fix(mobile): clear letter-spacing on joined-script Arabic labels

[iscekic]

Changelog for users

• The Arabic home overline, its action, and the bottom-nav labels render as joined words with no in-word gaps.
• The English home keeps its tracked uppercase overline, action, and tab labels.
• Latin runs inside an Arabic screen keep their letter-spacing.

Changelog for maintainers

• The shared Text component sets an inline letterSpacing: 0 for joined-script children, overriding any tracking-* class.
• rtl-text.ts adds JOINED_SCRIPT, containsJoinedScript, and textLetterSpacing for the Arabic, Arabic Supplement, Extended-A, and Presentation Forms blocks.
• Detection reads direct string children and recurses arrays; a nested Text element applies the rule in its own run.
• Latin, Hebrew, digits, and punctuation keep their tracking and their RTL writing direction.
• The override is placed before the caller style, relying on NativeWind merging the class style first.
• Look first at the style array in text.tsx and at any stand-alone tracking-* applied to Arabic text.
• Tests cover block ranges, array and element children, mounted Text at both isRTL values, and the section header label and action.

E2E proof

Owner request

kwf-fix: proof-05bd533

kwf-fix-pr: #6497

Prove the behaviour of PR #6497 with a live end-to-end run, and make no code change.
The pull request description carries no evidence: the '## E2E proof' section is empty.

Run the pull request's own scenarios on a device or a simulator, for every platform its diff touches. Capture the decisive log lines always, and a screenshot as well for a user-visible change. Never a recording: they are gone (owner, 2026-09-16).
Write the evidence under a '## E2E proof' heading in your own pull request description. Never run gh: the driver publishes.
If a scenario fails, name it and say why, and still change no product code: this section proves what the branch already carries.

The pull request description as it stands now

This is the live body; you never need to fetch it, and you must not edit it. The driver merges what your final summary says into it.

## Changelog for users

- The Arabic home overline, its action, and the bottom-nav labels render as joined words with no in-word gaps.
- The English home keeps its tracked uppercase overline, action, and tab labels.
- Latin runs inside an Arabic screen keep their letter-spacing.

## Changelog for maintainers

- The shared `Text` component sets an inline `letterSpacing: 0` for joined-script children, overriding any `tracking-*` class.
- `rtl-text.ts` adds `JOINED_SCRIPT`, `containsJoinedScript`, and `textLetterSpacing` for the Arabic, Arabic Supplement, Extended-A, and Presentation Forms blocks.
- Detection reads direct string children and recurses arrays; a nested `Text` element applies the rule in its own run.
- Latin, Hebrew, digits, and punctuation keep their tracking and their RTL writing direction.
- The override is placed before the caller `style`, relying on NativeWind merging the class style first.
- Look first at the style array in `text.tsx` and at any stand-alone `tracking-*` applied to Arabic text.
- Tests cover block ranges, array and element children, mounted `Text` at both `isRTL` values, and the section header label and action.

## E2E proof



## Open findings (not fixed here)
- and y121-135/y120-136 (e3-scene.log); the pixel-level letter join is the visual reviewer's on e3.png."},{"name":"e4","result":"pass","evidence":"/Users/igor/.local/share/kwf/ios/runs/explorer-home-arabic-the-arabic-overline-and-bottom-nav-l-f3af6-f2a3/evidence/e4-scene.log","note":"ios; …[truncated]
- =1). --- shard 1 ---
VERDICT {"verdict":"passed","scenarios":[{"name":"e1","result":"pass","evidence":"/Users/igor/.local/share/kwf/ios/runs/explorer-home-arabic-the-arabic-overline-and-bottom-nav-l-f3af6-f2a3/evidence/e1-scene.log","note":"ios; signed-in English home digest shows 'LIVE NOW', 'See a
- not proved live: e1.png is no longer on the host that took it, so no publish can carry it
- not proved live: e2-home-ar.png is no longer on the host that took it, so no publish can carry it
- not proved live: e6-acct-tap.png is no longer on the host that took it, so no publish can carry it
- not proved live: e6-en.png is no longer on the host that took it, so no publish can carry it
- not proved live: e6-prefs.png is no longer on the host that took it, so no publish can carry it
- not proved live: e6.png is no longer on the host that took it, so no publish can carry it
- not proved live: home-arabic.png is no longer on the host that took it, so no publish can carry it
- the '## E2E proof' section is empty

![e1-probe](https://github.com/user-attachments/assets/3f72aff3-72f4-45e2-8164-5bb35dda237b)

<details>
<summary>Owner request</summary>

> Explorer finding: home-arabic: The Arabic overline and bottom-nav labels inherit the tracked all-caps letter-spacing, which splits connected script mid-word (e.g. الوكلاء renders as 'الوكلا ء' and الجلسات الجارية الآن breaks into separate letters).
> 
> The user-agent explorer found this while using the app like a user.
> One finding per item; the explorer never edits product code.
> 
> Flow: home-arabic
> Found on revision: f2181ae79
> 
> Repro:
> 1. set this state first: reviews 4; credits 25; seed app:github-account e2e-mobile-cloud-android@example.com
> 2. open the app on 1F6F1503-9C26-4120-82D4-5F8768CBE42F
> 3. reach home-arabic
> 4. the capture shows the defect named below
> 
> Observed: The Arabic overline and bottom-nav labels inherit the tracked all-caps letter-spacing, which splits connected script mid-word (e.g. الوكلاء renders as 'الوكلا ء' and الجلسات الجارية الآن breaks into separate letters).
> Expected: the screen renders without this defect
> 
> Evidence (from the device run):

</details>

## Open findings (not fixed here) - and y121-135/y120-136 (e3-scene.log); the pixel-level letter join is the visual reviewer's on e3.png."},{"name":"e4","result":"pass","evidence":"/Users/igor/.local/share/kwf/ios/runs/explorer-home-arabic-the-arabic-overline-and-bottom-nav-l-f3af6-f2a3/evidence/e4-scene.log","note":"ios; …[truncated] - =1). --- shard 1 --- VERDICT {"verdict":"passed","scenarios":[{"name":"e1","result":"pass","evidence":"/Users/igor/.local/share/kwf/ios/runs/explorer-home-arabic-the-arabic-overline-and-bottom-nav-l-f3af6-f2a3/evidence/e1-scene.log","note":"ios; signed-in English home digest shows 'LIVE NOW', 'See a - not proved live: e1.png is no longer on the host that took it, so no publish can carry it - not proved live: e2-home-ar.png is no longer on the host that took it, so no publish can carry it - not proved live: e6-acct-tap.png is no longer on the host that took it, so no publish can carry it - not proved live: e6-en.png is no longer on the host that took it, so no publish can carry it - not proved live: e6-prefs.png is no longer on the host that took it, so no publish can carry it - not proved live: e6.png is no longer on the host that took it, so no publish can carry it - not proved live: home-arabic.png is no longer on the host that took it, so no publish can carry it - the '## E2E proof' section is empty

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Executive Summary

The rewritten joined-script letter-spacing fix is correct at bf5959b08: LTR Arabic children get an inline letterSpacing: 0, RTL keeps the prior unconditional reset, and the detection/fallback logic introduces no regression on the changed lines.

Files Reviewed (5 files)

• apps/mobile/src/components/home/section-header.mounted.test.tsx
• apps/mobile/src/components/ui/text.mounted.test.tsx
• apps/mobile/src/components/ui/text.tsx
• apps/mobile/src/lib/rtl-text.test.ts
• apps/mobile/src/lib/rtl-text.ts

The previously reviewed SHA (6aa171d93) is no longer an ancestor of the branch head, so the review was re-run against the full current PR diff. JOINED_SCRIPT is stateless (no g flag) and spans Arabic, Arabic Supplement, Extended-A, and both Presentation Forms blocks; containsJoinedScript recurses arrays and treats numbers, elements, and empty children as non-strings. In text.tsx, ownStyles falls back to RTL_NO_LETTER_SPACING when no joined script is present, so an RTL interface resets tracking for every run exactly as the base did, while LTR adds letterSpacing: 0 only for joined-script children and leaves a Latin LTR style undefined. The pre-existing text.rtl-tracking.mounted.test.tsx assertions still hold (deep equality, so the separate but structurally identical NATURAL_LETTER_SPACING satisfies them). No memory leaks (no effects, listeners, or subscriptions added). The mounted tests mock react-native, so they assert the emitted style/className props rather than NativeWind's className-vs-inline precedence on device, and the PR body states the visual E2E proof is not live; the on-device visual result remains unverified here.

---

Reviewed at bf5959b08.

Previous Review Summaries (4 snapshots, latest commit 6aa171d)

Current summary above is authoritative. Previous snapshots are kept for context only.

Previous review (commit 6aa171d)

Status: No Issues Found | Recommendation: Merge

Executive Summary

The rewritten joined-script letter-spacing fix (6aa171d93) is correct: it restores the unconditional RTL letterSpacing: 0 reset that the prior revision dropped and adds the LTR-only override for Arabic children, with no regression on the changed lines.

Files Reviewed (8 files)

• apps/mobile/src/components/code-reviewer/manual-review-screen.mounted.test.tsx
• apps/mobile/src/components/home/section-header.mounted.test.tsx
• apps/mobile/src/components/ui/segmented-control.tsx
• apps/mobile/src/components/ui/text.mounted.test.tsx
• apps/mobile/src/components/ui/text.rtl-tracking.mounted.test.tsx
• apps/mobile/src/components/ui/text.tsx
• apps/mobile/src/lib/rtl-text.test.ts
• apps/mobile/src/lib/rtl-text.ts

The branch was rewritten (05bd5339c is no longer an ancestor of 6aa171d93), so the review was re-run against the current PR diff. JOINED_SCRIPT is stateless (no g flag) and covers Arabic, Arabic Supplement, Extended-A, and the two Presentation Forms blocks; containsJoinedScript recurses arrays and treats numbers, elements, and empty children as non-strings (a nested Text is its own run). In text.tsx, ownStyles now falls back to RTL_NO_LETTER_SPACING when no joined script is present, so an RTL interface resets tracking for every run exactly as the base did, while the LTR branch adds letterSpacing: 0 only for joined-script children and leaves Latin LTR style undefined. SectionHeader and TabBarLabel both pass the Arabic label as a direct string child to this Text, so the overline, action, and bottom-nav paths are reached. The segmented-control.tsx change only removes a duplicate numberOfLines={1} prop. No memory leaks (no effects, listeners, or subscriptions added). The mounted tests mock react-native, so they assert the emitted style/className props rather than NativeWind's className-vs-inline precedence on device; the PR body states the visual E2E proof is not live, so the on-device visual result remains unverified here.

---

Reviewed at 6aa171d93.

Previous review (commit 05bd533)

Status: No Issues Found | Recommendation: Merge

Executive Summary

Incremental re-verification of the joined-script letter-spacing fix at the current head found no new issues on the changed lines.

Files Reviewed (5 files)

• apps/mobile/src/components/home/section-header.mounted.test.tsx
• apps/mobile/src/components/ui/text.mounted.test.tsx
• apps/mobile/src/components/ui/text.tsx
• apps/mobile/src/lib/rtl-text.test.ts
• apps/mobile/src/lib/rtl-text.ts

Re-verified at 05bd5339c: JOINED_SCRIPT is stateless (no g flag) and covers the intended Arabic, Supplement, Extended-A, and Presentation Forms blocks; containsJoinedScript recurses arrays and treats nested Text, numbers, elements, and empty children as non-string (inner runs apply the rule themselves so Latin runs keep their tracking); and the ownStyles array (letterSpacing: 0, then RTL_WRITING_DIRECTION, then the caller style) preserves the prior LTR/Latin behavior while overriding a tracking-* class. TabBarLabel and SectionHeader both render through this Text, so the bottom-nav label path named in the PR scope is reached. The mounted tests mock react-native, so they assert the emitted style/className props rather than NativeWind's className-vs-inline precedence on device, and the PR body states the visual E2E proof is not live; the on-device visual result remains unverified here.

Previous review (commit ce38481)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (5 files)

• apps/mobile/src/components/home/section-header.mounted.test.tsx
• apps/mobile/src/components/ui/text.mounted.test.tsx
• apps/mobile/src/components/ui/text.tsx
• apps/mobile/src/lib/rtl-text.test.ts
• apps/mobile/src/lib/rtl-text.ts

Re-verified the current head (ce384817d) after the branch rewrite: JOINED_SCRIPT covers the intended blocks and is stateless (no g flag), containsJoinedScript handles string/array/undefined/number/element children, and the ownStyles array (letterSpacing: 0 then RTL_WRITING_DIRECTION, then the caller style) preserves the previous RTL/Latin behavior. The eyebrow variant and SectionHeader action retain their LTR-only tracking, so the joined-script override is the only new path. The mounted tests mock react-native and therefore assert the emitted style/className props rather than NativeWind's className-vs-inline precedence on device; the PR body also states the visual E2E proof is not live, so the on-device visual result remains unverified here.

Previous review (commit 0847a09)

Status: No Issues Found | Recommendation: Merge

Files Reviewed (5 files)

• apps/mobile/src/components/home/section-header.mounted.test.tsx
• apps/mobile/src/components/ui/text.mounted.test.tsx
• apps/mobile/src/components/ui/text.tsx
• apps/mobile/src/lib/rtl-text.test.ts
• apps/mobile/src/lib/rtl-text.ts

Reviewed the joined-script letter-spacing override: the regex block ranges, the ReactNode string/array recursion, the ownStyles merge order relative to props.style, and the RTL-writing-direction behavior are all consistent with the tests and preserve the prior LTR/non-Arabic behavior. Note that the mounted tests mock react-native, so they assert the emitted style prop but do not exercise NativeWind's className-vs-inline-style precedence on device; the PR body also states the visual E2E proof is not live.

---

_{Reviewed by deepseek-v4.1-flash · Input: 0 · Output: 0 · Cached: 0}

_{Review guidance: REVIEW.md from base branch main}

[iscekic]

This description names a scenario the proof did not capture:

• not proved live: e1.png is no longer on the host that took it, so no publish can carry it

A repeated proof run rebuilds the same evidence, so no proof run is dispatched for a named gap. Merging with this gap open is your decision.