- From: Philip Taylor <pjt47@cam.ac.uk>
- Date: Mon, 01 Mar 2010 19:10:48 +0000
- To: Kris Krueger <krisk@microsoft.com>
- CC: "'public-html-testsuite@w3.org'" <public-html-testsuite@w3.org>
Kris Krueger wrote: > [...] > Maybe you can review some of the cases I have submitted recently as well? > > -Thx I assume these are the ones inside http://dev.w3.org/cvsweb/html5/tests/submission/Microsoft/ ? Looking at windowobject/security_location_0.htm: The test itself seems to be correct and matches the spec. Using "http://www.w3.org/" as the iframe src looks dangerous, because these test cases might get hosted on that domain, in which case it will no longer be testing cross-origin as intended. Perhaps "http://example.com/" and "http://example.org/" would be suitable vendor-neutral pages to use instead. Ideally there would be a W3C-managed domain (that is not www.w3.org or dev.w3.org) for cross-origin tests - does any such thing exist yet? It's slightly confusing that the page says "FAIL" (with no explanation why) while it's still busy loading. Perhaps the description should say "Test passes if the word PASS appears below once the page has finished loading". It may also be better for testresult to say "FAIL (script did not start)", and change it to "FAIL (script did not complete)" at the beginning of RunTest, and change it to "FAIL" or "PASS" at the end of RunTest, so that it's easier to tell what stage it has reached. I think it'd be useful for <link rel="help"> to link to the closest fragment id in the spec (rather than just the page), which in this case is #security-2 (but that looks like an unstable autogenerated id - http://www.w3.org/Bugs/Public/show_bug.cgi?id=9173). Looking briefly at htmldom/ and comparing with the appropriate spec section, they all appear to match the spec: anchor01: OK - http://dev.w3.org/html5/spec/#dom-accesskey anchor02: OK though obsolete - http://dev.w3.org/html5/spec/#dom-a-charset anchor03: OK though obsolete - http://dev.w3.org/html5/spec/#dom-a-coords anchor04: OK - http://dev.w3.org/html5/spec/#dom-a-href and http://dev.w3.org/html5/spec/#attr-hyperlink-href anchor05: OK - http://dev.w3.org/html5/spec/#dom-a-type anchor06: OK though obsolete - http://dev.w3.org/html5/spec/#dom-a-shape area01: OK - http://dev.w3.org/html5/spec/#dom-area-coords area02: OK though obsolete - http://dev.w3.org/html5/spec/#dom-area-nohref area03: OK - http://dev.w3.org/html5/spec/#dom-tabindex area04: OK - http://dev.w3.org/html5/spec/#dom-accesskey HTMLAreaElement01: OK - http://dev.w3.org/html5/spec/#dom-accesskey HTMLAreaElement02: OK - http://dev.w3.org/html5/spec/#dom-area-alt HTMLAreaElement03: OK though obsolete - http://dev.w3.org/html5/spec/#dom-a-coords HTMLAreaElement04: OK - http://dev.w3.org/html5/spec/#dom-a-href and http://dev.w3.org/html5/spec/#attr-hyperlink-href HTMLAreaElement05: OK though obsolete - http://dev.w3.org/html5/spec/#dom-area-nohref HTMLAreaElement06: OK though obsolete and the lowercasing is unnecessary - http://dev.w3.org/html5/spec/#dom-a-shape HTMLAreaElement07: OK - http://dev.w3.org/html5/spec/#dom-tabindex HTMLAreaElement08: OK - http://dev.w3.org/html5/spec/#dom-area-target As a general comment, these look far from sufficient for testing content-to-IDL attribute reflection - e.g. the noHref tests will pass in a browser that doesn't actually implement noHref at all and always returns undefined, and the shape tests don't check what's returned for <area shape="Circ">, etc. (IE6 says "CIRCLE", IE8 says "circle", Firefox/WebKit/Opera say "Circ", HTML5 says "circle" for <area shape> and "Circ" for <a shape>, so that'd be a relatively useful thing to test). These are better than having no tests at all, but a decent comprehensive test suite would probably have to be written from scratch, in a methodical way based on the IDL fragments in the spec, to make sure it covers all the cases. -- Philip Taylor pjt47@cam.ac.uk
Received on Monday, 1 March 2010 19:11:18 UTC