Re: Format guidelines for tests (particularly HTML5)

On 04/18/2012 08:05 AM, Linss, Peter wrote:
>
> On Apr 17, 2012, at 11:29 PM, Aryeh Gregor wrote:
>
>> On Wed, Apr 18, 2012 at 4:58 AM, fantasai<fantasai.lists@inkedblade.net>  wrote:
>>> The reason XML serializations are recommended and HTML serializations
>>> were not is because of the last clause about reserialization, which
>>> you have now attached to HTML<  5 and not to HTML5, even though it
>>> applies equally to HTML5.
>>
>> So HTML files are reserialized but XHTML files are not?  I thought all
>> tests were generated in both HTML and XHTML forms, so all tests will
>> get parsed and reserialized.
>>
>
> All tests get re-serialized, regardless of source or output format. We do this so that the build process can inject markup into the metadata without causing unrelated diffs in the output. (That in turn is necessary for the test harness import process which looks for diffs but knows to ignore certain metadata markup diffs.)
>
> I already updated the text in the wiki and also added a note about encoding:
> http://wiki.csswg.org/test/format#acceptable-test-formats
>
> It occurred to me that for markup parsing tests we need to be able to flag the test source in such a way that the build process will simply copy the source file as is. It can copy all the metadata (including any metadata it needs to inject) into a sidecar file (.meta). Any suggestions for the name of that flag?

I'd make it a per-testsuite option. None of the CSS tests should ever
need that flag. But some of the HTML ones would.

~fantasai

Received on Wednesday, 18 April 2012 18:39:03 UTC