Repository navigation
Recover macOS .NET restore killed on catalogue main #677
Description
🤖 Generated by the Agentic Engineer
Description
Shared automation main is red because the macOS .NET action fixture was killed during package restore. This blocks a reliable catalogue validation result; no production application outage or source defect has been established.
Reproduction and evidence
- Current main:
0c7b6c3509b73adbc7537e030935bb01987e625b. - CI run 39495567912, attempt 1, failed job
118879874548,test-run-dotnet-tests (macos-latest). - The action's restore process exits 137 about three seconds after starting, before a package, compiler or test diagnostic. SDK setup succeeds. Linux and Windows fixtures pass at the same revision.
- Complete job pagination contains 247 unique jobs: 194 successful, 51 skipped, this failure and the dependent required-check reducer failure.
Exit 137 establishes a killed process, not its cause. There is no evidence yet that memory pressure, a runner problem or repository code caused it. A single failed-job recovery attempt is appropriate after rebinding current main and ownership; a repeated failure requires diagnosis before another retry.
Acceptance criteria
- Obtain executed macOS restore and fixture success at the same source revision, or deliver and validate a root-cause correction if the failure repeats.
- Verify the required CI aggregate succeeds with complete job and exact-head evidence.
- Preserve the fixture, credential boundary, scanner scope, timeouts and required checks; record the bounded outcome without claiming an unproven cause.
Planning
High priority because default-branch CI is red. Initial effort is Small for bounded recovery and evidence collection; reassess if the failure repeats. This is immediate operate work under the live-breakage exception, not new feature intake.
Part of #287
Metadata
Metadata
Assignees
Labels
Type
Fields
Priority
Effort
Projects
- Status
Activity
chore: agent-claim #677 nonce=ef9ff08e02607789b77599c23c53b66f
chore: agent-claim #677 nonce=1d90403d75ef4bea244d3e7f3da15ea2
chore: agent-claim #677 nonce=c643af44f196ce85e0e01d61806f6e8b
chore: agent-claim #677 nonce=3d9f0a743c26bbf47ac982c04b5b7598
chore: agent-claim #677 nonce=e7b3a0221c05a6cbfc6ac8bc4bc2a4a5
chore: agent-claim #677 nonce=632b3fc3c27a78278275ea66db0a767d
chore: agent-claim #677 nonce=7b563fa9c145135fe55fc307e9aa92a9
chore: agent-claim #677 nonce=24ed46345d421a98f835807c8ee19cb6
chore: agent-claim #677 nonce=7c02c672fbf9d470381d315529e19a66
chore: agent-claim #677 nonce=a2373344bc0d00cdc5d6fc218d2615ae
chore: agent-claim #677 nonce=83f85156c6793eed07e6e549ea424cf8
devantler commentedon Oct 20, 2026
Recovery is verified at unchanged main
0c7b6c3509b73adbc7537e030935bb01987e625b. The single failed-job retry was accepted at 21:15:46 UTC. Run 39495567912, attempt 2 completed successfully at 21:17:42 UTC.The macOS fixture job
118891100857actually restored both fixture projects and ran all 18 tests successfully. Required aggregate job118891668999executed and passed. Complete pagination contains 494 unique job records across both attempts; choosing the latest attempt for each job identity yields all 247 expected jobs, with 196 successful and 51 skipped, zero failures or unfinished jobs. Both recovery jobs independently bind to the same run, attempt and source SHA. The current default branch still equals that SHA.This establishes recovery, not a cause for the earlier exit 137. No source, fixture, credential boundary, scanner scope, timeout or required-check setting changed. If the failure repeats, investigate it before another retry.
Actual first work was 20 October. Expedite is a recorded WIP exception for confirmed default-branch breakage; the planned Start date was not changed. All three acceptance criteria are met by the executed native result, so this recovery incident can be completed. The roadmap parent remains open.
chore: agent-claim #677 nonce=cd4d0318eda3bbfbbac648615c5b4816
chore: agent-claim #677 nonce=5944c3a8e6024769ed15216883063378