annator-command-center / tests /scripts /flaky_test_detector.py
techprotrade's picture
Deploy ATOM FastAPI command center runtime (part 9)
81e3673 verified
Raw
History Blame Contribute Delete
20.4 kB
#!/usr/bin/env python
"""
Flaky Test Detection Script for Atom Test Suite
This script identifies flaky tests by running tests multiple times with different
random seeds and recording which tests fail intermittently.
Flaky tests are those that:
- Fail in some runs but pass in others (inconsistent behavior)
- Often indicate race conditions, timing issues, or shared state problems
Usage:
python flaky_test_detector.py --runs 3 --update-json
python flaky_test_detector.py --help
Exit Codes:
0: No flaky tests detected
1: Flaky tests found
2: Error in execution
"""
import argparse
import json
import os
import subprocess
import sys
from datetime import datetime
from pathlib import Path
from typing import Dict, List, Set, Tuple, Optional
# Import FlakyTestTracker for database integration
try:
from tests.scripts.flaky_test_tracker import FlakyTestTracker
except ImportError:
FlakyTestTracker = None
def run_tests_with_seed(seed: int, test_path: str = "tests/", verbose: bool = False) -> Set[str]:
"""
Run pytest with a specific random seed and return failed test names.
Args:
seed: Random seed for test order randomization
test_path: Path to tests directory
verbose: Enable verbose output
Returns:
Set of failed test names
"""
cmd = [
"python3", "-m", "pytest",
test_path,
"-q",
"--random-order-seed", str(seed),
"--tb=no",
"--no-header"
]
if verbose:
print(f"\nRunning: {' '.join(cmd)}")
result = subprocess.run(
cmd,
capture_output=True,
text=True,
cwd=Path(__file__).parent.parent.parent
)
# Parse failed tests from output
failed_tests = set()
# pytest output format: "FAILED tests/test_module.py::test_function"
for line in result.stdout.split('\n'):
if line.startswith('FAILED '):
test_name = line.split(' ', 1)[1].strip()
failed_tests.add(test_name)
if verbose:
print(f"Failed tests (seed={seed}): {len(failed_tests)}")
for test in failed_tests:
print(f" - {test}")
return failed_tests
def run_test_multiple_times(
test_path: str,
runs: int = 10,
pytest_args: Optional[List[str]] = None,
verbose: bool = False
) -> Tuple[int, List[bool], Dict[str, any]]:
"""
Run a test multiple times to detect flakiness.
Args:
test_path: Test identifier (e.g., tests/test_module.py::test_function)
runs: Number of times to run the test
pytest_args: Additional pytest arguments
verbose: Enable verbose output
Returns:
(failure_count, failure_list, report_dict)
"""
pytest_args = pytest_args or []
failures = []
for i in range(runs):
cmd = [
"python3", "-m", "pytest",
test_path,
"-v",
"--tb=no",
"--no-header"
] + pytest_args
if verbose:
print(f"Run {i+1}/{runs}: {' '.join(cmd)}")
result = subprocess.run(
cmd,
capture_output=True,
text=True,
cwd=Path(__file__).parent.parent.parent
)
failed = result.returncode != 0
failures.append(failed)
if verbose and failed:
print(f" -> FAILED")
failure_count = sum(failures)
flaky_rate = failure_count / runs if runs > 0 else 0.0
# Classify flakiness
if failure_count == 0:
classification = "stable"
elif failure_count == runs:
classification = "broken"
elif 0 < failure_count < runs:
classification = "flaky"
report = {
"test_path": test_path,
"total_runs": runs,
"failures": failure_count,
"flaky_rate": round(flaky_rate, 3),
"classification": classification,
"failure_details": [
{"run": i, "failed": failed}
for i, failed in enumerate(failures)
]
}
return failure_count, failures, report
def classify_flakiness(failure_count: int, total_runs: int) -> Tuple[str, float]:
"""
Classify test flakiness based on failure patterns.
Args:
failure_count: Number of test failures
total_runs: Total number of test runs
Returns:
(classification, flaky_rate)
- classification: "stable", "flaky", or "broken"
- flaky_rate: Failure rate (0.0 to 1.0)
"""
if total_runs == 0:
return "stable", 0.0
flaky_rate = failure_count / total_runs
if failure_count == 0:
classification = "stable"
elif failure_count == total_runs:
classification = "broken"
elif 0 < failure_count < total_runs:
classification = "flaky"
return classification, round(flaky_rate, 3)
def parse_test_results(output: str) -> Set[str]:
"""
Parse pytest output to extract failed test names.
Args:
output: Pytest stdout/stderr combined output
Returns:
Set of failed test names
"""
failed_tests = set()
for line in output.split('\n'):
if line.startswith('FAILED '):
test_name = line.split(' ', 1)[1].strip()
failed_tests.add(test_name)
return failed_tests
def compare_results(results_list: List[Set[str]]) -> Dict[str, int]:
"""
Compare test results across multiple runs to count failures.
Args:
results_list: List of failed test sets from each run
Returns:
Dictionary mapping test name to failure count
"""
failure_counts = {}
for failed_set in results_list:
for test_name in failed_set:
if test_name not in failure_counts:
failure_counts[test_name] = 0
failure_counts[test_name] += 1
return failure_counts
def identify_flaky(failure_counts: Dict[str, int], total_runs: int) -> Dict[str, float]:
"""
Identify flaky tests from failure counts.
Flaky tests fail in at least one run but not all runs.
They exhibit inconsistent behavior across multiple runs.
Args:
failure_counts: Dictionary of test -> failure count
total_runs: Total number of test runs
Returns:
Dictionary mapping flaky test to failure frequency (0-1)
"""
flaky_tests = {}
for test_name, failures in failure_counts.items():
# Flaky: fails in some runs but not all (0 < failures < total_runs)
if 0 < failures < total_runs:
frequency = failures / total_runs
flaky_tests[test_name] = frequency
return flaky_tests
def update_health_json(flaky_tests: Dict[str, float], phase: str = "090", plan: str = "02") -> None:
"""
Update test_health.json with flaky test entries.
Args:
flaky_tests: Dictionary of flaky test -> failure frequency
phase: Current phase number
plan: Current plan number
"""
health_file = Path(__file__).parent.parent / "coverage_reports" / "metrics" / "test_health.json"
# Load existing health data or create new structure
if health_file.exists():
try:
with open(health_file, 'r') as f:
health_data = json.load(f)
except (json.JSONDecodeError, IOError):
health_data = {}
else:
health_data = {}
# Ensure structure exists
if "flaky_tests" not in health_data:
health_data["flaky_tests"] = []
if "metadata" not in health_data:
health_data["metadata"] = {}
# Add current flaky test detection results
timestamp = datetime.now().isoformat()
for test_name, frequency in flaky_tests.items():
entry = {
"test_name": test_name,
"failure_frequency": round(frequency, 2),
"detected_date": timestamp,
"phase": phase,
"plan": plan
}
health_data["flaky_tests"].append(entry)
# Update metadata
health_data["metadata"]["format_version"] = 1
health_data["metadata"]["last_flaky_scan"] = timestamp
# Write back to file
health_file.parent.mkdir(parents=True, exist_ok=True)
with open(health_file, 'w') as f:
json.dump(health_data, f, indent=2)
def print_summary(
flaky_tests: Dict[str, float],
total_runs: int,
failure_counts: Dict[str, int],
verbose: bool = False
) -> None:
"""
Print formatted summary of flaky test detection.
Args:
flaky_tests: Dictionary of flaky test -> failure frequency
total_runs: Total number of test runs
failure_counts: All failure counts (including stable failures)
verbose: Enable verbose output
"""
print("\n" + "="*70)
print("FLAKY TEST DETECTION")
print("="*70)
print(f"\nTest Runs: {total_runs}")
print(f"Total Failed Tests (across all runs): {len(failure_counts)}")
print(f"Flaky Tests (inconsistent failures): {len(flaky_tests)}")
if flaky_tests:
print("\n" + "-"*70)
print("FLAKY TESTS DETECTED:")
print("-"*70)
# Sort by failure frequency (most frequent first)
sorted_tests = sorted(
flaky_tests.items(),
key=lambda x: x[1],
reverse=True
)
for test_name, frequency in sorted_tests:
failure_pct = frequency * 100
failure_count = int(frequency * total_runs)
print(f"\n {test_name}")
print(f" Failed {failure_count}/{total_runs} times ({failure_pct:.0f}%)")
print("\n" + "="*70)
print("STATUS: FLAKY TESTS FOUND ✗")
print("="*70)
print("\nRECOMMENDED ACTIONS:")
print(" 1. Investigate race conditions or timing dependencies")
print(" 2. Check for shared state between tests")
print(" 3. Add proper mocks for external dependencies")
print(" 4. Use unique_resource_name fixture for parallel isolation")
print(" 5. Mark with @pytest.mark.flaky as TEMPORARY workaround")
print("="*70 + "\n")
else:
if failure_counts:
print("\n" + "-"*70)
print("STABLE FAILURES (not flaky):")
print("-"*70)
for test_name in failure_counts.keys():
print(f" - {test_name}")
print("-"*70)
print("\n" + "="*70)
print("STATUS: NO FLAKY TESTS ✓")
print("="*70 + "\n")
if verbose and failure_counts:
print("\nVerbose Output:")
print("All Test Failures by Frequency:")
for test_name, count in sorted(failure_counts.items(), key=lambda x: x[1], reverse=True):
print(f" {test_name}: {count}/{total_runs} failures")
print()
def export_flaky_tests_json(
flaky_tests_data: List[Dict],
total_tests_scanned: int,
output_path: Path
) -> None:
"""Export flaky test results to JSON file.
Args:
flaky_tests_data: List of flaky test records with details
total_tests_scanned: Total number of tests scanned
output_path: Path to output JSON file
"""
# Calculate summary statistics
flaky_count = sum(1 for t in flaky_tests_data if t['classification'] == 'flaky')
broken_count = sum(1 for t in flaky_tests_data if t['classification'] == 'broken')
stable_count = total_tests_scanned - flaky_count - broken_count
output_data = {
"scan_date": datetime.now().isoformat(),
"detection_runs": len(set(t['total_runs'] for t in flaky_tests_data)) if flaky_tests_data else 0,
"flaky_tests": flaky_tests_data,
"summary": {
"total_tests_scanned": total_tests_scanned,
"flaky_count": flaky_count,
"broken_count": broken_count,
"stable_count": stable_count
}
}
# Ensure output directory exists
output_path.parent.mkdir(parents=True, exist_ok=True)
# Write JSON output
with open(output_path, 'w') as f:
json.dump(output_data, f, indent=2)
print(f"\nJSON export written to: {output_path}")
def record_to_quarantine_db(
flaky_tests: Dict[str, float],
total_runs: int,
db_path: Path,
platform: str
) -> None:
"""Record flaky tests to SQLite quarantine database.
Args:
flaky_tests: Dictionary of test -> flaky_rate
total_runs: Total number of runs
db_path: Path to SQLite database
platform: Platform name
"""
if FlakyTestTracker is None:
print("WARNING: FlakyTestTracker not available, skipping database recording")
return
tracker = FlakyTestTracker(db_path)
try:
for test_path, flaky_rate in flaky_tests.items():
failure_count = int(flaky_rate * total_runs)
# Generate failure history
failure_history = []
for i in range(total_runs):
# Distribute failures based on flaky_rate
if i < failure_count:
failure_history.append({"run": i, "failed": True})
else:
failure_history.append({"run": i, "failed": False})
# Classify flakiness
classification, _ = classify_flakiness(failure_count, total_runs)
# Record in database
tracker.record_flaky_test(
test_path,
platform,
total_runs,
failure_count,
classification,
failure_history,
quarantine_reason=f"Detected via flaky_test_detector.py"
)
print(f"\nRecorded {len(flaky_tests)} flaky tests to quarantine database: {db_path}")
finally:
tracker.close()
def main():
"""Main entry point for flaky test detection."""
parser = argparse.ArgumentParser(
description="Detect flaky tests by running multiple times with random seeds",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog="""
Examples:
python flaky_test_detector.py --runs 3
python flaky_test_detector.py --runs 2 --update-json --verbose
python flaky_test_detector.py --runs 3 --test-path tests/unit/
Exit Codes:
0: No flaky tests detected
1: Flaky tests found
2: Error in execution
How it works:
1. Runs the test suite N times with different random seeds
2. Records which tests fail in each run
3. Identifies tests that fail inconsistently (not 0 or N failures)
4. Updates test_health.json with flaky test entries
Flaky tests indicate:
- Race conditions in parallel execution
- Timing dependencies without proper mocking
- Shared state between tests
- Non-deterministic test data
Multi-run mode:
--multi-run: Run specific test N times to detect flakiness
--runs 10 --test-path tests/test_module.py::test_function
"""
)
parser.add_argument(
"--runs",
type=int,
default=3,
help="Number of test runs (default: 3)"
)
parser.add_argument(
"--test-path",
type=str,
default="tests/",
help="Path to tests directory or specific test (default: tests/)"
)
parser.add_argument(
"--update-json",
action="store_true",
help="Update test_health.json with flaky test entries"
)
parser.add_argument(
"--verbose",
action="store_true",
help="Enable verbose output"
)
parser.add_argument(
"--phase",
type=str,
default="151",
help="Current phase number for health tracking (default: 151)"
)
parser.add_argument(
"--plan",
type=str,
default="01",
help="Current plan number for health tracking (default: 01)"
)
parser.add_argument(
"--multi-run",
action="store_true",
help="Enable multi-run verification mode (run single test N times)"
)
parser.add_argument(
"--quarantine-db",
type=str,
default=None,
help="Path to SQLite quarantine database (default: None, no database tracking)"
)
parser.add_argument(
"--platform",
type=str,
default="backend",
choices=["backend", "frontend", "mobile", "desktop"],
help="Platform name for quarantine tracking (default: backend)"
)
parser.add_argument(
"--output",
type=str,
default=None,
help="Path to JSON export file (default: None, no export)"
)
args = parser.parse_args()
if args.runs < 2:
print("ERROR: --runs must be at least 2 for flaky test detection")
sys.exit(2)
# Multi-run mode: Run single test multiple times
if args.multi_run:
print("="*70)
print(f"FLAKY TEST DETECTION: Multi-run verification ({args.runs} runs)")
print(f"Test: {args.test_path}")
print("="*70)
failure_count, failures, report = run_test_multiple_times(
args.test_path,
args.runs,
verbose=args.verbose
)
print("\n" + "="*70)
print("MULTI-RUN VERIFICATION RESULTS")
print("="*70)
print(f"\nTest: {report['test_path']}")
print(f"Total Runs: {report['total_runs']}")
print(f"Failures: {report['failures']}")
print(f"Flaky Rate: {report['flaky_rate']}")
print(f"Classification: {report['classification'].upper()}")
if args.verbose:
print("\nFailure Details:")
for detail in report['failure_details']:
status = "FAILED" if detail['failed'] else "PASSED"
print(f" Run {detail['run']}: {status}")
print("="*70)
# Return exit code based on classification
if report['classification'] == 'flaky':
sys.exit(1)
elif report['classification'] == 'broken':
sys.exit(1)
else:
sys.exit(0)
# Original mode: Run full test suite with random seeds
print("="*70)
print(f"FLAKY TEST DETECTION: {args.runs} runs with random seeds")
print("="*70)
# Run tests multiple times with different seeds
results_list = []
for i in range(args.runs):
seed = i * 1000 # Use different seeds: 0, 1000, 2000, ...
print(f"\nRun {i+1}/{args.runs} (seed={seed})...", end=" ")
failed_tests = run_tests_with_seed(seed, args.test_path, args.verbose)
results_list.append(failed_tests)
print(f"{len(failed_tests)} failed")
# Compare results across runs
failure_counts = compare_results(results_list)
# Identify flaky tests (inconsistent failures)
flaky_tests = identify_flaky(failure_counts, args.runs)
# Print summary
print_summary(flaky_tests, args.runs, failure_counts, args.verbose)
# Update health JSON if requested
if args.update_json and flaky_tests:
update_health_json(flaky_tests, phase=args.phase, plan=args.plan)
if args.verbose:
print(f"Updated test_health.json with {len(flaky_tests)} flaky tests\n")
# Record to quarantine database if requested
if args.quarantine_db and flaky_tests:
db_path = Path(args.quarantine_db)
record_to_quarantine_db(flaky_tests, args.runs, db_path, args.platform)
# Export to JSON if requested
if args.output:
# Build flaky tests data for export
flaky_tests_data = []
for test_path, flaky_rate in flaky_tests.items():
failure_count = int(flaky_rate * args.runs)
classification, _ = classify_flakiness(failure_count, args.runs)
flaky_tests_data.append({
"test_path": test_path,
"platform": args.platform,
"total_runs": args.runs,
"failure_count": failure_count,
"flaky_rate": round(flaky_rate, 3),
"classification": classification,
"failure_details": [
{"run": i, "failed": i < failure_count}
for i in range(args.runs)
]
})
# Count total unique tests across all runs
total_tests_scanned = len(failure_counts)
export_path = Path(args.output)
export_flaky_tests_json(flaky_tests_data, total_tests_scanned, export_path)
# Return exit code
if flaky_tests:
sys.exit(1)
else:
sys.exit(0)
if __name__ == "__main__":
main()