Under Xcode 26.x, a suite of many trivially fast value-checks run through the swift-testing library intermittently aborts swiftpm-testing-helper with signal 6 and no failing test. Observed on CI run 30796756743 (PR #280): the XCTest suites passed, then the swift-testing harness died ~0.16s in, before the new suite reported at all.
Current workaround (documented in-file): such suites use XCTest — ShellEscapeTests.swift, BracketedPasteTests.swift. There are also .disabled("...crashes the swift-testing runner...") tests elsewhere from the same bug.
This contradicts CLAUDE.md's "Swift tests use swift-testing" convention, and automated reviewers flag it on every PR that touches those files. Worth either (a) confirming against a newer Xcode and reverting the workaround, or (b) writing the exception into CLAUDE.md so it stops being re-litigated.
Under Xcode 26.x, a suite of many trivially fast value-checks run through the swift-testing library intermittently aborts
swiftpm-testing-helperwith signal 6 and no failing test. Observed on CI run 30796756743 (PR #280): the XCTest suites passed, then the swift-testing harness died ~0.16s in, before the new suite reported at all.Current workaround (documented in-file): such suites use XCTest —
ShellEscapeTests.swift,BracketedPasteTests.swift. There are also.disabled("...crashes the swift-testing runner...")tests elsewhere from the same bug.This contradicts CLAUDE.md's "Swift tests use swift-testing" convention, and automated reviewers flag it on every PR that touches those files. Worth either (a) confirming against a newer Xcode and reverting the workaround, or (b) writing the exception into CLAUDE.md so it stops being re-litigated.