Re: [css-writing-modes-3] text-combine-upright-inherit-all-001 and text-combine-upright-inherit-all-002

Thank you for the feedback, Gérard. Copied to https://github.com/w3c/csswg-test/issues/803


> On Jul 17, 2015, at 15:04, Gérard Talbot <css21testsuite@gtalbot.org> wrote:
> 
> Koji,
> 
> [test]
> http://test.csswg.org/suites/css-writing-modes-3_dev/nightly-unstable/html/text-combine-upright-inherit-all-001.htm

> 
> [reference file]
> http://test.csswg.org/suites/css-writing-modes-3_dev/nightly-unstable/html/reference/text-combine-upright-inherit-all-001.htm

> 
> When I take such test with Firefox 42.0a1 (all buildID), the reference file *_matches perfectly_* the test while the pass-fail-conditions sentence clearly describes that Firefox 42 *_fails_* that test.
> 
> I have no idea how to improve that test right now.
> 
> As I am sure I said before, presumably to Hajime, testing properties that inherits by default is very difficult…

Yeah, these days I’m not too worried about that; lots of tests in writing-mode fall into this category, but we will know once the UA starts implementing the property even partially.

> [test]
> http://test.csswg.org/suites/css-writing-modes-3_dev/nightly-unstable/html/text-combine-upright-inherit-all-002.htm

> 
> [reference file]
> http://test.csswg.org/suites/css-writing-modes-3_dev/nightly-unstable/html/reference/text-combine-upright-inherit-all-002.htm

> 
> 1-
> line 8:  <meta charset="utf-8">
> 
> The charset meta should be between lines 2 and 3; we want browsers to know soon about the character encoding.

Right, I just took over this abandoned test, fixed.

> 2-
>    <meta content="Two white triangles does not pointing to the right. Pointing up." name="assert">
> 
> Obviously, this needs updating.
> 
> The test must obviously be testing inheritance through the containment hierarchy and must also be testing interrupted (by a box boundary) text runs..

Fixed.

> 3-
> When I take such test with Firefox 42.0a1 (all buildID), the reference file *_matches perfectly_* the test while the pass-fail-conditions sentence clearly describes that Firefox 42 *_fails_* that test. So, this must be a problem.
> 
> 
> One way I see to overcome a layout problem is to define the pass-fail-conditions sentence in a way that the tester will not consider the width of digits and letter-spacing between digits as a criteria for passing the test ... like in
> 
> http://test.csswg.org/source/css-writing-modes-3/text-orientation-011.xht

> 
> "
> different advances (letter spacings) are considered as PASS.
> “

That’s possible, but then the test is no longer a ref test. I prefer a test being a ref test over false-positive for UA that doesn’t implement it at all.

If an UA doesn’t implement a property at all, other tests should discover that anyway. All tests to test the existence of the property doesn’t make sense, does it?

/koji

Received on Friday, 17 July 2015 08:58:37 UTC