iOS selector click taps a point resolved from a pre-tap capture, so a layout shift between capture and tap retargets it silently
メンテナーはふだん 1 日以内に返信
評価
- 難易度
- 4/5
- 見積もり時間
- 3〜5日
- 初心者へのやさしさ
- 25/100
- issue の種類
- バグ
- 明瞭さ
- おおむね明確
- 活発さ
- 停滞
- 技術スタック
- ios, typescript
- 領域
- mobile
調査の方向性
Start with settleAppleOpen in packages/platform-apple/src/open-policy.ts, where the fixed post-open sleep fallback lives, and with the runtime-selector tap path that takes its point from the pre-tap capture. The issue asks you to pick one owning fix, so first settle the design, then update the guarantee cell in packages/contracts/src/interaction-guarantees.ts. Done means a fixture or live run where a moved target either gets tapped at its new location or fails with a typed reason, with warm click timing unchanged. Linked PR #3356 is already open, so check it before starting.
索引モデルが issue の本文から書いたものです。
説明
Purpose
An iOS selector click taps the point it resolved from its own pre-tap snapshot. If the layout moves between that capture and the tap, the tap lands on whatever is now under the old point, and the click still reports success. Nothing re-checks where the target is at dispatch time.
Evidence
Main run 37815426076 (2026-10-08, iOS 26 lane), 01-settings.ad. Times are from the session runner.log:
open postOpenObservation: unobservable -> fixed POST_OPEN_SETTLE_MS sleep
17:41:31.77 COMMAND_ACCEPTED command=snapshot (click's pre-tap capture)
17:41:34.57 COMMAND_COMPLETED command=snapshot ok=1 (2.80 s capture)
17:41:35.56 SYNTHESIZED_DISPATCH kind=tap point=(201.0,319.0) attempt 1 -> following wait failed
17:42:10.96 SYNTHESIZED_DISPATCH kind=tap point=(201.0,406.0) attempt 2 -> passed
General was at y=319 in a capture taken about 0.5 s after open returned. On the settled root list it is at y=406, 87 pt lower. Attempt 1's tap probably hit the Apple Account band.
This alone does not prove the tap missed. In #2936 a failing first attempt tapped the settled (201,406) and its wait still failed, so a stale point is one cause among others. Diagnostic history: #2948.
Where it happens
settleAppleOpen(packages/platform-apple/src/open-policy.ts): when the bridge can't read the app, the open falls back to a fixed sleep. When the bridge does read it,observableonly means the tree was readable, not that the layout is stable.- The runtime-selector path takes its tap point from the rect in the pre-tap capture. Neither the
offscreen/nonHittablechecks nor anything in the runner confirm that the node is still at that point when the tap goes out.
Required behavior
- Choose one owning fix and justify it against the evidence. Candidates:
- re-resolve or re-check the target's frame in the runner right before tapping;
- require the target's frame to be stable across captures before tapping;
- gate the first post-open interaction on the layout settling.
- A bounded check must have a budget shorter than the click itself and must not add a capture to warm clicks whose layout is already stable.
- A target that moved and can't be confirmed must fail with a typed reason, not tap the old point.
- Update the guarantee cell this changes in
packages/contracts/src/interaction-guarantees.ts.
Completion
- A fixture or live run where the target moves between capture and tap either taps the new location or fails with a typed reason.
- Warm click p50 on
01-settings.ad(about 2.6 s locally) does not regress.
Related: #3335 (click reports dispatch, not landing), #2948, #2491/#3328 (wait observation stall).
- 主要言語
- TypeScript
- スター
- 4.9k
- フォーク
- 328
- 平均マージ
- 12時間 18分
- マージ済み PR(30日)
- 538
環境構築
- Dockerfile・Docker Compose ファイルなし
- プルリクエストのテンプレートなし
- コントリビューションガイドを読む
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
callstack/agent-device のほかの issue
-
難易度 1/5 1時間未満 初心者へのやさしさ 88/100
callstack/agent-device#3353 ·
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
callstack/agent-device#1869 ·
メンテナーはふだん 1 日以内に返信
-
難易度 3/5 1〜2日 初心者へのやさしさ 45/100
callstack/agent-device#3392 ·
メンテナーはふだん 1 日以内に返信
-
settings permission --app writes a display name as the bundle id instead of resolving it対応中かも @thymikee が今日担当しました。 オープン
難易度 3/5 1〜2日 初心者へのやさしさ 22/100
callstack/agent-device#3384 ·
メンテナーはふだん 1 日以内に返信
-
iOS: a canceled request releases the session lock while its runner command still runs on the device対応中かも @thymikee が今日担当しました。 オープン
難易度 4/5 3〜5日 初心者へのやさしさ 22/100
callstack/agent-device#3383 ·
メンテナーはふだん 1 日以内に返信
callstack/agent-device の issue をすべて見る
似ている issue
-
chore v2
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
modelcontextprotocol/servers#5115 ·
メンテナーはふだん 1 日以内に返信
-
beginner bug good first issue
難易度 1/5 1時間未満 初心者へのやさしさ 85/100
philaconvalley/website#168 ·
メンテナーはふだん 1 日以内に返信
-
bug frontend good first issue
難易度 2/5 1〜3時間 初心者へのやさしさ 86/100
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 82/100
oss-slu/lrda_mobile#294 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
hatchet-dev/hatchet#5179 ·
メンテナーはふだん 1 日以内に返信