[Fleet Monitor] petry-projects/.github-private — .github/workflows/lint.yml

[don-petry]

🟠 DEGRADED: petry-projects/.github-private / .github/workflows/lint.yml

Metric	Value
Failure Rate	31.8% (54 / 170 runs)
Status	DEGRADED
p50 Duration	129s
p95 Duration	318s
Successful Runs	116
Cancelled Runs	0

Workflow: .github/workflows/lint.yml

Last updated by Fleet Monitor run on 2026-06-14.
Threshold: failure rate > 10% over the monitored window.

[don-petry]

Dev-Lead Implementation Plan

Issue: #695 — [Fleet Monitor] petry-projects/.github-private — .github/workflows/lint.yml

Root-cause analysis

I reviewed the 33 recent lint.yml failures. The majority are CI correctly catching real
problems in PRs (e.g. shellcheck SC2034 on token_report.sh, a gh aw compile error,
PR-branch-specific bats failures whose line numbers no longer exist on main). Those are the
workflow doing its job and are not flakiness to be "fixed".

The one clear infra-level reliability defect is in the bats job: it runs a hardcoded list of
test files, and bats aborts the entire job running zero tests when any listed file is absent on
the branch (confirmed in run 27390275169: Error: Test file ".../test_initiative_planner_redispatch.bats" does not exist). The list has also drifted — 5 test files exist in tests/ but are not listed,
so they never run. The inline shell in the YAML is also untestable, which blocks a TDD approach.

Scope

Extract bats test execution into a small, unit-tested scripts/run-bats.sh that runs every listed
test file that exists, emits a GitHub ::warning:: for any missing one instead of aborting, and fails
only on a real test failure or when no test files are found. Wire lint.yml to call it.

Implementation Checklist

• scripts/run-bats.sh — resilient bats runner (skip+warn on missing file, fail on real failure / zero files)
• .github/workflows/lint.yml — bats job Run tests step calls bash scripts/run-bats.sh <list>
• .github/workflows/lint.yml — add tests/test_run_bats.bats to the maintained list (per AGENTS.md)

Tests to Write

• tests/test_run_bats.bats — all-present files run; missing file is skipped with a warning while others still run; zero existing files exits non-zero; a genuinely failing test propagates non-zero exit

Test Command

bats tests/*.bats (CI runs the maintained list; new test added to it)

[don-petry]

Dev-Lead: Implementation Complete

Plan Execution

• Root-caused the lint.yml failures across 33 recent runs
• Added scripts/run-bats.sh — resilient bats runner (skip+warn on missing file; fail only on real failure / zero files)
• Wired the bats job Run tests step to call bash scripts/run-bats.sh <list>
• Added tests/test_run_bats.bats to the maintained list (per AGENTS.md)
• TDD: wrote tests first (red), implemented to green
• Verified shellcheck, dev-lead-lint, full CI test list, YAML parse

Root cause

Most of the 33 recent failures are CI correctly catching real PR problems (e.g. shellcheck
SC2034 on token_report.sh, a gh aw compile error, PR-branch-specific bats failures whose line
numbers no longer exist on main) — those are the workflow doing its job, not flakiness.

The one clear infra-level reliability defect: the bats job ran a hardcoded list via
bats <file> ..., and bats aborts the entire job (running zero tests) the moment one listed
file is absent on the branch. Confirmed in run 27390275169:
Error: Test file ".../test_initiative_planner_redispatch.bats" does not exist. The inline YAML
shell was also untestable, blocking TDD.

scripts/run-bats.sh runs every listed file that exists, emits a GitHub ::warning:: for any that is
missing instead of aborting, and fails only on a genuine test failure or when none of the listed
files exist.

Test Results

# New unit tests (tests/test_run_bats.bats)
1..5
ok 1 runs all listed files when they all exist and pass (exit 0)
ok 2 missing listed file is skipped with a warning, present files still run (exit 0)
ok 3 exits non-zero when none of the listed files exist
ok 4 exits non-zero when called with no arguments
ok 5 a genuinely failing test propagates a non-zero exit

# Full CI test list via scripts/run-bats.sh: 264 tests, 0 failures (exit 0)
# Resilience check (missing file in list):
#   ::warning::run-bats: test file 'tests/DOES_NOT_EXIST.bats' not found — skipping
#   → remaining files still ran, exit 0 (previously: abort, exit 1, zero tests)

# shellcheck --severity=warning -x scripts/run-bats.sh  → OK
# .dev-lead/scripts/dev-lead-lint.sh                    → all checks passed

Files Changed

• scripts/run-bats.sh (new): resilient bats wrapper; skips+warns on missing files, fails on real failures / zero files.
• tests/test_run_bats.bats (new): unit tests for the wrapper.
• .github/workflows/lint.yml: bats job now calls scripts/run-bats.sh; added tests/test_run_bats.bats to the list.

Notes

• Kept the explicit, maintained test list (AGENTS.md mandates it; not switched to a glob), so the
  template-sync exception still applies — the new test file was appended to the list.
• Out of scope / not touched: pre-existing failures in tests/test_batch_fallback.bats (currently
  not in the CI list) and other repo test/script bugs — unrelated to lint.yml reliability.
• The ::warning:: for a missing listed file is intentional graceful degradation; a stale/renamed
  entry now warns loudly rather than turning the whole job red with no test signal.