summaryrefslogtreecommitdiffstats
path: root/Tools/Scripts/webkitpy/tool/bot/commitqueuetask.py
diff options
context:
space:
mode:
Diffstat (limited to 'Tools/Scripts/webkitpy/tool/bot/commitqueuetask.py')
-rw-r--r--Tools/Scripts/webkitpy/tool/bot/commitqueuetask.py220
1 files changed, 220 insertions, 0 deletions
diff --git a/Tools/Scripts/webkitpy/tool/bot/commitqueuetask.py b/Tools/Scripts/webkitpy/tool/bot/commitqueuetask.py
new file mode 100644
index 0000000..1d82ea8
--- /dev/null
+++ b/Tools/Scripts/webkitpy/tool/bot/commitqueuetask.py
@@ -0,0 +1,220 @@
+# Copyright (c) 2010 Google Inc. All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are
+# met:
+#
+# * Redistributions of source code must retain the above copyright
+# notice, this list of conditions and the following disclaimer.
+# * Redistributions in binary form must reproduce the above
+# copyright notice, this list of conditions and the following disclaimer
+# in the documentation and/or other materials provided with the
+# distribution.
+# * Neither the name of Google Inc. nor the names of its
+# contributors may be used to endorse or promote products derived from
+# this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+# "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+# LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+# A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+# LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+# DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+# THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+from webkitpy.common.system.executive import ScriptError
+from webkitpy.common.net.layouttestresults import LayoutTestResults
+
+
+class CommitQueueTaskDelegate(object):
+ def run_command(self, command):
+ raise NotImplementedError("subclasses must implement")
+
+ def command_passed(self, message, patch):
+ raise NotImplementedError("subclasses must implement")
+
+ def command_failed(self, message, script_error, patch):
+ raise NotImplementedError("subclasses must implement")
+
+ def refetch_patch(self, patch):
+ raise NotImplementedError("subclasses must implement")
+
+ def layout_test_results(self):
+ raise NotImplementedError("subclasses must implement")
+
+ def report_flaky_tests(self, patch, flaky_tests):
+ raise NotImplementedError("subclasses must implement")
+
+
+class CommitQueueTask(object):
+ def __init__(self, delegate, patch):
+ self._delegate = delegate
+ self._patch = patch
+ self._script_error = None
+
+ def _validate(self):
+ # Bugs might get closed, or patches might be obsoleted or r-'d while the
+ # commit-queue is processing.
+ self._patch = self._delegate.refetch_patch(self._patch)
+ if self._patch.is_obsolete():
+ return False
+ if self._patch.bug().is_closed():
+ return False
+ if not self._patch.committer():
+ return False
+ # Reviewer is not required. Missing reviewers will be caught during
+ # the ChangeLog check during landing.
+ return True
+
+ def _run_command(self, command, success_message, failure_message):
+ try:
+ self._delegate.run_command(command)
+ self._delegate.command_passed(success_message, patch=self._patch)
+ return True
+ except ScriptError, e:
+ self._script_error = e
+ self.failure_status_id = self._delegate.command_failed(failure_message, script_error=self._script_error, patch=self._patch)
+ return False
+
+ def _clean(self):
+ return self._run_command([
+ "clean",
+ ],
+ "Cleaned working directory",
+ "Unable to clean working directory")
+
+ def _update(self):
+ # FIXME: Ideally the status server log message should include which revision we updated to.
+ return self._run_command([
+ "update",
+ ],
+ "Updated working directory",
+ "Unable to update working directory")
+
+ def _apply(self):
+ return self._run_command([
+ "apply-attachment",
+ "--no-update",
+ "--non-interactive",
+ self._patch.id(),
+ ],
+ "Applied patch",
+ "Patch does not apply")
+
+ def _build(self):
+ return self._run_command([
+ "build",
+ "--no-clean",
+ "--no-update",
+ "--build-style=both",
+ ],
+ "Built patch",
+ "Patch does not build")
+
+ def _build_without_patch(self):
+ return self._run_command([
+ "build",
+ "--force-clean",
+ "--no-update",
+ "--build-style=both",
+ ],
+ "Able to build without patch",
+ "Unable to build without patch")
+
+ def _test(self):
+ return self._run_command([
+ "build-and-test",
+ "--no-clean",
+ "--no-update",
+ # Notice that we don't pass --build, which means we won't build!
+ "--test",
+ "--non-interactive",
+ ],
+ "Passed tests",
+ "Patch does not pass tests")
+
+ def _build_and_test_without_patch(self):
+ return self._run_command([
+ "build-and-test",
+ "--force-clean",
+ "--no-update",
+ "--build",
+ "--test",
+ "--non-interactive",
+ ],
+ "Able to pass tests without patch",
+ "Unable to pass tests without patch (tree is red?)")
+
+ def _failing_tests_from_last_run(self):
+ results = self._delegate.layout_test_results()
+ if not results:
+ return None
+ return results.failing_tests()
+
+ def _land(self):
+ # Unclear if this should pass --quiet or not. If --parent-command always does the reporting, then it should.
+ return self._run_command([
+ "land-attachment",
+ "--force-clean",
+ "--ignore-builders",
+ "--non-interactive",
+ "--parent-command=commit-queue",
+ self._patch.id(),
+ ],
+ "Landed patch",
+ "Unable to land patch")
+
+ def _report_flaky_tests(self, flaky_tests):
+ self._delegate.report_flaky_tests(self._patch, flaky_tests)
+
+ def _test_patch(self):
+ if self._patch.is_rollout():
+ return True
+ if self._test():
+ return True
+
+ first_failing_tests = self._failing_tests_from_last_run()
+ if self._test():
+ self._report_flaky_tests(first_failing_tests)
+ return True
+
+ second_failing_tests = self._failing_tests_from_last_run()
+ if first_failing_tests != second_failing_tests:
+ # We could report flaky tests here, but since run-webkit-tests
+ # is run with --exit-after-N-failures=1, we would need to
+ # be careful not to report constant failures as flaky due to earlier
+ # flaky test making them not fail (no results) in one of the runs.
+ # See https://bugs.webkit.org/show_bug.cgi?id=51272
+ return False
+
+ if self._build_and_test_without_patch():
+ raise self._script_error # The error from the previous ._test() run is real, report it.
+ return False # Tree must be red, just retry later.
+
+ def run(self):
+ if not self._validate():
+ return False
+ if not self._clean():
+ return False
+ if not self._update():
+ return False
+ if not self._apply():
+ raise self._script_error
+ if not self._build():
+ if not self._build_without_patch():
+ return False
+ raise self._script_error
+ if not self._test_patch():
+ return False
+ # Make sure the patch is still valid before landing (e.g., make sure
+ # no one has set commit-queue- since we started working on the patch.)
+ if not self._validate():
+ return False
+ # FIXME: We should understand why the land failure occured and retry if possible.
+ if not self._land():
+ raise self._script_error
+ return True