Test results that follow the run

Swift Testing bundles now report every test and what actually became of it. The testing library keeps a record of the run, and Once turns that record into its normalized results instead of listing what the sources declare and crediting each one with the run's exit status. A test that was filtered out, skipped, or never reached is reported as such rather than as a pass, and an issue a test marks as known no longer counts as a failure.

Where results still have to come from the XCTest host, that host reports the run's outcome rather than each test's. Once lists the cases it finds so a shard can address them, but lists them without a verdict, so a green run never carries verdicts nothing produced.

Test bundles that depend on a Swift macro can now import it. Code behind canImport of a macro module used to compile away, which quietly emptied the test targets that exercise macro implementations. Such a bundle builds for the host along with its dependencies, matching how a macro itself is built.