Re: bidi tests assume undefined letter-spacing behavior

On 01/07/2011 03:20 PM, Arron Eicholz wrote:
> On Sunday, December 05, 2010 1:43 AM fantasai wrote:
>> On 10/13/2010 09:48 PM, L. David Baron wrote:
>>> The following tests:
>>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/bidi-005.xht
>>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/bidi-006.xht
>>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/bidi-007.xht
>>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/bidi-008.xht
>>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/bidi-009.xht
>>> http://test.csswg.org/suites/css2.1/20101001/xhtml1/bidi-010.xht
>>> http://test.csswg.org/suites/css2.1/20101001/html4/bidi-005.htm
>>> http://test.csswg.org/suites/css2.1/20101001/html4/bidi-006.htm
>>> http://test.csswg.org/suites/css2.1/20101001/html4/bidi-007.htm
>>> http://test.csswg.org/suites/css2.1/20101001/html4/bidi-008.htm
>>> http://test.csswg.org/suites/css2.1/20101001/html4/bidi-009.htm
>>> http://test.csswg.org/suites/css2.1/20101001/html4/bidi-010.htm
>>>
>>> assume letter-spacing behavior that is, as far as I can tell,
>>> undefined in CSS 2.1.
>>>
>>> While I agree Gecko's current behavior is suboptimal (and probably
>>> best fixed by implementing what is required by css3-text (in bug
>>> 299943)), I don't think these tests are valid as tests for CSS 2.1.
>>
>> Which element generates letter-spacing at an element boundary is not
>> defined in CSS2.1. But the letter-spacing here is consistent throughout the
>> paragraph. Gecko is failing here because the inter-character spacing is
>> inconsistent.
>>
>> You might be right that the test is dependent on undefined behavior at
>> element boundaries, but in that case I'd split the test into two halves: one
>> that checks the location of element borders, and another that tests
>> interaction with letter-spacing.
>> (I'm having trouble right now coming up with a letter-spacing behavior that
>> would require this split, though; i.e that would pass with the split, but fail
>> with the current combination.
>> But if you think I should split the tests, then let me know and I'll split them.)
>
> I think the cases should be split since we do not define the placement
> of the letter spacing space.

We do, however, require that the space between the letters must be
consistent: there is no allowance for varying inter-letter spacing
at bidi boundaries. So do we have an implementation that has
consistent space between the letters *and* fails the test? Because
if not, there's no practical reason for me to split them.

> For CSS3 we should define the letter spacing space to be placed after
> the character in the current text run direction. This would match most
> implementations. However, for CSS 2.1 this should be just undefined
> and we can just correct the tests.

CSS3 already has a better definition.

~fantasai

Received on Friday, 7 January 2011 23:50:36 UTC