[Webkit-unassigned] [Bug 167626] New: run-perf-tests timeout settings broken

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Mon Jan 30 19:31:40 PST 2017


https://bugs.webkit.org/show_bug.cgi?id=167626

            Bug ID: 167626
           Summary: run-perf-tests timeout settings broken
    Classification: Unclassified
           Product: WebKit
           Version: WebKit Nightly Build
          Hardware: Unspecified
                OS: Unspecified
            Status: NEW
          Severity: Normal
          Priority: P2
         Component: Tools / Tests
          Assignee: webkit-unassigned at lists.webkit.org
          Reporter: clopez at igalia.com
                CC: bugs-noreply at webkitgtk.org, lforschler at apple.com,
                    ossy at webkit.org, rniwa at webkit.org

Not sure when this started happening, but it seems that the timeout feature of run-perf-tests is broken.

The Script run-perf-tests accepts a parameter --time-out-ms with default value of 600 seconds (600 * 1000).

But this parameter is completely ignored because in webkitpy/performance_tests/perftest.py the driver is created with no_timeout=True unconditionally.

def _create_driver(self):
   return self._port.create_driver(worker_number=0, no_timeout=True)

And this causes that the tests run without any timeout deadline: See Driver.run_test() at webkitpy/port/driver.py when self._no_timeout is true.


So, I think we should:

 * Make the --time-out-ms parameter work back as expected.
 * Add a --no-timeout parameter for testing purposes.
 * Keep the default 600 second time out value for for each performance (sub)test.


This means that:

* Some tests on the perf bots may break if we do what I propose, as they are running now without any timeout value. But I think this tests should be fixed to not take so much time to run. 
   *  There is a general 1200 second timeout value for the whole perf-test step (command timed out: 1200 seconds without output), but when this triggers it kills the whole step rather than stopping only the running test.
   *  If some port (Apple/EFL/GTK+) has any reason to want different defaults, they can override this for their own Perf bots in the master.cfg file at Tools/BuildSlaveSupport/build.webkit.org-config/

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20170131/47b87718/attachment.html>


More information about the webkit-unassigned mailing list