Skip to content

Code review · T7-retry, Nextflow retry and public runner limits

review-code on autopilot/T7-retry, range origin/autopilot/integration...a29fb74, 2026-10-28. The step applies triage findings SPK-T7-05 (retry a failed Nextflow entry once with -resume) and SPK-T7-02 (public runner limits from the measurement in nightly run 40649538349). Reviewers: code-reviewer and security-reviewer, independent, in parallel. Format: README.md.

Round Head code-reviewer security-reviewer
1 a29fb74 approve (0 blockers, 0 majors, 3 minors) approve (0 blockers, 0 majors, 3 minors)

Round 1 approved with minors. The fixes are in the commit that adds this page, on autopilot/T7-retry, and were not reviewed again. code-simplifier was not run. Test command:

python3 -m unittest tests/pipelines/real/test_run_corpus.py
CODE  step=T7-retry  review=approve b/M/m=0/0/3  security=approve b/M/m=0/0/3  rounds=1  simplify=notrun -0/+0  tests=pass
Id Severity Finding Status
CR-T7R-01 minor PASS (retried) in the status column broke the alignment of the summary columns fixed in the commit that adds this page: the status stays PASS/FAIL, and (retried) follows the seconds column
CR-T7R-02 minor An nf-core entry could put -resume in its args, which the harness now owns fixed in the commit that adds this page: validate() rejects -resume in nfcore[].args, with a unit test
CR-T7R-03 minor The T7 Progress page, the pipelines-real.yml header and the research note still said public runners are not measured fixed in the commit that adds this page: a dated resolution line in T7 Findings item 3, the workflow header names runner_limits.<visibility>, and a dated note in corpus-ci-2026-10-27 §1
SEC-T7R-01 minor A retried pass is green, so a flaky first attempt left no visible trace; a run killed by a signal was retried too fixed in the commit that adds this page: result.json records first_exit (null when not retried); a retry prints ::warning title=corpus retry::<entry> exit <rc> on first attempt, <pass\|fail> on -resume; no retry when the exit code is negative (killed by a signal); unit tests for each
SEC-T7R-02 minor A retry may double an entry's runtime against the 120-min job timeout deferred: when the job times out, the cancel path still tails run.log (if: failure() \|\| cancelled() in pipelines-real.yml), so the log of both attempts survives
SEC-T7R-03 minor The spec did not say which failures are retried, nor how a retried pass is flagged fixed in the commit that adds this page: one clause in dev-env §10.2 (any non-zero exit except a signal kill; a retried pass is green, flagged by the annotation and first_exit)