[takemiyamakoto]

Summary

Fixes every test failure from the latest fearlessTests runs: GitHub's build job on master (dce8790cd) and on #1320, and the local full runs. None came from #1320's change. The full suite now passes locally at this branch's head: 3,599 tests, 0 failures.

Japanese Welcome text drew Chinese glyph forms (GitHub: 3 OnboardingStartTests)

• The Welcome screen draws Japanese in the system font, and CoreText picks fallback glyphs for kanji by the device language. On a device not set to Japanese, such as GitHub's English simulator, they came out in PingFang SC, the Chinese forms. Local simulators inherit this Mac's ja-JP, so they passed.
• FWWelcomeView and FWWelcomeActionButton now tag their text with the app language (kCTLanguageAttributeName). The reference test tags its samples the same way.

Bold Text left the HUD at its normal weight (GitHub: FWDesignSystemTests)

• The bundled JetBrains Mono family has only Regular and Medium. withSymbolicTraits(.traitBold) returns nil, so a bold HUD fell back to Medium, which is its normal weight. Outside Japanese and Chinese, Bold Text did nothing to the HUD.
• A bold HUD now uses the system monospaced face at Semibold, as the existing code already does when the bundled font is missing.
• The bold check runs for explicit English and Japanese locales.

PIN change could navigate after its screen was popped (local: PinPersistenceTests, long known as "fails under load")

• popViewController takes the PIN screen off the stack and detaches it at once. UIKit calls its viewWillDisappear, which cancels the presenter, only at the next layout pass, which I measured with a probe test. A save acknowledged in between still called showMain.
• PinSetupViewController.willMove(toParent: nil) runs during the pop itself and now cancels the create/change presenter.
• Results from repeated runs of the original test: 3/20, 6/50 and 7/50 failed before; 100/100 pass after.

Profile PIN prompt race (GitHub, on #1320 only: ProfileAuthorizationLifecycleTests)

• After an authorization screen leaves, the wireframe releases its pending request a moment later. By design, it ignores a Change PIN tap until then, and the test tapped as soon as the screen had gone.
• Every open in that class now waits for the release. That wait still fails a test whose dismissed request is never released.

Clipboard hang and a relaunched, inactive test host (local: 4 SendDependencyErrorSurfacingTests)

• A read of the system pasteboard blocked the host's main thread for 2 minutes; the spindump shows -[_UIConcretePasteboard items] in a semaphore wait. Content another app wrote, such as the Mac clipboard that Simulator syncs, waits for a paste-permission answer.
• XCTest then relaunched the host, and the next Send form tests ran before its scene was active.
• SendPresenter and WalletSendConfirmPresenter now take a PasteboardStringAccessing, which defaults to UIPasteboard.general. The tests pass an in-memory one, and the Solana fixtures wait for an active scene.

Run configuration only (local: 13 tests, no code change)

• 11 keychain and app-group tests fail when CODE_SIGNING_ALLOWED=NO drops the simulator entitlements. GitHub's job and scripts/test-matrix.sh build signed, and so do the runs below.
• 2 AccountConfirmPresentationTests cases take about 127 s at this Mac's load, past a 2-minute allowance an earlier run set. The runs below allow 5 minutes.
• AGENTS.md notes these traps, how to match GitHub's English runs (-testLanguage en -testRegion US), and system-pasteboard reads in tests.

Transaction-builder audit (GitHub: the build job stopped at this pre-build gate on this branch)

• scripts/audit-transaction-builder-tests.sh required the literal UIPasteboard.general.string = recovery.messageHashHex, which the pasteboard injection changed.
• It now requires pasteboard.string = recovery.messageHashHex and the UIPasteboard.general default.
• Its harness has a 141st destructive fixture, which fails a presenter whose default is redirected.

Evidence

A CoreText probe run inside the simulators gives the same result on iOS 18.4, 26.2 and 26.5:

• Untagged 新しいウォレットを作成 in the system font gives .HiraKakuInterface-W6 under ja-JP. Under English it gives .PingFangUIDisplaySC-Semibold and .CJKSymbolsFallbackSC-Semibold, the names in GitHub's failure messages.
• Tagged ja, it gives Hiragino under both languages.
• JetBrainsMono-Medium with .traitBold is nil.

On master's code the font tests fail as GitHub shows: run 41823644385 (master push) and run 42376841255 (#1320), on iOS 26.2 in English.

Tests

All runs are local, with a signed simulator build and a 300 s per-test allowance. No hosted CI is required.

Run	Commit	Simulator	Language	Result
Full fearlessTests	1a2437833 (head's app and test code; 893c7b9fd only changes scripts/)	iOS 26.5 iPhone 17	ja-JP	3,599 tests: 3,525 passed, 74 skipped, 0 failed
The changed classes plus the PIN classes	1a2437833	iOS 26.2 iPhone 16e (GitHub's runtime)	-testLanguage en -testRegion US	234/234 passed
testActualNavigationPopFencesPendingPinChangeCompletion ×100	1a2437833	iOS 26.5	ja-JP	100/100 passed
GitHub's 26 audit and contract steps (scripts/*-audit*.sh, remote config, native source, XCM boundary)	893c7b9fd (head)	—	—	all pass
11 keychain/app-group tests + AccountConfirmPresentationTests + the changed classes	d09ce3f2c content	iOS 26.5	ja-JP	236/236 passed

Not covered:

• The iOS 18.x check that AGENTS.md asks for UIKit text changes could not run. This Mac's Xcode 27.2 Debug build doesn't launch below iOS 26 (dyld: Library not loaded: /usr/lib/swift/libswiftWebKit.dylib; see below). The CoreText probe above did run on iOS 18.4.
• The iOS-native Authorized RPC contract suite and the Iroha SDK download check were not rerun: this branch doesn't touch Native/, and the download check needs the network.
• Two full-suite attempts were cut short by this Mac's CoreSimulatorService, which was restarted at 04:34 and sent SIGTERM at 06:03, neither by this session. The suite was rerun from the start each time.

Not in this PR

• Other screens don't tag their text, so their Japanese kanji still follow the device language.
• Other test classes still read the system pasteboard; ExactTonMnemonicExportTests does so on purpose.
• The Xcode 27.2 Debug build declares iOS 15.0 as its minimum, yet its debug dylib strongly links /usr/lib/swift/libswiftWebKit.dylib and @rpath/libswiftCompatibilitySpan.dylib, and the app embeds neither. Release archives were not checked.

Risk and rollback

• Welcome text gets a language attribute.
• Bold HUD text changes face only when Bold Text is on and the language is neither Japanese nor Chinese.
• The PIN create/change screen cancels at pop time instead of one layout pass later. Authorization mode is unchanged.
• The pasteboard parameter defaults to the system pasteboard.
• Rollback: revert the commits.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 95.55556% with 2 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (master@dce8790). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...WalletSend/SendDependencyErrorSurfacingTests.swift	92.59%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff            @@
##             master    #1322   +/-   ##
=========================================
  Coverage          ?   78.19%           
=========================================
  Files             ?      285           
  Lines             ?   142070           
  Branches          ?        0           
=========================================
  Hits              ?   111088           
  Misses            ?    30982           
  Partials          ?        0           

Files with missing lines	Coverage Δ
...lessTests/Common/Helpers/FWDesignSystemTests.swift	100.00% <100.00%> (ø)
...Modules/OnboardingStart/OnboardingStartTests.swift	98.59% <100.00%> (ø)
...lessTests/Modules/Pincode/PinContinuityTests.swift	95.25% <100.00%> (ø)
...WalletSend/SendDependencyErrorSurfacingTests.swift	97.71% <92.59%> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[takemiyamakoto]

Folded into #1326, which brings every open iOS PR into master in one change, so one code-owner approval covers it. This PR's branch is kept.