[webkit-dev] 答复: 答复: About the layout test result.

[Eric Seidel]

My earlier statement was inaccurate.  It is not the *port* which
requires the fonts, but rather the tests themselves (many of which
come from the W3C or other publicly available test suites).  The tests
make specific request for fonts (like Ahem, Times, Verdana, Helvetica)
which are not provided by default on windows systems.  To mitigate
this, folks set the WEBKIT_TESTFONTS environment variable and point it
at a directory containing windows versions of the required fonts.  As
I pointed out in my previous mail, it is possible to find windows
versions of these fonts on the interwebs.  Although whether or not you
have license to use these fonts is for you to determine.  IANAL. :)

Safari for Windows ships with a certain set of licensed fonts (which
is not a fully overlapping set with the set of fonts used/required-by
the Layout Tests), many of these fonts Apple bundles with Safari for
Windows ship as part of Mac OS X and happen to be some of the same
fonts used by these test suites.

My complaint earlier was that we don't have any results checked in to
webkit.org's svn repository which use the "default" windows set of
fonts.  I now realize that that's such results wouldn't necessarily be
very useful.  If you don't have the required fonts for the test suites
on your machine, that means webkit would use whatever fallback fonts
are specified in the tests.  That would result in less-repeatable test
results, since if you happened to have some of the "required" fonts
installed, you'd get results which neither match the Apple-WebKit-Win
results or some theoretical
Apple-WebKit-using-only-windows-default-fonts results.

My apologies for any confusion my statements may have caused.

-eric

2008/10/7 Eric Seidel <eric at webkit.org>:
> If you search the web for WEBKIT_TESTFONTS you may be able to find a
> copy of the fonts you need.
>
> It's kinda lame that the current set of windows layout tests requires
> these fonts.  Eventually I would like to see a set of results checked
> in which do not require Apple-provided fonts, but right now there is
> no WebKit windows port shipping out of webkit.org which does not use
> those Apple fonts.
>
> -eric
>
> 2008/10/7 Dan Bernstein <mitz at apple.com>:
>>
>> On Oct 6, 2008, at 10:56 PM, Bely wrote:
>>
>> Thanks Dan!
>>
>> But what fonts should I install on windows to run the tests and where can I
>> get them?
>>
>> The initialize() function in
>> WebKitTools/DumpRenderTree/win/DumpRenderTree.cpp lists all of them,
>> alongside other fonts (Ahem and the WebKit* fonts) that are in the WebKit
>> source tree. The Times family alone would probably suffice for the majority
>> of the tests. I do not know how or even whether any or all of the required
>> fonts are available for installation on Windows.
>> -- Dan
>>
>>
>> Thanks!
>>
>>
>> -- Bely
>> ________________________________
>> 发件人: Dan Bernstein [mailto:mitz at apple.com]
>> 发送时间: 2008年10月7日 11:36
>> 收件人: Bely
>> 抄送: 'Mark Rowe'; webkit-dev at lists.webkit.org
>> 主题: Re: [webkit-dev] 答复: About the layout test result.
>>
>>
>> On Oct 6, 2008, at 7:55 PM, Bely wrote:
>>
>>
>>
>> Hi Mark,
>>
>>     Thanks for your reply!
>>
>>
>>
>> I am running the tests on windows, with cygwin installed. After build it in
>> VC8 using "debug"
>>
>> configuration, I execute the tests by invoking run-webkit-tests in cygwin
>> console like:
>>
>>
>>
>> ./run-webkit-tests
>>
>>
>>
>> First time it says WebKitGUID.lib is missing, I rename WebKitGUID_debug.lib
>> to WebKitGUID.lib,
>>
>> And later it says WTF_debug.lib is missing, again I rename WTF_debug.lib to
>> WTF.lib, then I began to run the tests and
>>
>> Got the result.
>>
>>
>>
>> Is there anything wrong with my process?
>>
>> Hi Bely,
>>
>> Many layout tests' results are dependent on the fonts that are available on
>> the system. When you run the tests on Windows, the results for most tests
>> are compared against the expected results in LayoutTests/platform/mac, so it
>> is expected that the results will mismatch, because a typical installation
>> of Windows is missing many fonts that are available on Mac OS X.
>>
>> Thanks,
>> -- Dan
>>
>>
>>
>>
>> Thanks!
>>
>>
>>
>> - Bely
>>
>>
>>
>> -----邮件原件-----
>> 发件人: Mark Rowe [mailto:mrowe at apple.com]
>> 发送时间: 2008年10月7日 10:07
>> 收件人: Bely
>> 抄送: webkit-dev at lists.webkit.org
>> 主题: Re: [webkit-dev] About the layout test result.
>>
>>
>>
>>
>>
>> On 2008-10-05, at 23:16, Bely wrote:
>>
>>
>>
>>> Hi all,
>>
>>>          I am new to webkit. Recently I run the layout test and got
>>
>>> the following result :
>>
>>>          [table]
>>
>>> Test Feature
>>
>>> Total
>>
>>> passed/ incorrect/timedout/crashed/stderr
>>
>>> Pass rate
>>
>>> Layout Testing
>>
>>> 9085
>>
>>> 4853
>>
>>> 4198
>>
>>> 15
>>
>>> 5
>>
>>> 14
>>
>>> 53.4%
>>
>>>
>>
>>>
>>
>>>
>>
>>>
>>
>>>
>>
>>>
>>
>>>
>>
>>>
>>
>>>          [text]
>>
>>>          Total:        9085
>>
>>>          Passed: 4853
>>
>>>          Incorrect:  4198
>>
>>>          Timedout:          15
>>
>>>          Crashed:  5
>>
>>>          Stderr:       14
>>
>>>
>>
>>> Why the pass rate is so low?  Is this result reasonable? Or I
>>
>>> wrongly executed the test?  Can someone help me?
>>
>>
>>
>> Which platform did you run tests on?  Which port of WebKit did you
>>
>> use?  The buildbot at <http://build.webkit.org/waterfall> shows that
>>
>> WebKit on Mac is currently passing 100% of the regression tests.
>>
>> Results for ports other than Mac and Windows are likely to be
>>
>> significantly worse as the developers of other ports have not spent
>>
>> much time focussing on passing the layout tests.
>>
>>
>>
>> - Mark
>>
>> _______________________________________________
>> webkit-dev mailing list
>> webkit-dev at lists.webkit.org
>> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>>
>>
>>
>> _______________________________________________
>> webkit-dev mailing list
>> webkit-dev at lists.webkit.org
>> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>>
>>
>