Re: Requirements for (level >=3) tests

On 02/16/2012 10:37 PM, Aryeh Gregor wrote:
> On Thu, Feb 16, 2012 at 1:52 PM, Chris Lilley<chris@w3.org>  wrote:
>> I think we should also allow HTML5 (in the XML serialisation, so existing toolchain can consume it).
>
> Does it need to be actual XHTML5, or is it sufficient for it to be
> HTML5 that's well-formed XML?  What language is the toolchain written
> in, and could the XML parser be swapped out for an HTML5 parser?

I'll have to defer to Peter Linss on this point, as he's maintaining
the build scripts right now.

> Not that I object to annotations, but I don't think they have to be a
> hard requirement.  For instance, a test that particular transforms are
> equivalent to particular matrices should be self-explanatory to anyone
> who's familiar with the spec at all.

The "help" link annotations are used:
   - to direct the reviewer to the appropriate section of the spec
   - to allow easier analysis of test coverage of the spec
   - to generate reports for CR exit
   - to annotate implementation data into each part of the spec

The assert is optional because some people find it onerous. However,
it does help the reviewer in understanding the test writer's original
intent and thus for finding false positives. (Gérard's found a number
of those in the 2.1 test suite.)

> 2012/2/16 "Gérard Talbot"<css21testsuite@gtalbot.org>:
>> So the author name in that reftest is not important, useful? Who are web
>> authors supposed to contact if there is a problem with a reftest then?
>
> They can look in the revision history.  There's no need to duplicate
> that information in metadata.  Also, the test is the responsibility of
> the Working Group, not the original author -- issues with tests should
> be filed as bugs or sent to www-style, not privately sent to the
> author.

The author data is also used to generate a list of credits when we publish
the test suite.

~fantasai

Received on Thursday, 16 February 2012 22:58:07 UTC