testland/lcov-analysis
Use when a coverage gate must work across languages from LCOV .info or Cobertura XML files (gcov, llvm-cov, coverage.py, JaCoCo, Jest, nyc, coverlet). Parses per-file line, function and branch metrics, compares with a baseline and returns per-file pass or fail verdicts.
88%
Does it follow best practices?
Run evals on this skill
Adds up to 20 points to the overall score
Passed
No findings from the security scan
format-diff-and-ci.mdreferences/
LCOV record format, baseline diff, and CI wiring
Deep reference for the lcov-analysis SKILL.md. Consult for the full .info
record-keyword table, the baseline-diff and gate reference implementation, and
the PR coverage-gate CI job.
.info record-keyword reference
Per lcov-readme, the LCOV coverage data format uses these record types:
| Keyword | Meaning |
|---|---|
TN:<test> | Test name (often empty for whole-suite captures). |
SF:<path> | Source file path (one record set per source file). |
FN:<line>,<name> | Function declared at <line> named <name>. |
FNDA:<count>,<name> | Function <name> was called <count> times. |
FNF:<n> | Functions found in this file. |
FNH:<n> | Functions hit at least once. |
BRDA:<line>,<block>,<branch>,<taken> | Branch coverage data. |
BRF:<n> | Branches found. |
BRH:<n> | Branches hit. |
DA:<line>,<count> | Line <line> was executed <count> times. |
LH:<n> | Lines hit. |
LF:<n> | Lines found. |
end_of_record | Marks completion of the current source file's data. |
Per record, BRDA's fourth value <taken> is the hit count for
that branch arm or - if the branch was never reached (the
preceding line wasn't executed).
Diff vs baseline
def coverage_diff(current, baseline):
"""For each file, compute (line%_now - line%_then), (branch%_now - branch%_then)."""
base_by_path = {f['path']: f for f in baseline}
out = []
for f in current:
b = base_by_path.get(f['path'])
line_now = pct(f['lh'], f['lf'])
line_then = pct(b['lh'], b['lf']) if b else None
branch_now = pct(f['brh'], f['brf'])
branch_then = pct(b['brh'], b['brf']) if b else None
out.append({
'path': f['path'],
'line_now': line_now, 'line_then': line_then,
'branch_now': branch_now, 'branch_then': branch_then,
'is_new': b is None,
})
return out
def pct(num, denom):
return None if denom == 0 else round(100 * num / denom, 1)The interesting outputs are drops (line_now < line_then) and new files with sub-threshold coverage (is_new and line_now < gate).
Gate
def gate(diff, whole_drop_max=0.5, file_drop_max=5.0, new_file_min=80.0):
failures = []
for f in diff:
if f['is_new'] and f['line_now'] is not None and f['line_now'] < new_file_min:
failures.append((f['path'], 'new file below threshold', f['line_now']))
elif f['line_then'] is not None and f['line_now'] is not None:
drop = f['line_then'] - f['line_now']
if drop > file_drop_max:
failures.append((f['path'], f'line% dropped {drop:.1f}pp', drop))
# Whole-repo drop:
sum_then_lh = sum(f['line_then'] for f in diff if f['line_then'] is not None)
sum_now_lh = sum(f['line_now'] for f in diff if f['line_now'] is not None)
return failuresCI shape
- name: Run tests with LCOV reporter
run: npm test -- --coverage --coverageReporters=lcov
- name: Download baseline
uses: actions/download-artifact@v4
with:
name: lcov-main
path: baseline/
- name: Parse + diff + gate
run: |
python scripts/parse_lcov.py coverage/lcov.info > current.json
python scripts/parse_lcov.py baseline/lcov.info > baseline.json
python scripts/coverage_gate.py current.json baseline.json
- name: Upload current LCOV (becomes next PR's baseline when on main)
if: github.ref == 'refs/heads/main'
uses: actions/upload-artifact@v4
with:
name: lcov-main
path: coverage/lcov.info
retention-days: 90Cache main's LCOV as an artifact so PRs diff against it rather than
recomputing the merge base's coverage twice.