this same
 * placeholder through their own markup (a `.cta-link-wrapper .cta-link-button:first-child a`
 * chain in the source, or -- for testimonials -- the group selector `%cta-button, .cta-button >
 * a, input.cta-button, a.cta-button` directly), and each unit's own build raised the same
 * promotion as it joined the list ("the third unit to need these is the trigger to raise
 * promoting them", then again at the fourth, then again at the fifth) without taking it,
 * because a unit's own build may lengthen a shared list but may not decide a theme-wide rule.
 * This pass is that decision, confirmed against 7 built patterns that reference `cta-button`/
 * `cta-link` (cta-band, text-image-carousel, centered-text, testimonials, product-teaser-card,
 * inner-section-content) plus the header CTA below, rather than taken on the unit count alone.
 *
 * `isolation: isolate` ON THE OUTER `.wp-block-button` IS LOAD-BEARING WHEREVER AN ANCESTOR
 * PAINTS A BACKGROUND (a texture, or a plain colour), AND HARMLESS EVERYWHERE ELSE -- which is
 * exactly why it is safe to make unconditional. `z-index: -1` puts the plate in step 3 of its
 * nearest ancestor STACKING CONTEXT's paint order, below the step-4 background of any
 * non-positioned block between; without a stacking context on the button wrapper itself, an
 * ancestor's own background paints over the plate and it disappears. This was confirmed twice
 * over on the units that carried it before promotion -- `section-content` (a background
 * texture) and `inner-section-content` (a plain white `backgroundColor`) -- and a button with
 * no intervening background loses nothing by carrying an isolation nobody's stacking order
 * needs.
 *
 * THE HOVER SKEW -- `&:hover { &:after { transform: skew(15deg) } }`
 * (scss/base/_common.scss:48-52) -- is the whole of the filled CTA's hover treatment: the
 * source changes NO property on the anchor itself, which is why there is still no `:hover`
 * under theme.json's `styles.elements.button` and adding one would be inventing a state the
 * source does not have. The plate is a pseudo-element and `transform` is not a theme.json key
 * -- nor, per wp-block-themes' own theme-json.md, is the WP 6.9 button-pseudo-class addition
 * able to reach one; that feature covers hover/focus on the same declarative border/color/
 * spacing/typography set a theme.json style always could, not a `::after` or a `transform` --
 * so this is the correct and only level for it. The transition is declared on the plate AT
 * REST, not on the hover rule (_common.scss:46), which is what makes the skew ease in and out
 * again rather than only in.
 *
 * THE HEADER'S OWN CTA ("Contact Us" in the nav) IS THE SAME MECHANISM, REGISTERED SEPARATELY
 * IN navigation.css, AND CANNOT MOVE HERE. It is a styled `core/navigation-link`, not a
 * `core/button` -- a different block, with no selector in common with anything in this file --
 * and this file is only ever enqueued where a `core/button` is on the page
 * (`functions.php`'s `washington_mills_block_stylesheets()`, which calls
 * `wp_enqueue_block_style( 'core/button', … )`, confirmed by reading that map directly). The
 * header's nav is on every page including ones with no `core/button` at all (confirmed: neither
 * `patterns/header.php` nor `patterns/footer.php` places one), so folding its selector into this
 * rule would silently drop the header's fill-plate on any such page -- worse than the
 * two-copies problem it would be fixing. The two rules are kept in sync by hand instead, cross-
 * referenced from both files; navigation.css carries, in addition, the header CTA's own base
 * styling (colour, background, padding, font) that `core/button` gets for free from core support
 * and theme.json's `styles.elements.button` and that `core/navigation-link` has no equivalent
 * for at all -- `core/navigation-link`'s block.json declares no colour support whatsoever.
 *
 * WHAT THIS PASS DID NOT TOUCH. The unfilled CTA's own promotion (the red-square-and-chevron
 * pair below, `%cta-link`) is the same shape of overdue promotion, flagged in this file's own
 * history the same way ("Five units share the filled-plate rules and four share the outline
 * pair"), and is a separate finding from the one this pass was dispatched to fix -- left as four
 * unit-scoped selector lists.
 */
.wp-block-button:not(.is-style-outline) {
	isolation: isolate;
}

/*
 * THE HOVER SKEW LEFT A 1px STEP AT THE BUTTON'S BOTTOM-LEFT -- FIXED HERE, ONCE, 2026-09-20.
 * Operator QA: "the button on hover gets weird broken at bottom". MEASURED (Playwright, hover,
 * screenshot pixel columns, DPR 1 / 1.25 / 2): the anchor is 71.5px tall at a fractional y, so its
 * own background snaps to the pixel grid one way while the transformed ::after plate -- promoted
 * to its own compositor layer once the skew starts -- snaps the other, leaving the plate 1px lower
 * than the anchor along the bottom edge (bottom-most red row 74 at the anchor's left edge, 75 under
 * the plate). THE SOURCE HAS THE SAME STEP at some fractional offsets (same 71.5px box, same skew),
 * so it is not a transcription error and not the skew itself: the skew shape and its 0.2679 matrix
 * match the source. `will-change: transform` on the anchor puts anchor and plate on the same
 * compositing path; re-measured at all three DPRs the two bottom edges coincide. It moves no box.
 */
.wp-block-button:not(.is-style-outline) > .wp-block-button__link {
	position: relative;
	will-change: transform;
}

.wp-block-button:not(.is-style-outline) > .wp-block-button__link::after {
	content: "";
	position: absolute;
	top: 0;
	left: 30px;
	z-index: -1;
	width: 100%;
	height: 100%;
	background-color: var(--wp--preset--color--primary);
	transition: all 0.3s ease-in-out;
}

.wp-block-button:not(.is-style-outline) > .wp-block-button__link:hover::after {
	transform: skew(15deg);
}

/*
 * .wm-section-content: the unfilled CTA's RED SQUARE and CHEVRON.
 *
 * The contract's `chrome.cta-link-bullet` and `chrome.cta-link-chevron` rows -- the
 * `%cta-link` placeholder's `:before` and `:after` (scss/base/_common.scss:98-126), reached
 * by `.cta-link-wrapper .cta-link-button:nth-child(2) a` (_paragraph.scss:226-234). Both
 * items carry the pair, so this is two instances per item and four on the unit.
 *
 * THE SPACE FOR BOTH ALREADY EXISTS AND IS NOT TOUCHED HERE. theme.json's
 * `styles.blocks.core/button.variations.outline` sets the anchor's padding to
 * `0 20px 0 25px`, which is the source's measured 25px left and 20px right reserve -- the
 * 14px square plus 11px of clearance, and the 14px chevron plus 6px. That is why the defect
 * read as missing ink rather than as shifted text.
 *
 * NEITHER IS AN ICON BLOCK, AND THE SQUARE IS NOT AN ICON AT ALL: `content: ''` with a flat
 * `background: $highlight-1`. The chevron IS an icon-font glyph (icomoon U+E900, `$icon-arrow`)
 * and `core/icon` still cannot carry it -- not for kses reasons, its path is fill-drawn and
 * would survive -- but because there is nowhere to put it: `core/buttons` declares
 * `allowedBlocks: ["core/button"]` and `core/button` has no inner-blocks slot at all, its
 * label being the `text` attribute.
 *
 * THE FONT IS NOT SHIPPED, following the decision already recorded on `units/footer.yml`'s
 * `core/social-links` node for the same icomoon set. The glyph is reproduced from its own
 * path (fonts/selection.json, a single `<path>` on a 1024-unit em square) as a MASK, with
 * the colour supplied by `background-color` so it stays on the token rather than baked into
 * a data URI. THE TRAP THE OTHER CHOICE SPRINGS: a rule naming `font-family: icomoon` with
 * no `@font-face` renders U+E900 from a fallback font, where it is a private-use codepoint
 * with no glyph -- a blank or a tofu box, and nothing reports it.
 *
 * BOTH COLOURS ARE ONE TOKEN. `$highlight-1` is `#db240b` (scss/_settings.scss:59), which is
 * this theme's `primary` -- the same rgb(219, 36, 11) the filled CTA beside it measures. The
 * chevron declares no colour of its own and inherits the anchor's, which is the same token.
 *
 * `top` AND THE GLYPH'S SIZE MOVE AT FOUNDATION'S `xlarge`, 1300px: 6px/15px at and above
 * it, 4px/14px below (_common.scss:108-125). That band is not one WordPress has -- this
 * theme's two viewport bands are already spent on the source's 640/1124 split -- but a
 * scoped stylesheet writes its own media query, so chrome escapes that limit. The narrow
 * values are the base here and the 1300px query carries the wide pair, because three of the
 * four captured viewports are below 1300.
 *
 * `margin-right: 10px` IS IN THE SOURCE'S RULE AND IS INERT -- an absolutely positioned box
 * with `left: 0` and no `right` ignores its right margin. Not transcribed, and named here so
 * nobody reproduces it as a 10px gap that does not exist.
 */
.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link {
	position: relative;

	/* The source's anchor: `transition: all 0.3s ease-in-out` (computed on both hosts, 2026-09-21).
	   Only padding changes on hover. Without it the label snapped where the source eases. */
	transition: padding 0.3s ease-in-out;
}

.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after {
	content: "";
	position: absolute;
	top: 4px;
	display: inline-block;
	width: 14px;
	height: 14px;
	background-color: var(--wp--preset--color--primary);
}

.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::before {
	left: 0;

	/* _common.scss:108, declared on the square at rest. */
	transition: all 0.4s ease-in-out;
}

.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::after,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after {
	right: 0;
	mask-image: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 1024 1024'%3E%3Cpath d='M328.692 1024l-139.062-139.062 379.259-372.938-379.259-372.938 139.062-139.062 505.679 512z'/%3E%3C/svg%3E");
	mask-position: center;
	mask-repeat: no-repeat;
	mask-size: 14px 14px;

	/* _common.scss:121, declared on the chevron at rest. `right` is what animates. */
	transition: right 0.4s ease-in-out;
}

/*
 * THE UNFILLED CTA's HOVER: the chevron travels the width of the label and lands as a
 * 3px bar just past the red square. `%cta-link:hover` (scss/base/_common.scss:129-138):
 *
 *     &:hover {
 *       padding-left: 28px; padding-right: 0;
 *       &:after { width: 3px; background: $highlight-1; overflow: hidden;
 *                 right: calc(100% - 21px); }
 *     }
 *
 * `mask-image: none` IS HOW THE BAR IS MADE HERE, AND IT IS A MECHANISM DIFFERENCE, NOT A
 * SECOND OPINION ABOUT THE RESULT. In the source the ::after is an icomoon GLYPH with a
 * `background` behind it, so `width: 3px` + `overflow: hidden` clips the glyph away and
 * leaves the background painting a solid 3px bar. Here the chevron is not a glyph: it is
 * `background-color` clipped BY a mask, so narrowing the box to 3px would leave a 3px
 * vertical slice through the middle of the chevron -- mask-position is `center` -- rather
 * than a solid bar. Dropping the mask lets the same background-color paint the whole 3px
 * box, which is the shape the source ends up with. `overflow: hidden` has nothing to clip
 * once there is no glyph and is not transcribed.
 *
 * THE PADDING PAIR OVERRIDES theme.json's `variations.outline`, which is the rest state this
 * hover moves away from: the left pad grows as the square makes room, THE LABEL MOVES 3px
 * RIGHT, and the right pad gives back exactly those 3px.
 *
 * DECIDED DEVIATION FROM THE SOURCE, 2026-09-21 (operator): THE HOVER MUST NOT MOVE ANYTHING
 * ELSE. The source's `%cta-link:hover` sets `padding-right: 0`, so the anchor's outer box
 * narrows 17px (255 -> 238 on /contact-us "Browse Products") and, because the button row is
 * centred, the NEIGHBOURING "Browse Locations" button slides 6-8px sideways while the pointer
 * is on a different control (MEASURED, Playwright 1440x900, both hosts: sibling x 444 -> 450 on
 * the source at 600ms; 444 -> 452.5 on the build before this change). Operator: "it shouldn't
 * shift the button when you're interacting with something else." The visible effect is kept --
 * label 3px right, square/chevron bar geometry unchanged, and both ease over the source's own
 * 0.3s ease-in-out (`transition` on the anchor, above; the build had none and snapped) -- but
 * the padding pair is a SUM-PRESERVING trade against the rest state, so the anchor's width is
 * identical at every frame. The pseudo-elements are positioned against the padding box, not the
 * padding, so the chevron's travel to the bar is unaffected. Revisit if the source is fixed.
 *
 * UPDATED 2026-09-25: the rest-state left pad snapped from a measured 25px to
 * `spacing|30` (20px, spacing-standard.md §1 item 3) when theme.json's `variations.outline`
 * left padding was tokenized. The +3px/-3px trade is unchanged and still sums correctly
 * against the new rest state (20+20 = 23+17 = 40); the fixed 3px shift itself is this
 * component's own hover-geometry constant (item 4's "component's own size"), not a scale
 * step -- the label must move exactly 3px to clear the square, not to the nearest preset.
 */
.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link:hover,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link:hover,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link:hover,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link:hover,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link:hover {
	padding-left: calc(var(--wp--preset--spacing--30) + 3px);
	padding-right: calc(var(--wp--preset--spacing--30) - 3px);
}

.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link:hover::after,
.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link:hover::after,
.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link:hover::after,
.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link:hover::after,
.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link:hover::after {
	right: calc(100% - 21px);
	width: 3px;
	mask-image: none;
}

@media (min-width: 1300px) {
	.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::before,
	.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::before,
	.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::before,
	.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::before,
	.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::before,
	.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after {
		top: 6px;
	}

	.wm-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-centered-text .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-cta-band .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-page-hero__content .wp-block-button.is-style-outline > .wp-block-button__link::after,
	.wm-inner-section-content .wp-block-button.is-style-outline > .wp-block-button__link::after {
		mask-size: 15px 15px;
	}
}

/*
 * .wm-inner-section-content: THE CTA WRAPPER'S 40px LEFT INDENT, which lines the two CTAs up with
 * the list's text above them.
 *
 * THE MEASUREMENT. The contract's `core/buttons` node puts the wrapper at 555x102 at x=750 and its
 * two children at x=790 and x=1040, and its `core/list-item` nodes put every list item at x=790 --
 * the `ul`'s own 40px `padding-inline-start`. So the CTAs are indented to the list's text, and the
 * 40px is the `.cta-link-wrapper`'s own padding rather than anything the column contributes.
 *
 * WHY IT IS NOT A BLOCK ATTRIBUTE. The contract's `chrome.cta-button-plate` row records where the
 * indent comes from: the nested copy of the placeholder pair "adds only `padding-left: 40px` on the
 * wrapper at `large`". Foundation 6's `large` is 1024px. This theme's two viewport bands are
 * 639.98px and 1123.98px (theme.json `settings.viewport`), so a `style["@tablet"]` spelling of it
 * would drop the indent from 1123.98px down and render 0 at the CAPTURED 1024x768 viewport, where
 * the source still has 40. A scoped stylesheet writes its own media query and can use the source's
 * own edge, which is the only spelling that is right at all four captured viewports.
 *
 * THE EDGE RESTS ON A STATED CONTRACT VALUE, NOT ON A MEASUREMENT OF IT. 40px at 1440 is measured;
 * that it arrives at 1024 rather than at Foundation's `xlarge` 1200 is the contract's `why` prose
 * plus Foundation 6's documented default breakpoint map. The projection carries no per-viewport x
 * for the buttons wrapper, so nothing here can distinguish the two. Named in the hand-back as a
 * discovery defect; if the real edge is 1200, the band 1024-1199 carries a 40px indent the source
 * does not have.
 *
 * WHY IT LIVES IN THIS FILE, which is attached to `core/button` and not to `core/buttons`: a
 * `core/buttons` block always contains a `core/button`, so `wp_enqueue_block_style( 'core/button' )`
 * has already fired on every page this selector can match. Adding a `core/buttons` row to the map
 * in functions.php would buy nothing and would touch a file four other packages are writing.
 *
 * Written as a min-width rule with no inline counterpart, so there is no inline `padding-left` to
 * beat and no `!important` anywhere. Its companion `padding-top: 30px` IS a block attribute, on
 * the buttons block in patterns/inner-section-content.php, because it holds at every viewport.
 */
@media (min-width: 1024px) {
	.wm-inner-section-content .wp-block-buttons {
		padding-left: var(--wp--preset--spacing--50);
	}
}

/*
 * .wm-news-article, .wm-product-detail: THE BACK LINK. A THIRD CTA MECHANISM, joining none of the
 * lists above.
 *
 * Its contract carries exactly one chrome row -- `back-button-chevron` -- and says in the row's own
 * words that it "is a THIRD mechanism, not the `%cta-button` plate or the `%cta-link` pair the other
 * CTAs in this definition use, and that is what makes it its own row". So `.wm-news-article` is
 * deliberately absent from every selector list above: it has no offset fill plate, no red square
 * bullet, and no right-hand chevron. It has one LEFT-pointing chevron inside the anchor, and the
 * source draws it by flipping the same glyph: "the icon set has no left arrow, so the right one is
 * flipped" (`%back-button`'s `&:before`, scss/base/_common.scss:72-79).
 *
 * `.wm-product-detail` JOINS THIS SELECTOR GROUP RATHER THAN GETTING ITS OWN COPY, added when
 * `templates/single-product.html` gained its own "Back to Products"/"View All Products" button
 * (`washington-mills/product-back-link`, functions.php). It is the SAME source component, not a
 * lookalike: both anchors render `class="back-button"` in the DOM (confirmed on
 * ai/pages/calcium-aluminate.yml's own capture, `<a href="/products" class="back-button">View All
 * Products</a>`), i.e. one SCSS placeholder, `%back-button`, used by two content types. Every number
 * below is news-article's own measured figure, cited rather than re-measured -- product-detail's own
 * capture never reached this element (it sits past `product-detail.yml`'s `truncated` cap, in the
 * documents-wrapper this contract's own `why` names but does not model), so there is nothing of its
 * own to measure against. Flagged in the build hand-back rather than asserted as independently
 * confirmed.
 *
 * THE SOURCE'S DECLARATIONS, transcribed: `@extend %icomoon` with `content: $icon-arrow`,
 * `position: absolute`, `top: 22px`, `left: 13px`, `font-size: rem-calc(18)` and
 * `transform: scaleX(-1)`, over a host anchor measured `position: relative`, 229x68 at x=235 y=403,
 * `padding: 15px 25px 15px 40px`, `font-size: 21px`, `color: rgb(219, 36, 11)` -- and the same
 * 229x68 at 1024x768, 768x1024 and 390x844, because `%inverted-button` sets a flat `rem-calc(21)`
 * with no `xlarge` branch and neither `top` nor `left` carries a breakpoint.
 *
 * THE GLYPH IS REPRODUCED AS A MASK, NOT AS AN ICON FONT, following the decision already recorded on
 * `units/footer.yml`'s `core/social-links` node for this same icomoon set and on the chevron rules
 * above. icomoon U+E900, name `arrow`: one fill-drawn `<path>` on a 1024-unit em square, taken from
 * the contract's own transcription of web/themes/washington_mills/fonts/selection.json. THE TRAP THE
 * OTHER CHOICE SPRINGS, and the contract names it too: a rule naming `font-family: icomoon` with no
 * `@font-face` renders U+E900 from a fallback font, where it is a private-use codepoint with no
 * glyph -- a blank or a tofu box, and nothing reports it.
 *
 * THE BOX IS 18x18 BECAUSE THAT IS WHAT `font-size: 18px` MAKES AN em-SQUARE GLYPH, and the contract
 * records no box for this pseudo-element on purpose: "The rule declares no `width` and no `height`,
 * so the generated box is the glyph's own advance at 18px icomoon ... Recording a number here would
 * be arithmetic presented as a measurement." An 18px em square with the path's own 189.6..834.4
 * horizontal extent paints 11.3px of ink, centred, which is what the source paints. Contrast the
 * chevron rules above, where the source declares 14x14 and the number is a transcription.
 *
 * `currentColor` RATHER THAN THE TOKEN, and that is the contract's own instruction: `%inverted-button`
 * sets no colour, the anchor measures `color: rgb(219, 36, 11)` = `$highlight-1` = this theme's
 * `primary`, and the `:before` declares none, so the glyph inherits it. "The scoped rule writes
 * `currentColor` so it follows the button rather than pinning a second copy of the token."
 *
 * `core/icon` CANNOT CARRY IT, and not for the kses reason: the path is fill-drawn and would survive
 * `WP_Icons_Registry::sanitize_icon_content()` intact. There is nowhere to put it. `core/buttons`
 * declares `allowedBlocks: ["core/button"]` and `core/button` has no inner-blocks slot at all -- its
 * label is the `text` attribute -- read off this install's own
 * web/wp-includes/blocks/{buttons,button}/block.json. So no block can be placed inside or beside the
 * anchor and the chevron is a scoped rule on the built button's own hook class.
 *
 * THE FILL-ON-HOVER IS REPRODUCED, at the foot of this file (2026-09-13). It was not, and this
 * paragraph used to say so — being "named so it is not lost" is what made it findable. `&:hover
 * { background: $highlight-1; color: $white }` (_common.scss:81-84) repaints the ANCHOR, not the
 * pseudo-element. Nothing in this definition captures hover on this unit and nothing can: a DOM
 * capture records a resting state and no pseudo-state at all, so the rule is transcribed from the
 * source's own SCSS. The ::before chevron follows the fill because it is painted in currentColor.
 *
 * THE 40px LEFT PADDING THAT RESERVES THE SPACE IS NOT HERE -- it is a block attribute on the button
 * in patterns/news-article.php, with the rest of the measured 15px 25px 15px 40px. theme.json's
 * `variations.outline` padding (0 20px 0 25px) is the reserve for the square and chevron the OTHER
 * units carry and is wrong for this one; an inline padding beats it with no `!important`.
 *
 * TWO DECLARATIONS BELOW ARE NOT CHROME AT ALL AND ARE HERE BECAUSE THE THEME WOULD OTHERWISE MAKE
 * THE MEASURED BOX UNREACHABLE:
 *
 *   `display: inline-block` RESTORES CORE'S OWN VALUE AGAINST THIS FILE'S FIRST RULE. The unscoped
 *   `.wp-block-button.is-style-outline > .wp-block-button__link { display: inline }` at the top of
 *   this file is right for the inline-box CTAs it was measured on and wrong here: an inline box
 *   ignores vertical padding, so 15px top and bottom would contribute nothing and the measured 68px
 *   height could not be produced at all. This selector is 0,3,1 against that rule's 0,2,1.
 *
 *   `font-size: 21px` PINS WHAT theme.json MAKES FLUID. `styles.elements.button` uses the
 *   `body-text` preset and this theme's own `styles.css` redefines that preset to 1.125rem below
 *   1123.98px, so three of the four captured viewports would render this label 18px and move the
 *   box -- where the contract measures the same 229x68 at all four.
 *
 *   `line-height: 32px` IS ARITHMETIC, NOT A READING, AND IS THE ONE NUMBER HERE THAT NOTHING
 *   MEASURED. RECOMPUTED when the border correction below landed: the projection carries no
 *   line-height for this anchor, but it does carry a 68px height, a `padding: 15px 25px 15px 40px`
 *   AND, since the border-width correction, a real 3px border on all four sides of the SAME 68px
 *   box -- so the line box the padding and border leave is 68 - 15 - 15 - 3 - 3 = 32, not the 38
 *   an earlier pass computed while the border was still wrongly believed to be 0. 21px at this
 *   theme's root `lineHeight: 1.5` is 31.5px, which built the anchor 62px tall before any of padding,
 *   border or this declaration existed -- measured on this install with tooling/extract/probe.js.
 *   So the 32 stands on the contract's own box arithmetic, exactly as `.wm-news-teaser-card`'s 8px
 *   title margin does in post-title.css, and the mechanism behind it in the source is NOT inferred
 *   here: a declared `line-height: 1.7` would give 67.7 and a declared 32px would give 68, and
 *   nothing in the projection distinguishes them. Written in px rather than as a ratio so the number
 *   is the arithmetic and not a guess at what the author typed. Raised in the build hand-back as a
 *   contract gap, both times.
 *
 * WHAT WAS DIFFERENT AND IS NOW MASKED BY COINCIDENCE, NOT FIXED: the label's own width. Before the
 * border correction the anchor measured 229 wide in the source and 223 here -- 6px of text advance
 * over a 40px + 25px padding pair both sides transcribe exactly, because the source sets this label
 * in its own face and this theme sets it in Jura. Adding the real 3px border on both left and right
 * (content-box, the browser default this theme never overrides for buttons) adds exactly 3 + 3 = 6px
 * to the rendered width, which lands the box back on 229 -- THE SAME NUMBER, FOR AN UNRELATED REASON.
 * The font substitution is not fixed: swap the border out for a route that adds no left/right width
 * (a background-only treatment, say) and the label reverts to 223. The date's `<time>` shows the
 * underlying font gap on its own, with no border to mask it (135 built against 142 measured). That
 * remains a theme typography question and not a rule this unit may write; named here so the next
 * reader does not credit the border fix with closing it.
 */
.wm-news-article .wp-block-button.is-style-outline > .wp-block-button__link,
.wm-product-detail .wp-block-button.is-style-outline > .wp-block-button__link {
	position: relative;
	display: inline-block;
	font-size: 21px;
	line-height: 32px;
}

.wm-news-article .wp-block-button.is-style-outline > .wp-block-button__link::before,
.wm-product-detail .wp-block-button.is-style-outline > .wp-block-button__link::before {
	content: "";
	position: absolute;
	top: 22px;
	left: 13px;
	width: 18px;
	height: 18px;
	background-color: currentColor;
	mask-image: url("data:image/svg+xml,%3Csvg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 1024 1024'%3E%3Cpath d='M328.692 1024l-139.062-139.062 379.259-372.938-379.259-372.938 139.062-139.062 505.679 512z'/%3E%3C/svg%3E");
	mask-position: center;
	mask-repeat: no-repeat;
	mask-size: 18px 18px;
	transform: scaleX(-1);
}

/*
 * .wm-news-article's BACK BUTTON FILLS ON HOVER. `%back-button:hover { background:
 * $highlight-1; color: $white }` (scss/base/_common.scss:81-84) -- the outlined box
 * inverts to a solid primary one with a white label, and the chevron drawn by the
 * ::before above follows because it is painted in `currentColor`.
 *
 * WHY THIS IS NOT theme.json's `styles.blocks.core/button.variations.outline[":hover"]`,
 * which is where a site-wide button state would otherwise belong. Two different source
 * components share this theme's ONE outline variation: `%back-button` (this one, a
 * bordered box that fills) and `%cta-link` (the red-square-and-chevron text CTA used by
 * section-content, centered-text and the rest). Their hovers are not the same treatment
 * and are not even the same kind -- `%cta-link:hover` shifts its own padding and morphs
 * its arrow into a bar (_common.scss:129-138), it does not fill. Writing the fill on the
 * variation would repaint every `%cta-link` CTA on the site with a hover the source does
 * not give them. Same reasoning wp-theme-scaffold gives for `styles.elements.h5`: the
 * shared key is the site's answer for every instance, so a difference that belongs to one
 * component goes on that component's own hook class.
 *
 * `%cta-link`'s OWN HOVER IS NOT WRITTEN HERE and is named so it is not lost. It is a
 * padding shift plus a geometry change on the ::after chevron, and it belongs with the
 * rules that draw that chevron, per unit. Reported in the hand-back.
 */
.wm-news-article .wp-block-button.is-style-outline > .wp-block-button__link:hover,
.wm-product-detail .wp-block-button.is-style-outline > .wp-block-button__link:hover {
	background-color: var(--wp--preset--color--primary);
	color: var(--wp--preset--color--white);
}

/*
 * FOCUS -- THE WHITE RECTANGLE THROUGH THE SKEWED PLATE, FIXED AT ITS ROOT, 2026-09-21.
 * Operator: "when i click the buttons, there's a white border cutting thru the right part before
 * the curved edge". MEASURED (Playwright, CDP CSS.getMatchedStylesForNode, mouse click on
 * "Browse Locations", /contact-us): the outline came from the theme's own style.css/style.min.css
 * `:where(.wp-site-blocks *:focus) { outline: 2px solid }` -- a Twenty Twenty-Five inheritance --
 * which fires on MOUSE focus, and paints a rectangle on the anchor's box over the skewed plate.
 * The source computes `outline: none` there. That rule now reads `:focus-visible`, so a click
 * paints nothing, site-wide, and keyboard focus keeps the outline.
 *
 * A plain outline on a keyboard-focused filled button would still cut through the plate, because
 * the plate is a 30px-wider skewed pseudo-element that an outline on the anchor's rectangle does
 * not follow. `filter: drop-shadow` follows the RENDERED alpha shape (anchor plus plate), so a
 * two-tone 2px ring (white inside, black outside: visible on light, dark and red grounds) hugs
 * the real silhouette. WCAG 2.4.7 binds what this toolkit authors; the source has no keyboard
 * indicator at all.
 */
.wp-block-button:not(.is-style-outline) > .wp-block-button__link:focus-visible {
	outline: none;
	filter: drop-shadow(2px 0 0 #fff) drop-shadow(-2px 0 0 #fff) drop-shadow(0 2px 0 #fff) drop-shadow(0 -2px 0 #fff)
		drop-shadow(2px 0 0 #000) drop-shadow(-2px 0 0 #000) drop-shadow(0 2px 0 #000) drop-shadow(0 -2px 0 #000);
}
