Error auto-installing the irclib.py package while running ui_tests

[Paweł Hajdan, Jr.]

I was running ui_tests locally on Mac OS X 10.6, and got the following error message for AppCacheUITest.FLAKY_AppCacheLayoutTests_NoPHP:

Traceback (most recent call last):

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/new-run-webkit-httpd", line 49, in <module>

    import port

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/port/__init__.py", line 34, in <module>

    from test import unit_test_filesystem

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/port/test.py", line 36, in <module>

    from webkitpy.tool import mocktool

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/tool/mocktool.py", line 35, in <module>

    from webkitpy.common.net.bugzilla import Bug, Attachment

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/common/net/bugzilla/__init__.py", line 5, in <module>

    from .bugzilla import Bugzilla, parse_bug_id

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/common/net/bugzilla/bugzilla.py", line 48, in <module>

    from webkitpy.thirdparty.autoinstalled.mechanize import Browser

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/thirdparty/__init__.py", line 85, in <module>

    installer.install(url="http://downloads.sourceforge.net/project/python-irclib/python-irclib/0.4.8/python-irclib-0.4.8.zip", url_subpath="irclib.py")

  File "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/common/system/autoinstall.py", line 482, in install

    raise Exception(message)

Exception: Error auto-installing the irclib.py package to:

 "/Users/ph/chromium-svn/src/third_party/WebKit/Tools/Scripts/webkitpy/thirdparty/autoinstalled/irc/irclib.py"

 --> Inner message: HTTP Error code 502

It seems bad that the test infrastructure tries to autoinstall something. Also, the following error messages are very similar to what we can see on the flakiness dashboard. Do you know why that happens and what can be done to fix it?

[41239:263:1578219673430975:WARNING:render_view.cc(2182)] didStartLoading called while loading

/Users/ph/chromium-svn/src/chrome/test/ui/ui_test.cc:780: Failure

Failed

Timeout reached in WaitUntilCookieNonEmpty

Google Test trace:

/Users/ph/chromium-svn/src/chrome/test/ui/ui_layout_test.cc:179: 404-manifest.html

/Users/ph/chromium-svn/src/chrome/test/ui/ui_layout_test.cc:256: Failure

Value of: value.c_str()

  Actual: "

"

Expected: expected_result_value.c_str()

Which is: "Test that subresources can be loaded if manifest is not available. Should say SUCCESS.

SUCCESS

"

Google Test trace:

/Users/ph/chromium-svn/src/chrome/test/ui/ui_layout_test.cc:179: 404-manifest.html

[41239:263:1578235287739982:WARNING:render_view.cc(2182)] didStartLoading called while loading

/Users/ph/chromium-svn/src/chrome/test/ui/ui_test.cc:780: Failure

Failed

Timeout reached in WaitUntilCookieNonEmpty

Google Test trace:

/Users/ph/chromium-svn/src/chrome/test/ui/ui_layout_test.cc:179: 404-resource.html

/Users/ph/chromium-svn/src/chrome/test/ui/ui_layout_test.cc:256: Failure

Value of: value.c_str()

  Actual: "

"

Expected: expected_result_value.c_str()

Which is: "Test that a missing explicit resource makes caching fail.

Should say SUCCESS:

SUCCESS

"

[Tony Chang]

Can we lazy install these libs?  new-run-webkit-httpd doesn't need irclib.py.

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

[Nicolas Sylvain]

Last week we had a bunch of bots turn red because of that too.  All our machines had to do a full checkouts multiple times in a row (bad DEPS change), and every time they had to re-download all those packages. I think our 200+ machines doing downloads after downloads on sourceforge was too much and we started seeing a lot of errors.

Can someone at webkit be convinced that doing that is evil ?

(Note: this is not only a problem for the bots running webkit tests. Since the flakiness dashboard depends on webkitpy, every gtest-based tests running on our buildbot end up calling into webkitpy and downloading all those dependencies we don't need)

Nicolas

[Adam Barth]

We should be able to download them lazily. The problem is that the
licenses aren't compatible with WebKit, so they can't be checked in as
such. We can host them on a more reliable server (e.g.,
code.google.com) if that would help.

Adam

[Nicolas Sylvain]

On Tue, Feb 1, 2011 at 1:54 PM, Adam Barth <aba...@chromium.org> wrote:

We should be able to download them lazily.  The problem is that the
licenses aren't compatible with WebKit, so they can't be checked in as
such.  We can host them on a more reliable server (e.g.,
code.google.com) if that would help.

I think this is a good step forward.   code.google.com is not more reliable, but at least if all dependencies are

all at the same place it would help reduce our chances for failures. And if it's downloaded lazily on top of that, then

this should make it a lot less of a problem for us.

Thanks

Nicolas