new-run-webkit-tests on Windows

52 views
Skip to first unread message

Kenneth Russell

unread,
Oct 14, 2010, 9:43:58 PM10/14/10
to Chromium-dev
I'm trying to run new-run-webkit-tests on Windows (Vista 64-bit) and
finding that it always reports "thread Thread-1 (...) is wedged" and
hangs. It's running within Python 2.5.2 in Cygwin. I've tried the
following configurations:

new-run-webkit-tests --chromium --platform chromium-win-vista --use-drt --debug
new-run-webkit-tests --chromium --platform chromium-gpu --use-drt --debug

Running the built DumpRenderTree.exe against a few individual test
cases seems to work.

I followed all of the setup instructions on
http://www.chromium.org/developers/testing/webkit-layout-tests . Is
anything else necessary?

Thanks,

-Ken

Mihai Parparita

unread,
Oct 14, 2010, 9:50:15 PM10/14/10
to k...@chromium.org, Chromium-dev
Have you tried using --child_processes=1? NRWT has deadlock problems
on the Mac (https://bugs.webkit.org/show_bug.cgi?id=38553), perhaps
you're also experiencing that.

Mihai

> --
> Chromium Developers mailing list: chromi...@chromium.org
> View archives, change email options, or unsubscribe:
>    http://groups.google.com/a/chromium.org/group/chromium-dev
>

Dirk Pranke

unread,
Oct 15, 2010, 1:16:14 AM10/15/10
to k...@chromium.org, mih...@chromium.org, Chromium-dev
Mihai's suggestion should be a rock-solid workaround.

I would also recommend you try to run it in from a dos prompt; cygwin
seems to aggravate the threading issues. Although I haven't actually
tried the DRT version from the command prompt, I think it works (I
think that's what the bot does).

If you send me the log file from running with --verbose, I can see if
there's anything else weird going on (and/or send me the stack trace
you get from the wedged thread).
Although the wedging issues are sadly over-frequent, I have yet to see
a completely reproduceable test case, so it would be interesting if
you had one.

-- Dirk

Kenneth Russell

unread,
Oct 18, 2010, 6:26:34 PM10/18/10
to Dirk Pranke, mih...@chromium.org, Chromium-dev
Thanks for the tips. Things seem to be working with the following
changes from my previous configuration:

- Running with a Python 2.7 from http://www.python.org/ rather than
Cygwin's version of Python. The cygwin-style paths seem to be what was
causing the harness to fail to find pixel test results, think that DRT
had crashed and get wedged. I put Cygwin last in my PATH only to pick
up Ruby.
- Running from the command prompt.

I'd like to put some of these suggestions somewhere permanent. Since
http://www.chromium.org/developers/testing/webkit-layout-tests still
mostly talks about test_shell, should I add a section to the end on
new-run-webkit-tests?

-Ken

Kenneth Russell

unread,
Oct 18, 2010, 6:44:29 PM10/18/10
to Dirk Pranke, mih...@chromium.org, Chromium-dev
I went ahead and added a new section to the bottom of
http://www.chromium.org/developers/testing/webkit-layout-tests .

Dirk Pranke

unread,
Oct 18, 2010, 6:50:38 PM10/18/10
to Kenneth Russell, mih...@chromium.org, Chromium-dev
On Mon, Oct 18, 2010 at 3:26 PM, Kenneth Russell <k...@chromium.org> wrote:
> Thanks for the tips. Things seem to be working with the following
> changes from my previous configuration:
>
>  - Running with a Python 2.7 from http://www.python.org/ rather than
> Cygwin's version of Python. The cygwin-style paths seem to be what was
> causing the harness to fail to find pixel test results, think that DRT
> had crashed and get wedged. I put Cygwin last in my PATH only to pick
> up Ruby.
>  - Running from the command prompt.
>

I am about to check in a fix that should make everything work with the
cygwin Python version from a bash prompt, as well. However, it is (at
least for me) notably slower.

> I'd like to put some of these suggestions somewhere permanent. Since
> http://www.chromium.org/developers/testing/webkit-layout-tests still
> mostly talks about test_shell, should I add a section to the end on
> new-run-webkit-tests?
>

Sure, that would be great --- oops, you've already done it :)

-- Dirk

John Abd-El-Malek

unread,
Oct 20, 2010, 2:48:43 AM10/20/10
to dpr...@google.com, Kenneth Russell, mih...@chromium.org, Chromium-dev
I just synced and I can't run run-webkit-tests under cygwin.  Any other workarounds?

jabdelmalek@JABDELMALEK0-W /cygdrive/a/chrome2/src/webkit/tools/layout_tests
$ ./run_webkit_tests.sh plugins/mouse-move-over-pluginn-in-frame.html
A:\chrome2\src\webkit\tools\layout_tests\run_webkit_tests.py
/usr/bin/python: can't open file 'third_party/WebKit/WebKitTools/Scripts/new-run-webkit-tests': [Errno 2] No such file or directory

Dirk Pranke

unread,
Oct 20, 2010, 10:54:22 PM10/20/10
to John Abd-El-Malek, Kenneth Russell, mih...@chromium.org, Chromium-dev
This works fine if you run it from the src directory. That said, I did
break something and I've uploaded a patch ...

-- Dirk

Michael Nordman

unread,
Nov 10, 2010, 8:05:00 PM11/10/10
to Chromium-dev, John Abd-El-Malek, Kenneth Russell, mih...@chromium.org, dpr...@google.com
Dirk walked me thru making the following edits to executive.py to get
the http tests working too...

> src/third_party/WebKit/WebKitTools/Scripts/webkitpy/common/system

$ svn diff
Index: executive.py
===================================================================
--- executive.py (revision 71759)
+++ executive.py (working copy)
@@ -201,7 +201,7 @@
return
raise

- def _win32_check_running_pid(self):
+ def _win32_check_running_pid(self, pid):

class PROCESSENTRY32(ctypes.Structure):
_fields_ = [("dwSize", ctypes.c_ulong),
@@ -246,7 +246,7 @@
except OSError:
return False
elif sys.platform == 'win32':
- return self._win32_check_running_pid()
+ return self._win32_check_running_pid(pid)

assert(False)

Dirk Pranke

unread,
Nov 11, 2010, 1:46:24 AM11/11/10
to Michael Nordman, Chromium-dev, John Abd-El-Malek, Kenneth Russell, mih...@chromium.org
Thanks for sending this out. I have a patch up for review to fix this
in the tree, but in the meantime it should help anyone who is current
stuck.

The previous brokenness was fixed a couple weeks ago, shortly after it
was filed.

-- Dirk

Reply all
Reply to author
Forward
0 new messages