Fix mobile thread tail and iOS channel header

[klopez4212]

Summary

• Hide the thread Latest control once the lazy tail is fully reached.
• Polish the mobile channel header and use native iOS liquid glass for Back.
• Keep the two-line header aligned and safe at larger text sizes.

Validation

• just mobile-check
• flutter test (1,553 tests)
• Signed iPhone build installed and launched

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d58650344a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[klopez4212]

Carl reviewing on Kenny Lopez's behalf.

Requesting changes for two release-blocking regressions in the new native iOS navigation control:

1. P1 — the pre-iOS-26 fallback can render critical controls effectively invisible. NavigationGlassButtonPlatformView forces secondarySystemBackground, while the camera, inline picker, and media viewer explicitly request a white glyph. On the supported iOS 16–25 light appearance, that is white on a nearly white surface. This is captured in Fix mobile thread tail and iOS channel header #6399 (comment).
2. P2 — the native actionable area is only 40×40 points. The Flutter wrappers are 44–58 points, but taps outside the constrained UIButton land on the inert container, regressing the prior 44–48-point hit targets and falling below the intended minimum. This is captured in Fix mobile thread tail and iOS channel header #6399 (comment).

Both findings affect UI/interaction or accessibility. Per Kenny's instruction, I have not changed them pending explicit confirmation. The thread-tail state change itself is bounded and covered by the added regression assertion; I found no separate blocker there. Mobile CI passes at reviewed head d58650344a0b86c892626ef09ce55d494f678405.

Non-blocking cleanup: cupertino_icons is newly declared but not imported anywhere under mobile/; it can be removed unless a follow-up intends to use it.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: eac4268d13

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[jedwards27]

:bot: Jude’s code review agent

Verdict: REQUEST CHANGES

Reviewed: e5d1dfef7bf24ad527c9c8c1785b613abad574f7..eac4268d13aeab857971de66035e6b73fa50ce74 (exact live head eac4268d13aeab857971de66035e6b73fa50ce74)

Risk: high — this changes virtualized thread-tail state plus a Flutter↔UIKit navigation control across supported iOS versions and Android header layout.

Behavior/contracts traced: explicit Latest navigation and lazy viewport growth; stale completion/drag cancellation; automatic reply-follow separation; iOS platform gating and availability fallback; method-channel/view lifecycle and hit testing; appearance updates; Android/iOS header geometry.

Blocking findings

1. P2 — supported light accents lose required contrast on the pre-iOS-26 navigation fallback. IosGlassNavigationButton sends context.colors.primary by default (mobile/lib/shared/widgets/ios_glass_navigation_button.dart:54-56), while usesDarkFallbackSurface only recognizes near-white grayscale and otherwise retains secondarySystemBackground (mobile/ios/Runner/JumpToLatestGlassButton.swift:258-289). In light appearance, supported Cyan #06B6D4, Green #22C55E, and Lilac #C0A2F1 (mobile/lib/shared/theme/accent_colors.dart:34-68) contrast against #F2F2F7 at approximately 2.18:1, 2.04:1, and 1.94:1. That is below the 3:1 non-text UI threshold for critical Back/Close glyphs on iOS 16–25. Choose the fallback foreground/background from measured contrast for every supported accent, and parameterize native tests across the accent catalog and light/dark appearances; current native coverage exercises only white and system blue (mobile/ios/RunnerTests/RunnerTests.swift:28-47).

2. P2 — pushed Android channel headers receive iOS-only title spacing. Native leading is correctly gated by usesNativeIosGlassBackButton (mobile/lib/features/channels/channel_detail_page.dart:268-270,328-337), but title padding independently checks platform-agnostic Navigator.canPop (:341-345). On Android, the existing 48dp Flutter Back therefore gets an extra iOS spacer, shifting the title and causing earlier truncation for long channel names. The thread header already uses the correct platform-gated condition (mobile/lib/features/channels/thread_detail_page.dart:839-843). Gate this spacer on usesNativeIosGlassBackButton and cover narrow/long-title Android and iOS layouts.

Non-blocking coverage gap

The new expanding-lazy-tail test proves repeated correction but not the extentAfter fallback at thread_detail_page.dart:400-409: removing || reachedScrollExtent left thread Latest settles across expanding lazy scroll extents green. The explicit Latest latch itself is mutation-proven, so this is not an additional blocker, but the fallback needs a discriminating regression row if it is intended to be load-bearing.

Validation at the exact clean head

• PASS just mobile-check (format 428 files / 0 changed; analyze clean).
• PASS full just mobile-test (1,557 tests).
• PASS changed mobile test set, including the complete 193-test channel-detail suite (193/193).
• PASS flutter build ios --simulator --debug --no-pub; produced Buzz.app and compiled the Flutter↔Swift registration/availability path.
• PASS git diff --check.
• Mutation: removing the explicit-navigation visibility latch caused thread shows Latest after browsing history and returns to tail to fail at line 9131 as expected; restoring the exact head returned the tree clean.
• CI Mobile is green at this head. One Desktop Smoke shard is failed and Desktop Core remains in progress; neither establishes this mobile boundary.

Manual/native evidence: no reviewable exact-head signed-device journey was provided. “Signed iPhone build installed and launched” does not identify device/runtime/artifact or demonstrate Back/Latest interaction, VoiceOver, fallback appearance, narrow layout, or largest Dynamic Type.

Residual risk: physical signed-device behavior; VoiceOver focus/activation; current-iOS glass and iOS 16–25 fallback appearance across accents; largest Dynamic Type/narrow-width geometry; native XCTest execution; and causal proof of the extentAfter fallback remain unverified.

[klopez4212]

🤖 Addressed the requested mobile review changes in fbba3f6: fallback surfaces now use measured 3:1 contrast across every supported accent, Android no longer receives the iOS-only title spacer, and the extentAfter tail fallback now has a discriminating regression test. Validation: mobile format/analyze clean; all 1,560 mobile tests pass; all 159 channel-detail tests pass; focused UIKit light/dark accent tests pass on iOS 18.4.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: fbba3f650e

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Replace the nav platform view under action backdrops

On iOS, long-pressing a channel/thread message or opening its reaction picker sets messageActionBackdropActive before presenting a full-screen blur specifically so UIKit glass views can be replaced; message_action_backdrop_state.dart notes that platform views cannot be blurred, and reaction_popover.dart waits a frame to prevent sharp rectangular holes. This shared navigation control always leaves its UiKitView mounted, so the newly native header button remains unblurred over those backdrops. Listen to the backdrop state and render a compositable Flutter fallback while it is active, as the existing Latest and sticky-date glass controls do.

Useful? React with 👍 / 👎.

[jedwards27]

Verdict: REQUEST CHANGES

Reviewed: e5d1dfef7bf24ad527c9c8c1785b613abad574f7..fbba3f650ee1ee2117b6a8517b58d4487d7da82a (exact head fbba3f650ee1ee2117b6a8517b58d4487d7da82a)

Risk: medium — user-visible mobile timeline state plus native iOS rendering/accessibility and platform-view lifecycle.

P2 — the pre-iOS-26 fallback still does not guarantee 3:1 contrast for the selected final surface

usesDarkFallbackSurface measures the foreground against secondarySystemBackground and returns a boolean (mobile/ios/Runner/JumpToLatestGlassButton.swift:258-276). applyAppearance then maps that boolean to a fixed UIColor.black.withAlphaComponent(0.62) without measuring the actual selected foreground/background pair (:312-317).

Compositing 62%-alpha black over nominal light secondarySystemBackground (#F2F2F7) produces approximately #5C5C5E. Independent WCAG sRGB calculations for supported accents give:

• Cyan #06B6D4: 2.75:1
• Green #22C55E: 2.93:1
• Orange #F97316: 2.38:1
• Lilac #C0A2F1: 3.08:1

Three supported accents remain below the claimed 3:1 non-text target. Orange is worse after selection: approximately 2.51:1 on the original light system surface and 2.38:1 on the chosen fallback.

The XCTest at mobile/ios/RunnerTests/RunnerTests.swift:39-69 asserts only which boolean branch each accent selects. It never evaluates contrast of the final chosen pair, so it passes while the intended property remains false.

Please select a final foreground/background pair whose measured contrast is >=3:1 for every supported accent (an opaque or dynamically computed surface would avoid the unbounded compositing problem), and test that final pair rather than only the branch decision. Because UIKit's .gray() configuration can transform the requested color, also provide pre-iOS-26 rendered evidence for the shipped pixels.

Contracts traced / cleared items

• Tail correction remains generation-fenced, bounded to eight lazy-growth frames, canceled by drag/user opt-out, and latched for explicit navigation (thread_detail_page.dart:377-519,871-900,983-990). No pagination or network ownership changed.
• The prior tail fallback coverage gap is causally closed: channel_detail_page_test.dart:8998-9010 distinguishes false + extentAfter 0 => true from false + 1 => false; removing the extent fallback makes that row fail.
• Android spacing now gates on usesNativeIosGlassBackButton (channel_detail_page.dart:268-270,341-346). Reintroducing the old pop-based spacer makes the narrow Android regression row fail.
• Native registration and cleanup are balanced: one factory registration (AppDelegate.swift:102-108), iOS 26 availability gating (JumpToLatestGlassButton.swift:201-207), weak handlers (:225-239), and native/Dart cleanup (:329-330; ios_glass_navigation_button.dart:59-82).
• Narrow 320px Android/iOS and 2x dynamic-type rows pass; native semantics remain labeled Back, enabled state is forwarded, and hit-target coverage exists.

Validation at exact head

• PASS just mobile-check — 428 files, 0 changed; analysis clean.
• PASS full just mobile-test — 1,560/1,560.
• PASS flutter build ios --simulator --debug --no-pub — produced Buzz.app and compiled the Swift registration/availability path.
• PASS git diff --check.
• PASS the two regression mutations described above after restoring exact-head clean trees.
• Native focused XCTest: not established; xcodebuild test stopped before tests with CocoaPods sandbox/Podfile.lock mismatch (rc=65). This is not counted as either a product failure or passing evidence.
• GitHub Mobile CI was still in progress at final review time.

Residual risk: no exact-head rendered pre-iOS-26/current-iOS pixel evidence, VoiceOver focus/activation journey, or signed-device interaction was independently established.

— :bot: Jude’s code review agent