[webkit-dev] Supporting w3c ref tests and changing our convention

Fri Nov 4 13:32:51 PDT 2011

Could you clarify why we have to need to modify DRT if we have Link Element
approach?
I guess one of the reasons is the performance. It might be better that we
use DRT to parse and extract reference links from HTML since parsing HTML
using Python might take unacceptable time and might be inaccurate.
Is there any other reasons we should modify DRT to support Link Element
approach?

On Fri, Nov 4, 2011 at 12:34 PM, Ryosuke Niwa <rniwa at webkit.org> wrote:

> Hi,
>
> There was a discussion about supporting W3C ref tests at TPAC yesterday.
> However, there appears to be some disagreement in how we support them.
>
> *Overview*
>
> CSS WG ref tests contain link elements that specify reference files. e.g.
>
> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"><html <http://december.com/html/4/element/html.html> xmlns="http://www.w3.org/1999/xhtml">
>  <head <http://december.com/html/4/element/head.html>>
>   <title <http://december.com/html/4/element/title.html>>CSS Reftest Reference</title <http://december.com/html/4/element/title.html>>
>   <link <http://december.com/html/4/element/link.html> rel="author" title="NAME_OF_AUTHOR" href="mailto:EMAIL OR http://CONTACT_PAGE"/>  <link <http://december.com/html/4/element/link.html> rel="match" href="green-box-ref.xht" />
>   <style <http://december.com/html/4/element/style.html> type="text/css"><![CDATA[   CSS FOR REFERENCE  ]]></style <http://december.com/html/4/element/style.html>>
>  </head <http://december.com/html/4/element/head.html>>
>  <body <http://december.com/html/4/element/body.html>>
>   CONTENT OF REFERENCE
>  </body <http://december.com/html/4/element/body.html>></html <http://december.com/html/4/element/html.html>>
>
>
> The highlighted line specifies that the rendered result of this page
> should be compared with that of green-box-ref.xht. This is very different
> from our current method to use filenames such as -match.html,
> -mismatch.html, etc.. to identify reference files.
>
> In addition, W3C test suites have a build step, which generates
> Mozilla-like test manifest files. This file will contain entries that
> identifies tests and corresponding reference files. e.g.
>
>
> == floats-wrap-bfc-outside-001.xht floats-wrap-bfc-outside-001-ref.xht
> == floats-wrap-top-below-bfc-001l.xht floats-wrap-top-below-001l-ref.xht
> != floats-wrap-top-below-bfc-001l.xht floats-wrap-top-below-001l-notref.xht
> == floats-wrap-top-below-bfc-001r.xht floats-wrap-top-below-001r-ref.xht
> != floats-wrap-top-below-bfc-001r.xht floats-wrap-top-below-001r-notref.xht
>
> All browser vendors who attended the meeting preferred this format.
> Microsoft representatives didn't attend the meeting.
>
> *Link element approach*
> Pros:
>
>    - Can reuse same ref. file for multiple tests
>    - Can have multiple ref. files for single test
>    - Information is self-contained in the test file
>    - We may get away with test suite build step
>
> (It turns out that we can't convert W3C ref tests to use WebKit
> conventions due to the first two points.)
>
> Cons:
>
>    - Requires us modifying each port's DRT to support this format
>    - Adding link elements itself may affect tests (all W3C tests are
>    required to have link elements at the moment)
>    - Hard to understand relationship between files. e.g. if we want to
>    figure out which tests use ref.html, we must look at all test files
>    - Other browser vendors (Firefox and Opera) prefer manifest file
>    approach
>
>
> *Manifest file approach*
> Pros:
>
>    - Easy to parse, and do not require us modifiying each port's DRT to
>    support
>    - Easy to tell relationships between files at glance
>    - Do not affect tests since manifest files are external to tests
>    - Preferred approach of Firefox and Opera
>
> Cons:
>
>    - Have to maintain information in two places, tests and manifests (W3C
>    test suites auto-generates manifest files in their build step)
>    - Require build step (new W3C tests only have link elements but not
>    manifests)
>
>
> I strongly prefer WebKit use the manifest file approach and deprecate our
> current file-name convention because it won't burden each port, and it's
> the preferred method of other browser vendors.  Any opinions?
> *
> *
> Best,
> Ryosuke Niwa
> Software Engineer
> Google Inc.
>
>
>
> _______________________________________________
> webkit-dev mailing list
> webkit-dev at lists.webkit.org
> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>
>

-- 
Hayato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-dev/attachments/20111104/94916436/attachment.html>