|
|
|
#!/usr/bin/env python2.7
|
|
|
|
# Copyright (c) 2015 - present Facebook, Inc.
|
|
|
|
# All rights reserved.
|
|
|
|
#
|
|
|
|
# This source code is licensed under the BSD style license found in the
|
|
|
|
# LICENSE file in the root directory of this source tree. An additional grant
|
|
|
|
# of patent rights can be found in the PATENTS file in the same directory.
|
|
|
|
|
|
|
|
from __future__ import absolute_import
|
|
|
|
from __future__ import division
|
|
|
|
from __future__ import print_function
|
|
|
|
from __future__ import unicode_literals
|
|
|
|
|
|
|
|
import json
|
|
|
|
import os
|
|
|
|
import shutil
|
|
|
|
import subprocess
|
|
|
|
import sys
|
|
|
|
import tempfile
|
|
|
|
import unittest
|
|
|
|
|
|
|
|
SCRIPTS_DIRECTORY = os.path.dirname(os.path.realpath(__file__))
|
|
|
|
sys.path.insert(0,
|
|
|
|
os.path.join(SCRIPTS_DIRECTORY,
|
|
|
|
os.pardir, 'infer', 'lib', 'python'))
|
|
|
|
|
|
|
|
from inferlib import issues, utils
|
|
|
|
|
|
|
|
|
|
|
|
CURRENT_DIR = os.getcwd()
|
|
|
|
REPORT_JSON = 'report.json'
|
|
|
|
|
|
|
|
INFER_EXECUTABLE = 'infer'
|
|
|
|
|
|
|
|
RECORD_ENV = 'INFER_RECORD_INTEGRATION_TESTS'
|
|
|
|
|
|
|
|
REPORT_FIELDS = [
|
|
|
|
issues.JSON_INDEX_FILENAME,
|
|
|
|
issues.JSON_INDEX_PROCEDURE,
|
|
|
|
issues.JSON_INDEX_TYPE,
|
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
def should_record_tests():
|
|
|
|
return RECORD_ENV in os.environ and os.environ[RECORD_ENV] == '1'
|
|
|
|
|
|
|
|
|
|
|
|
def quote(s):
|
|
|
|
return '\"%s\"' % s
|
|
|
|
|
|
|
|
|
|
|
|
def string_of_error(e):
|
|
|
|
line = ''
|
|
|
|
if issues.JSON_INDEX_LINE in e:
|
|
|
|
line = ' on line %s ' % e[issues.JSON_INDEX_LINE]
|
|
|
|
msg = '%s in file %s, procedure %s%s' % (
|
|
|
|
e[issues.JSON_INDEX_TYPE],
|
|
|
|
quote(e[issues.JSON_INDEX_FILENAME]),
|
|
|
|
quote(e[issues.JSON_INDEX_PROCEDURE]),
|
|
|
|
line,
|
|
|
|
)
|
|
|
|
return msg
|
|
|
|
|
|
|
|
|
|
|
|
def save_report(reports, filename):
|
|
|
|
# sorting to avoid spurious differences between two lists of reports
|
|
|
|
reports.sort()
|
|
|
|
|
|
|
|
def filter_report(report):
|
|
|
|
return dict((k, v) for (k, v) in report.items() if k in REPORT_FIELDS)
|
|
|
|
|
|
|
|
def should_report(report):
|
|
|
|
return len(report) > 0
|
|
|
|
|
|
|
|
filtered = filter(should_report, map(filter_report, reports))
|
|
|
|
utils.dump_json_to_path(filtered, filename,
|
|
|
|
separators=(',', ': '), sort_keys=True)
|
|
|
|
|
|
|
|
|
|
|
|
def run_analysis(root, clean_cmd, build_cmd, analyzer):
|
|
|
|
os.chdir(root)
|
|
|
|
|
|
|
|
subprocess.check_call(clean_cmd)
|
|
|
|
|
|
|
|
temp_out_dir = tempfile.mkdtemp(suffix='_out', prefix='infer_')
|
|
|
|
infer_cmd = ['infer', '-a', analyzer, '-o', temp_out_dir, '--'] + build_cmd
|
|
|
|
|
|
|
|
with tempfile.TemporaryFile(
|
|
|
|
mode='w',
|
|
|
|
suffix='.out',
|
|
|
|
prefix='analysis_') as analysis_output:
|
|
|
|
subprocess.check_call(infer_cmd, stdout=analysis_output)
|
|
|
|
|
|
|
|
json_path = os.path.join(temp_out_dir, REPORT_JSON)
|
|
|
|
found_errors = utils.load_json_from_path(json_path)
|
|
|
|
shutil.rmtree(temp_out_dir)
|
|
|
|
os.chdir(CURRENT_DIR)
|
|
|
|
|
|
|
|
return found_errors
|
|
|
|
|
|
|
|
|
|
|
|
def match_pattern(f, p):
|
|
|
|
for key in p.keys():
|
|
|
|
if f[key] != p[key]:
|
|
|
|
return False
|
|
|
|
return True
|
|
|
|
|
|
|
|
|
|
|
|
def is_expected(e, patterns):
|
|
|
|
for p in patterns:
|
|
|
|
if match_pattern(e, p):
|
|
|
|
return True
|
|
|
|
return False
|
|
|
|
|
|
|
|
|
|
|
|
def is_missing(p, errors):
|
|
|
|
for e in errors:
|
|
|
|
if match_pattern(e, p):
|
|
|
|
return False
|
|
|
|
return True
|
|
|
|
|
|
|
|
|
|
|
|
def unexpected_errors(errors, patterns):
|
|
|
|
return [e for e in errors if not is_expected(e, patterns)]
|
|
|
|
|
|
|
|
|
|
|
|
def missing_errors(errors, patterns):
|
|
|
|
return [p for p in patterns if is_missing(p, errors)]
|
|
|
|
|
|
|
|
|
|
|
|
def check_results(errors, patterns):
|
|
|
|
unexpected = unexpected_errors(errors, patterns)
|
|
|
|
if unexpected != []:
|
|
|
|
utils.stderr('\nInfer found the following unexpected errors:')
|
|
|
|
for e in unexpected:
|
|
|
|
utils.stderr('\t{}\n'.format(string_of_error(e)))
|
|
|
|
missing = missing_errors(errors, patterns)
|
|
|
|
if missing != []:
|
|
|
|
utils.stderr('\nInfer did not find the following errors:')
|
|
|
|
for p in missing:
|
|
|
|
utils.stderr('\t{}\n'.format(string_of_error(p)))
|
|
|
|
assert unexpected == []
|
|
|
|
assert missing == []
|
|
|
|
|
|
|
|
|
|
|
|
def is_tool_available(cmd):
|
|
|
|
try:
|
|
|
|
subprocess.call(cmd)
|
|
|
|
except OSError as e:
|
|
|
|
if e.errno == os.errno.ENOENT:
|
|
|
|
return False
|
|
|
|
else:
|
|
|
|
raise
|
|
|
|
return True
|
|
|
|
|
|
|
|
|
|
|
|
def do_test(errors, expected_errors_filename):
|
|
|
|
if should_record_tests():
|
|
|
|
save_report(errors, expected_errors_filename)
|
|
|
|
return
|
|
|
|
else:
|
|
|
|
patterns = utils.load_json_from_path(expected_errors_filename)
|
|
|
|
check_results(errors, patterns)
|
|
|
|
|
|
|
|
|
|
|
|
class BuildIntegrationTest(unittest.TestCase):
|
|
|
|
|
|
|
|
def test_ant_integration(self):
|
|
|
|
if is_tool_available(['ant', '-version']):
|
|
|
|
print('\nRunning Gradle integration test')
|
|
|
|
root = os.path.join(CURRENT_DIR, 'infer', 'tests')
|
|
|
|
errors = run_analysis(
|
|
|
|
root,
|
|
|
|
['ant', 'clean'],
|
|
|
|
['ant', 'compile'],
|
|
|
|
INFER_EXECUTABLE)
|
|
|
|
do_test(errors, os.path.join(root, 'ant_report.json'))
|
|
|
|
else:
|
|
|
|
print('\nSkipping Ant integration test')
|
|
|
|
assert True
|
|
|
|
|
|
|
|
def test_gradle_integration(self):
|
|
|
|
if is_tool_available(['gradle', '--version']):
|
|
|
|
print('\nRunning Gradle integration test')
|
|
|
|
root = os.path.join(CURRENT_DIR, 'examples', 'android_hello')
|
|
|
|
errors = run_analysis(
|
|
|
|
root,
|
|
|
|
['gradle', 'clean'],
|
|
|
|
['gradle', 'build'],
|
|
|
|
INFER_EXECUTABLE)
|
|
|
|
do_test(errors, os.path.join(root, 'gradle_report.json'))
|
|
|
|
else:
|
|
|
|
print('\nSkipping Gradle integration test')
|
|
|
|
assert True
|
|
|
|
|
|
|
|
def test_buck_integration(self):
|
|
|
|
if is_tool_available(['buck', '--version']):
|
|
|
|
print('\nRunning Buck integration test')
|
|
|
|
root = CURRENT_DIR
|
|
|
|
errors = run_analysis(
|
|
|
|
root,
|
|
|
|
['buck', 'clean'],
|
|
|
|
['buck', 'build', 'infer'],
|
|
|
|
INFER_EXECUTABLE)
|
|
|
|
report_path = os.path.join(
|
|
|
|
CURRENT_DIR, 'infer', 'tests', 'buck_report.json')
|
|
|
|
do_test(errors, report_path)
|
|
|
|
else:
|
|
|
|
print('\nSkipping Buck integration test')
|
|
|
|
assert True
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
unittest.main() # run all the tests
|