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@lists.webkit.org http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
