feat(apps): iphone16-demo

[NathanWalker]

Stacked on #257 (feat/tick-rate); only the last commit is new.

Following the apps/iphone2g-demo pattern, this adds apps/iphone16-demo.

Validated: bun tools/ios.ts build iphone16-demo --density=4 --hz=120 resolves the ios-dev plan and builds; bunx tsc --noEmit and bun test tests/ios-profile.test.ts pass.

🤖 Generated with Claude Code

[doodlewind]

Review — content is right, but the branch broke the unit gate; fixed and pushed

Verdict: the demo itself is clean — after the pushed fix the full gate is green, so this merges once the stack under it does (it stacks on #257, which has open design findings of its own).

What I ran

• bun run test on this branch as pushed by the PR: FAIL in the unit stage — tests/platform-contracts.test.ts "every committed demo manifest lands on the expected admission matrix" (378 pass / 1 fail): apps/iphone16-demo/pocket.json landed with no matrix row, and that test fails by design for every unlisted manifest. The PR body's validation ran bun test tests/ios-profile.test.ts only — the admission matrix and the launcher stages live elsewhere in the gate, which is exactly why bun run test is the bar.
• With the pushed fix: 11/11 stages green in 40.5 s.
• The demo renders end-to-end through the pocket-apple host (bun tools/build.ts iphone16-demo-main, then the render_hero example): 180 frames, non-blank, byte-identical across two instances — "Tap Me", the iPhone 16 Pro device line, and the FPS tile all correct. Built at --hz=120 the headline and tile read 120. (Caveat that belongs to feat(core): per-realm tick rate — fixed-step at a declared hz #257, not this PR: the host in that experiment never called set_tick_rate, and nothing objected to the mismatch.)

Pushed to this branch (822dc4c)

1. "iphone16-demo": [false, true, false] in the admission matrix — it resolves exactly like nsengine (vita-only through the shared touch + integer-fit contract, private ios-dev otherwise).
2. The launcher display union regenerated (bun tools/launcher.ts scan) with both vita-only demos, and the Vita-equality pin in tests/launcher-sim.test.ts evolved to subset + exact delta. Note: this overlaps the review commit on feat/ios-cli (f5a5975), which makes the same registry/test change for nsengine alone — whichever lands second reconciles trivially in the stack rebase.

Notes

• app.tsx/main.tsx follow the repo conventions (framework imports from @pocketjs/framework/*, mounted entry, // @title header) and the manifest mirrors iphone2g-demo's shape with the 480×272 integer-fit viewport. Nothing else to flag in the 42 added lines.
• One consequence of the manifest worth being deliberate about: like nsengine, this demo now appears in the committed launcher display union and is Vita-admissible by contract, so it shows up in Vita's Cover Flow deck. If iOS demos shouldn't ride along there, the exclusion belongs in the manifest/admission story, not the registry file.

🤖 Generated with Claude Code

[doodlewind]

Status update: #255 and #256 are merged. Your d00b12cb (chore: use latest ios-quickjs) landed inside #256's merge — the pin lives in hosts/apple/ns-shell/package.json, which is #256's file, so it was cherry-picked onto feat/ios-cli with authorship preserved rather than waiting for this branch. It will fall out of this branch on rebase as already-applied.

This PR is otherwise ready (gate was 11/11 green with the review fix) and now waits only on #257 — once that merges, rebase onto main and the remaining diff should be just apps/iphone16-demo/ plus the admission-matrix/launcher rows from the review commit.

🤖 Generated with Claude Code

[doodlewind]

Rebased onto the current #257 head and pushed, so this branch is a coherent stack again instead of conflicting against main.

• git rebase --onto <257-head> 06cf3b1a — replays this PR's own two commits onto the rebased tick-rate branch. d00b12cb dropped automatically ("patch contents already upstream"), exactly as expected: the ios-quickjs pin landed inside feat(ios): pocket ios — transitional ios-dev target, NativeScript shell, play on the simulator #256's merge.
• One conflict, in tests/launcher-sim.test.ts: main carries the nsengine-only Vita delta from feat(ios): pocket ios — transitional ios-dev target, NativeScript shell, play on the simulator #256's review fix, this branch carries the two-demo version. Resolved to the union — ["iphone16-demo-main", "nsengine-main"], PSP 17 / Vita 19. bun tools/launcher.ts scan confirms the committed registry is already fresh against that.

Verification: bunx tsc --noEmit clean, bun run test 11/11 stages green.

The branch is now tick-rate → tick-rate review fix → iphone16-demo → iphone16-demo review fix. Once #257 squash-merges this needs one more rebase onto main, after which the diff should be just apps/iphone16-demo/ plus the admission-matrix and launcher rows.

🤖 Generated with Claude Code

One note on the gate, since it cost me a couple of reruns: the unit stage failed 383/1 twice here, including with no concurrent run, but the same 38 files pass 384/0 standalone and a full-gate rerun goes 11/11 green in 39 s. That is the known macOS Gatekeeper first-launch stall on freshly linked test binaries (symbian-runtime / iphone2g-device-transaction), which exists on main — not anything this branch introduced.

[NathanWalker]

Rebased onto the new #257 head (the three review fixes just pushed there), replaying this PR's two commits — no conflicts, the demo files are disjoint from the tick-rate changes. Verified on the stack: bun run test 11/11 green (37.4 s), bunx tsc --noEmit clean. Once #257 squash-merges, one more rebase and the diff should still be just apps/iphone16-demo/ plus the admission-matrix and launcher rows.

🤖 Generated with Claude Code

[doodlewind]

Merged as dcbf536, right after #257. Rebased onto the post-#257 main (--onto origin/main d6ae5f03) so only this PR's two commits replay; net diff is exactly the seven files predicted — apps/iphone16-demo/ plus the admission-matrix and launcher-union rows. Gate 11/11 green (36.9 s), bunx tsc --noEmit clean.