[thymikee]

Summary

Behavior-preserving simplification of the runner native bridge and the shared geometry header. These Objective-C files ship byte-for-byte in the npm package, so their narration is shipped size.

• Deleted RunnerResolveSoleScreenDisplayID: a renaming wrapper that discarded the resolved screen. RunnerCreateEventRecord now calls RunnerResolveSoleScreen at the same point, so the dispatch-time sole-screen admission recheck keeps the same call and timing.
• Fixed the stale RunnerSynthesizedGesture.h comment: a nil resolvedWindow is the sole-screen route, not the first window.
• Inlined the single-use RunnerDragProgressFunction / RunnerSampledDragPointerPath into RunnerControlledScrollPointerPath, its only caller, which passed a fixed progress function. Same event sequence.
• Replaced AgentDeviceForegroundOwnershipMatches (an inline a == b) with == at both callers; deleted the dead kAttributeFrame branch in SnapshotBridgeRuntime.m (it returned nil into a return nil;).
• Trimmed ObjC narration to one sentence per exported symbol plus the constraints code cannot show (frame read before screen; exists required; class absence cached, never geometry; phone idiom skips the probe). History notes (#2728, "measured on a live Duo") are gone.

7 files, two commits. No wire, CLI, or public type changed; the shared header's protocol, tolerance, and measurable/fills predicates are untouched.

Validation

Tested SHA 25bc6b2d5f (rebased onto origin/main a16d437237; 1ba2e5fa74 changed only one header comment).

• Bridge native tests (cache-native-runtime.test.ts, cache.test.ts, cache-identity.test.ts): 58 passed, including the -Werror bridge compile.
• Mutant (disposable checkout): drop the bridge ownership check → cache-native-runtime.test.ts 3 failed / 46 passed (baseline 55 passed).
• Runner unit suites (bridge, synthesized interaction, scroll gesture, gesture) built with AGENT_DEVICE_XCUITEST_INCLUDE_UNIT_TESTS=1: iPhone 17 Pro 36 passed / 0 failed; iPad Air 13-inch (M4) 38 passed / 0 failed (counts after the rebase; the pre-rebase run was 37 and 39).
• Mutant: remove the sole-screen fills-screen refusal → the nightly-iPad selection fails 4/4 (baseline 4 passed).
• pnpm check:xctest-selection passed. pnpm check:affected --run --head 25bc6b2d5f: all runnable checks passed, exactCommit=true; swift-runner lanes are GitHub-authoritative.
• XCTest Nightly on the final head (the iPad sole-screen route changed): https://github.com/callstack/agent-device/actions/runs/40403407186 — success; iPhone full suite 498 executed, iPad sole-screen leg 4 executed.

Size Report

Metric	Base	Current	Diff
Installed (including dependencies)	5.57 MB	5.56 MB	-5.1 kB
Package (unpacked)	5.56 MB	5.56 MB	-5.1 kB
Package (download)	1.69 MB	1.69 MB	-1.8 kB

🤖 Generated with Claude Code

[github-actions]

Size Report

Metric	Base	Current	Diff
Installed (including dependencies)	5.57 MB	5.56 MB	-5.1 kB
Package (unpacked)	5.56 MB	5.56 MB	-5.1 kB
Package (download)	1.69 MB	1.69 MB	-1.8 kB

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	40.0 ms	38.9 ms	-1.1 ms
CLI --help	79.2 ms	81.6 ms	+2.4 ms

[cubic-dev-ai]

All reported issues were addressed across 7 files

<sub>Reply with feedback, questions, or to request a fix.

View guided diff | Turn on auto-fix | Re-trigger cubic</sub>

[thymikee]

PR 3976 is ready at 1ba2e5f. Not blocking: the local id soleScreen = nil; at RunnerSynthesizedGesture.m:327 is never read and only fills the out-parameter, which is an acceptable price for one entry point, and you could drop the second comment line ("Re-asking here..."); take or leave both.

All 16 checks pass on this commit and there are no conflicts. I did not rerun the simulator runner suites, the bridge -Werror compile or the mutants, so those results come from the PR body. I read the diff and every non-comment hunk is an inline, a deleted pass-through or a dead branch, so I expect no device behavior change. The header comment edits change the runner source fingerprints, so installed users rebuild once, which is expected.

The one bot thread (the RunnerResolveWindowDisplayID header wording) is fixed at this commit, so you can resolve it: #3976 (comment)

Before merge, confirm the XCTest Nightly run on this head is green.

[thymikee]

I found no problems in 25bc6b2. The new header comment in RunnerXCTestEventBridge.h now says the window route returns a refusal message, and RunnerXCTestEventBridge.m returns an NSString there, so the earlier doc thread is fixed (#3976 (comment)); please resolve it.

All 16 checks pass. The change since the last review is a header comment only, so no route behavior changed. I did not build or run the Objective-C or runner suites. The simulator and -Werror results at the rebased head come from the PR body, and I did not re-run them. No conflicts are known, and nothing further is needed before merge.

[thymikee]

Thanks — taking both notes on r4381544807 as-is: the discarded soleScreen out-parameter is the price of the single entry point, and the "Re-asking here" line states the constraint that keeps the dispatch-time recheck from being folded into the earlier admission read, which the code cannot show. No push for them.

The doc thread r4381544807 is already resolved, and nothing changed since: head stays 25bc6b2. XCTest Nightly is green at that head — run https://github.com/callstack/agent-device/actions/runs/40403407186 (iPhone full suite 498 executed, iPad sole-screen leg 4 executed, both legs success).