ios-swift-ci

Files (1)

• SKILL.md

SKILL.md (first lines)

# iOS/Swift CI Triage

Use this for `SwiftLint`, `Swift Code Coverage`, `iOS Build`, Swift package,
WHEP, or `XCTestRunner Simulator Tests` failures.

1. Resolve the exact PR head and current main base. Do not trust a prior green
   head after a rebase or merge queue update.
2. Identify whether the check is required. `XCTestRunner Simulator Tests` is
   advisory (nightly, or on a PR labelled `run-ios-sim`); the required iOS checks are `SwiftLint`,
   `Swift Code Coverage`, and `iOS Build`.
3. Read the failed job by run and job ID with
   `gh run view <run-id> --job <job-id> --log-failed`. If it is incomplete,
   retrieve the job log or failure artifact before assigning a cause.
4. Classify the failure as PR-caused (with a local command), PR-caused without
   a local command, stale-base/main-red, runner infrastructure, test flake, or
   advisory roll-up fan-out. Do not retry merely because a simulator job is red.
5. For simulator jobs, preserve ownership: one explicit simulator UDID per job,
   used serially. Never share a simulator between parallel jobs/runners.
6. Recognize this week's known simulator signatures: five-minute CtrlProxy UI
   test timeout; CtrlProxy surviving forced teardown; video recording `simctl
   list` state timeout; hierarchy UI test over 90 seconds. Gather new evidence
   before treating any of them as the same bug.
7. For a Swift change, run `scripts/prepush-ios.sh`. It checks the pinned
   SwiftFormat 0.54.6, SwiftLint's error rules, `swift build` in XCTestRunner,
   and only simulator-free XCTestRunner tests.
8. Before merge, refresh the exact head on latest main; a stale base can hide a
   main-red period.

From kaeawc/auto-mobile, Apache-2.0.

More from kaeawc/auto-mobile