def artifacts(self, tree, job, artifact_job_class, rev):
# Grab the second part of the repo name, which is generally how things
# are indexed. Eg: 'integration/mozilla-inbound' is indexed as
# 'mozilla-inbound'
tree = tree.split('/')[1] if '/' in tree else tree
# PGO builds are now known as "shippable" for all platforms but Android.
# For macOS and linux32 shippable builds are equivalent to opt builds and
# replace them on some trees. Additionally, we no longer produce win64
# opt builds on integration branches.
if not job.startswith('android-'):
if job.endswith('-pgo') or job in ('macosx64-opt', 'linux-opt',
'win64-opt'):
tree += '.shippable'
if job.endswith('-pgo'):
job = job.replace('-pgo', '-opt')
namespace = '{trust_domain}.v2.{tree}.revision.{rev}.{product}.{job}'.format(
trust_domain=artifact_job_class.trust_domain,
rev=rev,
tree=tree,
product=artifact_job_class.product,
job=job,
)
self.log(logging.INFO, 'artifact',
{'namespace': namespace},
'Searching Taskcluster index with namespace: {namespace}')
try:
taskId = find_task_id(namespace)
except KeyError:
# Not all revisions correspond to pushes that produce the job we
# care about; and even those that do may not have completed yet.
raise ValueError('Task for {namespace} does not exist (yet)!'.format(namespace=namespace))
return taskId, list_artifacts(taskId)
def get_failures(task_id):
"""Returns a dict containing properties containing a list of
directories containing test failures and a separate list of
individual test failures from the errorsummary.log artifact for
the task.
Calls the helper function munge_test_path to attempt to find an
appropriate test path to pass to the task in
MOZHARNESS_TEST_PATHS. If no appropriate test path can be
determined, nothing is returned.
"""
re_test = re.compile(r'"test": "([^"]+)"')
re_bad_test = re.compile(r'(Last test finished|'
r'Main app process exited normally|'
r'[(]SimpleTest/TestRunner.js[)]|'
r'remoteautomation.py|'
r'unknown test url|'
r'https?://localhost:\d+/\d+/\d+/.*[.]html)')
re_extract_tests = [
re.compile(r'(?:^[^:]+:)?(?:https?|file):[^ ]+/reftest/tests/([^ ]+)'),
re.compile(r'(?:^[^:]+:)?(?:https?|file):[^:]+:[0-9]+/tests/([^ ]+)'),
re.compile(r'xpcshell-[^ ]+\.ini:(.*)'),
]
def munge_test_path(test_path):
if re_bad_test.search(test_path):
return None
for r in re_extract_tests:
m = r.match(test_path)
if m:
test_path = m.group(1)
break
return test_path
dirs = set()
tests = set()
artifacts = list_artifacts(task_id)
for artifact in artifacts:
if 'name' in artifact and artifact['name'].endswith('errorsummary.log'):
stream = get_artifact(task_id, artifact['name'])
if stream:
# Read all of the content from the stream and split
# the lines out since on macosx and windows, the first
# line is empty.
for line in stream.read().split('\n'):
line = line.strip()
match = re_test.search(line)
if match:
test_path = munge_test_path(match.group(1))
if test_path:
tests.add(test_path)
test_dir = os.path.dirname(test_path)
if test_dir:
dirs.add(test_dir)
return {'dirs': sorted(dirs), 'tests': sorted(tests)}
def install_from_task(self, taskId, distdir):
artifacts = list_artifacts(taskId)
urls = []
for artifact_name in self._artifact_job.find_candidate_artifacts(artifacts):
# We can easily extract the task ID from the URL. We can't easily
# extract the build ID; we use the .ini files embedded in the
# downloaded artifact for this.
url = get_artifact_url(taskId, artifact_name)
urls.append(url)
if not urls:
raise ValueError('Task {taskId} existed, but no artifacts found!'.format(taskId=taskId))
for url in urls:
if self.install_from_url(url, distdir):
return 1
return 0
def artifact_urls(self, tree, job, rev, download_symbols):
try:
artifact_job = get_job_details(job,
log=self._log,
download_symbols=download_symbols)
except KeyError:
self.log(logging.INFO, 'artifact', {'job': job},
'Unknown job {job}')
raise KeyError("Unknown job")
# Grab the second part of the repo name, which is generally how things
# are indexed. Eg: 'integration/mozilla-inbound' is indexed as
# 'mozilla-inbound'
tree = tree.split('/')[1] if '/' in tree else tree
namespace = 'gecko.v2.{tree}.revision.{rev}.{product}.{job}'.format(
rev=rev,
tree=tree,
product=artifact_job.product,
job=job,
)
self.log(logging.DEBUG, 'artifact', {'namespace': namespace},
'Searching Taskcluster index with namespace: {namespace}')
try:
taskId = find_task_id(namespace)
except Exception:
# Not all revisions correspond to pushes that produce the job we
# care about; and even those that do may not have completed yet.
raise ValueError(
'Task for {namespace} does not exist (yet)!'.format(
namespace=namespace))
artifacts = list_artifacts(taskId)
urls = []
for artifact_name in artifact_job.find_candidate_artifacts(artifacts):
# We can easily extract the task ID from the URL. We can't easily
# extract the build ID; we use the .ini files embedded in the
# downloaded artifact for this. We could also use the uploaded
# public/build/buildprops.json for this purpose.
url = get_artifact_url(taskId, artifact_name)
urls.append(url)
if not urls:
raise ValueError(
'Task for {namespace} existed, but no artifacts found!'.format(
namespace=namespace))
return urls
def artifact_urls(self, tree, job, rev, download_symbols):
try:
artifact_job = get_job_details(job, log=self._log, download_symbols=download_symbols)
except KeyError:
self.log(logging.INFO, 'artifact',
{'job': job},
'Unknown job {job}')
raise KeyError("Unknown job")
# Grab the second part of the repo name, which is generally how things
# are indexed. Eg: 'integration/mozilla-inbound' is indexed as
# 'mozilla-inbound'
tree = tree.split('/')[1] if '/' in tree else tree
namespace = 'gecko.v2.{tree}.revision.{rev}.{product}.{job}'.format(
rev=rev,
tree=tree,
product=artifact_job.product,
job=job,
)
self.log(logging.DEBUG, 'artifact',
{'namespace': namespace},
'Searching Taskcluster index with namespace: {namespace}')
try:
taskId = find_task_id(namespace)
except Exception:
# Not all revisions correspond to pushes that produce the job we
# care about; and even those that do may not have completed yet.
raise ValueError('Task for {namespace} does not exist (yet)!'.format(namespace=namespace))
artifacts = list_artifacts(taskId)
urls = []
for artifact_name in artifact_job.find_candidate_artifacts(artifacts):
# We can easily extract the task ID from the URL. We can't easily
# extract the build ID; we use the .ini files embedded in the
# downloaded artifact for this. We could also use the uploaded
# public/build/buildprops.json for this purpose.
url = get_artifact_url(taskId, artifact_name)
urls.append(url)
if not urls:
raise ValueError('Task for {namespace} existed, but no artifacts found!'.format(namespace=namespace))
return urls
def get_failures(task_id):
"""Returns a dict containing properties containing a list of
directories containing test failures and a separate list of
individual test failures from the errorsummary.log artifact for
the task.
Calls the helper function munge_test_path to attempt to find an
appropriate test path to pass to the task in
MOZHARNESS_TEST_PATHS. If no appropriate test path can be
determined, nothing is returned.
"""
re_bad_tests = [
re.compile(r'Last test finished'),
re.compile(r'LeakSanitizer'),
re.compile(r'Main app process exited normally'),
re.compile(r'ShutdownLeaks'),
re.compile(r'[(]SimpleTest/TestRunner.js[)]'),
re.compile(r'automation.py'),
re.compile(r'https?://localhost:\d+/\d+/\d+/.*[.]html'),
re.compile(r'jsreftest'),
re.compile(r'leakcheck'),
re.compile(r'mozrunner-startup'),
re.compile(r'pid: '),
re.compile(r'remoteautomation.py'),
re.compile(r'unknown test url'),
]
re_extract_tests = [
re.compile(
r'"test": "(?:[^:]+:)?(?:https?|file):[^ ]+/reftest/tests/([^ "]+)'
),
re.compile(
r'"test": "(?:[^:]+:)?(?:https?|file):[^:]+:[0-9]+/tests/([^ "]+)'
),
re.compile(r'xpcshell-?[^ "]*\.ini:([^ "]+)'),
re.compile(r'/tests/([^ "]+) - finished .*'),
re.compile(r'"test": "([^ "]+)"'),
re.compile(
r'"message": "Error running command run_test with arguments '
'[(]<wptrunner[.]wpttest[.]TestharnessTest ([^>]+)>'),
re.compile(r'"message": "TEST-[^ ]+ [|] ([^ "]+)[^|]*[|]')
]
def munge_test_path(line):
test_path = None
for r in re_bad_tests:
if r.search(line):
return None
for r in re_extract_tests:
m = r.search(line)
if m:
test_path = m.group(1)
break
return test_path
dirs = set()
tests = set()
artifacts = list_artifacts(task_id)
for artifact in artifacts:
if 'name' not in artifact or not artifact['name'].endswith(
'errorsummary.log'):
continue
stream = get_artifact(task_id, artifact['name'])
if not stream:
continue
# The number of tasks created is determined by the
# `times` value and the number of distinct tests and
# directories as: times * (1 + len(tests) + len(dirs)).
# Since the maximum value of `times` specifiable in the
# Treeherder UI is 100, the number of tasks created can
# reach a very large value depending on the number of
# unique tests. During testing, it was found that 10
# distinct tests were sufficient to cause the action task
# to exceed the maxRunTime of 1800 seconds resulting in it
# being aborted. We limit the number of distinct tests
# and thereby the number of distinct test directories to a
# maximum of 5 to keep the action task from timing out.
for line in stream.read().split('\n'):
test_path = munge_test_path(line.strip())
if test_path:
tests.add(test_path)
test_dir = os.path.dirname(test_path)
if test_dir:
dirs.add(test_dir)
if len(tests) > 4:
break
return {'dirs': sorted(dirs), 'tests': sorted(tests)}
def get_failures(task_id):
"""Returns a dict containing properties containing a list of
directories containing test failures and a separate list of
individual test failures from the errorsummary.log artifact for
the task.
Calls the helper function munge_test_path to attempt to find an
appropriate test path to pass to the task in
MOZHARNESS_TEST_PATHS. If no appropriate test path can be
determined, nothing is returned.
"""
re_test = re.compile(r'"test": "([^"]+)"')
re_bad_test = re.compile(r'(Last test finished|'
r'Main app process exited normally|'
r'[(]SimpleTest/TestRunner.js[)]|'
r'remoteautomation.py|'
r'unknown test url|'
r'https?://localhost:\d+/\d+/\d+/.*[.]html)')
re_extract_tests = [
re.compile(r'(?:^[^:]+:)?(?:https?|file):[^ ]+/reftest/tests/([^ ]+)'),
re.compile(r'(?:^[^:]+:)?(?:https?|file):[^:]+:[0-9]+/tests/([^ ]+)'),
re.compile(r'xpcshell-[^ ]+\.ini:(.*)'),
]
def munge_test_path(test_path):
if re_bad_test.search(test_path):
return None
for r in re_extract_tests:
m = r.match(test_path)
if m:
test_path = m.group(1)
break
return test_path
dirs = set()
tests = set()
artifacts = list_artifacts(task_id)
for artifact in artifacts:
if 'name' in artifact and artifact['name'].endswith(
'errorsummary.log'):
stream = get_artifact(task_id, artifact['name'])
if stream:
# Read all of the content from the stream and split
# the lines out since on macosx and windows, the first
# line is empty.
for line in stream.read().split('\n'):
if len(tests) > 4:
# The number of tasks created is determined by
# the `times` value and the number of distinct
# tests and directories as:
# times * (1 + len(tests) + len(dirs)).
# Since the maximum value of `times`
# specifiable in the Treeherder UI is 100, the
# number of tasks created can reach a very
# large value depending on the number of
# unique tests. During testing, it was found
# that 10 distinct tests were sufficient to
# cause the action task to exceed the
# maxRunTime of 1800 seconds resulting in it
# being aborted. We limit the number of
# distinct tests and thereby the number of
# distinct test directories to a maximum of 5
# to keep the action task from timing out.
break
line = line.strip()
match = re_test.search(line)
if match:
test_path = munge_test_path(match.group(1))
if test_path:
tests.add(test_path)
test_dir = os.path.dirname(test_path)
if test_dir:
dirs.add(test_dir)
break
return {'dirs': sorted(dirs), 'tests': sorted(tests)}
def get_failures(task_id):
"""Returns a dict containing properties containing a list of
directories containing test failures and a separate list of
individual test failures from the errorsummary.log artifact for
the task.
Calls the helper function munge_test_path to attempt to find an
appropriate test path to pass to the task in
MOZHARNESS_TEST_PATHS. If no appropriate test path can be
determined, nothing is returned.
"""
def re_compile_list(*lst):
# Ideally we'd just use rb"" literals and avoid the encode, but
# this file needs to be importable in python2 for now.
return [re.compile(s.encode("utf-8")) for s in lst]
re_bad_tests = re_compile_list(
r"Last test finished",
r"LeakSanitizer",
r"Main app process exited normally",
r"ShutdownLeaks",
r"[(]SimpleTest/TestRunner.js[)]",
r"automation.py",
r"https?://localhost:\d+/\d+/\d+/.*[.]html",
r"jsreftest",
r"leakcheck",
r"mozrunner-startup",
r"pid: ",
r"RemoteProcessMonitor",
r"unknown test url",
)
re_extract_tests = re_compile_list(
r'"test": "(?:[^:]+:)?(?:https?|file):[^ ]+/reftest/tests/([^ "]+)',
r'"test": "(?:[^:]+:)?(?:https?|file):[^:]+:[0-9]+/tests/([^ "]+)',
r'xpcshell-?[^ "]*\.ini:([^ "]+)',
r'/tests/([^ "]+) - finished .*',
r'"test": "([^ "]+)"',
r'"message": "Error running command run_test with arguments '
r"[(]<wptrunner[.]wpttest[.]TestharnessTest ([^>]+)>",
r'"message": "TEST-[^ ]+ [|] ([^ "]+)[^|]*[|]',
)
def munge_test_path(line):
test_path = None
for r in re_bad_tests:
if r.search(line):
return None
for r in re_extract_tests:
m = r.search(line)
if m:
test_path = m.group(1)
break
return test_path
dirs = set()
tests = set()
artifacts = list_artifacts(task_id)
for artifact in artifacts:
if "name" not in artifact or not artifact["name"].endswith("errorsummary.log"):
continue
stream = get_artifact(task_id, artifact["name"])
if not stream:
continue
# The number of tasks created is determined by the
# `times` value and the number of distinct tests and
# directories as: times * (1 + len(tests) + len(dirs)).
# Since the maximum value of `times` specifiable in the
# Treeherder UI is 100, the number of tasks created can
# reach a very large value depending on the number of
# unique tests. During testing, it was found that 10
# distinct tests were sufficient to cause the action task
# to exceed the maxRunTime of 1800 seconds resulting in it
# being aborted. We limit the number of distinct tests
# and thereby the number of distinct test directories to a
# maximum of 5 to keep the action task from timing out.
# We handle the stream as raw bytes because it may contain invalid
# UTF-8 characters in portions other than those containing the error
# messages we're looking for.
for line in stream.read().split(b"\n"):
test_path = munge_test_path(line.strip())
if test_path:
tests.add(test_path.decode("utf-8"))
test_dir = os.path.dirname(test_path)
if test_dir:
dirs.add(test_dir.decode("utf-8"))
if len(tests) > 4:
break
return {"dirs": sorted(dirs), "tests": sorted(tests)}
