def test_set_bug_duplicate(failure_lines, classified_failures, test_matcher):
classified_failures[0].bug_number = 1234
classified_failures[0].save()
match = failure_lines[0].matches.all()[0]
match.score = 0.7
match.save()
# Add a FailureMatch that will have the same (failure_line_id, classified_failure_id)
# as another FailureMatch when classified_failure[1] is replaced by classified_failure[0]
duplicate_match = FailureMatch(failure_line=failure_lines[0],
classified_failure=classified_failures[1],
matcher=test_matcher.db_object,
score=0.8)
duplicate_match.save()
assert len(failure_lines[0].matches.all()) == 2
rv = classified_failures[1].set_bug(1234)
assert rv == classified_failures[0]
assert rv.bug_number == 1234
for item in failure_lines:
item.refresh_from_db()
match.refresh_from_db()
# Check that we updated the best classification that previously pointed
# to the now-defunct classified_failures[0]
assert failure_lines[1].best_classification == classified_failures[0]
# Check that we only have one match for the first failure line
matches = failure_lines[0].matches.all()
assert len(matches) == 1
# Check we picked the better of the two scores for the new match.
assert matches[0].score == Decimal("0.8")
# Ensure we deleted the ClassifiedFailure on which we tried to set the bug
assert len(
ClassifiedFailure.objects.filter(id=classified_failures[1].id)) == 0
def classified_failures(request, jm, eleven_jobs_stored, initial_data, failure_lines):
from treeherder.model.models import ClassifiedFailure, FailureMatch, Matcher
from treeherder.autoclassify import detectors
job_1 = jm.get_job(1)[0]
class TreeherderUnitTestDetector(detectors.Detector):
def __call__(self, failure_lines):
pass
test_matcher = Matcher.objects.register_detector(TreeherderUnitTestDetector)
def finalize():
Matcher._detector_funcs = {}
Matcher._matcher_funcs = {}
request.addfinalizer(finalize)
classified_failures = []
for failure_line in failure_lines:
if failure_line.job_guid == job_1["job_guid"]:
classified_failure = ClassifiedFailure()
classified_failure.save()
match = FailureMatch(failure_line=failure_line,
classified_failure=classified_failure,
matcher=test_matcher.db_object,
score=1.0,
is_best=True)
match.save()
classified_failures.append(classified_failure)
return classified_failures
def test_set_bug_duplicate(failure_lines, classified_failures, test_matcher):
classified_failures[0].bug_number = 1234
classified_failures[0].save()
match = failure_lines[0].matches.all()[0]
match.score = 0.7
match.save()
# Add a FailureMatch that will have the same (failure_line_id, classified_failure_id)
# as another FailureMatch when classified_failure[1] is replaced by classified_failure[0]
duplicate_match = FailureMatch(
failure_line=failure_lines[0],
classified_failure=classified_failures[1],
matcher=test_matcher.db_object,
score=0.8)
duplicate_match.save()
assert len(failure_lines[0].matches.all()) == 2
rv = classified_failures[1].set_bug(1234)
assert rv == classified_failures[0]
assert rv.bug_number == 1234
for item in failure_lines:
item.refresh_from_db()
match.refresh_from_db()
# Check that we updated the best classification that previously pointed
# to the now-defunct classified_failures[0]
assert failure_lines[1].best_classification == classified_failures[0]
# Check that we only have one match for the first failure line
matches = failure_lines[0].matches.all()
assert len(matches) == 1
# Check we picked the better of the two scores for the new match.
assert matches[0].score == Decimal("0.8")
# Ensure we deleted the ClassifiedFailure on which we tried to set the bug
assert len(ClassifiedFailure.objects.filter(id=classified_failures[1].id)) == 0
def classified_failures(request, jm, eleven_jobs_stored, failure_lines):
from treeherder.model.models import ClassifiedFailure, FailureMatch, Matcher
from treeherder.autoclassify import detectors
job_1 = jm.get_job(1)[0]
class TreeherderUnitTestDetector(detectors.Detector):
def __call__(self, failure_lines):
pass
test_matcher = Matcher.objects.register_detector(
TreeherderUnitTestDetector)
def finalize():
Matcher._detector_funcs = {}
Matcher._matcher_funcs = {}
request.addfinalizer(finalize)
classified_failures = []
for failure_line in failure_lines:
if failure_line.job_guid == job_1["job_guid"]:
classified_failure = ClassifiedFailure()
classified_failure.save()
match = FailureMatch(failure_line=failure_line,
classified_failure=classified_failure,
matcher=test_matcher.db_object,
score=1.0,
is_best=True)
match.save()
classified_failures.append(classified_failure)
return classified_failures
def match_errors(repository, job_guid):
unmatched_failures = FailureLine.objects.unmatched_for_job(
repository, job_guid)
if not unmatched_failures:
return
all_matched = set()
for matcher in Matcher.objects.registered_matchers():
matches = matcher(unmatched_failures)
for match in matches:
match.failure_line.matches.add(
FailureMatch(score=match.score,
matcher=matcher.db_object,
classified_failure=match.classified_failure))
match.failure_line.save()
logger.info("Matched failure %i with intermittent %i" %
(match.failure_line.id, match.classified_failure.id))
all_matched.add(match.failure_line)
if all_lines_matched(unmatched_failures):
break
for failure_line in all_matched:
# TODO: store all matches
best_match = failure_line.best_match(AUTOCLASSIFY_CUTOFF_RATIO)
if best_match:
best_match.is_best = True
best_match.save()
def _update(self, data, email, many=True):
by_project = defaultdict(list)
ids = []
failure_line_ids = set()
classification_ids = set()
for item in data:
line_id = int(item.get("id"))
if line_id is None:
return "No failure line id provided", 400
failure_line_ids.add(line_id)
if "best_classification" not in item:
return "No classification id provided", 400
classification_id = item.get("best_classification")
if classification_id is not None:
classification_ids.add(classification_id)
ids.append((line_id, classification_id))
failure_lines = as_dict(
FailureLine.objects.prefetch_related('classified_failures').filter(
id__in=failure_line_ids), "id")
if len(failure_lines) != len(failure_line_ids):
missing = failure_line_ids - set(failure_lines.keys())
return "No failure line with id: {0}".format(
", ".join(missing)), 404
classifications = as_dict(
ClassifiedFailure.objects.filter(id__in=classification_ids), "id")
if len(classifications) != len(classification_ids):
missing = classification_ids - set(classifications.keys())
return "No classification with id: {0}".format(
", ".join(missing)), 404
for line_id, classification_id in ids:
failure_line = failure_lines[line_id]
if classification_id is not None:
classification = classifications[classification_id]
else:
classification = None
by_project[failure_line.repository.name].append(
failure_line.job_guid)
failure_line.best_classification = classification
failure_line.best_is_verified = True
failure_line.save()
if (classification is not None and classification
not in failure_line.classified_failures.all()):
manual_detector = Matcher.objects.get(name="ManualDetector")
match = FailureMatch(failure_line=failure_line,
classified_failure=classification,
matcher=manual_detector,
score=1.0)
match.save()
for project, job_guids in by_project.iteritems():
with JobsModel(project) as jm:
jobs = jm.get_job_ids_by_guid(job_guids)
for job in jobs.values():
jm.update_after_verification(job["id"], email)
# Force failure line to be reloaded, including .classified_failures
rv = FailureLine.objects.prefetch_related(
'classified_failures').filter(id__in=failure_line_ids)
if not many:
rv = rv[0]
return serializers.FailureLineNoStackSerializer(rv,
many=many).data, 200
def _update(self, data, email, many=True):
by_project = defaultdict(list)
ids = []
failure_line_ids = set()
classification_ids = set()
for item in data:
line_id = int(item.get("id"))
if line_id is None:
return "No failure line id provided", 400
failure_line_ids.add(line_id)
if "best_classification" not in item:
return "No classification id provided", 400
classification_id = item.get("best_classification")
if classification_id is not None:
classification_ids.add(classification_id)
ids.append((line_id, classification_id))
failure_lines = as_dict(
FailureLine.objects.prefetch_related('classified_failures').filter(
id__in=failure_line_ids), "id")
if len(failure_lines) != len(failure_line_ids):
missing = failure_line_ids - set(failure_lines.keys())
return "No failure line with id: {0}".format(", ".join(missing)), 404
classifications = as_dict(
ClassifiedFailure.objects.filter(id__in=classification_ids), "id")
if len(classifications) != len(classification_ids):
missing = classification_ids - set(classifications.keys())
return "No classification with id: {0}".format(", ".join(missing)), 404
for line_id, classification_id in ids:
failure_line = failure_lines[line_id]
if classification_id is not None:
classification = classifications[classification_id]
else:
classification = None
by_project[failure_line.repository.name].append(failure_line.job_guid)
failure_line.best_classification = classification
failure_line.best_is_verified = True
failure_line.save()
if (classification is not None and
classification not in failure_line.classified_failures.all()):
manual_detector = Matcher.objects.get(name="ManualDetector")
match = FailureMatch(failure_line=failure_line,
classified_failure=classification,
matcher=manual_detector,
score=1.0)
match.save()
for project, job_guids in by_project.iteritems():
with JobsModel(project) as jm:
jobs = jm.get_job_ids_by_guid(job_guids)
for job in jobs.values():
jm.update_after_verification(job["id"], email)
# Force failure line to be reloaded, including .classified_failures
rv = FailureLine.objects.prefetch_related('classified_failures').filter(
id__in=failure_line_ids)
if not many:
rv = rv[0]
return serializers.FailureLineNoStackSerializer(rv, many=many).data, 200
