Repository navigation
Release UI: signup passkey screen, Welcome A/B test, signup/KYC/card fixes - #4083
Conversation
A KYC start could leave the drawer locked on "Loading" for minutes: the transport timeout stopped when headers arrived, so a stalled body, auth readiness or the native module import could wait forever. - The transport budget now covers the response body and error telemetry. - The whole KYC start has one 30 s deadline; late results are fenced so they cannot reopen the SDK. - Manteca Add Money and Withdraw pass the start error to the drawer and keep it open instead of closing silently. - After the Sumsub SDK closes without a submit, refresh identity once so the backend can heal a missing applicant link (TASK-24858). TASK-24163, TASK-24858
A US resident in a state Rain does not cover (Georgia) was told the card "isn't available for residents of your country". Rain issues cards across the US, so for a US resident a geo-blocked card is always a state block: name the state instead.
…save - TASK-24847: the automatic iOS passkey sign-in ran on every Landing mount and could outlive Landing, so its sheet kept popping up over Claim username. It now runs once per app launch, never after Sign up or Log in, and leaving Landing retires it and frees the native slot. A ceremony retired while its result settles commits no session. - TASK-24856: a failed profile fetch on /setup is "unknown", with a retry, not "signed out" (no more signup for a logged-in user). The signup email step tells a protected email replacement (step-up, verification code, replacement disabled) from a retryable failure, keeps the confirmed email and lets signup continue instead of failing the same request again. - TASK-24857: login verify 429/5xx are typed and retryable, keep the existing session and honour Retry-After; the login hydration budget now covers the per-request timeout; setup initialization retries once before the generic error screen. TASK-24847, TASK-24856, TASK-24857
- Transport: once response headers arrive, a body-read failure or budget expiry never resends a mutating request (no native/WebView fallback, no retry); a POST that got 201 is sent once. Statement downloads skip the body drain so a large export is not cut off and regenerated. - /setup: a profile-fetch error is "unknown" only when no profile is loaded, so a failed focus refetch no longer replaces the current step; a successful recovery resumes the post-login routing. - Email step: keep the current email only when the account has one; otherwise the save failure shows as before. - Login: once the passkey login has committed its token, retiring the automatic attempt no longer cancels it half way; it finishes hydration and key setup. - KYC: the refresh after the SDK closes sends the same scope as the poll and skips document, action and residence-change sessions; the Manteca drawers close when the SDK opens or demo succeeds. - Retry-After is capped at 60 s. TASK-24847, TASK-24856, TASK-24857, TASK-24163, TASK-24858
Round 2 skipped the identity refresh for action sessions, which also skipped clearing the user-initiated flag, so a delayed APPROVED after closing an unsubmitted action (supplement:bridge) called onKycSuccess and moved FeatureAction to feature-ready without the upload. Clear the flag on every close without a submit; the refresh stays skipped. TASK-24858
…ailed login The "Need to recover a Peanut account?" link sat on the first screen for everyone, new users included. Show it only after a Log in attempt fails: any login error, or a second dismissal of the passkey sheet (the drawer's existing "can't log in" signal). One dismissal or the silent automatic iOS attempt does not count.
…y residence Chip review: "isn't available in your state yet" promised issuance that product/card.md does not plan (Rain refuses 19 states, DC and the territories). Drop "yet" in en, es-419 and pt-BR. A verified US user with a pending residence change elsewhere can be refused by that change (strictest residence), so name the state only when every residence on file is US.
…ens the SDK Chip review: only the ordinary initiate branch closed the entry drawer, so for a restart-identity or fixable rejection the drawer stayed open under the Sumsub SDK and came back after it, and Add Money kept showing the document-request page instead of the amount step. The restart and fixable handlers now report whether they opened the SDK; both Manteca flows close the drawer on success and keep it, with the error, on a failed start. TASK-24163
…ogress Chip review: a start that finds verification already in progress opens the progress modal without the SDK, and the Manteca entry drawer stayed over it. Both Manteca flows now close the drawer whenever the status modal opens, whichever start handed off; a failed start keeps it with its error. TASK-24163
…y release Conflict: src/hooks/useSumsubKycFlow.ts, where main (#4070) added a restart fence (restartGeneration / restartPendingRef / captureSessionFence) and residence-review state while the branch added the start deadline, sdkSubmittedRef and the close-time identity refresh. Kept both: - unmount cleanup bumps restart, residence and start generations; - handleClose releases a pending restart (main) and then runs the branch's unsubmitted-close refresh and userInitiated reset; - restartIdentity keeps main's fence and the branch's Promise<boolean> contract: every stale-fence exit returns false, a resumed replacement returns true only when the fence still holds. Other overlapping files (MantecaAddMoneyFlow, ClosedRowDrawer, useMultiPhaseKycFlow, i18n) merged cleanly.
The Screen library only captured PRs into main when the head was dev (release PRs). Any same-repository, non-draft PR into main (hotfix and cherry-pick branches included) now gets the before/after capture and the review page, using the release path: before = the PR's base SHA, after = its head. Forks and drafts stay excluded; publish provenance checks are unchanged and now key on the PR head ref. One main PR runs eight capture jobs (four locales before and after); a compatible baseline can skip the before builds. Release port of the Screen Library main-PR capture (origin/innolope/screen-library-main-prs, dev-based). Applied cleanly onto main; no conflicts, no dev prerequisites. (cherry picked from commit efd4add)
…creen New order: Welcome → sell screens → Residence → Payment plan → Claim username → Invite → Passkey → Email → Notifications → Control. The account is now created on a dedicated Passkey screen (the old SetupPasskey behaviour: the button opens the phone's passkey sheet; "Learn more" opens the passkeys help drawer), not when the username is claimed. - Claim username validates the name and reserves it on the server (POST /users/username/reservations, x-username-hold device token); an unavailable endpoint, network error or rate limit falls back to the server's registration-time collision check. - Answers given before the passkey (residence, nationality, plan, username, invite selection, sell-screen country, screen) live in a versioned on-device draft for up to 30 days and survive restarts. Registration binds the draft to the new account and flushes it; failed saves are retried until acknowledged and never re-register. - Invite stores the selection before an account exists and accepts it after registration; Skip clears only the user's own selection. - Back is locked from Email on. Invite links start at the first screen after Welcome. Users already registered on the old order finish where they are; a registered user missing residence or plan is asked before Email. An unrelated login discards the anonymous draft. - SIGNUP_FLOW_VERSION 2; passkey events fire from the passkey screen. - Copy in en, es-419, pt-BR and the es-AR overlay; screen catalogue, fixtures, unit and e2e expectations updated. Release port of the signup reorder (origin/innolope/signup-passkey-later, dev-based, commit ce38683 on dev 23da334). The new POST /users/username/reservations endpoint ships with a separate backend release. Conflicts and resolutions (dev prerequisites NOT imported): - src/app/(setup)/setup/page.tsx: kept main/#4077's landing auto-login disable, profileFetchFailed recovery and initializationAttempt retry, and added the item's draft restore/bind, feature-screen pinning, legacy recovery boundary and draft-driven entry. Dropped dev's INVITE_TYPE_PARAM / PAYMENT_LINK browser-handoff code (TASK-24498, dev #3999), which the item only carried as context; main stashes a bare code as DIRECT. - src/components/Setup/Views/Passkey.tsx (new): removed dev's PasskeyProviderUnsupportedDrawer (TASK-24498, dev #3999, not on main). Main's useSignupPasskey reports an unsupported provider as an inline error, which the screen already shows; Passkey.test.tsx asserts that instead of the drawer's actions. - Signup.tsx / Signup.test.tsx: took the item's Claim screen (passkey sheet, help drawer and taken-username Log in move to Passkey). - Landing.tsx / Landing.test.tsx: kept #4077's landing-auto-login import next to the item's draft/setup-flow imports. - page.test.tsx, passkey-auth-capture.test.ts: kept both sides' tests. - scripts/screens/adapter.test.mjs: added only the item's reservation fixture test; the neighbouring Avenia demo-route test is dev-only. - e2e/flows/setup-top-bar.spec.ts: main's openSetupSurface helper gains a fixture argument so the passkey surface uses setup-passkey. (cherry picked from commit ce38683)
PostHog experiment welcome-headline-copy (4 variants, 25% each): today's copy as control, plus 'One account. Every border.', 'The whole world in one account.' and 'Money that travels free.', in en, es-419, pt-BR and the es-AR overlay. Only new users on the native apps are exposed: the flag is read once, after waiting at most 1.5 s for PostHog, and never for known devices or the web download page. A timeout or unknown value shows control without exposure. The variant is added to signup_signup_clicked as welcome_variant and registered in the dev screens flag registry. Release port of the Welcome headline A/B test (origin/innolope/welcome-headline-experiment, dev-based, commit d31d2a6 on dev 23da334). Conflicts: SetupWrapper.tsx and Landing.tsx, only because main has since changed the same lines (main's notification-permission min-h-0 classes from 62e46d5, and #4077's landing-auto-login next to the signup-reorder imports). Wrapped main's panel in WelcomeHeadlineExperimentProvider and kept main's classes; Landing gets the item's variant capture and pending cancel on top of the reorder/auto-login changes. No dev prerequisites. (cherry picked from commit d31d2a6)
Android passkey login and signup time out on ~6-8% of attempts and the analytics cannot tell which part of the 60 s guard is slow. Add per-attempt phase timings to the existing webauthn_ceremony event: options_ms, native_ms, verify_ms, total_ms, backgrounded_ms, an attempt_id, and on a timeout the phase that was in flight (timeout_phase). When an Android request settles after the app gave up, the existing late_settlement lifecycle event now carries the same attempt_id and native_settled_after_timeout: true so the two can be joined. No behaviour change: same timeouts, errors, UI and retries. Telemetry is per attempt, wrapped so it can never throw into the ceremony, and sends no challenge, credential id, username or origin. Release port of passkey phase timing, part 1 (f37c643 from origin/innolope/passkey-phase-timing, dev-based). Conflicts and resolutions: - passkeyCeremony.utils.ts: wrapped main's guard (including #4077's "retired login must not commit a session" re-check after the race) in the item's timing begin/activate/operation/timeout/finish calls. - passkey-auth-capture.ts: added the item's telemetry import; dropped the PasskeyProviderUnsupportedError import, which belongs to dev's TASK-24498 (#3999) register/verify handling that main does not have. - webauthn-ceremony-telemetry.ts: took the item's timing-aware error path without dev's describeRefusal (challengeOffset/backupEligible, #3999). Dev prerequisite brought in: the 3-line currentCeremonyPurpose() helper from dev #4071 (TASK-24840), which the item's timing code calls; it only names the existing purposeStack expression. #4071's signature check is not imported. (cherry picked from commit f37c643)
A registration whose options request failed before navigator.credentials ran committed a fallback record with the whole attempt as its duration, and the signup failure handler published that as ceremony_ms (OS-sheet time). Fallback records are now marked synthetic and latestCeremony only returns real credential calls; the attempt outcome still reports options_ms, total_ms and timeout_phase. Release port of passkey phase timing, part 2 (5dfefb2 from origin/innolope/passkey-phase-timing, dev-based). Conflict: useSignupPasskey.test.tsx mocks. Took the item's requireActual-backed telemetry mock (needed by its real-telemetry suite) and the Sentry mock with captureMessage, on top of the signup-reorder version of the suite. The isAndroidNativeBridge mock stays out: it only served dev's TASK-24498 drawer. (cherry picked from commit 5dfefb2)
…tity now returns main's refresh-owner tests (6c71651) hold handleRestartIdentity()'s promise as Promise<void>; ui#4077 (08b70af and follow-ups) made restartIdentity resolve to a boolean, so after both merged tsc reported TS2322 three times. Only the annotation changes. Port fix for the #4077 merge: main's useSumsubKycFlow.refreshOwner test (from #4070) typed restartIdentity as Promise<void>; with #4077 it returns Promise<boolean>, so tsc failed on this branch the same way it did on the dev-wide integration. (cherry picked from commit df2c7fe)
…atalogue The Welcome headline experiment registered three welcome-headline-copy variants in flag-variants.ts, which adds three 01-a-landing flag captures to the screen catalogue, but capture-catalogue.json was not regenerated, so capture-cost-integration's 'trusted generated catalogue matches' test failed. Regenerated from src/dev/screens/catalogue.ts; only the three new entries are added. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit 0b6aba9)
…nd stub its flag read The Welcome headline experiment (d31d2a6) adds welcome_variant to signup_signup_clicked and has SetupWrapper read PostHog flags on the native landing. Landing.test still expected the bare event (3 failures) and SetupWrapper.test's posthog mock had only capture, so the Android status-bar test threw on featureFlags.hasLoadedFlags. Both fail on the feature branch alone. Landing.test now expects { welcome_variant: 'control' }; the SetupWrapper mock reports flags loaded with no variant, which resolves to control. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit 4cae02d)
After the passkey-later merge Landing imports signup-draft, whose import chain (invite-stash -> ... -> zerodev.consts) calls isCapacitor() while modules load. Landing.welcome.test's capacitor mock reads mockNative, which is not initialised yet, so the suite failed to run (TDZ ReferenceError). Mock signup-draft with no stored draft, as Landing.test does on the passkey branch. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit ce1942b)
…t an inline style d31d2a6 hid the non-shown headline arms with style={{ visibility }}, which raised the ds-lint inlineStyle ratchet from 22 to 23 and fails 'ds-lint-counts --check'. Tailwind's visible/invisible set the same visibility property; the rendered result is unchanged. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit 575e79d)
…suite ce38683 added mockUseEnv, which calls useSyncExternalStore and runs only inside the mocked hooks. It has to keep its mock prefix to be used in jest.mock factories, so react-hooks/rules-of-hooks flagged it: the only eslint error in the repo. Disable that rule on the line, with the reason. Also drop a no-var disable directive eslint reports as unused (the config has no no-var rule). Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit 68c596b)
…diness baseline d31d2a6 adds three headline/description arms (setup.steps.landing. variants.*). The wordiness scorer counts the heaviest arm on every screen that reaches the Welcome title block, so verify/page.tsx went 398 -> 417 and setup/finish/page.tsx 82 -> 101, failing 'screen-wordiness --check' (the same failure happens on the feature branch alone). Only one arm shows at a time and the copy is the experiment itself, so the growth is deliberate: raise just those two entries. Revert this commit to make the experiment cut copy instead. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing). Conflict: the baseline differs between dev and main. Applied only this commit's two entries (verify 398->417, setup/finish 82->101) to main's baseline; `node scripts/screen-wordiness.mjs --check` passes. (cherry picked from commit 79b0ca2)
…hose profile read failed c968603 (#4077, TASK-24856) treats any failed /users/me on /setup as "account unknown" and shows Something went wrong + Try again instead of the entry step. On web the readable jwt-token cookie is the only credential the API sees (apiFetch sends no cookies), so a visitor with no token is signed out whatever the request did. With the API down, offline, or absent (the press-contract harness) they never reached Welcome: press-contract timed out waiting for Sign up. profileFetchFailed now also requires mayHoldSession(): a web token, or any native app, where a legacy session can live in the OS cookie jar that JS cannot read (see api-fetch). Signed-in web sessions and every native case keep the TASK-24856 recovery screen. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit 63642ad)
…p the setup press spec same-origin The app saves keep-web with saveToCookie (JSON, URI-encoded) and reads it with JSON.parse. Both specs set the raw token, the parse failed, and web signup stayed closed. The press spec only passed by clicking the preview "Sign up on web" link, which the offline toast covers for ~57 s (the test took 1.1 min of its 90 s budget). It now reaches the real Sign up and runs in ~6 s. The setup press test also aborts third-party requests, as drawer-shadow-clip already does: a country from ipapi.co with no API to supply restrictions hides both sell screens, so the step after Welcome depended on the runner's IP and ipapi's rate limit. No assertion changed. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit 83d64c5)
… unit run The CI unit job on #4081 failed on this one test: the child node --test run of scripts/screens/adapter.test.mjs was killed at the 30 s execFileSync timeout (32.3 s, stdout only "TAP version 13"). The suite takes ~17-19 s alone and passed in the screen-tests job. Child limit is now 90 s, Jest's 95 s, and a kill names its signal. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit a42cfd7)
…ame refusal The API now refuses new usernames that contain "peanut" or equal official, verified, helpdesk or customercare, with 400 USERNAME_INVALID and reason "This username is reserved". - usernameRules (Claim hint) adds the rule in en, es-419, es-AR, pt-BR. Curly quotes match how these catalogues quote terms. - reserveUsername returns 'refused' for that reason. Claim shows "This username is reserved. Try a different one." (all four locales) instead of the generic invalid error; Passkey sends a refused name back to Claim like a taken one. Other USERNAME_INVALID reasons are unchanged. Follow-up from the dev-wide integration (origin/innolope/passkey-phase-timing), applied cleanly. (cherry picked from commit b0e3757)
… again The Create passkey points titled each item in Label/M (12/16, 800), smaller than the 14px description under it. They now use the DS BulletList, the list for unordered claims, with Body/S semibold titles and the descriptions in its secondary Body/S. The username field's placeholder goes back to "Username" (es-419 "Usuario", pt-BR "Usuário"), as before 47b6d1b. Since then a bold "batman" sat in the field for four seconds and then turned into the animated, regular weight "batman?", so the first idea looked like it jumped. The separate accessible label from that commit stays.
|
…outcome unknown" The transport budget now covers the response body, so a send, withdraw or pot POST that got its headers and then stalled surfaced as a plain ConnectionTimeoutError: the UI said it failed and Sentry logged a connectivity failure although the server had already committed. fetchWithSentry now throws MutationResponseTimeoutError (a ConnectionTimeoutError with the HTTP status, responseReceived and outcomeUnknown) for that case, reported under its own fingerprint at warning level with response_received/request.outcome tags. Direct send, request payments, pot contributions and crypto withdrawal re-read the known charge and show the existing confirming view instead of an error; the payment recorder no longer turns it into a failure message. Manteca bank withdrawal, Bridge, Rain and deposit recovery keep their current handling: they have no charge to re-read yet.
…ore failing The 30 s start deadline raced every call against one shared timer without cancelling anything: a late initiation still created or reopened the session server-side, the UI showed errorInitiateFailed, and a second tap started another one. A response that did arrive was dropped by the wall-clock post-check. Each stage (plan read, paperwork route, resume, initiate) now has its own bound and passes an AbortSignal through serverFetch, so a timeout stops the request before dispatch or in flight. Arrived results are kept. Before showing a failure the flow re-reads the KYC status and known session and resumes it, and an unknown outcome stays fenced against another initiation on re-tap. authReady() can still park inside callApi; the hook bounds that wait and blocks late dispatch.
|
…ersion callApi awaited authReady() and the cached binary version before dispatch without looking at options.signal, so a caller that aborted (the KYC start deadline) left its request parked there and could still dispatch once auth settled. Each readiness wait now races the caller's signal: an abort rejects that caller with the signal's reason and removes its listener, while the shared gates stay untouched for everyone else. An already aborted signal never dispatches, and a cancel still reports no timing failure.
|
…age in the wordiness baseline 4137f28 lets crypto withdrawal re-read a charge whose record request timed out after the server answered. When that charge comes back failed, the screen shows payment.errors.externalPaymentNotConfirmed ("This payment could not be confirmed. If money left your wallet, contact support."), 13 words it never showed before. The growth is deliberate: it is the only place that tells the user what to do if funds may have moved. The update also locks in three deposit-account screens that got shorter.
|
… and errors Audit of the sign-up flow against the "every control reports" rule (48372ec). Each registered step already sent signup_step_viewed; the gaps were errors, drawers and a few controls: - Create passkey: reservation, preflight, ceremony (incl. cancel and timeout), username-taken and log-in errors, and the help drawer opening. - Welcome: login error categories, and the settled welcome_variant on later captures. - Residence: the refused-issuer warning and the compare sheet open/dismiss. - Claim username and Invite: a repeated error is reported again after an edit clears it. - Email: every error goes through trackError. - Notifications (both variants): the toggles, Continue, Not now, and denied/dismissed permission and preference-save errors. - Final confirmation, celebration, the standalone finish page and the setup-error recovery: their step views and errors. - Docs drawers (passkeys, terms, privacy, recovery): article-link taps; login recovery: drawer open. No new event family; drawers keep their parent screen id.
|
bcbb949 passes an AbortSignal as initiateSumsubKyc's third argument so a timed-out KYC start cancels its request. The two correction-flow tests still expected a single argument and failed on CI's full unit run; they now assert the signal, like useSumsubKycFlow.test.ts does.
|
Features
welcome-headline-copy, a draft until launched).main, hotfixes included, gets a before/after capture and review page. The page is posted bydev's copy of the publisher, so it starts working once peanut-ui#4084 merges intodev. Until then this PR captures both sides but publishes no page.What's not here
main: the TASK-24702 paperwork fixes (fix(TASK-24702): recover paperwork questions and blocked setup #4063, via fix(TASK-24702): recover paperwork questions and blocked setup #4064) and the Add Money, bank setup, residence review and card issuance work (fix: Add Money picker and analytics, bank setup, residence review, card issuance, and signup layout #4065, via fix: Add Money picker and analytics, bank setup, residence review, card issuance, and signup layout #4070).dev's Avenia Pix work, which isn't in this release, so Pix stays on the Manteca rail as onmaintoday.Checks
CI runs typecheck, lint, unit, e2e and the native export on this PR. Earlier runs of the same changes on a clean Linux box:
press-contracte2e spec passed.Replaces
#4069, #4077, #4066, #4063, #4065 and the earlier all-of-dev draft #4081.
Screen Library copy
The three changed strings no state showed now have their own capture states (5c7e9e8). The copy baseline also lists 15 older
mainkeys that no state shows yet:addMoney.crypto.warningTitle,card.status.stillCreatingTitle,depositAccounts.list.countriesPitch,identity.reasons.verified-name-missing,kyc.documents.bank,kyc.documents.local,kyc.paperworkVerification.addLater.doneTitle.bank,kyc.paperworkVerification.addLater.doneTitle.local,kyc.paperworkVerification.addLater.title.bank,kyc.paperworkVerification.addressMinimum,setup.landing.recoverWallet,setup.notifications.changeAnytime,setup.notifications.emailDescription,setup.notifications.emailLabel,setup.notifications.offersDescription.Screen Library copy exception: 15 older main copy keys have no capture state yet; the three strings this release changes each have one.
Merge order: this PR and the backend release peanut-api-ts#2138 can merge in either order. Username holds need a backend endpoint that isn't in that release (it goes to
devin peanut-api-ts#2134). Until it reachesmain, signup continues without a hold, as it does today, and the server still refuses duplicate usernames at registration.Screen wordiness
Crypto withdraw grows from 127 to 140 (6187f98). When a withdrawal's record request times out after the server answered and the charge then reads back as failed, the screen now shows "This payment could not be confirmed. If money left your wallet, contact support." It is the only place that tells the user what to do if funds may have moved. Three deposit-account screens got shorter and are locked in.