PR32 remote check snapshot
Observed at 2026-10-09T23:52:59Z, head
3c291e511cb1c5457a5bd1945c45681f263bd064, source commit
62e7f53f237ceaf0f5037a396e79ff70f2c735cd, parent
764d732ac38fb9116c4cf525a6d1bb463d409a06.
- CI run 38006379058:
8 completed failures, 17 queued jobs; workflow status
queued, no conclusion. - Docs Preview 38006378962:
completed,
failure; deployment and comment steps skipped.
ci-final.json / docs-final.json preserve exact API status and step results.
classification.json maps every job to its status, reason, URL and available log.
No job was cancelled, restarted or modified while collecting this evidence.
Completed failures
Seven CI jobs (MSRV, wasm, cargo-deny, fmt, semver, Clippy, docs) and Docs Preview
fail because cyber-hemera is a sibling path dependency whose manifest is absent:
/home/runner/work/radio/hemera/rs/Cargo.toml, or /github/hemera/rs/Cargo.toml
inside cargo-deny's container. These are manifest-loading/setup failures, before
the changed exporter can compile or its tests run. Cargo-make's fmt wrapper exits
105 after metadata failure; codespell exits65; normal Cargo setup failures exit101.
The full completed-job logs are retained as job-<id>.log.
The relevant workflows and manifests are byte-unchanged versus the parent.
iroh-relay/Cargo.toml:93 still declares
hemera = { package = "cyber-hemera", version = "0.3", path = "../../hemera/rs" };
the observed workflows check out radio alone. This identifies an unchanged
dependency setup gap, not a test of the corrected exporter.
Codespell reports 68 findings in 16 paths. codespell-file-comparison.json
records exact before/after Git blobs: all 16 are identical to the parent, none is
a touched file. This includes the older traversal receipt's copied diagnostic
text. No changed Rust source or new exporter receipt path appears in this run's
codespell output. The raw log retains the actual diagnostics.
Incomplete coverage
All 17 queued jobs request self-hosted runners in the pinned workflows: ordinary test matrices, minimal crates, cross/Android builds and Netsim. They have no executed steps in the snapshot. Runner availability is not inferred from this queue; these jobs have neither passed nor failed. The CI run is incomplete.
The failure classification does not turn red jobs green or replace local gates. It finds no executed changed-code failure in the completed remote jobs because the source checks were blocked by setup, while the spelling findings are in unchanged files. Later runs at another head are separate evidence and cannot retroactively change these conclusions.
commands.json pins the read-only collection and comparison commands; successful
log/API retrieval has exit0 and is separate from the job conclusions above.