Bug 85459 – REGRESSION: run-webkit-tests failing on Chromium Mac

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Ryosuke Niwa]

http://build.webkit.org/builders/Chromium%20Mac%20Release%20%28Tests%29/builds/16131/steps/layout-test/logs/stdio
  layout_tests/controllers/manager_worker_broker.py:290 (in run)
    self._worker_connection.run_message_loop()
  layout_tests/controllers/manager_worker_broker.py:243 (in run_message_loop)
    self._broker.run_message_loop(self._run_topic, self._client, delay_secs)
  layout_tests/controllers/manager_worker_broker.py:175 (in run_message_loop)
    self._run_loop(topic_name, client, block=True, delay_secs=delay_secs)
  layout_tests/controllers/manager_worker_broker.py:189 (in _run_loop)
    self._dispatch_message(msg, client)
  layout_tests/controllers/manager_worker_broker.py:198 (in _dispatch_message)
    message_handler(message.src, *optargs)
  layout_tests/controllers/worker.py:144 (in handle_test_list)
    self._run_test(test_input)
  layout_tests/controllers/worker.py:171 (in _run_test)
    result = self.run_test_with_timeout(test_input, test_timeout_sec)
  layout_tests/controllers/worker.py:213 (in run_test_with_timeout)
    return self._run_test_in_this_thread(test_input)
  layout_tests/controllers/worker.py:298 (in _run_test_in_this_thread)
    return self.run_single_test(self._driver, test_input)
  layout_tests/controllers/worker.py:302 (in run_single_test)
    test_input, driver, self._name)
  layout_tests/controllers/single_test_runner.py:46 (in run_single_test)
    return runner.run()
  layout_tests/controllers/single_test_runner.py:104 (in run)
    return self._run_compare_test()
  layout_tests/controllers/single_test_runner.py:107 (in _run_compare_test)
    driver_output = self._driver.run_test(self._driver_input())
  layout_tests/port/driver.py:217 (in run_test)
    return self._running_drivers[cmd_line_key].run_test(driver_input)
  layout_tests/port/chromium.py:542 (in run_test)
    return super(ChromiumDriver, self).run_test(driver_input)
  layout_tests/port/webkit.py:557 (in run_test)
    text, audio = self._read_first_block(deadline)  # First block is either text or audio
  layout_tests/port/webkit.py:534 (in _read_first_block)
    block = self._read_block(deadline)
  layout_tests/port/webkit.py:622 (in _read_block)
    out_line, err_line = self._server_process.read_either_stdout_or_stderr_line(deadline)
  layout_tests/port/server_process.py:173 (in read_either_stdout_or_stderr_line)
    return_value = self._read(deadline, retrieve_bytes_from_buffers)
  layout_tests/port/server_process.py:282 (in _read)
    self._wait_for_data_and_update_buffers_using_select(deadline)
  layout_tests/port/server_process.py:216 (in _wait_for_data_and_update_buffers_using_select)
    read_fds, _, _ = select.select(select_fds, [], select_fds, deadline - time.time())
Stopping HTTP server ...Traceback (most recent call last):
  File "/System/Library/Frameworks/Python.framework/Versions/2.6/lib/python2.6/multiprocessing/queues.py", line 242, in _feed
    send(obj)
IOError: [Errno 32] Broken pipe
                        Stopping WebSocket server ...                             Releasing server lock ...Failed to execute Tools/Scripts/new-run-webkit-tests at ./Tools/Scripts/run-webkit-tests line 122.
program finished with exit code 254

It seems like stdout is all messed up too.

[Dirk Pranke]

looking now ...

by "stdout is all messed up too", I assume you're referring to the stuff scrolling off to the right with the progress meter not working? Or is there something else?

[Ryosuke Niwa]

(In reply to comment #1)
> looking now ...
> 
> by "stdout is all messed up too", I assume you're referring to the stuff scrolling off to the right with the progress meter not working? Or is there something else?

Right, that. I'm not even sure why we're trying to use progress meter on the bot.

[Dirk Pranke]

(In reply to comment #2)
> (In reply to comment #1)
> > looking now ...
> > 
> > by "stdout is all messed up too", I assume you're referring to the stuff scrolling off to the right with the progress meter not working? Or is there something else?
> 
> Right, that. I'm not even sure why we're trying to use progress meter on the bot.

We're not supposed to be :).

NRWT has three modes of output - normal interactive/console mode, --verbose mode, and a hybrid where we're not running with --verbose but not talking to a console/tty either. In the latter case, we're supposed to log a progress update every ten seconds or so, so that the log isn't completely silent for minutes at a time if everything is passing. Looks like that logic is probably busted, but that's not (I don't think) the cause of the crash. I'll look at that as a separate issue once the crash is resolved (unless it turns out to be related, of course).

[Dirk Pranke]

Created attachment 140134 [details]
Patch

[Dirk Pranke]

hopefully fixed in http://trac.webkit.org/changeset/116050 .