Re: [RC6] block-in-inline-append-001 has several issues

On 11/29/11 7:44 PM, "Gérard Talbot" wrote:
> Both block-in-inline-append-001 and block-in-inline-append-001-ref have
> validation issues. The body element must have with a block container.
>
> <!ELEMENT BODY O O (%block;|SCRIPT)+ +(INS|DEL) -- document body -->

If you want to stick a <div> around the whole thing to shut up the 
validator, go for it.  I have to admit that complicating tests to 
appease arbitrary validation constraints is pretty low on my priority 
list...

> As strange as it is, Opera 11.52 (displays a rectangular dotted line
> which is 10em wide), Chrome 15.0.874.121 (no dotted outline shown) and
> Firefox 8.0 (there are vertical "whiskers" on the left hand side of the
> first "A") have 3 different renderings of the reftest. So, they all
> "pass" the test ... according to/with 3 different respective reftests
> that they render. The test and its reftest definitely fail at
> determining in a reliable manner what is a fail condition.

A fail condition for this test is a rendering for the test that differs 
from the reference, by definition of "reftest".  This is a test for 
correct handling of dynamic changes to the DOM, so the only way to pass 
it is if the final rendering does not depend on the history of DOM 
manipulations but only on the final state of the DOM.

Now what rendering would be correct for the _reference_ is a separate 
question.  I don't think I wrote any tests for that for outline because 
the behavior in the spec is not exactly completely defined.  The 
behavior that I suspect is desirable isn't one that we could test via a 
reftest in any case, by the way.

> And I think a 4th way of rendering the test and the reftest would be
> okay too: with a non-rectangular shape which would outline the outermost
> inline.

And the block contained inside it, possibly, yes.  This seems like it 
would be ideal as rendering.

-Boris

Received on Wednesday, 30 November 2011 01:16:11 UTC