Building and checking native Docs

The native editor uses SwiftUI with AppKit/UIKit text controls and a bounded WASM interpreter for the shared Yjs-compatible CRDT. It does not load a web editor.

Requirements: Xcode 26, XcodeGen, Swift 6.2+, Node/npm, and Rust with the wasm32-unknown-unknown, aarch64-apple-darwin, x86_64-apple-darwin, aarch64-apple-ios, aarch64-apple-ios-sim, and x86_64-apple-ios targets. The server gate additionally requires the Go version in go.mod.

Run the native gate from the repository root:

native/scripts/check-native.sh

It builds the CRDT WASM, runs Rust editor and bounded-host tests, checks native and browser updates in both directions (including every list-rule reversal delta and its authoritative snapshot), prepares the Apple XCFramework, and runs Swift adapter/actor tests. --core-only runs the Rust and wire checks on Linux. CI invokes this through .amadan/ci.d/20-native, alongside the existing server/web gate. --mutations additionally proves representative faulty changes are caught, restoring every changed source and saving logs under native/build/mutation-logs. Do not run mutations concurrently with builds.

CARGO can select a non-default Cargo executable; put its matching Rust toolchain on PATH. If that toolchain needs an explicit LLVM runtime lookup path, set DOCS_LLVM_LIBRARY_PATH. No generated native framework is committed.

Build individual applications:

native/scripts/build-mac.sh
native/scripts/build-iphone.sh simulator
OFICINA_DEVELOPMENT_TEAM=YOUR_TEAM native/scripts/build-iphone.sh device

The suite's build-native-apps.py prepares the native framework and Xcode projects once, then runs Docs and Sheets builds concurrently with separate platform logs and a combined results.json. Its default targets are Mac and iOS Simulator; --platforms mac simulator device also builds signed device applications when OFICINA_DEVELOPMENT_TEAM is set. It does not install or publish applications.

Native UI tests use fresh local credentials and isolated databases. See REVIEW.md for their runner, exact evidence, and untested behavior. Performance measurements and their limits are in benchmarks/README.md.