XCTest discovery support for non-Darwin platforms #499
Conversation
owenv
requested review from
aciidgh,
jakepetroules,
mirza-garibovic,
mhrawdon and
neonichu
as code owners
|
@swift-ci test |
| @@ -649,7 +649,7 @@ public final class XCTestBundleProductTypeSpec : BundleProductTypeSpec, @uncheck | |||
| var (tableOpt, warnings, errors) = super.overridingBuildSettings(scope, platform: platform) | |||
| var table = tableOpt ?? MacroValueAssignmentTable(namespace: scope.namespace) | |||
|
|
|||
| let isDeviceBuild = platform?.isDeploymentPlatform == true && platform?.identifier != "com.apple.platform.macosx" | |||
| let isDeviceBuild = platform?.isDeploymentPlatform == true && platform?.name != scope.evaluate(BuiltinMacros.HOST_PLATFORM) | |||
There was a problem hiding this comment.
As an aside, this seems fundamentally flawed. There's no guarantee you're running the resulting binaries on the same host that built them, so it seems the host platform should enable this too.
Why simulators are excluded I'm not sure.
| @@ -295,6 +295,7 @@ public enum PIF { | |||
| case executable = "com.apple.product-type.tool" | |||
| case hostBuildTool = "com.apple.product-type.tool.host-build" | |||
| case unitTest = "com.apple.product-type.bundle.unit-test" | |||
| case swiftpmTestRunner = "com.apple.product-type.tool.swiftpm-test-runner" | |||
There was a problem hiding this comment.
Why does this need to be called "SwiftPM" test runner?
There was a problem hiding this comment.
I called it that to differentiate it from other kinds of test runners like XCTRunner, but I'm not too attached to the name
| #if canImport(Testing) | ||
| import Testing | ||
| #endif | ||
|
|
||
| \(testObservationFragment(testOutputPath: "TODO")) |
There was a problem hiding this comment.
TODO still supposed to be here?
There was a problem hiding this comment.
Nope, fixed now so that this is provided via a CLI arg
|
|
||
| private func testObservationFragment(testOutputPath: String) -> String { | ||
| """ | ||
| #if !os(Windows) // Test observation is not supported on Windows |
There was a problem hiding this comment.
Why? Do we have a GitHub issue for this?
There was a problem hiding this comment.
This is pulled in verbatim from SwiftPM
| override func createTaskAction(_ cbc: CommandBuildContext, _ delegate: any TaskGenerationDelegate) -> (any PlannedTaskAction)? { | ||
| TestEntryPointGenerationTaskAction() | ||
| } | ||
|
|
||
| public func constructTasks(_ cbc: CommandBuildContext, _ delegate: any TaskGenerationDelegate, indexStorePaths: [Path], indexUnitBasePaths: [Path]) async { |
There was a problem hiding this comment.
Can we inject the necessary additional arguments without overriding constructTasks? Ever bare call to createTasks adds risk that we forget to update things as its signature or default behavior changes.
There was a problem hiding this comment.
I tried to do this at first, but I couldn't find an elegant way to differentiate the input types, especially since the handling of the index store is a bit unusual since the build system doesn't track the content inside it
owenv
force-pushed
the
owenv/test-discovery
branch
from
b20fe8d to
5c687f8
Compare
owenv
force-pushed
the
owenv/test-discovery
branch
from
8046285 to
2415461
Compare
|
@swift-ci test |
owenv
force-pushed
the
owenv/test-discovery
branch
from
2415461 to
7a9476d
Compare
owenv
force-pushed
the
owenv/test-discovery
branch
from
7a9476d to
a2a03a9
Compare
|
@swift-ci test |
owenv
force-pushed
the
owenv/test-discovery
branch
from
a2a03a9 to
0135fdb
Compare
|
@swift-ci test |
Add a new target type for SwiftPM test runner executables used on Linux/Windows. XCTest discovery generates an entry point based on index store data. The index store implementation is a lightly edited version of what's in TSC. This will require some adoption work to generate the new test runner target in PIF - we should think about the best way to minimize the diff between Darwin and non-Darwin here (fyi @cmcgee1024 )