/**
* Harvest — Buttons (C8).
*
* One button system, two surfaces: the theme's own `.hrv-btn*` classes (used by
* Harvest templates) AND Magento core button classes (`.action.primary`,
* `.action.tocart`, `.action.secondary`, and the generic `button`/`.action`)
* so real catalog / cart / checkout pages get the Harvest look without touching
* core templates. All colour comes from --hrv-* tokens (palette + dark safe).
*
* Comp reference: solid accent fill / white text / 8px radius (primary),
* transparent + accent border (secondary), 13–14px, weight 600.
*/
/* SPECIFICITY REPAIR — Blank was winning the corner radius.
Magento Blank's `styles-m.css` carries `a.action.primary { border-radius: 3px }` at
(0,2,1). The theme's own `.action.primary` below is (0,2,0), so on any button that
happens to be an Blank won and the theme's radius never applied. S7 caught it on
T2: the hero CTA "Shop the season" rendered at 3px beside three "Add to Cart"
buttons at 14px — same colour, same height, two different corner shapes on one page.
This is the same (0,2,1)-beats-(0,2,0) class that beat the theme on `width` on Home,
where `float` in the SAME declaration block won and the block read as fully applied.
A DECLARATION BLOCK IS NOT AN ALL-OR-NOTHING UNIT.
Adding the element qualifier makes the theme's own selector (0,2,1) and puts it
after Blank in source order, so it wins on both counts rather than relying on either.
NOT CHANGED HERE, deliberately: the comp specifies a 999px pill for these CTAs
(12 pill declarations against 9 at 14px, so the artboard genuinely mixes the two).
Choosing WHICH buttons become pills is a theme-wide design decision affecting every
one of the 434 pages, and I am not making it inside a T2 fix. This change restores
the radius the theme already declares; the pill question is filed separately. */
.hrv-btn,
a.hrv-btn,
.action.primary,
a.action.primary,
.action.tocart,
a.action.tocart,
.action.submit,
.action.submit.primary,
.action.select,
.action.secondary,
a.action.secondary,
button.action-primary,
.actions-toolbar .action.primary,
.actions-toolbar a.action.primary,
/* N-16: enumerated from the SERVED Blank sheets rather than patched one at a time as each
page review surfaced one. These are the Blank/Luma button rules that set border-radius
and outrank the theme block above:
styles-l.css .block.newsletter .action.subscribe (0,3,0) <- N-16, global footer
styles-m.css .cart.table-wrapper .actions-toolbar > .action (0,4,0)
styles-m.css a.pagebuilder-button-primary / -secondary (0,1,1)
styles-m.css .minicart-wrapper .action.showcart .counter.qty(0,4,0) (a badge, not a button)
Two instances had already been found by two reviews of ONE page, which is the signal to
enumerate rather than keep repairing. The theme block now names them. */
.block.newsletter .action.subscribe,
.cart.table-wrapper .actions-toolbar > .action,
a.pagebuilder-button-primary,
a.pagebuilder-button-secondary {
display: inline-flex;
align-items: center;
justify-content: center;
gap: 8px;
font-family: var(--hrv-font-body);
font-size: 14px;
font-weight: 600;
line-height: 1.2;
padding: 13px 28px;
border-radius: var(--hrv-radius-md);
border: 1px solid transparent;
cursor: pointer;
text-align: center;
text-decoration: none;
transition: background-color 0.15s ease, border-color 0.15s ease, color 0.15s ease;
}
/* Primary — solid accent.
------------------------------------------------------------------------------
`a.action.primary` AND `.actions-toolbar a.action.primary` ARE LOAD-BEARING HERE,
for the same reason `a.action.secondary` is on the secondary rule below. This is
the second half of that repair — filed as register #20, "register #1 is HALF-done".
THE BUG. The base rule at the top of this file lists BOTH `a.action.primary`
(0,2,1) and `.actions-toolbar a.action.primary` (0,3,1) and sets the SHORTHAND
`border: 1px solid transparent`. This rule used to list only the un-qualified
`.action.primary` (0,2,0) and `.actions-toolbar .action.primary` (0,3,0), so on an
the base won `border-color` and every anchor-based primary in the theme carried
a transparent border. `background` and `color` still landed, because the base rule
sets neither.
RETRACTED: "THE DEFECT CLASS HAS FOUR MEMBERS, NOT ONE" - the count was itself an
undercount, and this is the THIRD time in one task that this claim has been stated too
narrowly. The sequence: fixed `.primary` alone and wrote "FIXED THEME-WIDE"; two reviewers
measured three more; the corrected claim said FOUR; an S8 re-derivation enumerated the base
rule mechanically rather than reading the four named, and found EIGHT element-qualified
selectors of which two anchors still computed `rgba(0, 0, 0, 0)`.
What kept going wrong is not the arithmetic. Each pass enumerated from the previous pass's
list instead of from the base rule itself. The list is now derived from the rule.
MEASURED, by a probe proven to discriminate both ways (control A: a repaired member reads
repaired; control B: a forced-transparent member reads transparent):
repaired a.hrv-btn · a.action.primary · a.action.tocart · a.action.secondary
button.action-primary · .actions-toolbar a.action.primary
WAS BROKEN a.pagebuilder-button-secondary · a.pagebuilder-button-primary
.block.newsletter .action.subscribe <- ships on 5/5 routes
not proven .cart.table-wrapper .actions-toolbar > .action (cart empty; unreached live)
The pagebuilder pair rendered 0 times across /, /farms, /checkout/cart, /about and /contact,
so this was latent rather than visible - which is why a live census reading "10 anchor
buttons, 0 transparent" was true as scoped and still missed them. Scope was the defect.
The base rule element-qualifies
`a.hrv-btn` (0,1,1), `a.action.primary` (0,2,1), `a.action.tocart` (0,2,1) and
`a.action.secondary` (0,2,1) — every one of them inheriting `border: 1px solid
transparent`. Register #1 repaired the secondary; register #20 named only
`a.action.primary`. A review applied CLAUDE.md §3.18 ("how many instances does this
class have?") and the answer, MEASURED live on basil at 1440 with a `