[snissn]

Summary

Refs #5390.

Recreate the sparse-catalog TCP setup diagnostics on main 78e4de8642b4e177a7ec24b87e18c51c4e0bd684. This is a new test-only candidate, not a byte-for-byte recovery of the unavailable sealed candidate.

• Record child setup, catalog-leader discovery, catalog publication, all-node readiness and ready checkpoints, with at most six retained checkpoints.
• Observe only the existing Status results. For configured nodes, retain call counts and independent last-success/last-error text with their own phase and elapsed offsets.
• Emit only when setup unwinds with a failed test, before the original child cleanup. Elapsed values start at helper entry, not at the earlier caller-context creation.
• Cap retained status/error text at 4096/1024 bytes including the truncation marker, without retaining status payloads, error objects or oversized string backing storage. Full status formatting can allocate before truncation.
• Preserve the original shared 30-second context, RPC count and ordering, publication, readiness predicates, polling loops and sleeps, child ownership and checked cleanup, and every route/tamper assertion. No additional RPC, retry, timeout change or production change.
• Refresh CI discovery through the maintained staged-source generator. No CI policy, owner or workflow change.

Review and local validation

Independent source review accepted tree 04bdd47defca9b40d892ddf4816655030dda2b7b with no blocking findings. Removing the nine diagnostic insertions restores the entire original sparse-catalog contract file byte-for-byte.

Laptop qualification on macOS arm64, Go 1.26.0, with GOMEMLIMIT=4GiB, GOMAXPROCS=2, -p 1, shared existing Go caches and an isolated temporary directory:

• go test -json -timeout 5m -p 1 ./nativewire -run '^TestSparseCatalogSetupFailureDiagnosticsV1$': 6 RUN / 6 PASS; package 0.813s.
• go test -json -timeout 30m -p 1 ./nativewire -run '^TestSparseCatalog': 32 RUN / 32 PASS; package 6.559s.
• go test -json -race -p 1 -timeout 12m ./nativewire -run '^TestSparseCatalog': 32 RUN / 32 PASS; package 20.564s.

Commands ran from TreeDB. Both owning-scope runs include the real tampered-route parent and all nine tamper children, nonparticipant and stale/fresh route assertions, and checked original child cleanup. No failures, skips, race reports or success-path setup diagnostics were observed. All processes joined. Raw JSONL, stderr, exact commands and producer/source bindings are retained locally.

The maintained inventory refresh/check, gofmt and diff checks passed. The local commit hook ran go fmt ./... without changing the reviewed tree.

Remaining gates and limits

Windows qualification and required full latest-head CI/review/merge gates remain pending. These laptop checks do not replace them.

The original #5386 Windows failure remains unresolved and its cause remains inconclusive. This diagnostic change does not repair it, clear it or rewrite the retained historical failure.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn