Let Fuse views keep their @State and @Environment private

[Ceylo]

Paired with skip-fuse-ui: skiptools/skip-fuse-ui#151 (branch Ceylo/skip-fuse-ui:feature/private-state-bridging). That one merges first; this one needs its runtime entry points. A follow-up there then uncomments its private-state fixture and test, which need a skipstone release carrying this PR (its diff).

What

With this pair, a Fuse view can keep its @State, @FocusState, @GestureState, @AppStorage and @Environment properties private, as SwiftUI code normally does, instead of widening them to internal just so the bridge can reach them. Today a Fuse view with a private or fileprivate @State, @FocusState, @GestureState, @AppStorage or @Environment property fails to transpile with kotlinBridgeStatePrivate ("Private state property 'x' cannot be bridged to Android. Consider making this property internal"), because the generated bridge file syncs each property by name and cannot see a private one. With this PR such a view builds: its Evaluate asks SkipSwiftUI for the view's dynamic properties, found at runtime by field reflection (the paired skip-fuse-ui PR), and syncs each by index instead of by name.

Before / after

before	after
@State private var count = 0 → error: Private state property 'count' cannot be bridged to Android	builds; Evaluate loops over Swift_dynamicPropertyKinds(Swift_peer) and syncs each property through Swift_initDynamicState/Swift_syncDynamicState (or the AppStorage/Environment equivalents) by index

The choice is made per view, in addSwiftUIImplementation:

• No private state or environment property: today's per-name code, unchanged. All 142 existing BridgeToKotlinTests expectations are untouched and pass.
• Any private/fileprivate one: the whole view's Evaluate switches to the index-based loop. It gets seven fixed-name external functions: the kinds query, plus six index-taking init/sync functions forwarding to SkipSwiftUI's Java_*Dynamic* entry points. A generic view's type-erased peer is then given no per-name closures, since nothing refers to them.

kotlinBridgeStatePrivate is removed, since nothing emits it any more.

Side effects

• Existing views: none. The output for a view without private state is byte-identical, so saved-state keys, the number of JNI calls and generated symbols are unchanged. The shared Evaluate scaffolding was factored into one helper, and the unchanged expectations show it emits the same text.
• Making one property private changes the view's generated code. Its rememberSaveable calls move inside key(i) in a loop, so saved state keyed under the old per-name layout is not carried over across that one source change.
• Needs the paired skip-fuse-ui. The generated Swift calls Java_dynamicPropertyKinds and six sibling functions. Those exist only in skip-fuse-ui with the paired PR, so this skipstone must ship with (or after) it. A project with a newer skipstone and an older skip-fuse-ui gets a "cannot find 'Java_dynamicPropertyKinds'" build error in generated code, where it used to get the kotlinBridgeStatePrivate message. That can only happen to a view that previously failed to build anyway.

Known limits

• Private types are still out of reach. A private struct view or ViewModifier gets no bridge at all. This covers private properties of a type the bridge can name.
• There is no recursion into custom DynamicProperty structs. A State nested inside a user-defined DynamicProperty is not discovered. That's a candidate for a follow-up.
• The path is chosen at transpile time from the attributes skipstone already recognises (@State, @FocusState, @GestureState, @AppStorage, @Environment). A view switches to runtime discovery only when one of those attributes, written literally, marks a private/fileprivate property. Once a view is on that path, every conforming wrapper in it is found by type.

Verified

• swift build succeeds.

• swift test --parallel: 949 tests, 948 pass. The one failure, SkipRunnerTests/testSnippets, compares "1 KB" against a French-locale "1 ko" from ByteCountFormatter. It fails the same way on unmodified main on this machine and is unrelated.

• BridgeToKotlinTests: 148 pass, the 142 existing ones with unchanged expectations plus 6 new ones:

  • testPrivateState
  • testFileprivateEnvironment
  • testMixedInternalAndPrivateState (both properties end up on the dynamic path)
  • testGenericPrivateState (the type-erased peer gets no per-name closures)
  • testPrivateStateViewModifier
  • testPrivateAppStorageFocusStateAndGestureState

  Each expectation was filled from the transpiler's actual output, then read by hand.

• Mutation. With the private-property check forced off, testPrivateAppStorageFocusStateAndGestureState fails, so the new tests do exercise the switch.

• Cross-repo, host bridge. skip-fuse-ui's JVM-hosted FuseComposeUITests.testPrivateCounterIncrements (committed commented out in Brdige NSNumber and fix some newly-discovered bridging bugs #151 until this ships, uncommented for this run) exercised this generator's output end to end, built with this branch as SKIPLOCAL: it passed, and it failed on its own assertion when the State conformance was removed. Caveat: since skip-fuse-ui Make all @_cdecl functions public in order to prevent removal in release builds #134, those Compose tests skip on a macOS host unless run against a device. That run re-enabled the host main-actor hook locally, and the change is not part of either PR.

• Android emulator, in a real app. Also built into the Android port of FurAffinityApp (Skip Fuse; this patch on a local branch) and run on an API 37 emulator: 32 views with private state or environment properties took the new path. Their state interactions (toggles persisting across a restart, text-field focus, sheets, a deep-link highlight) behaved as before, and frame times over a scripted scroll stayed within stock skipstone's run-to-run spread, with the dynamic path at ≤ 0.035 % CPU.

• FuseComposeUITests on a device: blocked. An attempt against the same API 37 (Android 17) emulator (ANDROID_SERIAL) of the paired skip-fuse-ui PR's tests, built with this branch as SKIPLOCAL, could not run any FuseComposeUITests case, the existing testCounterIncrements included: each fails before composing with Espresso's NoSuchMethodException: android.hardware.input.InputManager.getInstance, a method API 37 no longer has. This CI's Fuse skip init … --bridged emulator job exercises the bridged output on a device.

---

Skip Pull Request Checklist:

• REQUIRED: I have signed the Contributor Agreement
• REQUIRED: I have tested my change locally with swift test
• OPTIONAL: I have tested my change on an iOS simulator or device (not applicable: the generated bridge code only compiles for Android)
• OPTIONAL: I have tested my change on an Android emulator or device

---

• AI was used to generate or assist with generating this PR. Please specify below how you used AI to help you, and what steps you have taken to manually verify the changes.

Written with Claude Code: the patch and its tests. Every claim above comes from a run: swift build, swift test --parallel (with BridgeToKotlinTests expectations read against the transpiler's output), the mutation run above, the paired skip-fuse-ui PR's Robolectric tests built with this branch as SKIPLOCAL, and the FurAffinityApp build and emulator runs.