run-webkit-tests run-webkit-tests for the second time gives error - wptserve not killed after run |
||||
Issue description
When I run the
./third_party/WebKit/Tools/Scripts/run-webkit-tests external/wpt/[any subfolder]
for the second time I get this error:
Starting WPTServe ...
ServerError raised: Port 8001 is already in use.
Traceback (most recent call last):
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/run_webkit_tests.py", line 67, in main
return run(port, options, args, stderr, stdout).exit_code
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/run_webkit_tests.py", line 580, in run
run_details = _run_tests(port, options, args, printer)
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/run_webkit_tests.py", line 571, in _run_tests
return manager.run(args)
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/controllers/manager.py", line 127, in run
self._start_servers(tests_to_run)
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/controllers/manager.py", line 365, in _start_servers
self._port.start_wptserve()
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/port/base.py", line 1172, in start_wptserve
server.start()
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/servers/server_base.py", line 100, in start
self._check_that_all_ports_are_available()
File "/usr/local/google/home/nzolghadr/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/servers/server_base.py", line 276, in _check_that_all_ports_are_available
raise ServerError('Port %d is already in use.' % port)
ServerError: Port 8001 is already in use.
The only workaround is to kill the process manually.
,
Jan 19 2017
I've reproduced this too, and I also think that wptserve is not killed properly sometimes. After killing the wptserve processes manually, running for a second time works again.
,
Jan 19 2017
Caused by http://crrev.com/2624093003, but if possible I'd like to avoid reverting that CL, since then we may have to disable wptserve again.
,
Jan 19 2017
Just confirmed locally that that undoing http://crrev.com/2624093003 does solve this on my workstation.
,
Jan 20 2017
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/896b0f4ea756385f66de7b31c52fccf728a02cb6 commit 896b0f4ea756385f66de7b31c52fccf728a02cb6 Author: qyearsley <qyearsley@chromium.org> Date: Fri Jan 20 01:46:13 2017 When killing wptserve, use Executive.interrupt on non-Windows platforms. Before http://crrev.com/2624093003, the test runner would hang in a busy loop while trying to kill wptserve. After that CL, though, when running run-webkit-tests locally on Linux, wptserve wouldn't be effectively killed before the next run. This CL is meant to make it so that it behaves more like it did before (using interrupt instead of kill for non-Windows platforms), but still only tries to kill wptserve for 20 seconds, with 1-second sleeps between attempts. I tried locally to see what would happen if I undid the change in http://crrev.com/2624093003 but only changed _stop_running_server to use kill_process on all platforms (not just Windows), and confirmed that this apparently doesn't kill the process properly on my workstation. BUG= 669802 , 682711 Review-Url: https://codereview.chromium.org/2649433002 Cr-Commit-Position: refs/heads/master@{#444933} [modify] https://crrev.com/896b0f4ea756385f66de7b31c52fccf728a02cb6/third_party/WebKit/Tools/Scripts/webkitpy/common/system/executive_mock.py [modify] https://crrev.com/896b0f4ea756385f66de7b31c52fccf728a02cb6/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/servers/wptserve.py
,
Jan 20 2017
This should now be fixed.
,
Jan 23 2017
Nice, thanks for keeping wptserve enabled! |
||||
►
Sign in to add a comment |
||||
Comment 1 by nzolghadr@chromium.org
, Jan 19 2017