Bug 1321787 - Autoclassify through the TextLogError rather than through the FailureLine (#2179)

tests/autoclassify/test_classify_failures.py

@@ -1,23 +1,19 @@
-from datetime import (datetime,
-                      timedelta)
-
 from treeherder.autoclassify.detectors import (ManualDetector,
                                                TestFailureDetector)
 from treeherder.autoclassify.matchers import (CrashSignatureMatcher,
                                               ElasticSearchTestMatcher,
-                                              PreciseTestMatcher,
+                                              PreciseTestMatcher)
-                                              time_window)
 from treeherder.autoclassify.tasks import autoclassify
 from treeherder.model.models import (BugJobMap,
                                      ClassifiedFailure,
                                      FailureMatch,
                                      JobNote,
                                      TextLogError,
+                                     TextLogErrorMatch,
                                      TextLogErrorMetadata)
 
 from .utils import (crash_line,
-                    create_failure_lines,
+                    create_lines,
-                    create_text_log_errors,
                     log_line,
                     register_detectors,
                     register_matchers,
@@ -37,16 +33,6 @@ def do_autoclassify(job, test_failure_lines, matchers, status="testfailed"):
         item.refresh_from_db()
 
 
-def create_lines(test_job, lines):
-    error_lines = create_text_log_errors(test_job, lines)
-    failure_lines = create_failure_lines(test_job, lines)
-
-    for error_line, failure_line in zip(error_lines, failure_lines):
-        TextLogErrorMetadata.objects.create(text_log_error=error_line,
-                                            failure_line=failure_line)
-    return error_lines, failure_lines
-
-
 def test_classify_test_failure(text_log_errors_failure_lines,
                                classified_failures,
                                test_job_2):
@@ -120,7 +106,7 @@ def test_autoclassify_update_job_classification(failure_lines, classified_failur
 
 
 def test_autoclassify_no_update_job_classification(test_job, test_job_2,
-                                                   failure_lines,
+                                                   text_log_errors_failure_lines,
                                                    classified_failures):
 
     lines = [(test_line, {})]
@@ -134,13 +120,18 @@ def test_autoclassify_no_update_job_classification(test_job, test_job_2,
     assert JobNote.objects.filter(job=test_job_2).count() == 0
 
 
-def test_autoclassified_after_manual_classification(test_user, test_job_2,
+def test_autoclassified_after_manual_classification(test_user,
-                                                    failure_lines, failure_classifications):
+                                                    test_job_2,
+                                                    text_log_errors_failure_lines,
+                                                    failure_classifications):
     register_detectors(ManualDetector, TestFailureDetector)
 
     lines = [(test_line, {})]
     test_error_lines, test_failure_lines = create_lines(test_job_2, lines)
 
+    BugJobMap.objects.create(job=test_job_2,
+                             bug_id=1234,
+                             user=test_user)
     JobNote.objects.create(job=test_job_2,
                            failure_classification_id=4,
                            user=test_user,
@@ -186,9 +177,9 @@ def test_autoclassified_no_update_after_manual_classification_2(test_user, test_
     register_detectors(ManualDetector, TestFailureDetector)
 
     # Too many failure lines
-    test_failure_lines = create_failure_lines(test_job_2,
+    test_error_lines, test_failure_lines = create_lines(test_job_2,
-                                              [(log_line, {}),
+                                                        [(log_line, {}),
-                                               (test_line, {"subtest": "subtest2"})])
+                                                         (test_line, {"subtest": "subtest2"})])
 
     JobNote.objects.create(job=test_job_2,
                            failure_classification_id=4,
@@ -214,9 +205,9 @@ def test_classify_skip_ignore(test_job_2,
     failure_lines[1].best_classification = None
     failure_lines[1].save()
 
-    test_failure_lines = create_failure_lines(test_job_2,
+    test_error_lines, test_failure_lines = create_lines(test_job_2,
-                                              [(test_line, {}),
+                                                        [(test_line, {}),
-                                               (test_line, {"subtest": "subtest2"})])
+                                                         (test_line, {"subtest": "subtest2"})])
 
     do_autoclassify(test_job_2, test_failure_lines, [PreciseTestMatcher])
 
@@ -231,14 +222,14 @@ def test_classify_skip_ignore(test_job_2,
 
 
 def test_classify_es(test_job_2, failure_lines, classified_failures):
-    test_failure_lines = create_failure_lines(test_job_2,
+    test_error_lines, test_failure_lines = create_lines(test_job_2,
-                                              [(test_line, {}),
+                                                        [(test_line, {}),
-                                               (test_line, {"message": "message2"}),
+                                                         (test_line, {"message": "message2"}),
-                                               (test_line, {"message": "message 1.2"}),
+                                                         (test_line, {"message": "message 1.2"}),
-                                               (test_line, {"message": "message 0x1F"}),
+                                                         (test_line, {"message": "message 0x1F"}),
-                                               (test_line, {"subtest": "subtest3"}),
+                                                         (test_line, {"subtest": "subtest3"}),
-                                               (test_line, {"status": "TIMEOUT"}),
+                                                         (test_line, {"status": "TIMEOUT"}),
-                                               (test_line, {"expected": "ERROR"})])
+                                                         (test_line, {"expected": "ERROR"})])
 
     do_autoclassify(test_job_2, test_failure_lines, [ElasticSearchTestMatcher])
 
@@ -253,9 +244,9 @@ def test_classify_es(test_job_2, failure_lines, classified_failures):
 
 
 def test_classify_multiple(test_job_2, failure_lines, classified_failures):
-    test_failure_lines = create_failure_lines(test_job_2,
+    test_error_lines, test_failure_lines = create_lines(test_job_2,
-                                              [(test_line, {}),
+                                                        [(test_line, {}),
-                                               (test_line, {"message": "message 1.2"})])
+                                                         (test_line, {"message": "message 1.2"})])
 
     expected_classified_precise = [test_failure_lines[0]]
     expected_classified_fuzzy = [test_failure_lines[1]]
@@ -273,20 +264,24 @@ def test_classify_multiple(test_job_2, failure_lines, classified_failures):
 
 
 def test_classify_crash(test_repository, test_job, test_job_2, test_matcher):
-    failure_lines_ref = create_failure_lines(test_job,
+    error_lines_ref, failure_lines_ref = create_lines(test_job,
-                                             [(crash_line, {})])
+                                                      [(crash_line, {})])
 
-    failure_lines = create_failure_lines(test_job_2,
+    error_lines, failure_lines = create_lines(test_job_2,
-                                         [(crash_line, {}),
+                                              [(crash_line, {}),
-                                          (crash_line, {"test": "test1"}),
+                                               (crash_line, {"test": "test1"}),
-                                          (crash_line, {"signature": "signature1"}),
+                                               (crash_line, {"signature": "signature1"}),
-                                          (crash_line, {"signature": None})])
+                                               (crash_line, {"signature": None})])
 
     classified_failure = ClassifiedFailure.objects.create()
     FailureMatch.objects.create(failure_line=failure_lines_ref[0],
                                 classified_failure=classified_failure,
                                 matcher=test_matcher.db_object,
                                 score=1.0)
+    TextLogErrorMatch.objects.create(text_log_error=error_lines_ref[0],
+                                     classified_failure=classified_failure,
+                                     matcher=test_matcher.db_object,
+                                     score=1.0)
     do_autoclassify(test_job_2, failure_lines, [CrashSignatureMatcher])
 
     expected_classified = failure_lines[0:2]
@@ -297,22 +292,3 @@ def test_classify_crash(test_repository, test_job, test_job_2, test_matcher):
 
     for item in expected_unclassified:
         assert item.classified_failures.count() == 0
-
-
-def test_classify_test_failure_window(failure_lines, classified_failures):
-    failure_lines[0].created = datetime.now() - timedelta(days=2)
-    failure_lines[0].save()
-
-    failure_matches = FailureMatch.objects.all()
-    failure_matches[1].score = 0.5
-    failure_matches[1].save()
-
-    best_match = time_window(FailureMatch.objects.all(), timedelta(days=1), 0,
-                             lambda x: x.score)
-
-    assert best_match == failure_matches[1]
-
-    best_match = time_window(FailureMatch.objects.all(), timedelta(days=1), None,
-                             lambda x: x.score)
-
-    assert best_match == failure_matches[1]

tests/autoclassify/utils.py

@@ -3,15 +3,31 @@ import datetime
 from mozlog.formatters.tbplformatter import TbplFormatter
 
 from treeherder.model.models import (FailureLine,
+                                     Job,
                                      MatcherManager,
                                      TextLogError,
+                                     TextLogErrorMetadata,
                                      TextLogStep)
 from treeherder.model.search import refresh_all
 
 test_line = {"action": "test_result", "test": "test1", "subtest": "subtest1",
              "status": "FAIL", "expected": "PASS", "message": "message1"}
 log_line = {"action": "log", "level": "ERROR", "message": "message1"}
-crash_line = {"action": "crash", "signature": "signature"}
+crash_line = {"action": "crash", "signature": "signature", "test": "test1"}
+
+
+def create_lines(test_job, lines):
+    error_lines = create_text_log_errors(test_job, lines)
+    failure_lines = create_failure_lines(test_job, lines)
+
+    for error_line, failure_line in zip(error_lines, failure_lines):
+        TextLogErrorMetadata.objects.create(text_log_error=error_line,
+                                            failure_line=failure_line)
+
+    test_job.autoclassify_status = Job.CROSSREFERENCED
+    test_job.save()
+
+    return error_lines, failure_lines
 
 
 def create_failure_lines(job, failure_line_list,
@@ -46,6 +62,8 @@ def get_data(base_data, updates):
     elif data["action"] == "log":
         if data["level"] not in ("ERROR", "CRITICAL"):
             return
+    elif data["action"] == "crash":
+        pass
     else:
         return
     return data

tests/model/test_classified_failure.py

@@ -52,13 +52,12 @@ def test_update_autoclassification_bug(test_job, test_job_2,
     # Job 1 has two failure lines so nothing should be updated
     assert test_job.update_autoclassification_bug(1234) is None
 
-    failure_lines = create_failure_lines(test_job_2,
+    lines = [(test_line, {})]
-                                         [(test_line, {})])
+    create_failure_lines(test_job_2, lines)
-    failure_lines[0].best_classification = classified_failures[0]
+    error_lines = create_text_log_errors(test_job_2, lines)
-    failure_lines[0].save()
+
-    classified_failures[0].bug_number = None
+    error_lines[0].mark_best_classification(classified_failures[0])
-    lines = [(item, {}) for item in FailureLine.objects.filter(job_guid=test_job_2.guid).values()]
+    assert classified_failures[0].bug_number is None
-    create_text_log_errors(test_job_2, lines)
 
     assert test_job_2.update_autoclassification_bug(1234) == classified_failures[0]
     classified_failures[0].refresh_from_db()

tests/webapp/api/test_jobs_api.py

@@ -433,7 +433,9 @@ def test_text_log_steps_and_errors(webapp, test_job):
                         'search_terms': ['failure 1'],
                         'bugs': {'open_recent': [], 'all_others': []}
                     },
-                    'metadata': None
+                    'metadata': None,
+                    'matches': [],
+                    'classified_failures': []
                 },
                 {
                     'id': 2,
@@ -444,7 +446,9 @@ def test_text_log_steps_and_errors(webapp, test_job):
                         'search_terms': ['failure 2'],
                         'bugs': {'open_recent': [], 'all_others': []}
                     },
-                    'metadata': None
+                    'metadata': None,
+                    'matches': [],
+                    'classified_failures': []
                 }
             ],
             'finished': '1970-01-01T00:03:20',
@@ -494,7 +498,9 @@ def test_text_log_errors(webapp, test_job):
                 'search_terms': ['failure 1'],
                 'bugs': {'open_recent': [], 'all_others': []}
             },
-            'metadata': None
+            'metadata': None,
+            'matches': [],
+            'classified_failures': []
         },
         {
             'id': 2,
@@ -505,7 +511,9 @@ def test_text_log_errors(webapp, test_job):
                 'search_terms': ['failure 2'],
                 'bugs': {'open_recent': [], 'all_others': []}
             },
-            'metadata': None
+            'metadata': None,
+            'matches': [],
+            'classified_failures': []
         }
     ]
 

tests/webapp/api/test_text_log_error.py

@@ -0,0 +1,628 @@
+from django.core.urlresolvers import reverse
+from rest_framework.test import APIClient
+
+from tests.autoclassify.utils import (create_failure_lines,
+                                      create_text_log_errors,
+                                      test_line)
+from treeherder.autoclassify.detectors import ManualDetector
+from treeherder.model.models import (BugJobMap,
+                                     ClassifiedFailure,
+                                     FailureLine,
+                                     Job,
+                                     JobNote,
+                                     Matcher,
+                                     MatcherManager,
+                                     TextLogError,
+                                     TextLogErrorMetadata)
+from treeherder.model.search import TestFailureLine
+
+
+def test_get_error(text_log_errors_failure_lines):
+    """
+    test getting a single failure line
+    """
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+
+    client = APIClient()
+    resp = client.get(
+        reverse("text-log-error-detail", kwargs={"pk": text_log_errors[0].id}))
+
+    assert resp.status_code == 200
+
+    data = resp.json()
+
+    assert isinstance(data, object)
+    exp_error_keys = ["id", "line", "line_number", "matches",
+                      "classified_failures", "bug_suggestions", "metadata"]
+
+    assert set(data.keys()) == set(exp_error_keys)
+
+    exp_meta_keys = ["text_log_error", "failure_line", "best_classification",
+                     "best_is_verified"]
+    assert set(data["metadata"].keys()) == set(exp_meta_keys)
+
+
+def test_update_error_verify(test_repository,
+                             text_log_errors_failure_lines,
+                             classified_failures,
+                             test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    body = {"best_classification": classified_failures[0].id}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified
+
+    es_line = TestFailureLine.get(failure_line.id, routing=failure_line.test)
+    assert es_line.best_classification == classified_failures[0].id
+    assert es_line.best_is_verified
+
+
+def test_update_error_replace(test_repository,
+                              text_log_errors_failure_lines,
+                              classified_failures,
+                              test_user):
+
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    body = {"best_classification": classified_failures[1].id}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification == classified_failures[1]
+    assert failure_line.best_is_verified
+    assert len(failure_line.classified_failures.all()) == 2
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[1]
+    assert error_line.metadata.best_is_verified
+
+    expected_matcher = Matcher.objects.get(name="ManualDetector")
+    assert failure_line.matches.get(classified_failure_id=classified_failures[1].id).matcher == expected_matcher
+    assert error_line.matches.get(classified_failure_id=classified_failures[1].id).matcher == expected_matcher
+
+
+def test_update_error_mark_job(test_job,
+                               text_log_errors_failure_lines,
+                               classified_failures,
+                               test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    classified_failures[1].bug_number = 1234
+    classified_failures[1].save()
+
+    for text_log_error, failure_line in zip(text_log_errors, failure_lines):
+
+        assert failure_line.best_is_verified is False
+        assert text_log_error.metadata.best_is_verified is False
+
+        body = {"best_classification": classified_failures[1].id}
+
+        resp = client.put(reverse("text-log-error-detail", kwargs={"pk": text_log_error.id}),
+                          body, format="json")
+
+        assert resp.status_code == 200
+
+        failure_line.refresh_from_db()
+        text_log_error.metadata.refresh_from_db()
+
+        assert failure_line.best_classification == classified_failures[1]
+        assert failure_line.best_is_verified
+        assert text_log_error.metadata.best_classification == classified_failures[1]
+        assert text_log_error.metadata.best_is_verified
+
+    assert test_job.is_fully_verified()
+
+    # should only be one, will assert if that isn't the case
+    note = JobNote.objects.get(job=test_job)
+    assert note.failure_classification.id == 4
+    assert note.user == test_user
+    job_bugs = BugJobMap.objects.filter(job=test_job)
+    assert job_bugs.count() == 1
+    assert job_bugs[0].bug_id == 1234
+
+
+def test_update_error_mark_job_with_human_note(test_job,
+                                               text_log_errors_failure_lines,
+                                               classified_failures, test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    JobNote.objects.create(job=test_job,
+                           failure_classification_id=4,
+                           user=test_user,
+                           text="note")
+
+    for error_line in text_log_errors:
+
+        body = {"best_classification": classified_failures[1].id}
+
+        resp = client.put(reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+                          body, format="json")
+
+        assert resp.status_code == 200
+
+    assert test_job.is_fully_verified()
+
+    # should only be one, will assert if that isn't the case
+    note = JobNote.objects.get(job=test_job)
+    assert note.failure_classification.id == 4
+    assert note.user == test_user
+
+
+def test_update_error_line_mark_job_with_auto_note(test_job,
+                                                   mock_autoclassify_jobs_true,
+                                                   text_log_errors_failure_lines,
+                                                   classified_failures,
+                                                   test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    JobNote.objects.create(job=test_job,
+                           failure_classification_id=7,
+                           text="note")
+
+    for text_log_error in text_log_errors:
+        body = {"best_classification": classified_failures[1].id}
+
+        resp = client.put(reverse("text-log-error-detail", kwargs={"pk": text_log_error.id}),
+                          body, format="json")
+
+        assert resp.status_code == 200
+
+    assert test_job.is_fully_verified()
+
+    notes = JobNote.objects.filter(job=test_job).order_by('-created')
+    assert notes.count() == 2
+
+    assert notes[0].failure_classification.id == 4
+    assert notes[0].user == test_user
+    assert notes[0].text == ''
+
+    assert notes[1].failure_classification.id == 7
+    assert not notes[1].user
+    assert notes[1].text == "note"
+
+
+def test_update_errors(mock_autoclassify_jobs_true,
+                       test_repository,
+                       text_log_errors_failure_lines,
+                       classified_failures,
+                       eleven_jobs_stored,
+                       test_user):
+
+    jobs = (Job.objects.get(id=1), Job.objects.get(id=2))
+
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    lines = [(test_line, {}),
+             (test_line, {"subtest": "subtest2"})]
+    new_failure_lines = create_failure_lines(jobs[1], lines)
+    new_text_log_errors = create_text_log_errors(jobs[1], lines)
+
+    for text_log_error, failure_line in zip(new_text_log_errors,
+                                            new_failure_lines):
+        TextLogErrorMetadata.objects.create(text_log_error=text_log_error,
+                                            failure_line=failure_line)
+
+    failure_lines = FailureLine.objects.filter(
+        job_guid__in=[job.guid for job in jobs]).all()
+    text_log_errors = TextLogError.objects.filter(
+        step__job__in=jobs).all()
+
+    for text_log_error, failure_line in zip(text_log_errors, failure_lines):
+        assert text_log_error.metadata.best_is_verified is False
+        assert failure_line.best_is_verified is False
+
+    body = [{"id": failure_line.id,
+             "best_classification": classified_failures[1].id}
+            for failure_line in failure_lines]
+    resp = client.put(reverse("text-log-error-list"), body, format="json")
+
+    assert resp.status_code == 200
+
+    for text_log_error, failure_line in zip(text_log_errors, failure_lines):
+        text_log_error.metadata.refresh_from_db()
+        failure_line.refresh_from_db()
+        assert failure_line.best_classification == classified_failures[1]
+        assert failure_line.best_is_verified
+        assert text_log_error.metadata.best_classification == classified_failures[1]
+        assert text_log_error.metadata.best_is_verified
+
+    for job in jobs:
+        assert job.is_fully_verified()
+
+        # will assert if we don't have exactly one job, which is what we want
+        note = JobNote.objects.get(job=job)
+        assert note.failure_classification.id == 4
+        assert note.user == test_user
+
+
+def test_update_error_ignore(test_job, text_log_errors_failure_lines,
+                             classified_failures, test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    MatcherManager.register_detector(ManualDetector)
+
+    text_log_error = text_log_errors[0]
+    failure_line = failure_lines[0]
+    assert text_log_error.metadata.best_classification == classified_failures[0]
+    assert text_log_error.metadata.best_is_verified is False
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+
+    body = {"project": test_job.repository.name,
+            "best_classification": None}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": text_log_error.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    failure_line.refresh_from_db()
+    text_log_error.metadata.refresh_from_db()
+
+    assert failure_line.best_classification is None
+    assert failure_line.best_is_verified
+    assert text_log_error.metadata.best_classification is None
+    assert text_log_error.metadata.best_is_verified
+
+
+def test_update_error_all_ignore_mark_job(test_job,
+                                          mock_autoclassify_jobs_true,
+                                          text_log_errors_failure_lines,
+                                          classified_failures,
+                                          test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    job_failure_lines = [line for line in failure_lines if
+                         line.job_guid == test_job.guid]
+    job_text_log_errors = [error for error in text_log_errors if
+                           error.step.job == test_job]
+
+    for error_line, failure_line in zip(job_text_log_errors, job_failure_lines):
+        error_line.best_is_verified = False
+        error_line.best_classification = None
+        failure_line.best_is_verified = False
+        failure_line.best_classification = None
+
+    assert JobNote.objects.count() == 0
+
+    for error_line, failure_line in zip(job_text_log_errors, job_failure_lines):
+
+        assert error_line.metadata.best_is_verified is False
+        assert failure_line.best_is_verified is False
+
+        body = {"best_classification": None}
+
+        resp = client.put(reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+                          body, format="json")
+
+        assert resp.status_code == 200
+
+        error_line.metadata.refresh_from_db()
+        failure_line.refresh_from_db()
+
+        assert error_line.metadata.best_classification is None
+        assert error_line.metadata.best_is_verified
+        assert failure_line.best_classification is None
+        assert failure_line.best_is_verified
+
+    assert test_job.is_fully_verified()
+
+    assert JobNote.objects.count() == 1
+
+
+def test_update_error_partial_ignore_mark_job(test_job,
+                                              mock_autoclassify_jobs_true,
+                                              text_log_errors_failure_lines,
+                                              classified_failures,
+                                              test_user):
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    MatcherManager.register_detector(ManualDetector)
+
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    for i, (error_line, failure_line) in enumerate(zip(text_log_errors, failure_lines)):
+
+        assert error_line.metadata.best_is_verified is False
+        assert failure_line.best_is_verified is False
+
+        body = {"best_classification": None if i == 0 else classified_failures[0].id}
+
+        resp = client.put(reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+                          body, format="json")
+
+        assert resp.status_code == 200
+
+        error_line.metadata.refresh_from_db()
+        failure_line.refresh_from_db()
+
+        if i == 0:
+            assert error_line.metadata.best_classification is None
+            assert failure_line.best_classification is None
+        else:
+            assert error_line.metadata.best_classification == classified_failures[0]
+            assert failure_line.best_classification == classified_failures[0]
+        assert failure_line.best_is_verified
+
+    assert test_job.is_fully_verified()
+
+    # will assert if we don't have exactly one note for this job, which is
+    # what we want
+    note = JobNote.objects.get(job=test_job)
+    assert note.failure_classification.id == 4
+    assert note.user == test_user
+
+
+def test_update_error_verify_bug(test_repository,
+                                 text_log_errors_failure_lines,
+                                 classified_failures,
+                                 test_user):
+
+    MatcherManager.register_detector(ManualDetector)
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    classified_failures[0].bug_number = 1234
+    classified_failures[0].save()
+
+    body = {"bug_number": classified_failures[0].bug_number}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified
+
+    es_line = TestFailureLine.get(failure_line.id, routing=failure_line.test)
+    assert es_line.best_classification == classified_failures[0].id
+    assert es_line.best_is_verified
+
+
+def test_update_error_verify_new_bug(test_repository,
+                                     text_log_errors_failure_lines,
+                                     classified_failures,
+                                     test_user):
+
+    MatcherManager.register_detector(ManualDetector)
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    assert 78910 not in [item.bug_number for item in classified_failures]
+    body = {"bug_number": 78910}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification not in classified_failures
+    assert failure_line.best_classification.bug_number == 78910
+    assert failure_line.best_is_verified
+    assert error_line.metadata.best_classification not in classified_failures
+    assert error_line.metadata.best_is_verified
+    assert error_line.metadata.best_classification.bug_number == 78910
+
+
+def test_update_error_verify_ignore_now(test_repository,
+                                        text_log_errors_failure_lines,
+                                        classified_failures,
+                                        test_user):
+
+    MatcherManager.register_detector(ManualDetector)
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    assert 78910 not in [item.bug_number for item in classified_failures]
+    body = {}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification is None
+    assert failure_line.best_is_verified
+    assert error_line.metadata.best_classification is None
+    assert error_line.metadata.best_is_verified
+
+
+def test_update_error_change_bug(test_repository,
+                                 text_log_errors_failure_lines,
+                                 classified_failures,
+                                 test_user):
+
+    MatcherManager.register_detector(ManualDetector)
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    assert 78910 not in [item.bug_number for item in classified_failures]
+    body = {"best_classification": classified_failures[0].id,
+            "bug_number": 78910}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    classified_failures[0].refresh_from_db()
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_classification.bug_number == 78910
+    assert failure_line.best_is_verified
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified
+    assert error_line.metadata.best_classification.bug_number == 78910
+
+
+def test_update_error_bug_change_cf(test_repository,
+                                    text_log_errors_failure_lines,
+                                    classified_failures,
+                                    test_user):
+
+    MatcherManager.register_detector(ManualDetector)
+
+    text_log_errors, failure_lines = text_log_errors_failure_lines
+    client = APIClient()
+    client.force_authenticate(user=test_user)
+
+    failure_line = failure_lines[0]
+    error_line = text_log_errors[0]
+    assert failure_line.best_classification == classified_failures[0]
+    assert failure_line.best_is_verified is False
+    assert error_line.metadata.failure_line == failure_line
+    assert error_line.metadata.best_classification == classified_failures[0]
+    assert error_line.metadata.best_is_verified is False
+
+    assert 78910 not in [item.bug_number for item in classified_failures]
+    classified_failures[1].bug_number = 78910
+    classified_failures[1].save()
+
+    body = {"best_classification": classified_failures[0].id,
+            "bug_number": 78910}
+
+    resp = client.put(
+        reverse("text-log-error-detail", kwargs={"pk": error_line.id}),
+        body, format="json")
+
+    assert resp.status_code == 200
+
+    classified_failures[1].refresh_from_db()
+    failure_line.refresh_from_db()
+    error_line.metadata.refresh_from_db()
+
+    assert failure_line.best_classification == classified_failures[1]
+    assert failure_line.best_classification.bug_number == 78910
+    assert failure_line.best_is_verified
+    assert error_line.metadata.best_classification == classified_failures[1]
+    assert error_line.metadata.best_is_verified
+    assert error_line.metadata.best_classification.bug_number == 78910
+    assert ClassifiedFailure.objects.count() == len(classified_failures) - 1

tests/webapp/api/test_text_log_summary_lines.py

@@ -4,6 +4,8 @@ from rest_framework.test import APIClient
 from treeherder.model.models import (BugJobMap,
                                      FailureLine,
                                      JobNote,
+                                     TextLogError,
+                                     TextLogErrorMetadata,
                                      TextLogSummary)
 
 
@@ -136,6 +138,7 @@ def test_put_verify_job(webapp, test_repository, test_job, text_summary_lines, t
     client = APIClient()
     client.force_authenticate(user=test_user)
 
+    TextLogErrorMetadata.objects.filter(text_log_error__step__job=test_job).update(best_is_verified=True)
     FailureLine.objects.filter(job_guid=test_job.guid).update(best_is_verified=True)
 
     text_summary_lines = TextLogSummary.objects.filter(job_guid=test_job.guid).get().lines.all()

treeherder/autoclassify/autoclassify.py

@@ -3,10 +3,12 @@ from collections import defaultdict
 
 from django.db.utils import IntegrityError
 
-from treeherder.model.models import (FailureLine,
+from treeherder.model.models import (ClassifiedFailure,
                                      FailureMatch,
+                                     Job,
                                      JobNote,
                                      Matcher,
+                                     TextLogError,
                                      TextLogErrorMatch)
 
 logger = logging.getLogger(__name__)
@@ -21,66 +23,85 @@ def match_errors(job):
     # Only try to autoclassify where we have a failure status; sometimes there can be
     # error lines even in jobs marked as passing.
 
+    if job.autoclassify_status < Job.CROSSREFERENCED:
+        logger.error("Tried to autoclassify job %i without crossreferenced error lines" % job.id)
+        return
+
+    if job.autoclassify_status == Job.AUTOCLASSIFIED:
+        logger.error("Tried to autoclassify job %i which was already autoclassified" % job.id)
+        return
+
     if job.result not in ["testfailed", "busted", "exception"]:
         return
 
-    unmatched_failures = set(FailureLine.objects.unmatched_for_job(job))
+    unmatched_errors = set(TextLogError.objects.unmatched_for_job(job))
 
-    if not unmatched_failures:
+    if not unmatched_errors:
+        logger.info("Skipping autoclassify of job %i because it has no unmatched errors" % job.id)
         return
 
-    matches, all_matched = find_matches(unmatched_failures)
+    try:
-    update_db(job, matches, all_matched)
+        matches, all_matched = find_matches(unmatched_errors)
+        update_db(job, matches, all_matched)
+    except:
+        logger.error("Autoclassification of job %s failed" % job.id)
+        job.autoclassify_status = Job.FAILED
+        raise
+    else:
+        logger.debug("Autoclassification of job %s suceeded" % job.id)
+        job.autoclassify_status = Job.AUTOCLASSIFIED
+    finally:
+        job.save(update_fields=['autoclassify_status'])
 
 
-def find_matches(unmatched_failures):
+def find_matches(unmatched_errors):
     all_matches = set()
 
     for matcher in Matcher.objects.registered_matchers():
-        matches = matcher(unmatched_failures)
+        matches = matcher(unmatched_errors)
         for match in matches:
-            logger.info("Matched failure %i with intermittent %i" %
+            logger.info("Matched error %i with intermittent %i" %
-                        (match.failure_line.id, match.classified_failure.id))
+                        (match.text_log_error.id, match.classified_failure_id))
             all_matches.add((matcher.db_object, match))
             if match.score >= AUTOCLASSIFY_GOOD_ENOUGH_RATIO:
-                unmatched_failures.remove(match.failure_line)
+                unmatched_errors.remove(match.text_log_error)
 
-        if not unmatched_failures:
+        if not unmatched_errors:
             break
 
-    return all_matches, len(unmatched_failures) == 0
+    return all_matches, len(unmatched_errors) == 0
 
 
 def update_db(job, matches, all_matched):
-    matches_by_failure_line = defaultdict(set)
+    matches_by_error = defaultdict(set)
-    for item in matches:
+    classified_failures = {item.id: item for item in
-        matches_by_failure_line[item[1].failure_line].add(item)
+                           ClassifiedFailure.objects.filter(
+                               id__in=[match.classified_failure_id for _, match in matches])}
+    for matcher, match in matches:
+        classified_failure = classified_failures[match.classified_failure_id]
+        matches_by_error[match.text_log_error].add((matcher, match, classified_failure))
 
-    for failure_line, matches in matches_by_failure_line.iteritems():
+    for text_log_error, matches in matches_by_error.iteritems():
-        for matcher, match in matches:
+        for (matcher, match, classified_failure) in matches:
             try:
-                FailureMatch.objects.create(
+                TextLogErrorMatch.objects.create(
                     score=match.score,
                     matcher=matcher,
-                    classified_failure=match.classified_failure,
+                    classified_failure=classified_failure,
-                    failure_line=failure_line)
+                    text_log_error=match.text_log_error)
-                if failure_line.error:
+                if match.text_log_error.metadata and match.text_log_error.metadata.failure_line:
-                    TextLogErrorMatch.objects.create(
+                    FailureMatch.objects.create(
                         score=match.score,
                         matcher=matcher,
-                        classified_failure=match.classified_failure,
+                        classified_failure=classified_failure,
-                        text_log_error=failure_line.error)
+                        failure_line=match.text_log_error.metadata.failure_line)
             except IntegrityError:
                 logger.warning(
-                    "Tried to create duplicate match for failure line %i with matcher %i and classified_failure %i" %
+                    "Tried to create duplicate match for TextLogError %i with matcher %i and classified_failure %i" %
-                    (failure_line.id, matcher.id, match.classified_failure.id))
+                    (text_log_error.id, matcher.id, classified_failure.id))
-        best_match = failure_line.best_automatic_match(AUTOCLASSIFY_CUTOFF_RATIO)
+        best_match = text_log_error.best_automatic_match(AUTOCLASSIFY_CUTOFF_RATIO)
         if best_match:
-            failure_line.best_classification = best_match.classified_failure
+            text_log_error.mark_best_classification(classified_failure)
-            failure_line.save(update_fields=['best_classification'])
-            if failure_line.error:
-                failure_line.error.metadata.best_classification = best_match.classified_failure
-                failure_line.error.metadata.save(update_fields=['best_classification'])
 
     if all_matched:
         if job.is_fully_autoclassified():

treeherder/autoclassify/detectors.py

@@ -25,11 +25,14 @@ class Detector(object):
 
 
 class TestFailureDetector(Detector):
-    def __call__(self, failure_lines):
+    def __call__(self, text_log_errors):
         rv = []
-        for i, failure in enumerate(failure_lines):
+        with_failure_lines = [(i, item) for (i, item) in enumerate(text_log_errors)
-            if (failure.action == "test_result" and failure.test and failure.status
+                              if item.metadata and item.metadata.failure_line]
-                    and failure.expected):
+        for i, text_log_error in with_failure_lines:
+            failure = text_log_error.metadata.failure_line
+            if (failure.action == "test_result" and failure.test and failure.status and
+                failure.expected):
                 rv.append(i)
         return rv
 
@@ -37,7 +40,7 @@ class TestFailureDetector(Detector):
 class ManualDetector(Detector):
     """Small hack; this ensures that there's a matcher object indicating that a match
     was by manual association, but which never automatically matches any lines"""
-    def __call__(self, failure_lines):
+    def __call__(self, text_log_errors):
         return []
 
 

treeherder/autoclassify/matchers.py

@@ -3,8 +3,6 @@ import time
 from abc import (ABCMeta,
                  abstractmethod)
 from collections import namedtuple
-from datetime import (datetime,
-                      timedelta)
 from difflib import SequenceMatcher
 
 from django.conf import settings
@@ -12,16 +10,15 @@ from django.db.models import Q
 from elasticsearch_dsl.query import Match as ESMatch
 
 from treeherder.autoclassify.autoclassify import AUTOCLASSIFY_GOOD_ENOUGH_RATIO
-from treeherder.model.models import (ClassifiedFailure,
+from treeherder.model.models import (MatcherManager,
-                                     FailureLine,
+                                     TextLogError,
-                                     FailureMatch,
+                                     TextLogErrorMatch)
-                                     MatcherManager)
 from treeherder.model.search import (TestFailureLine,
                                      es_connected)
 
 logger = logging.getLogger(__name__)
 
-Match = namedtuple('Match', ['failure_line', 'classified_failure', 'score'])
+Match = namedtuple('Match', ['text_log_error', 'classified_failure_id', 'score'])
 
 
 class Matcher(object):
@@ -37,83 +34,134 @@ class Matcher(object):
     def __init__(self, db_object):
         self.db_object = db_object
 
+    def __call__(self, text_log_errors):
+        rv = []
+        for text_log_error in text_log_errors:
+            match = self.match(text_log_error)
+            if match:
+                rv.append(match)
+        return rv
+
+    def match(self, text_log_error):
+        best_match = self.query_best(text_log_error)
+        if best_match:
+            classified_failure_id, score = best_match
+            logger.debug("Matched using %s" % self.__class__.__name__)
+            return Match(text_log_error,
+                         classified_failure_id,
+                         score)
+
     @abstractmethod
-    def __call__(self, failure_lines):
+    def query_best(self, text_log_error):
         pass
 
 
-ignored_line = (Q(failure_line__best_classification=None) &
+ignored_line = (Q(text_log_error___metadata__best_classification=None) &
-                Q(failure_line__best_is_verified=True))
+                Q(text_log_error___metadata__best_is_verified=True))
 
 
-def time_window(queryset, interval, time_budget_ms, match_filter):
+class id_window(object):
-    upper_cutoff = datetime.now()
+    def __init__(self, size, time_budget):
-    lower_cutoff = upper_cutoff - interval
+        self.size = size
-    matches = []
+        self.time_budget_ms = time_budget
 
-    time_budget = time_budget_ms / 1000. if time_budget_ms is not None else None
+    def __call__(self, f):
-    t0 = time.time()
+        outer = self
 
-    min_date = FailureLine.objects.order_by("id")[0].created
+        def inner(self, text_log_error):
+            queries = f(self, text_log_error)
+            if not queries:
+                return
 
-    count = 0
+            for item in queries:
-    while True:
+                if isinstance(item, tuple):
-        count += 1
+                    query, score_multiplier = item
-        window_queryset = queryset.filter(
+                else:
-            failure_line__created__range=(lower_cutoff, upper_cutoff))
+                    query = item
-        logger.debug("[time_window] Queryset: %s" % window_queryset.query)
+                    score_multiplier = (1, 1)
-        match = window_queryset.first()
+
-        if match is not None:
+                result = outer.run(query, score_multiplier)
-            matches.append(match)
+                if result:
-            if match.score >= AUTOCLASSIFY_GOOD_ENOUGH_RATIO:
+                    return result
+        inner.__name__ = f.__name__
+        inner.__doc__ = f.__doc__
+        return inner
+
+    def run(self, query, score_multiplier):
+        matches = []
+        time_budget = self.time_budget_ms / 1000. if self.time_budget_ms is not None else None
+        t0 = time.time()
+
+        upper_cutoff = (TextLogError.objects
+                        .order_by('-id')
+                        .values_list('id', flat=True)[0])
+
+        count = 0
+        while upper_cutoff > 0:
+            count += 1
+            lower_cutoff = max(upper_cutoff - self.size, 0)
+            window_queryset = query.filter(
+                text_log_error__id__range=(lower_cutoff, upper_cutoff))
+            logger.debug("[time_window] Queryset: %s" % window_queryset.query)
+            match = window_queryset.first()
+            if match is not None:
+                score = match.score * score_multiplier[0] / score_multiplier[1]
+                matches.append((match, score))
+                if score >= AUTOCLASSIFY_GOOD_ENOUGH_RATIO:
+                    break
+            upper_cutoff -= self.size
+
+            if time_budget is not None and time.time() - t0 > time_budget:
+                # Putting the condition at the end of the loop ensures that we always
+                # run it once, which is useful for testing
                 break
-        upper_cutoff = lower_cutoff
-        lower_cutoff = upper_cutoff - interval
-        if upper_cutoff < min_date:
-            break
 
-        if time_budget_ms is not None and time.time() - t0 > time_budget:
+        logger.debug("[time_window] Used %i queries" % count)
-            # Putting the condition at the end of the loop ensures that we always
+        if matches:
-            # run it once, which is useful for testing
+            matches.sort(key=lambda x: (-x[1], -x[0].classified_failure_id))
-            break
+            best = matches[0]
+            return best[0].classified_failure_id, best[1]
 
-    logger.debug("[time_window] Used %i queries" % count)
+        return None
-    if matches:
+
-        matches.sort(key=match_filter)
+
-        return matches[0]
+def with_failure_lines(f):
-    return None
+    def inner(self, text_log_errors):
+        with_failure_lines = [item for item in text_log_errors
+                              if item.metadata and item.metadata.failure_line]
+        return f(self, with_failure_lines)
+    inner.__name__ = f.__name__
+    inner.__doc__ = f.__doc__
+    return inner
 
 
 class PreciseTestMatcher(Matcher):
-    """Matcher that looks for existing failures with identical tests and identical error
+    """Matcher that looks for existing failures with identical tests and
-    message."""
+    identical error message."""
 
-    def __call__(self, failure_lines):
+    @with_failure_lines
-        rv = []
+    def __call__(self, text_log_errors):
-        for failure_line in failure_lines:
+        return super(PreciseTestMatcher, self).__call__(text_log_errors)
-            logger.debug("Looking for test match in failure %d" % failure_line.id)
 
-            if failure_line.action != "test_result" or failure_line.message is None:
+    @id_window(size=20000,
-                continue
+               time_budget=500)
+    def query_best(self, text_log_error):
+        failure_line = text_log_error.metadata.failure_line
+        logger.debug("Looking for test match in failure %d" % failure_line.id)
 
-            matching_failures = FailureMatch.objects.filter(
+        if failure_line.action != "test_result" or failure_line.message is None:
-                failure_line__action="test_result",
+            return
-                failure_line__test=failure_line.test,
-                failure_line__subtest=failure_line.subtest,
-                failure_line__status=failure_line.status,
-                failure_line__expected=failure_line.expected,
-                failure_line__message=failure_line.message).exclude(
-                    ignored_line | Q(failure_line__job_guid=failure_line.job_guid)
-                ).order_by("-score", "-classified_failure")
 
-            best_match = time_window(matching_failures, timedelta(days=7), 500,
+        return [(TextLogErrorMatch.objects
-                                     lambda x: (-x.score, -x.classified_failure_id))
+                 .filter(text_log_error___metadata__failure_line__action="test_result",
-            if best_match:
+                         text_log_error___metadata__failure_line__test=failure_line.test,
-                logger.debug("Matched using precise test matcher")
+                         text_log_error___metadata__failure_line__subtest=failure_line.subtest,
-                rv.append(Match(failure_line,
+                         text_log_error___metadata__failure_line__status=failure_line.status,
-                                best_match.classified_failure,
+                         text_log_error___metadata__failure_line__expected=failure_line.expected,
-                                best_match.score))
+                         text_log_error___metadata__failure_line__message=failure_line.message)
-        return rv
+                 .exclude(ignored_line |
+                          Q(text_log_error__step__job=text_log_error.step.job))
+                 .order_by("-score", "-classified_failure"))]
 
 
 class ElasticSearchTestMatcher(Matcher):
@@ -126,79 +174,68 @@ class ElasticSearchTestMatcher(Matcher):
         self.calls = 0
 
     @es_connected(default=[])
-    def __call__(self, failure_lines):
+    @with_failure_lines
-        rv = []
+    def __call__(self, text_log_errors):
-        self.lines += len(failure_lines)
+        return super(ElasticSearchTestMatcher, self).__call__(text_log_errors)
-        for failure_line in failure_lines:
+
-            if failure_line.action != "test_result" or not failure_line.message:
+    def query_best(self, text_log_error):
-                logger.debug("Skipped elasticsearch matching")
+        failure_line = text_log_error.metadata.failure_line
-                continue
+        if failure_line.action != "test_result" or not failure_line.message:
-            match = ESMatch(message={"query": failure_line.message[:1024],
+            logger.debug("Skipped elasticsearch matching")
-                                     "type": "phrase"})
+            return
-            search = (TestFailureLine.search()
+        match = ESMatch(message={"query": failure_line.message[:1024],
-                      .filter("term", test=failure_line.test)
+                                 "type": "phrase"})
-                      .filter("term", status=failure_line.status)
+        search = (TestFailureLine.search()
-                      .filter("term", expected=failure_line.expected)
+                  .filter("term", test=failure_line.test)
-                      .filter("exists", field="best_classification")
+                  .filter("term", status=failure_line.status)
-                      .query(match))
+                  .filter("term", expected=failure_line.expected)
-            if failure_line.subtest:
+                  .filter("exists", field="best_classification")
-                search = search.filter("term", subtest=failure_line.subtest)
+                  .query(match))
-            try:
+        if failure_line.subtest:
-                self.calls += 1
+            search = search.filter("term", subtest=failure_line.subtest)
-                resp = search.execute()
+        try:
-            except:
+            self.calls += 1
-                logger.error("Elastic search lookup failed: %s %s %s %s %s",
+            resp = search.execute()
-                             failure_line.test, failure_line.subtest, failure_line.status,
+        except:
-                             failure_line.expected, failure_line.message)
+            logger.error("Elastic search lookup failed: %s %s %s %s %s",
-                raise
+                         failure_line.test, failure_line.subtest, failure_line.status,
-            scorer = MatchScorer(failure_line.message)
+                         failure_line.expected, failure_line.message)
-            matches = [(item, item.message) for item in resp]
+            raise
-            best_match = scorer.best_match(matches)
+        scorer = MatchScorer(failure_line.message)
-            if best_match:
+        matches = [(item, item.message) for item in resp]
-                logger.debug("Matched using elastic search test matcher")
+        best_match = scorer.best_match(matches)
-                rv.append(Match(failure_line,
+        if best_match:
-                                ClassifiedFailure.objects.get(
+            return (best_match[1].best_classification, best_match[0])
-                                    id=best_match[1].best_classification),
-                                best_match[0]))
-        return rv
 
 
 class CrashSignatureMatcher(Matcher):
     """Matcher that looks for crashes with identical signature"""
 
-    def __call__(self, failure_lines):
+    @with_failure_lines
-        rv = []
+    def __call__(self, text_log_errors):
+        return super(CrashSignatureMatcher, self).__call__(text_log_errors)
 
-        for failure_line in failure_lines:
+    @id_window(size=20000,
-            if (failure_line.action != "crash" or failure_line.signature is None
+               time_budget=250)
-                    or failure_line.signature == "None"):
+    def query_best(self, text_log_error):
-                continue
+        failure_line = text_log_error.metadata.failure_line
-            matching_failures = FailureMatch.objects.filter(
-                failure_line__action="crash",
-                failure_line__signature=failure_line.signature).exclude(
-                    ignored_line | Q(failure_line__job_guid=failure_line.job_guid)
-                ).select_related('failure_line').order_by(
-                    "-score",
-                    "-classified_failure")
 
-            score_multiplier = 10
+        if (failure_line.action != "crash" or
-            matching_failures_same_test = matching_failures.filter(
+            failure_line.signature is None or
-                failure_line__test=failure_line.test)
+            failure_line.signature == "None"):
+            return
 
-            best_match = time_window(matching_failures_same_test, timedelta(days=7), 250,
+        matching_failures = (TextLogErrorMatch.objects
-                                     lambda x: (-x.score, -x.classified_failure_id))
+                             .filter(text_log_error___metadata__failure_line__action="crash",
-            if not best_match:
+                                     text_log_error___metadata__failure_line__signature=failure_line.signature)
-                score_multiplier = 8
+                             .exclude(ignored_line |
-                best_match = time_window(matching_failures, timedelta(days=7), 250,
+                                      Q(text_log_error__step__job=text_log_error.step.job))
-                                         lambda x: (-x.score, -x.classified_failure_id))
+                             .select_related('text_log_error',
+                                             'text_log_error___metadata')
+                             .order_by("-score", "-classified_failure"))
 
-            if best_match:
+        return [matching_failures.filter(text_log_error___metadata__failure_line__test=failure_line.test),
-                logger.debug("Matched using crash signature matcher")
+                (matching_failures, (8, 10))]
-                score = best_match.score * score_multiplier / 10
-                rv.append(Match(failure_line,
-                                best_match.classified_failure,
-                                score))
-        return rv
 
 
 class MatchScorer(object):

treeherder/log_parser/crossreference.py

@@ -6,6 +6,7 @@ from django.db import (IntegrityError,
 from mozlog.formatters.tbplformatter import TbplFormatter
 
 from treeherder.model.models import (FailureLine,
+                                     Job,
                                      TextLogError,
                                      TextLogErrorMetadata,
                                      TextLogSummary,
@@ -14,7 +15,6 @@ from treeherder.model.models import (FailureLine,
 logger = logging.getLogger(__name__)
 
 
-@transaction.atomic
 def crossreference_job(job):
     """Populate the TextLogSummary and TextLogSummaryLine tables for a
     job. Specifically this function tries to match the
@@ -27,9 +27,21 @@ def crossreference_job(job):
     """
 
     try:
-        return _crossreference(job)
+        if job.autoclassify_status >= Job.CROSSREFERENCED:
+            logger.debug("Job %i already crossreferenced" % job.id)
+            return (TextLogError.objects
+                    .filter(step__job=job)
+                    .exists() and
+                    FailureLine.objects
+                    .filter(job_guid=job.guid)
+                    .exists())
+        rv = _crossreference(job)
+        job.autoclassify_status = Job.CROSSREFERENCED
+        job.save(update_fields=['autoclassify_status'])
+        return rv
     except IntegrityError:
         logger.warning("IntegrityError crossreferencing error lines for job %s" % job.id)
+        return False
 
 
 @transaction.atomic
@@ -43,7 +55,7 @@ def _crossreference(job):
     text_log_errors = TextLogError.objects.filter(
         step__job=job).order_by('line_number')
 
-    # If we don't have failure lines and text log errors nothing will happen
+    # If we don't have both failure lines and text log errors nothing will happen
     # so return early
     if not (failure_lines.exists() and text_log_errors.exists()):
         return False

treeherder/log_parser/tasks.py

@@ -96,5 +96,8 @@ def crossreference_error_lines(job_id, priority):
         autoclassify.apply_async(
             args=[job_id],
             routing_key="autoclassify.%s" % priority)
+    elif not settings.AUTOCLASSIFY_JOBS:
+        job.autoclassify_status = Job.SKIPPED
+        job.save(update_fields=['autoclassify_status'])
     else:
         logger.debug("Job %i didn't have any crossreferenced lines, skipping autoclassify " % job_id)

treeherder/model/migrations/0005_autoclassify_status.py

@@ -0,0 +1,44 @@
+# -*- coding: utf-8 -*-
+from __future__ import unicode_literals
+
+from django.db import migrations, models
+
+# Original SQL:
+#
+# BEGIN;
+# --
+# -- Add field autoclassify_status to job
+# --
+# ALTER TABLE `job` ADD COLUMN `autoclassify_status` integer DEFAULT 0 NOT NULL;
+# ALTER TABLE `job` ALTER COLUMN `autoclassify_status` DROP DEFAULT;
+# COMMIT;
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ('model', '0004_duplicate_failure_classifications'),
+    ]
+
+    operations = [migrations.RunSQL(
+        sql="""
+--
+-- Add field autoclassify_status to job
+--
+SET FOREIGN_KEY_CHECKS=0;
+ALTER TABLE `job` ADD COLUMN `autoclassify_status` integer DEFAULT 0 NOT NULL;
+ALTER TABLE `job` ALTER COLUMN `autoclassify_status` DROP DEFAULT;
+""",
+        reverse_sql="""
+--
+-- Add field autoclassify_status to job
+--
+ALTER TABLE `job` DROP COLUMN `autoclassify_status` CASCADE;
+""",
+        state_operations=[
+            migrations.AddField(
+                model_name='job',
+                name='autoclassify_status',
+                field=models.IntegerField(default=0, choices=[(0, 'pending'), (1, 'crossreferenced'), (2, 'autoclassified'), (3, 'skipped'), (255, 'failed')]),
+            ),
+        ])]

treeherder/model/models.py

@@ -7,7 +7,6 @@ import time
 from collections import (OrderedDict,
                          defaultdict)
 from hashlib import sha1
-from itertools import chain
 
 from django.conf import settings
 from django.contrib.auth.models import User
@@ -667,6 +666,7 @@ class Job(models.Model):
     repository = models.ForeignKey(Repository)
     guid = models.CharField(max_length=50, unique=True)
     project_specific_id = models.PositiveIntegerField(null=True)
+    autoclassify_status = models.IntegerField(choices=AUTOCLASSIFY_STATUSES, default=PENDING)
 
     coalesced_to_guid = models.CharField(max_length=50, null=True,
                                          default=None)
@@ -736,45 +736,31 @@ class Job(models.Model):
                                       action="truncated").count() > 0:
             return False
 
-        classified_failure_lines_count = FailureLine.objects.filter(
+        classified_error_count = TextLogError.objects.filter(
-            best_classification__isnull=False,
+            _metadata__best_classification__isnull=False,
-            job_guid=self.guid).count()
+            step__job=self).count()
 
-        if classified_failure_lines_count == 0:
+        if classified_error_count == 0:
             return False
 
         from treeherder.model.error_summary import get_filtered_error_lines
 
-        return classified_failure_lines_count == len(get_filtered_error_lines(self))
+        return classified_error_count == len(get_filtered_error_lines(self))
 
     def is_fully_verified(self):
-        if FailureLine.objects.filter(job_guid=self.guid,
-                                      action="truncated").count() > 0:
-            logger.error("Job %s truncated storage of FailureLines" % self.guid)
-            return False
-
         # Line is not fully verified if there are either structured failure lines
         # with no best failure, or unverified unstructured lines not associated with
         # a structured line
 
-        unverified_failure_lines = FailureLine.objects.filter(
+        unverified_errors = TextLogError.objects.filter(
-            best_is_verified=False,
+            _metadata__best_is_verified=False,
-            job_guid=self.guid).count()
+            step__job=self).count()
 
-        if unverified_failure_lines:
+        if unverified_errors:
-            logger.error("Job %s has unverified FailureLines" % self.guid)
+            logger.error("Job %r has unverified TextLogErrors" % self)
             return False
 
-        unverified_text_lines = TextLogSummaryLine.objects.filter(
+        logger.info("Job %r is fully verified" % self)
-            verified=False,
-            failure_line=None,
-            summary__job_guid=self.guid).count()
-
-        if unverified_text_lines:
-            logger.error("Job %s has unverified TextLogSummary" % self.guid)
-            return False
-
-        logger.info("Job %s is fully verified" % self.guid)
         return True
 
     def update_after_verification(self, user):
@@ -798,13 +784,13 @@ class Job(models.Model):
 
     def get_manual_classification_line(self):
         """
-        Return the FailureLine from a job if it can be manually classified as a side effect
+        Return the TextLogError from a job if it can be manually classified as a side effect
         of the overall job being classified.
         Otherwise return None.
         """
         try:
-            failure_lines = [FailureLine.objects.get(job_guid=self.guid)]
+            text_log_error = TextLogError.objects.get(step__job=self)
-        except (FailureLine.DoesNotExist, FailureLine.MultipleObjectsReturned):
+        except (TextLogError.DoesNotExist, TextLogError.MultipleObjectsReturned):
             return None
 
         # Only propagate the classification if there is exactly one unstructured failure
@@ -815,19 +801,20 @@ class Job(models.Model):
 
         # Check that some detector would match this. This is being used as an indication
         # that the autoclassifier will be able to work on this classification
-        if not any(detector(failure_lines)
+        if not any(detector([text_log_error])
                    for detector in Matcher.objects.registered_detectors()):
             return None
 
-        return failure_lines[0]
+        return text_log_error
 
     def update_autoclassification_bug(self, bug_number):
-        failure_line = self.get_manual_classification_line()
+        text_log_error = self.get_manual_classification_line()
 
-        if failure_line is None:
+        if text_log_error is None:
             return
 
-        classification = failure_line.best_classification
+        classification = (text_log_error.metadata.best_classification if text_log_error.metadata
+                          else None)
         if classification and classification.bug_number is None:
             return classification.set_bug(bug_number)
 
@@ -975,18 +962,18 @@ class JobNoteManager(models.Manager):
 
         # Only insert bugs for verified failures since these are automatically
         # mirrored to ES and the mirroring can't be undone
-        classified_failures = ClassifiedFailure.objects.filter(
+        bug_numbers = set(ClassifiedFailure.objects
-            best_for_lines__job_guid=job.guid,
+                          .filter(best_for_errors__text_log_error__step__job=job,
-            best_for_lines__best_is_verified=True)
+                                  best_for_errors__best_is_verified=True)
+                          .exclude(bug_number=None)
+                          .values_list('bug_number', flat=True))
 
-        text_log_summary_lines = TextLogSummaryLine.objects.filter(
+        # Legacy
-            summary__job_guid=job.guid, verified=True).exclude(
+        bug_numbers |= set(TextLogSummaryLine.objects
-                bug_number=None)
+                           .filter(summary__job_guid=job.guid,
-
+                                   verified=True)
-        bug_numbers = {item.bug_number
+                           .exclude(bug_number=None)
-                       for item in chain(classified_failures,
+                           .values_list('bug_number', flat=True))
-                                         text_log_summary_lines)
-                       if item.bug_number}
 
         for bug_number in bug_numbers:
             BugJobMap.objects.get_or_create(job=job,
@@ -1045,12 +1032,34 @@ class JobNote(models.Model):
         self.job.save()
 
         # if a manually filed job, update the autoclassification information
-        if self.user:
+        if not self.user:
-            if self.failure_classification.name in [
+            return
-                    "intermittent", "intermittent needs filing"]:
+
-                failure_line = self.job.get_manual_classification_line()
+        if self.failure_classification.name not in [
-                if failure_line:
+                "intermittent", "intermittent needs filing"]:
-                    failure_line.update_autoclassification()
+            return
+
+        text_log_error = self.job.get_manual_classification_line()
+        if not text_log_error:
+            return
+        bug_numbers = set(BugJobMap.objects
+                          .filter(job=self.job)
+                          .values_list('bug_id', flat=True))
+
+        existing_bugs = set(ClassifiedFailure.objects
+                            .filter(error_matches__text_log_error=text_log_error)
+                            .values_list('bug_number', flat=True))
+
+        add_bugs = (bug_numbers - existing_bugs)
+        if not add_bugs:
+            return
+
+        manual_detector = Matcher.objects.get(name="ManualDetector")
+        for bug_number in add_bugs:
+            classification, _ = text_log_error.set_classification(manual_detector,
+                                                                  bug_number=bug_number)
+        if len(add_bugs) == 1 and not existing_bugs:
+            text_log_error.mark_best_classification_verified(classification)
 
     def save(self, *args, **kwargs):
         super(JobNote, self).save(*args, **kwargs)
@@ -1067,24 +1076,6 @@ class JobNote(models.Model):
                                         self.who)
 
 
-class FailureLineManager(models.Manager):
-    def unmatched_for_job(self, job):
-        return FailureLine.objects.filter(
-            job_guid=job.guid,
-            repository=job.repository,
-            classified_failures=None,
-        )
-
-    def for_jobs(self, *jobs, **filters):
-        failures = FailureLine.objects.filter(
-            job_guid__in=[item.guid for item in jobs],
-            **filters)
-        failures_by_job = defaultdict(list)
-        for item in failures:
-            failures_by_job[item.job_guid].append(item)
-        return failures_by_job
-
-
 class FailureLine(models.Model):
     # We make use of prefix indicies for several columns in this table which
     # can't be expressed in django syntax so are created with raw sql in migrations.
@@ -1129,9 +1120,6 @@ class FailureLine(models.Model):
     created = models.DateTimeField(auto_now_add=True)
     modified = models.DateTimeField(auto_now=True)
 
-    objects = FailureLineManager()
-    # TODO: add indexes once we know which queries will be typically executed
-
     class Meta:
         db_table = 'failure_line'
         index_together = (
@@ -1199,7 +1187,9 @@ class FailureLine(models.Model):
         return classification, new_link
 
     def mark_best_classification_verified(self, classification):
-        if classification not in self.classified_failures.all():
+        if (classification and
+            classification.id not in self.classified_failures.values_list('id', flat=True)):
+            logger.debug("Adding new classification to TextLogError")
             manual_detector = Matcher.objects.get(name="ManualDetector")
             self.set_classification(manual_detector, classification=classification)
 
@@ -1302,23 +1292,26 @@ class ClassifiedFailure(models.Model):
         # ON matches.classified_failure_id = <other.id> AND
         #    matches.failure_line_id = failure_match.failue_line_id
         delete_ids = []
-        for match in self.matches.all():
+        for Match, key, matches in [(TextLogErrorMatch, "text_log_error",
-            try:
+                                     self.error_matches.all()),
-                existing = FailureMatch.objects.get(classified_failure=other,
+                                    (FailureMatch, "failure_line",
-                                                    failure_line=match.failure_line)
+                                     self.matches.all())]:
-                if match.score > existing.score:
+            for match in matches:
-                    existing.score = match.score
+                kwargs = {key: getattr(match, key)}
-                    existing.save()
+                existing = Match.objects.filter(classified_failure=other, **kwargs)
-                delete_ids.append(match.id)
+                if existing:
-            except FailureMatch.DoesNotExist:
+                    for existing_match in existing:
-                match.classified_failure = other
+                        if match.score > existing_match.score:
-                match.save()
+                            existing_match.score = match.score
-        FailureMatch.objects.filter(id__in=delete_ids).delete()
+                            existing_match.save()
+                    delete_ids.append(match.id)
+                else:
+                    match.classified_failure = other
+                    match.save()
+            Match.objects.filter(id__in=delete_ids).delete()
         FailureLine.objects.filter(best_classification=self).update(best_classification=other)
         self.delete()
 
-    # TODO: add indexes once we know which queries will be typically executed
-
     class Meta:
         db_table = 'classified_failure'
 
@@ -1510,6 +1503,32 @@ class TextLogStep(models.Model):
                            'finished_line_number')
 
 
+class TextLogErrorManager(models.Manager):
+    def unmatched_for_job(self, job):
+        """Return a the text log errors for a specific job that have
+        no associated ClassifiedFailure.
+
+        :param job: Job associated with the text log errors"""
+        return TextLogError.objects.filter(
+            step__job=job,
+            classified_failures=None,
+        ).prefetch_related('step', '_metadata', '_metadata__failure_line')
+
+    def for_jobs(self, *jobs, **filters):
+        """Return a dict of {job: [text log errors]} for a set of jobs, filtered by
+        caller-provided django filters.
+
+        :param jobs: Jobs associated with the text log errors
+        :param filters: filters to apply to text log errors"""
+        error_lines = TextLogError.objects.filter(
+            step__job__in=jobs,
+            **filters)
+        lines_by_job = defaultdict(list)
+        for item in error_lines:
+            lines_by_job[item.step.job].append(item)
+        return lines_by_job
+
+
 class TextLogError(models.Model):
     """
     A detected error line in the textual (unstructured) log
@@ -1520,6 +1539,8 @@ class TextLogError(models.Model):
     line = models.TextField()
     line_number = models.PositiveIntegerField()
 
+    objects = TextLogErrorManager()
+
     class Meta:
         db_table = "text_log_error"
         unique_together = ('step', 'line_number')
@@ -1538,6 +1559,91 @@ class TextLogError(models.Model):
         from treeherder.model import error_summary
         return error_summary.bug_suggestions_line(self)
 
+    def best_automatic_match(self, min_score=0):
+        return (TextLogErrorMatch.objects
+                .filter(text_log_error__id=self.id,
+                        score__gt=min_score)
+                .order_by("-score",
+                          "-classified_failure_id")
+                .select_related('classified_failure')
+                .first())
+
+    def set_classification(self, matcher, classification=None, bug_number=None,
+                           mark_best=False):
+        with transaction.atomic():
+            if classification is None:
+                if bug_number:
+                    classification, _ = ClassifiedFailure.objects.get_or_create(
+                        bug_number=bug_number)
+                else:
+                    classification = ClassifiedFailure.objects.create()
+
+            new_link = TextLogErrorMatch(
+                text_log_error=self,
+                classified_failure=classification,
+                matcher=matcher,
+                score=1)
+            new_link.save()
+
+            if self.metadata and self.metadata.failure_line:
+                new_link_failure = FailureMatch(
+                    failure_line=self.metadata.failure_line,
+                    classified_failure=classification,
+                    matcher=matcher,
+                    score=1)
+                new_link_failure.save()
+
+            if mark_best:
+                self.mark_best_classification(classification)
+
+        return classification, new_link
+
+    def mark_best_classification(self, classification):
+        if self.metadata is None:
+            TextLogErrorMetadata.objects.create(
+                text_log_error=self,
+                best_classification=classification)
+        else:
+            self.metadata.best_classification = classification
+            self.metadata.save(update_fields=['best_classification'])
+
+            if self.metadata.failure_line:
+                self.metadata.failure_line.best_classification = classification
+                self.metadata.failure_line.save(update_fields=['best_classification'])
+
+                self.metadata.failure_line.elastic_search_insert()
+
+    def mark_best_classification_verified(self, classification):
+        if classification not in self.classified_failures.all():
+            manual_detector = Matcher.objects.get(name="ManualDetector")
+            self.set_classification(manual_detector, classification=classification)
+
+        if self.metadata is None:
+            TextLogErrorMetadata.objects.create(text_log_error=self,
+                                                best_classification=classification,
+                                                best_is_verified=True)
+        else:
+            self.metadata.best_classification = classification
+            self.metadata.best_is_verified = True
+            self.metadata.save()
+        if self.metadata.failure_line:
+            self.metadata.failure_line.best_classification = classification
+            self.metadata.failure_line.best_is_verified = True
+            self.metadata.failure_line.save()
+            self.metadata.failure_line.elastic_search_insert()
+
+    def update_autoclassification(self):
+        """
+        If a job is manually classified and has a single line in the logs matching a single
+        TextLogError, but the TextLogError has not matched any ClassifiedFailure, add a
+        new match due to the manual classification.
+        """
+
+        manual_detector = Matcher.objects.get(name="ManualDetector")
+
+        classification, _ = self.set_classification(manual_detector)
+        self.mark_best_classification_verified(classification)
+
 
 class TextLogErrorMetadata(models.Model):
     """Optional, mutable, data that can be associated with a TextLogError."""

treeherder/webapp/api/jobs.py

@@ -261,6 +261,9 @@ class JobsViewSet(viewsets.ViewSet):
         except ObjectDoesNotExist:
             pass
 
+        status_map = {k: v for k, v in Job.AUTOCLASSIFY_STATUSES}
+        resp["autoclassify_status"] = status_map[job.autoclassify_status]
+
         return Response(resp)
 
     def list(self, request, project):
@@ -542,6 +545,11 @@ class JobsViewSet(viewsets.ViewSet):
                             status=HTTP_404_NOT_FOUND)
         textlog_errors = (TextLogError.objects
                           .filter(step__job=job)
+                          .select_related("_metadata",
+                                          "_metadata__failure_line")
+                          .prefetch_related("classified_failures",
+                                            "matches",
+                                            "matches__matcher")
                           .order_by('id'))
         return Response(serializers.TextLogErrorSerializer(textlog_errors,
                                                            many=True,

treeherder/webapp/api/serializers.py

@@ -216,11 +216,16 @@ class FailureLineNoStackSerializer(serializers.ModelSerializer):
 
 
 class TextLogErrorMetadataSerializer(serializers.ModelSerializer):
+    failure_line = FailureLineNoStackSerializer(read_only=True)
+
     class Meta:
         model = models.TextLogErrorMetadata
+        fields = '__all__'
 
 
 class TextLogErrorSerializer(serializers.ModelSerializer):
+    matches = FailureMatchSerializer(many=True)
+    classified_failures = ClassifiedFailureSerializer(many=True)
     bug_suggestions = NoOpSerializer(read_only=True)
     metadata = TextLogErrorMetadataSerializer(read_only=True)
 

treeherder/webapp/api/text_log_error.py

@@ -0,0 +1,127 @@
+import logging
+
+from django.db import transaction
+from rest_framework import viewsets
+from rest_framework.response import Response
+from rest_framework.status import (HTTP_200_OK,
+                                   HTTP_400_BAD_REQUEST,
+                                   HTTP_404_NOT_FOUND)
+
+from treeherder.model.models import (ClassifiedFailure,
+                                     TextLogError)
+from treeherder.webapp.api import (pagination,
+                                   serializers)
+from treeherder.webapp.api.utils import as_dict
+
+logger = logging.getLogger(__name__)
+
+
+class TextLogErrorViewSet(viewsets.ModelViewSet):
+    serializer_class = serializers.TextLogErrorSerializer
+    queryset = TextLogError.objects.prefetch_related("classified_failures",
+                                                     "matches",
+                                                     "matches__matcher",
+                                                     "_metadata",
+                                                     "_metadata__failure_line").all()
+    pagination_class = pagination.IdPagination
+
+    @transaction.atomic
+    def _update(self, data, user, many=True):
+        ids = []
+        error_line_ids = set()
+        classification_ids = set()
+        bug_number_classifications = {}
+
+        for item in data:
+            line_id = item.get("id")
+            if line_id is None:
+                return "No text log error id provided", HTTP_400_BAD_REQUEST
+            try:
+                line_id = int(line_id)
+            except ValueError:
+                return "Text log error id was not an integer", HTTP_400_BAD_REQUEST
+
+            error_line_ids.add(line_id)
+
+            classification_id = item.get("best_classification")
+
+            if classification_id is not None:
+                classification_ids.add(classification_id)
+
+            bug_number = item.get("bug_number")
+
+            if (not classification_id and
+                bug_number is not None and
+                bug_number not in bug_number_classifications):
+                bug_number_classifications[bug_number], _ = (
+                    ClassifiedFailure.objects.get_or_create(bug_number=bug_number))
+
+            ids.append((line_id, classification_id, bug_number))
+
+        error_lines = as_dict(
+            TextLogError.objects
+            .prefetch_related('classified_failures')
+            .filter(id__in=error_line_ids), "id")
+
+        if len(error_lines) != len(error_line_ids):
+            missing = error_line_ids - set(error_lines.keys())
+            return ("No text log error with id: {0}".format(", ".join(missing)),
+                    HTTP_404_NOT_FOUND)
+
+        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)),
+                    HTTP_404_NOT_FOUND)
+
+        for line_id, classification_id, bug_number in ids:
+            logger.debug("line_id: %s, classification_id: %s, bug_number: %s" %
+                         (line_id, classification_id, bug_number))
+            error_line = error_lines[line_id]
+            if classification_id is not None:
+                logger.debug("Using classification id")
+                classification = classifications[classification_id]
+                if bug_number is not None and bug_number != classification.bug_number:
+                    logger.debug("Updating classification bug number")
+                    classification = classification.set_bug(bug_number)
+            elif bug_number is not None:
+                logger.debug("Using bug number")
+                classification = bug_number_classifications[bug_number]
+            else:
+                logger.debug("Using null classification")
+                classification = None
+
+            error_line.mark_best_classification_verified(classification)
+            error_line.step.job.update_after_verification(user)
+
+        # Force failure line to be reloaded, including .classified_failures
+        rv = (TextLogError.objects
+              .prefetch_related('classified_failures')
+              .filter(id__in=error_line_ids))
+
+        if not many:
+            rv = rv[0]
+
+        return (serializers.TextLogErrorSerializer(rv, many=many).data,
+                HTTP_200_OK)
+
+    def update(self, request, pk=None):
+        data = {"id": pk}
+        for k, v in request.data.iteritems():
+            if k not in data:
+                data[k] = v
+
+        body, status = self._update([data], request.user, many=False)
+        return Response(body, status=status)
+
+    def update_many(self, request):
+        body, status = self._update(request.data, request.user, many=True)
+
+        if status == HTTP_404_NOT_FOUND:
+            # 404 doesn't make sense for updating many since the path is always
+            # valid, so if we get an invalid id instead return 400
+            status = HTTP_400_BAD_REQUEST
+
+        return Response(body, status=status)

treeherder/webapp/api/urls.py

@@ -18,6 +18,7 @@ from treeherder.webapp.api import (artifact,
                                    resultset,
                                    runnable_jobs,
                                    seta,
+                                   text_log_error,
                                    text_log_summary,
                                    text_log_summary_line)
 
@@ -122,6 +123,9 @@ default_router.register(r'failure-line', failureline.FailureLineViewSet,
 default_router.register(r'classified-failure',
                         classifiedfailure.ClassifiedFailureViewSet,
                         base_name='classified-failure')
+default_router.register(r'text-log-error',
+                        text_log_error.TextLogErrorViewSet,
+                        base_name='text-log-error')
 default_router.register(r'text-log-summary',
                         text_log_summary.TextLogSummaryViewSet,
                         base_name='text-log-summary')