chore(ds): finish the button axis rename + drift fixes (secondary/ghost, attention, input size, press gate) (TASK-22817) by kushagrasarathe · Pull Request #3302 · peanutprotocol/peanut-ui

kushagrasarathe · 2026-09-21T17:09:51Z

Eight drift fixes in the design system: names that no longer match the board, three variants the board never had, a prop the board does not draw, two types nobody could import, a comment that named the wrong colour, and a contract test that could not fail a merge.

Everything except H is a rename, a deletion or a comment. H has one deliberate visual change, named below.

Stacks on #3301

Branched off fix/nav-circle-one-color-press, not dev, because #3301 touches the same button call sites. Merge #3301 first. After it lands, this PR shows only its own 7 commits.

A — the button variants are named after the board rows

The board (17802:61527) calls the three live rows primary, secondary and ghost. The code called two of them stroke and transparent.

ButtonVariant: stroke → secondary, transparent → ghost
CSS: .btn-stroke → .btn-secondary (the ghost row is styled inline, so there is no .btn-transparent to rename)
111 files, ~250 lines: product call sites, /dev pages, jest tests, the press-physics spec and the prose that named the old words
the legacy tail (transparent-light, transparent-dark, primary-soft) kept its names here, then H removes it outright
the frozen audit fixture under dev/ds/audit/ is untouched by design

B — the non-blocking limits state is 'attention'

LimitsWarningCard renders a Callout with priority="attention", but the props that chose it said warning.

LimitsWarningType: 'warning' → 'attention'
titleKind: 'warning' → 'attention'
the message key those build renames with them: limits.warningCard.warningTitle → .attentionTitle, in the 3 locales that carry it (es-AR has no title keys and falls back)

Same copy, same colour, same branch.

Analytics scope. type reaches PostHog: LimitsWarningCard captures it on limits_check_link_navigation, so that property's value becomes attention from this PR on. Nothing reads it — no saved insight or dashboard in project 138913 references the event at all (system.insights, checked; Chip found the same). The break is a value split in raw events only, so any later query on that property has to accept both words across the rename date.

C — BaseInput's height axis is `size`

sm/md are heights (40/48px). Calling the prop variant made it read like a look, so callers forced the height by className instead.

prop and local type renamed (BaseInputVariant → BaseInputSize)
13 call sites: 1 product (ProfileEditField) and 12 on /dev pages
the native size attribute is omitted from the props — nothing passed it
SearchInput moves to size="sm" and drops its h-10. It already overrode the horizontal padding with px-10, so the class list it renders is byte-identical.
CopyField keeps its className="h-10". sm is h-10 px-3, md is h-12 px-4 — the size does not differ in height alone, and switching CopyField would narrow its padding from 16px to 12px. That is a visual change, so it waits for a ruling.

D — `disableBackBtn` is gone

No product screen ever passed it. Its only caller was the /dev/ds page documenting it, so the doc page was the whole demand. A disabled back button is also not a state the navigation board draws; flows that wanted it use hideBackBtn.

Removed: the prop, its destructure, the disabled={...} it fed, and the two doc-page rows plus the live demo usage.

E — the shadow-size unions are exported

Both were private, so a caller building a props object had to retype the union. Card's was also called ShadowSize while holding a different set of members than Button's, so it becomes CardShadowSize. Type-only: no union changes, no call-site changes.

F — AVATAR_LINK_BG is the brand pink

The comment said "peanut pink". The same file defines a peanut_pink palette entry whose background is #FFD5F6. #FF90E8 is action-primary. One line.

G — the press-physics contract can now fail a merge

e2e/flows/button-press-physics.spec.ts ran only inside ds-shots, which is deliberately not in ci-success.needs. A change that broke the press state could merge green. The contract has broken twice, and a class-name unit test cannot catch it: the bug is which state wins in the cascade, readable only off computed style in a real browser.

New press-contract job: checkout with submodules, install, cached Playwright chromium, pnpm build, then that one spec. No screenshot or diff machinery. Added to ci-success.needs. ds-shots keeps its advisory role unchanged.

Prettier reflowed the needs: array when the entry pushed it past the print width — that is why the diff there is larger than one line.

H — `ButtonVariant` is `primary | secondary | ghost`

kush 2026-09-22: the union is the board. The three legacy variants had no board row and kept a fourth, fifth and sixth way to ask for a button the board already answers. All 10 call sites are migrated and the type, the buttonVariants record and the two CSS classes are gone.

primary-soft → secondary (4 sites, no visual change at rest.) primary-soft was bg-white with no shadow, so every call site paired it with shadowSize="4". --color-background-default is #ffffff and .btn-secondary already ships the same 4px shadow, so the swap is pixel-identical and the redundant shadowSize goes with it.

Profile — "Log out"
Success.link.send.view — "Cancel link"
AddMoneyBankDetails — the "Share details" ShareButton
CopyToClipboard — "Copy code" (button mode)

The disabled state changes: 4px → 1px. primary-soft carried no shadow of its own, so the whole shadow on these four came from shadowSize="4" → .btn-shadow-primary-4, which has no disabled: scope (globals.css). A disabled primary-soft therefore still painted the full 4px offset, only dimmed by .btn's disabled:opacity-40. .btn-secondary scopes its own disabled:shadow-[0.0625rem…], so the four now drop to the 1px residual instead. That is the board's disabled state, so it is a gain, not a regression — but the delta is 4px to 1px, not flat to 1px. ShareButton's own hard-coded shadowSize="4" stays — ProfileHeader passes it variant="ghost", which has no shadow of its own and would lose one if it were dropped.

transparent-dark → ghost (3 sites, no visual change.) SetupWrapper ("Skip") and the two BadgesRow scroll arrows. No className changes.

The fourth transparent-dark was not a button at all. The UnsupportedBrowserModal paste hint was an ActionModal cta with no onClick — a button that did nothing, styled down to h-2 to stop looking like one. It is now a <p className="text-body-xs text-foreground-secondary"> in the modal footer. Footer, not content: "Then paste it in your preferred browser." has to read after the copy cta, and content renders above the ctas. Net spacing below the button moves 20px → 24px. The test's ActionModal mock gained footer and lost its now-dead onClick ? button : p branch.

transparent-light → ghost (QRScanner, 2 sites) — the one visual change. The scanner overlay circles now use NAV_CIRCLE_BUTTON_CLASSES with border-white text-white fill-white, so they match every other icon-only nav button in the app. At rest: white ring, white glyph on the camera feed — same as today. On press: pink fill with a black glyph, which is the ruled ghost icon-only state and is what changes. The hard-coded fill="white" on the inner <Icon> goes, so the glyph now inherits from the button and can invert.

On a desktop pointer the circles also take the pink fill on hover, not only on press. NAV_CIRCLE_BUTTON_CLASSES paints hover and press with the same action-primary — one colour for both states is the 2026-09-21 ruling, so this is the ruled behaviour and not a second change. Touch has no hover, so the scanner's own users see press only.

Deleted: .btn-transparent-light and .btn-transparent-dark from globals.css; the three variants from ButtonVariant, buttonVariants and the Button docblock; the three from the /dev/ds/primitives/button playground and variants list. That list's usage counts were stale, so it now names each row's job instead of a number nobody regenerates. The frozen audit fixture under dev/ds/audit/ is untouched.

Separate commit — the dead UserHeader is deleted. It was the fifth primary-soft call site and had zero callers; the home header was rebuilt during the consolidation. All 12 imports of @/components/UserHeader pull VerifiedUserLabel, which stays in place with its import path unchanged.

Not done — each waits on a ruling

unifying the size vocabulary across primitives (Button uses small|medium|large, BaseInput uses sm|md)
Avatar defaults
font schemes
folding Field and FieldColumn into one primitive

Gates

prettier --check . · typecheck · TZ=UTC npm test (717 suites, 9133 tests) · npm run build · full pnpm test:e2e:regression against the production build (112 passed) — all green locally, re-run after H. The press-physics spec is unaffected and passes.

ds-shots should now report two classes of difference, both intended:

the QR scanner overlay circles (H), if the shot captures a pressed state
disabled-state shots of the four migrated primary-soft buttons, whose shadow drops from 4px to the 1px residual

Anything else is a regression.

Extends TASK-22817. The id is in the title only — the branch was already pushed and open, and renaming a branch closes its PR, so the second mention the linker wants is not available here. Follow-up fixes are in #3317.

🤖 Generated with Claude Code

The board (17802:61527) calls the three live rows primary, secondary and ghost. The code called two of them 'stroke' and 'transparent', so every review of a button had to translate between the two vocabularies, and the press-physics spec named a test after a word the board does not use. Renames the union members, the class (.btn-stroke -> .btn-secondary) and every call site. The legacy tail (transparent-light|dark, primary-soft) keeps its names: it has no board row and waits on a page rebuild. No visual change — the class bodies and the variant strings are untouched.

The card renders a Callout with priority="attention" — the board's name for the state — but the props that choose it were called 'warning'. Reading the file meant mapping one word onto the other, and a new caller had no way to guess which name the board uses. Renames the LimitsWarningType member and the titleKind value, plus the message key they build (warningCard.warningTitle -> .attentionTitle) in the three locales that carry it. Same copy, same colour, same branch.

sm/md are heights (40/48px), which is what every other primitive in the DS calls size. Calling it 'variant' made it read like a look, so call sites reached for className instead and forced the height by hand. Renames the prop and the local type, and moves SearchInput onto size="sm": it already overrode the horizontal padding with px-10, so the class list it renders is byte-identical. CopyField keeps its className h-10 — sm also narrows the padding to px-3, which would move its text 4px. The native `size` attribute is omitted from the props; nothing passed it.

No product screen has ever passed it. Its only caller was the /dev/ds page that documents it, so the doc page was the whole demand — and a documented prop reads as one somebody should use. A disabled back button is also not a state the navigation board draws: the flows that wanted it use hideBackBtn.

Both were private, so a caller building a props object for a Button or a Card had to retype the union or reach for `as const` and hope. Card's was also called ShadowSize while holding a different set of members than Button's — two names for two things is the rule, so Card's becomes CardShadowSize. Type-only change; the unions and every call site are untouched.

The comment said "peanut pink", and the file above it defines a peanut_pink palette entry whose background is #FFD5F6. Anyone matching the two would pick the wrong hex. #FF90E8 is action-primary.

The spec ran only inside ds-shots, which is deliberately not in ci-success.needs, so a change that broke the press state could merge green. The contract has broken twice and a class-name unit test cannot catch it: the bug is which state wins in the cascade, readable only off computed style in a real browser. Adds one lean job that builds and runs that single spec, and puts it in ci-success.needs. ds-shots keeps its advisory role.

vercel · 2026-09-21T17:09:57Z

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
peanut-wallet	Ready	Preview	Sep 22, 2026 8:02am UTC

coderabbitai · 2026-09-21T17:09:59Z

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository: peanutprotocol/peanut-ui/.coderabbit.yaml

Review profile: CHILL

Plan: Advanced

Run ID: 3e7bad1b-c94a-4c42-8633-6d281be69ff7

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

🔍 Trigger review

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

_{Comment @coderabbitai help to get the list of available commands.}

github-actions · 2026-09-21T17:11:37Z

Code-analysis diff

Painscore total: 8694.58 → 8705.18 (+10.6)
Findings: -2 net (+60 new, -62 resolved)

🆕 New findings (60)

high hotspot — src/components/Profile/index.tsx — 63 commits, +431/-339 lines since 6 months ago
high hotspot — src/components/Home/ActivationCTAs.tsx — 62 commits, +961/-421 lines since 6 months ago
high hotspot — src/components/Profile/views/UnlockPayments.view.tsx — 60 commits, +1714/-953 lines since 6 months ago
high hotspot — src/components/Global/QRScanner/index.tsx — 55 commits, +1079/-621 lines since 6 months ago
high hotspot — src/app/(setup)/setup/page.tsx — 49 commits, +689/-346 lines since 6 months ago
high hotspot — src/app/shhhhh/ShhhhhLandingPage.tsx — 49 commits, +1346/-1139 lines since 6 months ago
high complexity — src/components/0_Bruddle/Button.tsx — CC 49, MI 51.95, SLOC 82
high hotspot — src/components/AddWithdraw/DynamicBankAccountForm.tsx — 48 commits, +1196/-955 lines since 6 months ago
high complexity — src/components/Global/QRScanner/index.tsx — CC 47, MI 64.29, SLOC 182
high complexity — src/features/limits/components/LimitsWarningCard.tsx — CC 42, MI 59.55, SLOC 135
high complexity — src/components/Setup/components/SetupWrapper.tsx — CC 40, MI 61.91, SLOC 78
high method-complexity — src/components/0_Bruddle/Button.tsx:113 — CC 38 SLOC 46
medium react-long-component — src/app/(mobile-ui)/dev/ds/primitives/button/page.tsx:15 — ButtonPage is 362 lines — split it
medium high-mdd — src/app/(setup)/setup/page.tsx:59 — SetupPageContent: MDD 105.4 (uses across many lines from declarations)
medium high-mdd — src/components/Profile/index.tsx:32 — Profile: MDD 94.3 (uses across many lines from declarations)
medium high-mdd — src/components/Send/link/views/Success.link.send.view.tsx:26 — LinkSendSuccessView: MDD 64.5 (uses across many lines from declarations)
medium high-mdd — src/components/0_Bruddle/Button.tsx:113 — : MDD 54.1 (uses across many lines from declarations)
medium high-mdd — src/app/dev/devices/page.tsx:44 — Harness: MDD 39.9 (uses across many lines from declarations)
medium high-mdd — src/components/Badges/BadgesRow.tsx:38 — BadgesRow: MDD 35.8 (uses across many lines from declarations)
medium high-mdd — src/components/Global/QRScanner/index.tsx:176 — ScanRegionOverlay: MDD 31.3 (uses across many lines from declarations)

…and 40 more.

✅ Resolved (62)

src/components/Home/ActivationCTAs.tsx — 61 commits, +960/-420 lines since 6 months ago
src/components/Profile/index.tsx — 61 commits, +429/-336 lines since 6 months ago
src/components/Profile/views/UnlockPayments.view.tsx — 59 commits, +1713/-952 lines since 6 months ago
src/components/Global/QRScanner/index.tsx — 54 commits, +1070/-614 lines since 6 months ago
src/components/0_Bruddle/Button.tsx — CC 49, MI 51.59, SLOC 85
src/components/AddMoney/components/MantecaAddMoney.tsx — 48 commits, +375/-158 lines since 6 months ago
src/components/Global/TokenSelector/TokenSelector.tsx — 48 commits, +892/-774 lines since 6 months ago
src/hooks/useHomeCarouselCTAs.tsx — 48 commits, +534/-374 lines since 6 months ago
src/components/Global/QRScanner/index.tsx — CC 47, MI 64.3, SLOC 182
src/features/limits/components/LimitsWarningCard.tsx — CC 42, MI 59.56, SLOC 135
src/components/Setup/components/SetupWrapper.tsx — CC 40, MI 61.92, SLOC 78
src/components/0_Bruddle/Button.tsx:123 — CC 38 SLOC 46
src/app/(mobile-ui)/dev/ds/primitives/button/page.tsx:15 — ButtonPage is 383 lines — split it
src/app/(setup)/setup/page.tsx:59 — SetupPageContent: MDD 105.2 (uses across many lines from declarations)
src/components/Profile/index.tsx:32 — Profile: MDD 94.5 (uses across many lines from declarations)
src/components/Send/link/views/Success.link.send.view.tsx:26 — LinkSendSuccessView: MDD 64.9 (uses across many lines from declarations)
src/components/0_Bruddle/Button.tsx:123 — : MDD 54.1 (uses across many lines from declarations)
src/app/dev/devices/page.tsx:44 — Harness: MDD 38.9 (uses across many lines from declarations)
src/components/Badges/BadgesRow.tsx:38 — BadgesRow: MDD 36.6 (uses across many lines from declarations)
src/app/(mobile-ui)/dev/ds/_components/DocSidebar.tsx:12 — DocSidebar: MDD 33.5 (uses across many lines from declarations)

…and 42 more.

github-actions · 2026-09-21T17:14:28Z

🧪 UI test report — ✅ all green

Suites

✅ unit: 9168 ran, 0 failed, 0 skipped, 2.3m

📊 Coverage (unit)

metric	%
statements	81.2%
branches	70.0%
functions	76.1%
lines	82.3%

⏱ 10 slowest test cases

time	test
🐢 9.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › Network failure keeps loading while retries remain, then shows the generic error
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › MANTECA_USER_NOT_PROVISIONED fails fast with copy that names the real cause
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › MANTECA_MERCHANT_RECENT_REFUND fails fast with copy that names the real cause
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › MANTECA_SOURCE_OVER_MONTHLY_CAP fails fast with copy that names the real cause
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › User KYC not approved fails fast with copy that names the real cause
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › MANTECA_MERCHANT_VOLUME_NEAR_CAP fails fast with copy that names the real cause
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › a refused idempotency key tells the user to scan again, not to contact support
4.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › routes the KYC rejection on its wire code, and does not retry it
3.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › Going offline blames the connection, and reconnecting clears it for the recovered scan
3.0s	`src/app/(mobile-ui)/qr-pay/__tests__/qr-pay-states.test.tsx` › Scan that recovers on the retry lands on the payment screen, not an error

_{📍 Inline annotations are in the **Unit test report** check above. Coverage artifact: `coverage-unit`. Generated by `.github/workflows/tests.yml`.}

chip-peanut-bot

Chip review — no blocking findings — this is not an approval

Seven design-system drift fixes, all rename/delete/comment plus one new required CI gate. The mechanical sweep is complete and typechecked (nothing outside the intentionally frozen audit fixture still uses the old names), the SearchInput/CopyField padding calls were verified against twMerge conflict rules, the i18n key rename touches exactly the three locales that carry it, and no PostHog consumer reads the one capture value that changes. Two major findings: the mono design rulebook still records the opposite naming resolution for the button variants, and the new required press-contract job drops the apt-stall retry its own template documents.

Findings

MAJOR · src/components/0_Bruddle/Button.tsx:94 · DS rulebook still records stroke/transparent as the code names
The rename reverses a recorded resolution without updating the record. mono design/design.md's open-questions row for button variant names (TASK-22817) says "stroke=secondary and transparent=ghost keep their names", and its recipes still instruct variant="stroke" / variant="transparent" (design.md:144, 224, 489; components.md:17). The open mono PR #220 from this author corrects only the press colour, not the naming row. After this merges, the rulebook that agents build screens from contradicts the code union: misuse fails typecheck loudly, but every recipe line keeps teaching the old vocabulary — re-creating the two-vocabulary translation cost this PR exists to remove — and the reversal of a recorded ruling goes unrecorded (design.md law 6: never silently decide). Fix: extend mono #220 or open a mono PR updating the variant-names row, the recipes, and components.md.
MAJOR · .github/workflows/tests.yml:794 · press-contract's install-deps step drops the documented apt-stall retry
ds-shots documents three consecutive 6-minute npx playwright install-deps stalls on the runner's Azure mirrors (2026-08-19, PR #2755) and hardened its step with 3 attempts, a mirror fallback after the first failure, and a 9-minute cap (tests.yml ~509-533). press-contract copies the step without any of that. ds-shots is advisory, but press-contract is in ci-success.needs, so the same transient stall now fails a required check and blocks merges until a manual re-run — the exact failure mode the hardening exists for, moved onto a gate. Fix: copy ds-shots's retry block into this step.

Checked clean

Worktree HEAD verified at 9fd74aa; full 126-file diff vs base f9372c0 reviewed, including the two stacked #3301 nav commits it carries
Rename completeness: no variant="stroke"/"transparent" and no BaseInput variant prop survives outside the intentionally frozen dev/ds/audit fixture; no raw .btn-stroke class remains in product or e2e code
SearchInput size="sm" swap verified: twMerge yields the same final class list (h-10 + px-10) as the old h-10 override, matching the byte-identical claim; CopyField's h-10 correctly left alone to keep 16px padding
i18n: warningTitle -> attentionTitle in exactly the three locales carrying limits.warningCard; es-AR never had the key (same fallback as before); the unrelated crypto-deposit warningTitle untouched
PostHog production checked: no insight or dashboard filters on limits_check_link_navigation or its type property, so the 'warning'->'attention' capture value change breaks no consumer
CI at this head: press-contract green, plus unit, typecheck, eslint, ds-lint, screen-tests, format, native-export, human-authors, bot-approval; ci-success green; ds-shots still in flight (advisory, PR expects zero diffs)
e2e rewrite matches the 2026-09-21 one-colour nav-circle ruling (documented in open mono #220); SetupWrapper circles + NAV_CIRCLE_BUTTON_CLASSES twMerge precedence verified and the /setup/finish white-chip regression covered by the new test
AVATAR_LINK_BG comment verified against the token table: #FF90E8 is action-primary; the peanut_pink palette entry's lightShade is #FFD5F6
LimitsWarningCard: no caller still passes type="warning"; getLimitsWarningCardProps is the only type producer and now emits 'attention'

Security review by moonshotai/kimi-k3: 0 finding(s), marked with the model name. It reads the diff only and answers only security, privacy and money, so treat its findings as advice.

Third opinion by claude-opus: 0 finding(s), marked with the model name. It answers only product truth, missing tests and the cross-repo contract, so treat its findings as advice.

The usual first reviewer was out of plan, so this review was done by openrouter/z-ai/glm-5.3.

Exact head: 9fd74aaf7c8e · Context: repo, mono, notion, production · Took 13m

github-actions · 2026-09-21T17:27:15Z

🖼 Visual diff — 3 screens moved

5 of 146 shots changed · 141 identical · baseline 3e2ce1e → head d44fb85

worst %	screen	widths
48.91%	`early-user`	320
43.12%	`avatar-picker`	320, 430
1.48%	`guest-invite`	320, 430

job summary · before/after/diff images — artifact

_{Fixture screenshots, no backend. Advisory — this check never blocks a merge. Posted from the default branch by ds-shots-comment.yml; the report it renders is untrusted data.}

github-actions · 2026-09-21T21:46:41Z

English · Español · Español (Argentina) · Português (Brasil)

Open screen library dashboard

After merge: 3e2ce1e → b194ae8. Capture complete in all locales.

UserHeader had zero call sites — the home header was rebuilt during the consolidation and nothing renders it any more. All 12 imports of the module pull VerifiedUserLabel instead, which stays and keeps its import path. It was also the fifth primary-soft call site, so removing it is one less thing to migrate in the variant collapse.

kush 2026-09-22: the union is the board — primary/secondary/ghost. The three legacy variants had no board row and kept a fourth, fifth and sixth way to ask for a button that the board already answers. primary-soft -> secondary (4 sites). primary-soft was bg-white with no shadow, so every call site paired it with shadowSize="4"; --color-background-default is #ffffff and .btn-secondary already ships the same 4px shadow, so the swap is pixel-identical at rest and the redundant shadowSize goes. The sites gain secondary's 1px residual shadow when disabled, which is the board's disabled state. transparent-dark -> ghost (3 sites). The fourth was not a button at all: the UnsupportedBrowserModal paste hint was an ActionModal cta with no onClick — a button that did nothing. It is now a footer note, which is also where "Then paste it in your preferred browser." has to read, after the copy cta. transparent-light -> ghost (QRScanner, 2 sites). The only visual change. The scanner overlay circles now use NAV_CIRCLE_BUTTON_CLASSES with white border, text and fill, so they match every other icon-only nav button: white ring and white glyph at rest, pink fill with a black glyph on press. The hard-coded fill="white" on the inner Icon goes with it. Also drops .btn-transparent-light and .btn-transparent-dark from globals.css and the three variants from the /dev/ds button page, whose usage counts were stale anyway — the variants list now names each row's job instead.

…te must not fail on a transient mirror stall (chip review)

dev's new RequestCreatedView is the one semantic conflict: it was written against the old variant vocabulary while this branch renamed it. stroke -> secondary and transparent -> ghost, the same pure rename 526d911 applied to every other call site. No visual change — the class bodies are untouched. That single type error is what failed typecheck, native-export, press-contract, ds-shots, Deploy-Preview and ci-success: all five build steps type check, so all five inherited it from the one file.

dev landed the ButtonVariant collapse (#3302) while this branch was open, so the two sides met on the same two marketing files. Conflict: src/components/Jobs/Careers.tsx. dev's only edit to the Ready? section was the variant rename stroke -> secondary (526d911); this branch had rebuilt the whole section into the marketing CTA card (mascot behind the Card, centered, inline buttons). Union rule: keep this branch's composition, take dev's variant name on top of it. Nothing of either side is dropped. press/page.tsx auto-merged for the same reason — dev renamed the download pill's variant, this branch added its w-auto/whitespace-nowrap classes on the same element, and git kept both. Swept the branch for the other legacy variant names dev retired (stroke -> secondary, transparent -> ghost, transparent-light|dark -> ghost, primary-soft -> secondary). The only live site left was the careers CTA button, which this branch added after dev's migration ran — it is renamed here. The remaining hits are the /dev/ds audit datasets, which quote historical class names as data and are allowlisted out of ds-lint. ds-lint baseline: two debt metrics move up, both recorded with reasons. offScaleSpacing 85 -> 87 is this branch's own arrears — the Ready? rebuild added pt-24/md:pt-28 without a baseline bump, so the ratchet was already red before this merge. That padding is the mascot's overhang reservation, the geometry-tied-to-what-it-overlaps category the counter already exempts for inset offsets. hoverNoActiveFiles 20 -> 21 arrives with dev (UnsupportedBrowserModal/useCopyLinkActions.ts, added in the variant collapse); 21 is still under dev's own baseline of 22. Gates: prettier clean, tsc clean, 719/719 jest suites pass, ds-lint ratchet ok. Checked /careers and /en/content at 375px — the CTA still reads as the white outlined button it did under 'stroke', and the tab track keeps its weld, its 1px gutter and its single scroll axis.

notion-workspace · 2026-09-22T10:10:52Z

DS: rename Button purple variant → primary (naming debt) + confirm hover=flat figma spec

dev collapsed ButtonVariant to primary | secondary | ghost (#3302): stroke became secondary, transparent became ghost, and transparent-light, transparent-dark and primary-soft were deleted with their call sites migrated. BaseInput's height axis became `size`. This branch recreates real call sites on every primitive and pattern page, so the merge is not a text merge: each recreation had to be read against the file it cites. Three conflicts: - DocSidebar.tsx — ours. The old sidebar is gone; the file is now a 15-line shim over DocNavList, and the mobile drawer it used to hold lives in DocNavDrawer, mounted from the layout. dev's side only renamed variants inside the body this branch deleted, so nothing was left to port. - primitives/button/page.tsx — both. Takes dev's three board rows and their roles, keeps this branch's usage count beside each one, and re-counts them against the merged tree: primary 199 (97 explicit plus 102 that pass no variant), secondary 52, ghost 24, over 275 Button call sites with /dev and tests excluded. The header said 120+; it says 275. The rows for the three deleted variants are gone. - scripts/ds-lint-baseline.json — dev's, unchanged. No number was hand-merged and no metric ends above dev's value. Recreations reconciled against the file each cites: PaymentSuccessView, Landing, BridgeTosStep, LockCardModal and the drawer, modal, toast and divider trigger buttons all read secondary now; Profile's nav circle and the DS nav drawer read ghost; ProfileEditField and the nav search read size="sm". The clip-gate comment and the layout comment name .btn-secondary, which is what globals.css calls it. One spacing change: the progress-bar recreation paired its label and bar with a 6px gap copied from the real file. That is one off-scale site the app already counts once, and a second copy pushed offScaleSpacing past dev's 87. The showcase uses the on-scale step instead — the quoted code never showed the wrapper, so nothing it claims changed. The audit pages are left alone. They are a dated snapshot of the codebase and dev did not touch them either; rewriting a record of what was true in September would not make it truer. Gates: tsc clean, 117 dev/ds tests pass, ds-lint ratchet ok, prettier clean.

kushagrasarathe added 7 commits September 21, 2026 22:29

docs(ds): AVATAR_LINK_BG is the brand pink, not peanut_pink

258a355

The comment said "peanut pink", and the file above it defines a peanut_pink palette entry whose background is #FFD5F6. Anyone matching the two would pick the wrong hex. #FF90E8 is action-primary.

kushagrasarathe requested review from 0xkkonrad, Hugo0, abalinda, innolope-dev and jjramirezn as code owners September 21, 2026 17:09

vercel Bot deployed to Preview September 21, 2026 17:15 View deployment

chip-peanut-bot Bot reviewed Sep 21, 2026

View reviewed changes

Comment thread src/components/0_Bruddle/Button.tsx

Comment thread .github/workflows/tests.yml

kushagrasarathe added 3 commits September 22, 2026 12:21

ci: press-contract inherits ds-shots' apt-stall retry — a required ga…

53572dd

…te must not fail on a transient mirror stall (chip review)

vercel Bot deployed to Preview September 22, 2026 07:02 View deployment

vercel Bot deployed to Preview September 22, 2026 08:02 View deployment

kushagrasarathe merged commit b194ae8 into dev Sep 22, 2026
32 of 34 checks passed

kushagrasarathe mentioned this pull request Sep 22, 2026

fix(ds): two leftovers from the button/input axis renames (TASK-22817) #3317

Merged

kushagrasarathe changed the title ~~chore(ds): finish the button axis rename + drift fixes (secondary/ghost, attention, input size, press gate)~~ chore(ds): finish the button axis rename + drift fixes (secondary/ghost, attention, input size, press gate) (TASK-22817) Sep 22, 2026

kushagrasarathe deleted the chore/ds-drift-fixes branch September 22, 2026 11:23

Hugo0 mentioned this pull request Sep 26, 2026

release(ota): ship dev (TASK-23054) to main on the existing binaries #3512

Merged

Navigation Menu

Repository navigation

chore(ds): finish the button axis rename + drift fixes (secondary/ghost, attention, input size, press gate) (TASK-22817) - #3302

chore(ds): finish the button axis rename + drift fixes (secondary/ghost, attention, input size, press gate) (TASK-22817)#3302
kushagrasarathe merged 11 commits into
devfrom
chore/ds-drift-fixes

kushagrasarathe commented Sep 21, 2026 •

edited

Loading

Uh oh!

vercel Bot commented Sep 21, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Sep 21, 2026 •

edited

Loading

Review skipped

Uh oh!

github-actions Bot commented Sep 21, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Sep 21, 2026 •

edited

Loading

Uh oh!

chip-peanut-bot Bot left a comment

Uh oh!

Uh oh!

Uh oh!

github-actions Bot commented Sep 21, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Sep 21, 2026 •

edited

Loading

Uh oh!

Uh oh!

notion-workspace Bot commented Sep 22, 2026

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

kushagrasarathe commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Stacks on #3301

A — the button variants are named after the board rows

B — the non-blocking limits state is 'attention'

C — BaseInput's height axis is size

D — disableBackBtn is gone

E — the shadow-size unions are exported

F — AVATAR_LINK_BG is the brand pink

G — the press-physics contract can now fail a merge

H — ButtonVariant is primary | secondary | ghost

Not done — each waits on a ruling

Gates

Uh oh!

vercel Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

coderabbitai Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review skipped

Uh oh!

github-actions Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code-analysis diff

🆕 New findings (60)

✅ Resolved (62)

Uh oh!

github-actions Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🧪 UI test report — ✅ all green

Suites

📊 Coverage (unit)

Uh oh!

chip-peanut-bot Bot left a comment

Choose a reason for hiding this comment

Chip review — no blocking findings — this is not an approval

Findings

Checked clean

Uh oh!

Uh oh!

Uh oh!

github-actions Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🖼 Visual diff — 3 screens moved

Uh oh!

github-actions Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

notion-workspace Bot commented Sep 22, 2026

Uh oh!

This branch was successfully deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

kushagrasarathe commented Sep 21, 2026 •

edited

Loading

C — BaseInput's height axis is `size`

D — `disableBackBtn` is gone

H — `ButtonVariant` is `primary | secondary | ghost`

vercel Bot commented Sep 21, 2026 •

edited

Loading

coderabbitai Bot commented Sep 21, 2026 •

edited

Loading

github-actions Bot commented Sep 21, 2026 •

edited

Loading

github-actions Bot commented Sep 21, 2026 •

edited

Loading

github-actions Bot commented Sep 21, 2026 •

edited

Loading

github-actions Bot commented Sep 21, 2026 •

edited

Loading