def test_debug_sym_stack_2(self):
"""Tests debug build with a symbolized stack, with build type not explicitly
passed."""
self.assertEqual(
'[Command line] cmd_line_dbg\n\n' + self.start_seperator +
'Debug Build Stacktrace' + self.end_seperator + '\nsym_stack',
utils.get_crash_stacktrace_output('cmd_line_dbg', 'sym_stack'))
def test_beta_sym_stack(self):
"""Tests beta build with a symbolized stack, with build type not
explicitly passed."""
self.assertEqual(
'[Command line] cmd_line_beta\n\n' + self.start_seperator +
'Beta Build Stacktrace' + self.end_seperator + '\nsym_stack',
utils.get_crash_stacktrace_output('cmd_line_beta', 'sym_stack'))
def get_impact_on_build(build_type, current_version, testcase,
testcase_file_path):
"""Return impact and additional trace on a prod build given build_type."""
build = build_manager.setup_production_build(build_type)
if not build:
raise BuildFailedException('Build setup failed for %s' %
build_type.capitalize())
app_path = environment.get_value('APP_PATH')
if not app_path:
raise AppFailedException()
version = build.revision
if version == current_version:
return Impact(current_version, likely=False)
command = testcase_manager.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
result = testcase_manager.test_for_crash_with_retries(
testcase,
testcase_file_path,
environment.get_value('TEST_TIMEOUT'),
http_flag=testcase.http_flag)
if result.is_crash():
symbolized_crash_stacktrace = result.get_stacktrace(symbolized=True)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
stacktrace = utils.get_crash_stacktrace_output(
command, symbolized_crash_stacktrace,
unsymbolized_crash_stacktrace, build_type)
return Impact(version, likely=False, extra_trace=stacktrace)
return Impact()
def _check_fixed_for_custom_binary(testcase, job_type, testcase_file_path):
"""Simplified fixed check for test cases using custom binaries."""
revision = environment.get_value('APP_REVISION')
# Update comments to reflect bot information and clean up old comments.
testcase_id = testcase.key.id()
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(testcase, data_types.TaskState.STARTED)
build_manager.setup_build()
if not build_manager.check_app_path():
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(
testcase, data_types.TaskState.ERROR,
'Build setup failed for custom binary')
build_fail_wait = environment.get_value('FAIL_WAIT')
tasks.add_task(
'progression', testcase_id, job_type, wait_time=build_fail_wait)
return
test_timeout = environment.get_value('TEST_TIMEOUT', 10)
result = testcase_manager.test_for_crash_with_retries(
testcase, testcase_file_path, test_timeout, http_flag=testcase.http_flag)
_log_output(revision, result)
# Re-fetch to finalize testcase updates in branches below.
testcase = data_handler.get_testcase_by_id(testcase.key.id())
# If this still crashes on the most recent build, it's not fixed. The task
# will be rescheduled by a cron job and re-attempted eventually.
if result.is_crash():
app_path = environment.get_value('APP_PATH')
command = testcase_manager.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
symbolized_crash_stacktrace = result.get_stacktrace(symbolized=True)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
stacktrace = utils.get_crash_stacktrace_output(
command, symbolized_crash_stacktrace, unsymbolized_crash_stacktrace)
testcase.last_tested_crash_stacktrace = data_handler.filter_stacktrace(
stacktrace)
_update_completion_metadata(
testcase,
revision,
is_crash=True,
message='still crashes on latest custom build')
return
# Retry once on another bot to confirm our results and in case this bot is in
# a bad state which we didn't catch through our usual means.
if data_handler.is_first_retry_for_task(testcase, reset_after_retry=True):
tasks.add_task('progression', testcase_id, job_type)
_update_completion_metadata(testcase, revision)
return
# The bug is fixed.
testcase.fixed = 'Yes'
testcase.open = False
_update_completion_metadata(
testcase, revision, message='fixed on latest custom build')
def test_env_settings(self):
"""Tests that environment settings are added."""
os.environ['ASAN_OPTIONS'] = 'setting1=value1:setting2=value_2'
self.assertEqual(
'[Environment] ASAN_OPTIONS="setting1=value1:setting2=value_2"\n'
'[Command line] cmd_line\n\n' + self.start_seperator +
'Release Build Stacktrace' + self.end_seperator + '\nsym_stack',
utils.get_crash_stacktrace_output('cmd_line', 'sym_stack'))
def test_debug_sym_and_unsym_stacks(self):
"""Tests debug build with symbolized and unsymbolized stacks, with build
type explicitly passed."""
self.assertEqual(
'[Command line] cmd_line\n\n' + self.start_seperator +
'Debug Build Stacktrace' + self.end_seperator + '\nsym_stack\n\n' +
self.start_seperator + 'Debug Build Unsymbolized Stacktrace (diff)' +
self.end_seperator + '\n\nunsym_stack',
utils.get_crash_stacktrace_output(
'cmd_line', 'sym_stack', 'unsym_stack', build_type='debug'))
def test_release_sym_and_unsym_diff_stacks(self):
"""Tests release build with symbolized and unsymbolized stacks, having some
common frames, and build type not explicitly passed."""
self.assertEqual(
'[Command line] cmd_line\n\n' + self.start_seperator +
'Release Build Stacktrace' + self.end_seperator +
'\nc1\nc2\nc3\nsym_stack\nc4\nc5\nc6\n\n' + self.start_seperator +
'Release Build Unsymbolized Stacktrace (diff)' +
self.end_seperator + '\n\nc2\nc3\nunsym_stack\nc4\nc5',
utils.get_crash_stacktrace_output(
'cmd_line', 'c1\nc2\nc3\nsym_stack\nc4\nc5\nc6',
'c1\nc2\nc3\nunsym_stack\nc4\nc5\nc6'))
def _update_crash_result(testcase, crash_result, command):
"""Update testcase with crash result."""
min_state = crash_result.get_symbolized_data()
min_unsymbolized_crash_stacktrace = crash_result.get_stacktrace(
symbolized=False)
min_crash_stacktrace = utils.get_crash_stacktrace_output(
command, min_state.crash_stacktrace, min_unsymbolized_crash_stacktrace)
testcase.crash_type = min_state.crash_type
testcase.crash_address = min_state.crash_address
testcase.crash_state = min_state.crash_state
testcase.crash_stacktrace = data_handler.filter_stacktrace(
min_crash_stacktrace)
def get_impact_on_build(build_type, current_version, testcase,
testcase_file_path):
"""Return impact and additional trace on a prod build given build_type."""
# TODO(yuanjunh): remove es_enabled var after testing is done.
es_enabled = testcase.get_metadata('es_enabled', False)
if build_type == 'extended_stable' and not es_enabled:
return Impact()
build = build_manager.setup_production_build(build_type)
if not build:
raise BuildFailedException(
'Build setup failed for %s' % build_type.capitalize())
if not build_manager.check_app_path():
raise AppFailedException()
version = build.revision
if version == current_version:
return Impact(current_version, likely=False)
app_path = environment.get_value('APP_PATH')
command = testcase_manager.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
if es_enabled:
logs.log(
"ES build for testcase %d, command: %s" % (testcase.key.id(), command))
result = testcase_manager.test_for_crash_with_retries(
testcase,
testcase_file_path,
environment.get_value('TEST_TIMEOUT'),
http_flag=testcase.http_flag)
if result.is_crash():
symbolized_crash_stacktrace = result.get_stacktrace(symbolized=True)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
stacktrace = utils.get_crash_stacktrace_output(
command, symbolized_crash_stacktrace, unsymbolized_crash_stacktrace,
build_type)
return Impact(version, likely=False, extra_trace=stacktrace)
return Impact()
def execute_task(testcase_id, job_type):
"""Run analyze task."""
# Reset redzones.
environment.reset_current_memory_tool_options(redzone_size=128)
# Unset window location size and position properties so as to use default.
environment.set_value('WINDOW_ARG', '')
# Locate the testcase associated with the id.
testcase = data_handler.get_testcase_by_id(testcase_id)
if not testcase:
return
data_handler.update_testcase_comment(testcase, data_types.TaskState.STARTED)
metadata = data_types.TestcaseUploadMetadata.query(
data_types.TestcaseUploadMetadata.testcase_id == int(testcase_id)).get()
if not metadata:
logs.log_error(
'Testcase %s has no associated upload metadata.' % testcase_id)
testcase.key.delete()
return
is_lsan_enabled = environment.get_value('LSAN')
if is_lsan_enabled:
# Creates empty local blacklist so all leaks will be visible to uploader.
leak_blacklist.create_empty_local_blacklist()
# Store the bot name and timestamp in upload metadata.
bot_name = environment.get_value('BOT_NAME')
metadata.bot_name = bot_name
metadata.timestamp = datetime.datetime.utcnow()
metadata.put()
# Adjust the test timeout, if user has provided one.
if metadata.timeout:
environment.set_value('TEST_TIMEOUT', metadata.timeout)
# Adjust the number of retries, if user has provided one.
if metadata.retries is not None:
environment.set_value('CRASH_RETRIES', metadata.retries)
# Setup testcase and get absolute testcase path.
file_list, _, testcase_file_path = setup.setup_testcase(testcase)
if not file_list:
return
# Set up a custom or regular build based on revision.
build_manager.setup_build(testcase.crash_revision)
# Check if we have an application path. If not, our build failed
# to setup correctly.
app_path = environment.get_value('APP_PATH')
if not app_path:
data_handler.update_testcase_comment(testcase, data_types.TaskState.ERROR,
'Build setup failed')
if data_handler.is_first_retry_for_task(testcase):
build_fail_wait = environment.get_value('FAIL_WAIT')
tasks.add_task(
'analyze', testcase_id, job_type, wait_time=build_fail_wait)
else:
close_invalid_testcase_and_update_status(testcase, metadata,
'Build setup failed')
return
# Update initial testcase information.
testcase.absolute_path = testcase_file_path
testcase.job_type = job_type
testcase.binary_flag = utils.is_binary_file(testcase_file_path)
testcase.queue = tasks.default_queue()
testcase.crash_state = ''
# Set initial testcase metadata fields (e.g. build url, etc).
data_handler.set_initial_testcase_metadata(testcase)
# Update minimized arguments and use ones provided during user upload.
if not testcase.minimized_arguments:
minimized_arguments = environment.get_value('APP_ARGS') or ''
additional_command_line_flags = testcase.get_metadata(
'uploaded_additional_args')
if additional_command_line_flags:
minimized_arguments += ' %s' % additional_command_line_flags
environment.set_value('APP_ARGS', minimized_arguments)
testcase.minimized_arguments = minimized_arguments
# Update other fields not set at upload time.
testcase.crash_revision = environment.get_value('APP_REVISION')
data_handler.set_initial_testcase_metadata(testcase)
testcase.put()
# Initialize some variables.
gestures = testcase.gestures
http_flag = testcase.http_flag
test_timeout = environment.get_value('TEST_TIMEOUT')
# Get the crash output.
result = testcase_manager.test_for_crash_with_retries(
testcase,
testcase_file_path,
test_timeout,
http_flag=http_flag,
compare_crash=False)
# If we don't get a crash, try enabling http to see if we can get a crash.
# Skip engine fuzzer jobs (e.g. libFuzzer, AFL) for which http testcase paths
# are not applicable.
if (not result.is_crash() and not http_flag and
not environment.is_engine_fuzzer_job()):
result_with_http = testcase_manager.test_for_crash_with_retries(
testcase,
testcase_file_path,
test_timeout,
http_flag=True,
compare_crash=False)
if result_with_http.is_crash():
logs.log('Testcase needs http flag for crash.')
http_flag = True
result = result_with_http
# Refresh our object.
testcase = data_handler.get_testcase_by_id(testcase_id)
if not testcase:
return
# Set application command line with the correct http flag.
application_command_line = (
testcase_manager.get_command_line_for_application(
testcase_file_path, needs_http=http_flag))
# Get the crash data.
crashed = result.is_crash()
crash_time = result.get_crash_time()
state = result.get_symbolized_data()
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
# Get crash info object with minidump info. Also, re-generate unsymbolized
# stacktrace if needed.
crash_info, _ = (
crash_uploader.get_crash_info_and_stacktrace(
application_command_line, state.crash_stacktrace, gestures))
if crash_info:
testcase.minidump_keys = crash_info.store_minidump()
if not crashed:
# Could not reproduce the crash.
log_message = (
'Testcase didn\'t crash in %d seconds (with retries)' % test_timeout)
data_handler.update_testcase_comment(
testcase, data_types.TaskState.FINISHED, log_message)
# For an unreproducible testcase, retry once on another bot to confirm
# our results and in case this bot is in a bad state which we didn't catch
# through our usual means.
if data_handler.is_first_retry_for_task(testcase):
testcase.status = 'Unreproducible, retrying'
testcase.put()
tasks.add_task('analyze', testcase_id, job_type)
return
# In the general case, we will not attempt to symbolize if we do not detect
# a crash. For user uploads, we should symbolize anyway to provide more
# information about what might be happening.
crash_stacktrace_output = utils.get_crash_stacktrace_output(
application_command_line, state.crash_stacktrace,
unsymbolized_crash_stacktrace)
testcase.crash_stacktrace = data_handler.filter_stacktrace(
crash_stacktrace_output)
close_invalid_testcase_and_update_status(testcase, metadata,
'Unreproducible')
# A non-reproducing testcase might still impact production branches.
# Add the impact task to get that information.
task_creation.create_impact_task_if_needed(testcase)
return
# Update http flag and re-run testcase to store dependencies (for bundled
# archives only).
testcase.http_flag = http_flag
if not store_testcase_dependencies_from_bundled_testcase_archive(
metadata, testcase, testcase_file_path):
return
# Update testcase crash parameters.
testcase.crash_type = state.crash_type
testcase.crash_address = state.crash_address
testcase.crash_state = state.crash_state
# Try to guess if the bug is security or not.
security_flag = crash_analyzer.is_security_issue(
state.crash_stacktrace, state.crash_type, state.crash_address)
testcase.security_flag = security_flag
# If it is, guess the severity.
if security_flag:
testcase.security_severity = severity_analyzer.get_security_severity(
state.crash_type, state.crash_stacktrace, job_type, bool(gestures))
log_message = ('Testcase crashed in %d seconds (r%d)' %
(crash_time, testcase.crash_revision))
data_handler.update_testcase_comment(testcase, data_types.TaskState.FINISHED,
log_message)
# See if we have to ignore this crash.
if crash_analyzer.ignore_stacktrace(state.crash_stacktrace):
close_invalid_testcase_and_update_status(testcase, metadata, 'Irrelavant')
return
# Test for reproducibility.
one_time_crasher_flag = not testcase_manager.test_for_reproducibility(
testcase_file_path, state.crash_state, security_flag, test_timeout,
http_flag, gestures)
testcase.one_time_crasher_flag = one_time_crasher_flag
# Check to see if this is a duplicate.
project_name = data_handler.get_project_name(job_type)
existing_testcase = data_handler.find_testcase(
project_name, state.crash_type, state.crash_state, security_flag)
if existing_testcase:
# If the existing test case is unreproducible and we are, replace the
# existing test case with this one.
if existing_testcase.one_time_crasher_flag and not one_time_crasher_flag:
duplicate_testcase = existing_testcase
original_testcase = testcase
else:
duplicate_testcase = testcase
original_testcase = existing_testcase
metadata.status = 'Duplicate'
metadata.duplicate_of = existing_testcase.key.id()
duplicate_testcase.status = 'Duplicate'
duplicate_testcase.duplicate_of = original_testcase.key.id()
duplicate_testcase.put()
# Set testcase and metadata status if not set already.
if testcase.status != 'Duplicate':
testcase.status = 'Processed'
metadata.status = 'Confirmed'
# Add new leaks to global blacklist to avoid detecting duplicates.
# Only add if testcase has a direct leak crash and if it's reproducible.
if is_lsan_enabled:
leak_blacklist.add_crash_to_global_blacklist_if_needed(testcase)
# Add application specific information in the trace.
crash_stacktrace_output = utils.get_crash_stacktrace_output(
application_command_line, state.crash_stacktrace,
unsymbolized_crash_stacktrace)
testcase.crash_stacktrace = data_handler.filter_stacktrace(
crash_stacktrace_output)
# Update the testcase values.
testcase.put()
# Update the upload metadata.
metadata.security_flag = security_flag
metadata.put()
# Create tasks to
# 1. Minimize testcase (minimize).
# 2. Find regression range (regression).
# 3. Find testcase impact on production branches (impact).
# 4. Check whether testcase is fixed (progression).
# 5. Get second stacktrace from another job in case of
# one-time crashers (stack).
task_creation.create_tasks(testcase)
def find_fixed_range(testcase_id, job_type):
"""Attempt to find the revision range where a testcase was fixed."""
deadline = tasks.get_task_completion_deadline()
testcase = data_handler.get_testcase_by_id(testcase_id)
if not testcase:
return
if testcase.fixed:
logs.log_error('Fixed range is already set as %s, skip.' %
testcase.fixed)
return
# Setup testcase and its dependencies.
file_list, _, testcase_file_path = setup.setup_testcase(testcase)
if not file_list:
return
# Set a flag to indicate we are running progression task. This shows pending
# status on testcase report page and avoid conflicting testcase updates by
# triage cron.
testcase.set_metadata('progression_pending', True)
# Custom binaries are handled as special cases.
if build_manager.is_custom_binary():
_check_fixed_for_custom_binary(testcase, job_type, testcase_file_path)
return
release_build_bucket_path = environment.get_value(
'RELEASE_BUILD_BUCKET_PATH')
revision_list = build_manager.get_revisions_list(release_build_bucket_path,
testcase=testcase)
if not revision_list:
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.ERROR,
'Failed to fetch revision list')
tasks.add_task('progression', testcase_id, job_type)
return
# Use min, max_index to mark the start and end of revision list that is used
# for bisecting the progression range. Set start to the revision where noticed
# the crash. Set end to the trunk revision. Also, use min, max from past run
# if it timed out.
min_revision = testcase.get_metadata('last_progression_min')
max_revision = testcase.get_metadata('last_progression_max')
last_tested_revision = testcase.get_metadata('last_tested_crash_revision')
known_crash_revision = last_tested_revision or testcase.crash_revision
if not min_revision:
min_revision = known_crash_revision
if not max_revision:
max_revision = revisions.get_last_revision_in_list(revision_list)
min_index = revisions.find_min_revision_index(revision_list, min_revision)
if min_index is None:
raise errors.BuildNotFoundError(min_revision, job_type)
max_index = revisions.find_max_revision_index(revision_list, max_revision)
if max_index is None:
raise errors.BuildNotFoundError(max_revision, job_type)
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.STARTED,
'r%d' % max_revision)
# Check to see if this testcase is still crashing now. If it is, then just
# bail out.
result = _testcase_reproduces_in_revision(testcase, testcase_file_path,
job_type, max_revision)
if result.is_crash():
logs.log('Found crash with same signature on latest revision r%d.' %
max_revision)
app_path = environment.get_value('APP_PATH')
command = testcase_manager.get_command_line_for_application(
testcase_file_path,
app_path=app_path,
needs_http=testcase.http_flag)
symbolized_crash_stacktrace = result.get_stacktrace(symbolized=True)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
stacktrace = utils.get_crash_stacktrace_output(
command, symbolized_crash_stacktrace,
unsymbolized_crash_stacktrace)
testcase = data_handler.get_testcase_by_id(testcase_id)
testcase.last_tested_crash_stacktrace = data_handler.filter_stacktrace(
stacktrace)
_update_completion_metadata(
testcase,
max_revision,
is_crash=True,
message='still crashes on latest revision r%s' % max_revision)
# Since we've verified that the test case is still crashing, clear out any
# metadata indicating potential flake from previous runs.
task_creation.mark_unreproducible_if_flaky(testcase, False)
# For chromium project, save latest crash information for later upload
# to chromecrash/.
state = result.get_symbolized_data()
crash_uploader.save_crash_info_if_needed(testcase_id, max_revision,
job_type, state.crash_type,
state.crash_address,
state.frames)
return
# Don't burden NFS server with caching these random builds.
environment.set_value('CACHE_STORE', False)
# Verify that we do crash in the min revision. This is assumed to be true
# while we are doing the bisect.
result = _testcase_reproduces_in_revision(testcase, testcase_file_path,
job_type, min_revision)
if result and not result.is_crash():
testcase = data_handler.get_testcase_by_id(testcase_id)
# Retry once on another bot to confirm our result.
if data_handler.is_first_retry_for_task(testcase,
reset_after_retry=True):
tasks.add_task('progression', testcase_id, job_type)
error_message = (
'Known crash revision %d did not crash, will retry on another bot to '
'confirm result' % known_crash_revision)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.ERROR,
error_message)
_update_completion_metadata(testcase, max_revision)
return
_clear_progression_pending(testcase)
error_message = ('Known crash revision %d did not crash' %
known_crash_revision)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.ERROR,
error_message)
task_creation.mark_unreproducible_if_flaky(testcase, True)
return
# Start a binary search to find last non-crashing revision. At this point, we
# know that we do crash in the min_revision, and do not crash in max_revision.
while time.time() < deadline:
min_revision = revision_list[min_index]
max_revision = revision_list[max_index]
# If the min and max revisions are one apart this is as much as we can
# narrow the range.
if max_index - min_index == 1:
_save_fixed_range(testcase_id, min_revision, max_revision)
return
# Test the middle revision of our range.
middle_index = (min_index + max_index) // 2
middle_revision = revision_list[middle_index]
testcase = data_handler.get_testcase_by_id(testcase_id)
log_message = 'Testing r%d (current range %d:%d)' % (
middle_revision, min_revision, max_revision)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.WIP,
log_message)
try:
result = _testcase_reproduces_in_revision(testcase,
testcase_file_path,
job_type,
middle_revision)
except errors.BadBuildError:
# Skip this revision.
del revision_list[middle_index]
max_index -= 1
continue
if result.is_crash():
min_index = middle_index
else:
max_index = middle_index
_save_current_fixed_range_indices(testcase_id,
revision_list[min_index],
revision_list[max_index])
# If we've broken out of the loop, we've exceeded the deadline. Recreate the
# task to pick up where we left off.
testcase = data_handler.get_testcase_by_id(testcase_id)
error_message = ('Timed out, current range r%d:r%d' %
(revision_list[min_index], revision_list[max_index]))
data_handler.update_testcase_comment(testcase, data_types.TaskState.ERROR,
error_message)
tasks.add_task('progression', testcase_id, job_type)
def execute_task(testcase_id, job_type):
"""Run a test case with a different job type to see if they reproduce."""
testcase = data_handler.get_testcase_by_id(testcase_id)
if not testcase:
return
if environment.is_engine_fuzzer_job(
testcase.job_type) != environment.is_engine_fuzzer_job(job_type):
# We should never reach here. But in case we do, we should bail out as
# otherwise we will run into exceptions.
return
# Use a cloned testcase entity with different fuzz target paramaters for
# a different fuzzing engine.
original_job_type = testcase.job_type
testcase = _get_variant_testcase_for_job(testcase, job_type)
# Setup testcase and its dependencies.
file_list, _, testcase_file_path = setup.setup_testcase(testcase, job_type)
if not file_list:
return
# Set up a custom or regular build. We explicitly omit the crash revision
# since we want to test against the latest build here.
build_manager.setup_build()
# Check if we have an application path. If not, our build failed to setup
# correctly.
if not build_manager.check_app_path():
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(
testcase,
data_types.TaskState.ERROR,
"Build setup failed with job: " + job_type,
)
return
# Disable gestures if we're running on a different platform from that of
# the original test case.
use_gestures = testcase.platform == environment.platform().lower()
# Reproduce the crash.
app_path = environment.get_value("APP_PATH")
command = testcase_manager.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
test_timeout = environment.get_value("TEST_TIMEOUT", 10)
revision = environment.get_value("APP_REVISION")
result = testcase_manager.test_for_crash_with_retries(
testcase,
testcase_file_path,
test_timeout,
http_flag=testcase.http_flag,
use_gestures=use_gestures,
compare_crash=False,
)
if result.is_crash() and not result.should_ignore():
crash_state = result.get_state()
crash_type = result.get_type()
security_flag = result.is_security_issue()
gestures = testcase.gestures if use_gestures else None
one_time_crasher_flag = not testcase_manager.test_for_reproducibility(
testcase.fuzzer_name,
testcase.actual_fuzzer_name(),
testcase_file_path,
crash_state,
security_flag,
test_timeout,
testcase.http_flag,
gestures,
)
if one_time_crasher_flag:
status = data_types.TestcaseVariantStatus.FLAKY
else:
status = data_types.TestcaseVariantStatus.REPRODUCIBLE
crash_comparer = CrashComparer(crash_state, testcase.crash_state)
is_similar = (crash_comparer.is_similar()
and security_flag == testcase.security_flag)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
symbolized_crash_stacktrace = result.get_stacktrace(symbolized=True)
crash_stacktrace_output = utils.get_crash_stacktrace_output(
command, symbolized_crash_stacktrace,
unsymbolized_crash_stacktrace)
else:
status = data_types.TestcaseVariantStatus.UNREPRODUCIBLE
is_similar = False
crash_type = None
crash_state = None
security_flag = False
crash_stacktrace_output = "No crash occurred."
if original_job_type == job_type:
# This case happens when someone clicks 'Update last tested stacktrace using
# trunk build' button.
testcase = data_handler.get_testcase_by_id(testcase_id)
testcase.last_tested_crash_stacktrace = data_handler.filter_stacktrace(
crash_stacktrace_output)
testcase.set_metadata("last_tested_crash_revision",
revision,
update_testcase=True)
else:
# Regular case of variant analysis.
variant = data_handler.get_testcase_variant(testcase_id, job_type)
variant.status = status
variant.revision = revision
variant.crash_type = crash_type
variant.crash_state = crash_state
variant.security_flag = security_flag
variant.is_similar = is_similar
# Explicitly skipping crash stacktrace for now as it make entities larger
# and we plan to use only crash paramaters in UI.
variant.put()
def check_for_bad_build(job_type, crash_revision):
"""Return true if the build is bad, i.e. crashes on startup."""
# Check the bad build check flag to see if we want do this.
if not environment.get_value('BAD_BUILD_CHECK'):
return False
# Create a blank command line with no file to run and no http.
command = get_command_line_for_application(file_to_run='',
needs_http=False)
# When checking for bad builds, we use the default window size.
# We don't want to pick a custom size since it can potentially cause a
# startup crash and cause a build to be detected incorrectly as bad.
default_window_argument = environment.get_value('WINDOW_ARG', '')
if default_window_argument:
command = command.replace(' %s' % default_window_argument, '')
# TSAN is slow, and boots slow on first startup. Increase the warmup
# timeout for this case.
if environment.tool_matches('TSAN', job_type):
fast_warmup_timeout = environment.get_value('WARMUP_TIMEOUT')
else:
fast_warmup_timeout = environment.get_value('FAST_WARMUP_TIMEOUT')
# Initialize helper variables.
is_bad_build = False
build_run_console_output = ''
app_directory = environment.get_value('APP_DIR')
# Exit all running instances.
process_handler.terminate_stale_application_instances()
# Check if the build is bad.
return_code, crash_time, output = process_handler.run_process(
command,
timeout=fast_warmup_timeout,
current_working_directory=app_directory)
crash_result = CrashResult(return_code, crash_time, output)
# 1. Need to account for startup crashes with no crash state. E.g. failed to
# load shared library. So, ignore state for comparison.
# 2. Ignore leaks as they don't block a build from reporting regular crashes
# and also don't impact regression range calculations.
if (crash_result.is_crash(ignore_state=True)
and not crash_result.should_ignore() and
not crash_result.get_type() in ['Direct-leak', 'Indirect-leak']):
is_bad_build = True
build_run_console_output = utils.get_crash_stacktrace_output(
command, crash_result.get_stacktrace(symbolized=True),
crash_result.get_stacktrace(symbolized=False))
logs.log('Bad build for %s detected at r%d.' %
(job_type, crash_revision),
output=build_run_console_output)
# Exit all running instances.
process_handler.terminate_stale_application_instances()
# Any of the conditions below indicate that bot is in a bad state and it is
# not caused by the build itself. In that case, just exit.
build_state = data_handler.get_build_state(job_type, crash_revision)
if is_bad_build and utils.sub_string_exists_in(BAD_STATE_HINTS, output):
logs.log_fatal_and_exit(
'Bad bot environment detected, exiting.',
output=build_run_console_output,
snapshot=process_handler.get_runtime_snapshot())
# If none of the other bots have added information about this build,
# then add it now.
if (build_state == data_types.BuildState.UNMARKED
and not crash_result.should_ignore()):
data_handler.add_build_metadata(job_type, crash_revision, is_bad_build,
build_run_console_output)
return is_bad_build
def get_symbolized_stacktraces(testcase_file_path, testcase,
old_crash_stacktrace, expected_state):
"""Use the symbolized builds to generate an updated stacktrace."""
# Initialize variables.
app_path = environment.get_value("APP_PATH")
app_path_debug = environment.get_value("APP_PATH_DEBUG")
long_test_timeout = environment.get_value("WARMUP_TIMEOUT")
retry_limit = environment.get_value("FAIL_RETRIES")
symbolized = False
debug_build_stacktrace = ""
release_build_stacktrace = old_crash_stacktrace
# Symbolize using the debug build first so that the debug build stacktrace
# comes after the more important release build stacktrace.
if app_path_debug:
for _ in range(retry_limit):
process_handler.terminate_stale_application_instances()
command = testcase_manager.get_command_line_for_application(
testcase_file_path,
app_path=app_path_debug,
needs_http=testcase.http_flag,
)
return_code, crash_time, output = process_handler.run_process(
command, timeout=long_test_timeout, gestures=testcase.gestures)
crash_result = CrashResult(return_code, crash_time, output)
if crash_result.is_crash():
state = crash_result.get_symbolized_data()
if crash_analyzer.ignore_stacktrace(state.crash_stacktrace):
continue
unsymbolized_crash_stacktrace = crash_result.get_stacktrace(
symbolized=False)
debug_build_stacktrace = utils.get_crash_stacktrace_output(
command,
state.crash_stacktrace,
unsymbolized_crash_stacktrace,
build_type="debug",
)
symbolized = True
break
# Symbolize using the release build.
if app_path:
for _ in range(retry_limit):
process_handler.terminate_stale_application_instances()
command = testcase_manager.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
return_code, crash_time, output = process_handler.run_process(
command, timeout=long_test_timeout, gestures=testcase.gestures)
crash_result = CrashResult(return_code, crash_time, output)
if crash_result.is_crash():
state = crash_result.get_symbolized_data()
if crash_analyzer.ignore_stacktrace(state.crash_stacktrace):
continue
if state.crash_state != expected_state:
continue
# Release stack's security flag has to match the symbolized release
# stack's security flag.
security_flag = crash_result.is_security_issue()
if security_flag != testcase.security_flag:
continue
unsymbolized_crash_stacktrace = crash_result.get_stacktrace(
symbolized=False)
release_build_stacktrace = utils.get_crash_stacktrace_output(
command,
state.crash_stacktrace,
unsymbolized_crash_stacktrace,
build_type="release",
)
symbolized = True
break
stacktrace = release_build_stacktrace
if debug_build_stacktrace:
stacktrace += "\n\n" + debug_build_stacktrace
return symbolized, stacktrace
def execute_task(testcase_id, job_type):
"""Run a test case with a different job type to see if they reproduce."""
testcase = data_handler.get_testcase_by_id(testcase_id)
if not testcase:
return
if (environment.is_engine_fuzzer_job(testcase.job_type) !=
environment.is_engine_fuzzer_job(job_type)):
# We should never reach here. But in case we do, we should bail out as
# otherwise we will run into exceptions.
return
# Setup testcase and its dependencies.
fuzzer_override = builtin_fuzzers.get_fuzzer_for_job(job_type)
file_list, _, testcase_file_path = setup.setup_testcase(
testcase, fuzzer_override=fuzzer_override)
if not file_list:
return
# Set up a custom or regular build. We explicitly omit the crash revision
# since we want to test against the latest build here.
build_manager.setup_build()
# Check if we have an application path. If not, our build failed to setup
# correctly.
app_path = environment.get_value('APP_PATH')
if not app_path:
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(
testcase, data_types.TaskState.ERROR,
'Build setup failed with job: ' + job_type)
return
# Reproduce the crash.
command = testcase_manager.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
test_timeout = environment.get_value('TEST_TIMEOUT', 10)
revision = environment.get_value('APP_REVISION')
result = testcase_manager.test_for_crash_with_retries(
testcase,
testcase_file_path,
test_timeout,
http_flag=testcase.http_flag,
compare_crash=False)
if result.is_crash() and not result.should_ignore():
crash_state = result.get_state()
crash_type = result.get_type()
security_flag = result.is_security_issue()
one_time_crasher_flag = not testcase_manager.test_for_reproducibility(
testcase_file_path, crash_state, security_flag, test_timeout,
testcase.http_flag, testcase.gestures)
if one_time_crasher_flag:
status = data_types.TestcaseVariantStatus.FLAKY
else:
status = data_types.TestcaseVariantStatus.REPRODUCIBLE
crash_comparer = CrashComparer(crash_state, testcase.crash_state)
is_similar = (crash_comparer.is_similar()
and security_flag == testcase.security_flag)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
symbolized_crash_stacktrace = result.get_stacktrace(symbolized=True)
crash_stacktrace_output = utils.get_crash_stacktrace_output(
command, symbolized_crash_stacktrace,
unsymbolized_crash_stacktrace)
else:
status = data_types.TestcaseVariantStatus.UNREPRODUCIBLE
is_similar = False
crash_type = None
crash_state = None
security_flag = False
crash_stacktrace_output = 'No crash occurred.'
testcase = data_handler.get_testcase_by_id(testcase_id)
if testcase.job_type == job_type:
# This case happens when someone clicks 'Update last tested stacktrace using
# trunk build' button.
testcase.last_tested_crash_stacktrace = (
data_handler.filter_stacktrace(crash_stacktrace_output))
testcase.set_metadata('last_tested_crash_revision',
revision,
update_testcase=False)
else:
# Regular case of variant analysis.
variant = data_handler.get_testcase_variant(testcase_id, job_type)
variant.status = status
variant.revision = revision
variant.crash_type = crash_type
variant.crash_state = crash_state
variant.security_flag = security_flag
variant.is_similar = is_similar
# Explicitly skipping crash stacktrace for now as it make entities larger
# and we plan to use only crash paramaters in UI.
variant.put()
def execute_task(testcase_id, job_type):
"""Run a test case with a second job type to generate a second stack trace."""
# Locate the testcase associated with the id.
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.STARTED)
# Setup testcase and its dependencies.
file_list, _, testcase_file_path = setup.setup_testcase(testcase)
if not file_list:
return
# Initialize timeout values.
test_timeout = environment.get_value('TEST_TIMEOUT', 10)
# Set up a custom or regular build. We explicitly omit the crash revision
# since we want to test against the latest build here.
build_manager.setup_build()
# Check if we have an application path. If not, our build failed to setup
# correctly.
app_path = environment.get_value('APP_PATH')
if not app_path:
testcase = data_handler.get_testcase_by_id(testcase_id)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.ERROR,
'Build setup failed')
return
# TSAN tool settings (if the tool is used).
if environment.tool_matches('TSAN', job_type):
environment.set_tsan_max_history_size()
command = tests.get_command_line_for_application(
testcase_file_path, app_path=app_path, needs_http=testcase.http_flag)
result = tests.test_for_crash_with_retries(testcase,
testcase_file_path,
test_timeout,
http_flag=testcase.http_flag,
compare_crash=False)
# Get revision information.
revision = environment.get_value('APP_REVISION')
# If a crash occurs, then we add the second stacktrace information.
if result.is_crash():
state = result.get_symbolized_data()
security_flag = result.is_security_issue()
one_time_crasher_flag = not tests.test_for_reproducibility(
testcase_file_path, state.crash_state, security_flag, test_timeout,
testcase.http_flag, testcase.gestures)
# Attach a header to indicate information on reproducibility flag.
if one_time_crasher_flag:
crash_stacktrace_header = 'Unreliable'
else:
crash_stacktrace_header = 'Fully reproducible'
crash_stacktrace_header += (' crash found using %s job.\n\n' %
job_type)
unsymbolized_crash_stacktrace = result.get_stacktrace(symbolized=False)
stacktrace = utils.get_crash_stacktrace_output(
command, state.crash_stacktrace, unsymbolized_crash_stacktrace)
crash_stacktrace = data_handler.filter_stacktrace(
'%s%s' % (crash_stacktrace_header, stacktrace))
else:
crash_stacktrace = 'No crash found using %s job.' % job_type
# Decide which stacktrace to update this stacktrace with.
testcase = data_handler.get_testcase_by_id(testcase_id)
if testcase.last_tested_crash_stacktrace == 'Pending':
# This case happens when someone clicks 'Update last tested stacktrace using
# trunk build' button.
testcase.last_tested_crash_stacktrace = crash_stacktrace
testcase.set_metadata('last_tested_crash_revision',
revision,
update_testcase=False)
else:
# Default case when someone defines |SECOND_STACK_JOB_TYPE| in the job
# type. This helps to test the unreproducible crash with a different memory
# debugging tool to get a second stacktrace (e.g. running TSAN on a flaky
# crash found in ASAN build).
testcase.second_crash_stacktrace = crash_stacktrace
testcase.set_metadata('second_crash_stacktrace_revision',
revision,
update_testcase=False)
data_handler.update_testcase_comment(testcase,
data_types.TaskState.FINISHED)
