You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

535 lines
16 KiB

#!/usr/bin/env python
#
# Copyright (c) 2013-present Facebook. All rights reserved.
#
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from __future__ import unicode_literals
import argparse
import csv
import io
import json
import logging
import multiprocessing
import os
import platform
import re
import shutil
import stat
import subprocess
import sys
import tempfile
import time
import traceback
import zipfile
# Infer imports
import inferlib
import utils
ANALYSIS_SUMMARY_OUTPUT = 'analysis_summary.txt'
BUCK_CONFIG = '.buckconfig.local'
BUCK_CONFIG_BACKUP = '.buckconfig.local.backup_generated_by_BuckAnalyze'
DEFAULT_BUCK_OUT = os.path.join(os.getcwd(), 'buck-out')
DEFAULT_BUCK_OUT_GEN = os.path.join(DEFAULT_BUCK_OUT, 'gen')
INFER_REPORT = os.path.join(utils.BUCK_INFER_OUT, utils.CSV_REPORT_FILENAME)
INFER_STATS = os.path.join(utils.BUCK_INFER_OUT, utils.STATS_FILENAME)
INFERJ_SCRIPT = """\
#!/bin/sh
{0} {1} javac $@
"""
LOCAL_CONFIG = """\
[tools]
javac = %s
"""
def prepare_build(args):
"""Creates script that redirects javac calls to inferJ and a local buck
configuration that tells buck to use that script.
"""
inferJ_options = [
'--buck',
'--analyzer',
args.analyzer,
]
if args.debug:
inferJ_options.append('--debug')
if args.no_filtering:
inferJ_options.append('--no-filtering')
if args.analyzer_mode:
inferJ_options.append('--analyzer_mode')
inferJ_options.append(args.analyzer_mode)
# Create a temporary directory as a cache for jar files.
infer_cache_dir = os.path.join(args.infer_out, 'cache')
if not os.path.isdir(infer_cache_dir):
os.mkdir(infer_cache_dir)
inferJ_options.append('--infer_cache')
inferJ_options.append(infer_cache_dir)
temp_files = [infer_cache_dir]
try:
inferJ = utils.get_cmd_in_bin_dir('inferJ') + ' ' +\
' '.join(inferJ_options)
except subprocess.CalledProcessError as e:
logging.error('Could not find inferJ')
raise e
# Disable the use of buckd as this scripts modifies .buckconfig.local
logging.info('Disabling buckd: export NO_BUCKD=1')
os.environ['NO_BUCKD'] = '1'
# make sure INFER_ANALYSIS is set when buck is called
logging.info('Setup Infer analysis mode for Buck: export INFER_ANALYSIS=1')
os.environ['INFER_ANALYSIS'] = '1'
# Create a script to be called by buck
inferJ_script = None
with tempfile.NamedTemporaryFile(delete=False,
prefix='inferJ_',
suffix='.sh',
dir='.') as inferJ_script:
logging.info('Creating %s' % inferJ_script.name)
inferJ_script.file.write(
(INFERJ_SCRIPT.format(sys.executable, inferJ)).encode())
st = os.stat(inferJ_script.name)
os.chmod(inferJ_script.name, st.st_mode | stat.S_IEXEC)
# Backup and patch local buck config
patched_config = ''
if os.path.isfile(BUCK_CONFIG):
logging.info('Backing up %s to %s', BUCK_CONFIG, BUCK_CONFIG_BACKUP)
shutil.move(BUCK_CONFIG, BUCK_CONFIG_BACKUP)
with open(BUCK_CONFIG_BACKUP) as buckconfig:
patched_config = '\n'.join(buckconfig)
javac_section = '[tools]\n{0}javac = {1}'.format(
' ' * 4,
inferJ_script.name)
patched_config += javac_section
with open(BUCK_CONFIG, 'w') as buckconfig:
buckconfig.write(patched_config)
temp_files += [inferJ_script.name]
return temp_files
def java_targets():
target_types = [
'android_library',
'java_library',
]
try:
targets = subprocess.check_output([
'buck',
'targets',
'--type',
] + target_types).decode().strip().split('\n')
except subprocess.CalledProcessError as e:
logging.error('Could not compute java library targets')
raise e
return set(targets)
def is_alias(target):
return ':' not in target
def expand_target(target, java_targets):
if not is_alias(target):
return [target]
else:
try:
buck_audit_cmd = ['buck', 'audit', 'classpath', '--dot', target]
output = subprocess.check_output(buck_audit_cmd)
dotty = output.decode().split('\n')
except subprocess.CalledProcessError as e:
logging.error('Could not expand target {0}'.format(target))
raise e
targets = set()
edge_re = re.compile('.*"(.*)".*"(.*)".*')
for line in dotty:
match = re.match(edge_re, line)
if match:
for t in match.groups():
if t in java_targets:
targets.add(t)
return targets
def normalize_target(target):
if is_alias(target) or target.startswith('//'):
return target
else:
return '//' + target
def determine_library_targets(args):
""" Uses git and buck audit to expand aliases into the list of java or
android library targets that are parts of these aliases.
Buck targets directly passed as argument are not expanded """
args.targets = [normalize_target(t) for t in args.targets]
if any(map(is_alias, args.targets)):
all_java_targets = java_targets()
targets = set()
for t in args.targets:
targets.update(expand_target(t, all_java_targets))
args.targets = list(targets)
if args.verbose:
logging.debug('Targets to analyze:')
for target in args.targets:
logging.debug(target)
def init_stats(args, start_time):
"""Returns dictionary with target independent statistics.
"""
return {
'float': {},
'int': {
'cores': multiprocessing.cpu_count(),
'time': int(time.time()),
'start_time': int(round(start_time)),
},
'normal': {
'debug': str(args.debug),
'analyzer': args.analyzer,
'machine': platform.machine(),
'node': platform.node(),
'project': os.path.basename(os.getcwd()),
'revision': utils.vcs_revision(),
'branch': utils.vcs_branch(),
'system': platform.system(),
'infer_version': utils.infer_version(),
'infer_branch': utils.infer_branch(),
}
}
def store_performances_csv(infer_out, stats):
"""Stores the statistics about perfromances into a CSV file to be exported
to a database"""
perf_filename = os.path.join(infer_out, utils.CSV_PERF_FILENAME)
with open(perf_filename, 'w') as csv_file_out:
csv_writer = csv.writer(csv_file_out)
keys = ['infer_version', 'project', 'revision', 'files', 'lines',
'cores', 'system', 'machine', 'node', 'total_time',
'capture_time', 'analysis_time', 'reporting_time', 'time']
int_stats = list(stats['int'].items())
normal_stats = list(stats['normal'].items())
flat_stats = dict(int_stats + normal_stats)
values = []
for key in keys:
values.append(flat_stats[key])
csv_writer.writerow(keys)
csv_writer.writerow(values)
csv_file_out.flush()
def get_harness_code():
all_harness_code = '\nGenerated harness code:\n'
for filename in os.listdir(DEFAULT_BUCK_OUT_GEN):
if 'InferGeneratedHarness' in filename:
all_harness_code += '\n' + filename + ':\n'
with open(os.path.join(DEFAULT_BUCK_OUT_GEN,
filename), 'r') as file_in:
all_harness_code += file_in.read()
return all_harness_code + '\n'
def get_basic_stats(stats):
files_analyzed = '{0} files ({1} lines) analyzed in {2}s\n\n'.format(
stats['int']['files'],
stats['int']['lines'],
stats['int']['total_time'],
)
phase_times = 'Capture time: {0}s\nAnalysis time: {1}s\n\n'.format(
stats['int']['capture_time'],
stats['int']['analysis_time'],
)
to_skip = {
'files',
'lines',
'cores',
'time',
'start_time',
'capture_time',
'analysis_time',
'reporting_time',
'total_time',
}
bugs_found = 'Errors found:\n\n'
for key, value in sorted(stats['int'].items()):
if key not in to_skip:
bugs_found += ' {0:>8} {1}\n'.format(value, key)
basic_stats_message = files_analyzed + phase_times + bugs_found + '\n'
return basic_stats_message
def get_buck_stats():
trace_filename = os.path.join(
DEFAULT_BUCK_OUT,
'log',
'traces',
'build.trace'
)
ARGS = 'args'
SUCCESS_STATUS = 'success_type'
buck_stats = {}
try:
with open(trace_filename, 'r') as file_in:
trace = json.load(file_in)
for t in trace:
if SUCCESS_STATUS in t[ARGS]:
status = t[ARGS][SUCCESS_STATUS]
count = buck_stats.get(status, 0)
buck_stats[status] = count + 1
buck_stats_message = 'Buck build statistics:\n\n'
for key, value in sorted(buck_stats.items()):
buck_stats_message += ' {0:>8} {1}\n'.format(value, key)
return buck_stats_message
except IOError as e:
logging.error('Caught %s: %s' % (e.__class__.__name__, str(e)))
logging.error(traceback.format_exc())
return ''
class NotFoundInJar(Exception):
pass
def load_stats(opened_jar):
try:
return json.loads(opened_jar.read(INFER_STATS).decode())
except KeyError as e:
raise NotFoundInJar
def load_report(opened_jar):
try:
sio = io.StringIO(opened_jar.read(INFER_REPORT).decode())
return list(csv.reader(sio))
except KeyError as e:
raise NotFoundInJar
def rows_remove_duplicates(rows):
seen = {}
result = []
for row in rows:
t = tuple(row)
if t in seen:
continue
seen[t] = 1
result.append(row)
return result
def collect_results(args, start_time):
"""Walks through buck-gen, collects results for the different buck targets
and stores them in in args.infer_out/results.csv.
"""
buck_stats = get_buck_stats()
logging.info(buck_stats)
with open(os.path.join(args.infer_out, ANALYSIS_SUMMARY_OUTPUT), 'w') as f:
f.write(buck_stats)
all_rows = []
headers = []
stats = init_stats(args, start_time)
accumulation_whitelist = list(map(re.compile, [
'^cores$',
'^time$',
'^start_time$',
'.*_pc',
]))
expected_analyzer = stats['normal']['analyzer']
expected_version = stats['normal']['infer_version']
for root, _, files in os.walk(DEFAULT_BUCK_OUT_GEN):
for f in [f for f in files if f.endswith('.jar')]:
path = os.path.join(root, f)
try:
with zipfile.ZipFile(path) as jar:
# Accumulate integers and float values
target_stats = load_stats(jar)
found_analyzer = target_stats['normal']['analyzer']
found_version = target_stats['normal']['infer_version']
if (found_analyzer != expected_analyzer
or found_version != expected_version):
continue
else:
for type_k in ['int', 'float']:
items = target_stats.get(type_k, {}).items()
for key, value in items:
if not any(map(lambda r: r.match(key),
accumulation_whitelist)):
old_value = stats[type_k].get(key, 0)
stats[type_k][key] = old_value + value
rows = load_report(jar)
if len(rows) > 0:
headers.append(rows[0])
all_rows.extend(rows[1:])
# Override normals
stats['normal'].update(target_stats.get('normal', {}))
except NotFoundInJar:
pass
except zipfile.BadZipfile:
logging.warn('Bad zip file %s', path)
csv_report = os.path.join(args.infer_out, utils.CSV_REPORT_FILENAME)
bugs_out = os.path.join(args.infer_out, utils.BUGS_FILENAME)
if len(headers) == 0:
with open(csv_report, 'w'):
pass
logging.info('No reports found')
return
elif len(headers) > 1:
if any(map(lambda x: x != headers[0], headers)):
raise Exception('Inconsistent reports found')
# Convert all float values to integer values
for key, value in stats.get('float', {}).items():
stats['int'][key] = int(round(value))
# Delete the float entries before exporting the results
del(stats['float'])
with open(csv_report, 'w') as report:
writer = csv.writer(report)
writer.writerows([headers[0]] + rows_remove_duplicates(all_rows))
report.flush()
# export the CSV rows to JSON
utils.create_json_report(args.infer_out)
print('\n')
inferlib.print_errors(csv_report, bugs_out)
stats['int']['total_time'] = int(round(utils.elapsed_time(start_time)))
store_performances_csv(args.infer_out, stats)
stats_filename = os.path.join(args.infer_out, utils.STATS_FILENAME)
with open(stats_filename, 'w') as stats_out:
json.dump(stats, stats_out)
basic_stats = get_basic_stats(stats)
if args.print_harness:
harness_code = get_harness_code()
basic_stats += harness_code
logging.info(basic_stats)
with open(os.path.join(args.infer_out, ANALYSIS_SUMMARY_OUTPUT), 'a') as f:
f.write(basic_stats)
def cleanup(temp_files):
"""Removes the generated .buckconfig.local and the temporary inferJ script.
"""
for file in [BUCK_CONFIG] + temp_files:
try:
logging.info('Removing %s' % file)
if os.path.isdir(file):
shutil.rmtree(file)
else:
os.unlink(file)
except IOError:
logging.error('Could not remove %s' % file)
if os.path.isfile(BUCK_CONFIG_BACKUP):
logging.info('Restoring %s', BUCK_CONFIG)
shutil.move(BUCK_CONFIG_BACKUP, BUCK_CONFIG)
if __name__ == '__main__':
parser = argparse.ArgumentParser(parents=[inferlib.base_parser])
parser.add_argument('--verbose', action='store_true',
help='Print buck compilation steps')
parser.add_argument('--no-cache', action='store_true',
help='Do not use buck distributed cache')
parser.add_argument('--print-harness', action='store_true',
help='Print generated harness code (Android only)')
parser.add_argument('targets', nargs='*', metavar='target',
help='Build targets to analyze')
args = parser.parse_args()
utils.configure_logging(args.verbose)
timer = utils.Timer(logging.info)
temp_files = []
try:
start_time = time.time()
logging.info('Starting the analysis')
subprocess.check_call(
[utils.get_cmd_in_bin_dir('InferAnalyze'), '-version'])
if not os.path.isdir(args.infer_out):
os.mkdir(args.infer_out)
timer.start('Preparing build...')
temp_files += prepare_build(args)
timer.stop('Build prepared')
# TODO(t3786463) Start buckd.
timer.start('Computing library targets')
determine_library_targets(args)
timer.stop('%d targets computed', len(args.targets))
timer.start('Running buck...')
buck_cmd = ['buck', 'build']
if args.no_cache:
buck_cmd += ['--no-cache']
if args.verbose:
buck_cmd += ['-v', '2']
subprocess.check_call(buck_cmd + args.targets)
timer.stop('Buck finished')
timer.start('Collecting results...')
collect_results(args, start_time)
timer.stop('Done')
except KeyboardInterrupt as e:
timer.stop('Exiting')
sys.exit(0)
except Exception as e:
timer.stop('Failed')
logging.error('Caught %s: %s' % (e.__class__.__name__, str(e)))
logging.error(traceback.format_exc())
sys.exit(1)
finally:
cleanup(temp_files)
# vim: set sw=4 ts=4 et: