Gerrit Goossen

This commit introduces **build system support for Swift Testing** by conditionally including the new `_Testing_Foundation` framework for test builds targeting **Xcode 26.0 and later**. It **refactors** the existing **Apple test framework utilities** within `prelude/apple` to generalize their naming and functionality, moving away from XCTest-specific terminology. This **new capability** ensures that projects can leverage the modern Swift Testing framework, while also improving the modularity and future-proofing of the test bundle generation logic. The changes primarily affect `apple_bundle.bzl`, `apple_test.bzl`, and `apple_test_frameworks_utility.bzl`, including renaming the utility file and its main function `get_test_frameworks_bundle_parts`.

This commit introduces **build support for Swift Testing** within the Apple build system. It adds a new `swift_testing` boolean attribute to the **`apple_test` rule**, enabling the conditional linking of Swift Testing frameworks and libraries when set to true. This **new capability** simplifies the integration of Swift Testing into Apple targets by providing a declarative way to manage its dependencies. The changes primarily affect the `prelude/apple` build definitions, enhancing the testing infrastructure for Swift projects.

This commit **removes the explicit inclusion of Swift standard libraries and Swift Concurrency libraries** from **Apple tests**. It **refactors** the `apple_test_impl` rule within `prelude/apple/apple_test.bzl` to no longer conditionally include these libraries, instead passing `None` for their arguments. This **optimizes** the build process by eliminating redundant library bundling, as these components are already provided by iOS since versions 12 and 15 respectively. The change also removes an experimental configuration and a test that checked for the presence of these now-unnecessary libraries, **streamlining the Apple test setup** and ensuring correct architecture usage.

This commit introduces a **new debugging feature** for the **iOS `resource_broker`**, enabling the generation of replay scripts for executed commands. By setting the `DUMP_REPLAY` environment variable, the `resource_broker` will now create a shell script at `/tmp/resource_broker_replay.sh` that can re-execute the exact command. This **enhances the debuggability** of the `resource_broker` by providing a straightforward way to reproduce specific invocations. The change involves adding a new `prelude/apple/tools/resource_broker/debug.py` module with the `debug_dump_replay` function and integrating its call into the `main` function of `prelude/apple/tools/resource_broker/main.py`. This **developer tooling improvement** will significantly aid in diagnosing and fixing issues within the `resource_broker`'s operations.

This commit **fixes a bug** in the **iOS simulator resource broker** that prevented `simctl` from correctly identifying and creating simulators when runtimes included a patch version (e.g., 'iOS 26.0.1'). Previously, the system would pass only the runtime name, but `simctl` requires the `platform + runtime version` for accurate matching. The `_select_simulator_spec` function in `prelude/apple/tools/resource_broker/ios.py` is updated to use the full platform and version, supported by adding a `platform` field to the `XCSimRuntime` dataclass. This ensures **local testing simulator creation** functions reliably, resolving issues where simulators could not be provisioned due to runtime selection mismatches.

This commit **refactors** the **Apple test execution** configuration by **removing the `apple.test_execution` buck config** from `prelude/apple/apple_test.bzl`. The logic within `_get_test_info` no longer determines local or remote test execution based on this config, nor does it override the `ios-simulator` executor. This **performance improvement** prevents unnecessary Buck state invalidations and rebuilds, which previously caused build speed regressions when changing how tests run. Consequently, users will now manage **Apple test execution** via the new `--apple-test-execution` `tpx` flag, ensuring runtime option changes no longer trigger costly rebuilds.

This commit **introduces new command executor configurations** within the **`apple_test` build definitions** to support both local and remote **iOS simulator test execution**. It refactors the `_get_test_info` function in `prelude/apple/apple_test.bzl` to determine these execution properties, passing them as overrides to the `tpx` system. This **feature enhancement** aims to improve build performance by moving runtime test execution options from Buck config to `tpx` flags, preventing unnecessary build state invalidations. It's a foundational step, enabling `tpx` in subsequent changes to leverage these overrides for flexible test execution setup.

This commit introduces **new constants for Apple test device types** within the `prelude/apple` module, specifically by adding a `apple_test_device_types` struct in `prelude/apple/apple_test_device_types.bzl`. This **enhancement** aims to make the values for the `test_device_type` parameter more discoverable and less error-prone for developers. The existing `AppleTestDeviceType` enum is updated to utilize these new constants, thereby improving the **robustness and maintainability** of Apple test configurations. This change provides a clearer and safer way to specify test environments for Apple platforms.

This commit introduces a **new capability** for **Apple test execution** within Buck, enabling local test runs directly from an On-Demand (OD) environment. It adds a `local_from_od` option to the `apple.test_execution` configuration, which allows local execution *without* requiring a simulator, addressing a previous limitation where local runs in OD would fail. This change modifies the `_get_test_info` function in `prelude/apple/apple_test.bzl` to conditionally apply the `local_simulator_required` label based on this setting. This significantly improves the developer workflow by providing a prototype for running Apple tests locally from an OD, bypassing the need for remote execution or simulator setup.

This commit **refactors** the **Apple test infrastructure** by moving the logic for determining **iOS simulator** requirements into the build definitions. A new function, `tpx_needs_local_simulator`, is introduced in `prelude/apple/apple_test_device_types.bzl` to precisely evaluate if a given device type needs a local simulator. The `_get_test_info` function in `prelude/apple/apple_test.bzl` is updated to leverage this new logic, ensuring that local simulators are only provisioned when truly necessary. This **enhancement** prevents unnecessary resource allocation and improves the accuracy of **test environment setup** for Apple tests, particularly in nuanced execution scenarios.

This commit **enhances the iOS resource broker** by enabling it to identify and select simulators using **device type identifiers** in addition to traditional device names. Specifically, the `prelude/apple/tools/resource_broker/ios.py` module now supports filtering and choosing simulators based on this more reliable identifier. This **feature enhancement** addresses a problem where `buck test` invocations, particularly from VS Code, could fail to provision simulators if the simulator name did not precisely match the device name. By supporting device type identifiers, the **iOS simulator provisioning logic** becomes more robust, preventing test setup failures and improving the reliability of development workflows.

This commit **fixes a critical issue** preventing **watchOS logic tests** from running by modifying the `apple_test_impl` rule within `prelude/apple/apple_test.bzl`. Specifically, it **excludes the embedding of XCTest frameworks** in test host applications when targeting watchOS simulator SDKs. This **maintenance change** resolves `dyld` loading failures that previously occurred, thereby **enabling watchOS test bundles to load and execute correctly**.

This commit **fixes** a regression in the **Apple test build system** by refining an experimental optimization for **XCTest library exclusion**. It modifies the `apple_test_impl` rule in `prelude/apple/apple_test.bzl` to prevent the exclusion of testing frameworks when a test host app bundle is present. This **bug fix** ensures that **Apple tests utilizing a test host** can execute successfully, preventing failures that arose from the incorrect application of the exclusion. The change allows the experimental optimization to continue for tests without a host, maintaining performance benefits where applicable.

This commit introduces a **new capability** to the **Apple testing infrastructure**, allowing for the **exclusion of XCTest and Swift libraries** from test bundles. It modifies the `apple_test_impl` rule in `prelude/apple/apple_test.bzl` to include conditional logic based on new configuration flags. This **experimental feature** aims to facilitate the creation of **smaller test executables**, which can be beneficial for measuring performance advantages and optimizing testing workflows by reducing binary size.

This commit performs a **chore** by **updating the default iPad simulator device type** within the **Apple resource broker**. Specifically, it changes the default from 'iPad (7th generation)' to 'iPad (A16)' in `prelude/apple/tools/resource_broker/simulator.py`. This **maintenance update** ensures that the local development environment's default iPad device matches the one used in continuous integration (CI), promoting consistency and reducing potential configuration mismatches for simulator provisioning.

This commit introduces a **new capability** to explicitly pass the test device type to the **`tpx` test execution system** using a dedicated label. Previously, `tpx` would infer this information, often leading to incorrect device selections. By modifying `_get_test_info` in `prelude/apple/apple_test.bzl` to append a `tpx` label generated by the new `tpx_label_for_test_device_type` function, the system now leverages Buck's accurate knowledge of the test device. This change **improves the reliability of test device targeting** for Apple tests and **refactors** `tpx` by eliminating redundant device selection logic.

This commit introduces an **enhancement** to the **Apple test infrastructure's resource broker**, ensuring more consistent simulator provisioning. Specifically, the `main.py` script within the `resource_broker` now **requires a default device** for simulator setup when no specific device is requested by a test configuration. This change prevents issues where tests, particularly **screenshot tests**, might fail due to an arbitrary simulator device being selected. Consequently, this improves **test reliability and reproducibility** by aligning local test environments with CI behavior, unless a device is explicitly overridden.

This commit **fixes a critical bug** in the **`resource_broker`** that prevented it from correctly identifying and selecting appropriate simulator runtimes for specific Apple device types. Previously, the system would often default to the first available runtime, typically for iOS, hindering **iPad and Apple Watch testing**. The fix involves **refactoring** the device identifier matching logic within `prelude/apple/tools/resource_broker/simulator.py`, including renaming `matches_device_identifier` and updating its prefix-checking logic, and adjusting calls in `prelude/apple/tools/resource_broker/ios.py`. This ensures the `resource_broker` can now accurately select the correct runtime, **enabling proper testing for iPad and Apple Watch simulators**.

This commit **introduces support for iPad and Apple Watch simulator types** within the **Apple build system's test infrastructure**. It **adds new capabilities** by updating `apple_test` rules and the `resource_broker` to recognize and utilize these new simulator environments, enabling more comprehensive testing across Apple devices. Concurrently, it **refactors** existing simulator references, renaming 'ios' to 'iphone' in various definitions like `apple_test_extra_attrs` and the `SimulatorType` enum for improved clarity. This change allows developers to specify and run tests specifically targeting iPad and Apple Watch devices, enhancing test coverage and precision.

This commit introduces a **new capability** to the **resource broker**, enabling it to manage and provision **iPad and Apple Watch simulators** for testing. The `SimulatorType` enum in `prelude/apple/tools/resource_broker/simulator.py` is extended, and the core **iOS simulator management** logic in `prelude/apple/tools/resource_broker/ios.py` is updated to incorporate these new device types for selection and creation. Additionally, the command-line argument parser in `prelude/apple/tools/resource_broker/main.py` is updated to reflect these additions. This enhancement significantly expands the testing matrix, allowing for comprehensive application testing across a broader range of Apple devices.