Re: [css-writing-modes-3] text-combine-upright-inherit-all-001 and its reference file

Comments inline:

On Mon, Jul 6, 2015 at 1:54 AM, fantasai <fantasai.lists@inkedblade.net>
wrote:

> On 07/05/2015 12:50 AM, Koji Ishii wrote:
>
>> Inline:
>>
>>  On Jul 5, 2015, at 10:48 AM, Gérard Talbot <css21testsuite@gtalbot.org>
>>> wrote:
>>>
>>> Koji,
>>>
>>> This is just in case you are not already aware of this...
>>>
>>> [src]
>>>
>>> http://test.csswg.org/source/css-writing-modes-3/text-combine-upright-inherit-all-001.html
>>>
>>> [nightly-unstable]
>>>
>>> http://test.csswg.org/suites/css-writing-modes-3_dev/nightly-unstable/html/text-combine-upright-inherit-all-001.htm
>>>
>>> [reference src]
>>>
>>> http://test.csswg.org/source/css-writing-modes-3/reference/text-combine-upright-inherit-all-001.html
>>>
>>> [reference nightly-unstable]
>>>
>>> http://test.csswg.org/suites/css-writing-modes-3_dev/nightly-unstable/html/reference/text-combine-upright-inherit-all-001.htm
>>>
>>> The reference file should not use text-combine-horizontal property and
>>> no vendor prefix. As it is (right now), it's just a copy of the test.
>>>
>>
>> Tried that, but pixel images do not match due to some layouts involved by
>> text-combine-upright.
>>
>> It does pass on browsers that do not support text-combine-upright at all,
>> that’s unfortunate, but there are several other tests that tests the
>> support of the property, and since this test is specifically testing its
>> inheritability, I think that’s ok.
>>
>> If you have ideas to test inheritability that fails nicely for
>> non-supported browsers, that’s appreciated, but when it’s hard, I think
>> that’s ok, no?
>>
>
> Koji, the test is completely invalid. As Gérard wrote, the reference is
> *just a copy of the test*. Such a reftest will never, ever fail on any
> real implementation ever.
>
> The reference file must be the same effect by some *alternative* method.
> Otherwise you are just testing that two identical source files render
> identically--which is useless.
>
> If you are trying to test if
>   <div class=vrl><span class=tcy>66</span></div>
> renders the same as
>   <div class=vrl><span class=tcy><span>66</span></span></div>
> then your test file should use
>   <div class=vrl><span class=tcy><span>66</span></span></div>
> while your reference file uses
>   <div class=vrl><span class=tcy>66</span></div>
> Then you are actually testing if the two *different* sources render
> equivalently, and the test has some meaning.


I'm not sure if I follow, that's exactly what the test is as of now[1], and
the suggestion is to change the reference to:
  <div>66</div>

So are you saying the test should fix as suggested, or should not?

[1]
https://github.com/w3c/csswg-test/blob/master/css-writing-modes-3/reference/text-combine-upright-inherit-all-001.html

/koji