Re: submission of CSS3 testcases by east-tokyo [CSS Test: text-transform - large-kana - basic cases]

Hello Taka,
Here are my comments.

On 02/07/2011 09:02 PM, taka oshiyama wrote:
>
> 2. Use .xht file extension so that the file gets served as XHTML. (An
> HTML copy will appear on the website automatically.)
> Yes, changed to .xhl.

OK. You can remove the .html version, then. (You have to use
your subversion client to remove it, otherwise it stays in the
repository.)

> 3. The tests should be reftests if possible. This means
> a. Creating another, simpler XHTML page that does not use the feature
> you are testing but looks exactly the same as the indended rendering.
> b. Creating a reftest.list file in the directory that matches up the
> test and its reference.
> Okay, I understood it. In case of this test, I do not use it as I can make it self-explanatory.

That's not the point. The reftest format allows automation of the
test suite, so while it's good for the test to be self-explanatory
and human-checkable, it should also have a reftest reference.

(This allows layout engine developers to hook the tests into their
regression test framework, to run the tests automatically very
frequently to make sure recent changes have not broken anything.)

> Here are my initial comments on text-transform-006:
>
> 1. The test should be comprehensive, that is, it should test every single
> small kana character in Unicode. (There aren't many, so this is not
> unreasonable.) In addition, it should test a small sample of other
> characters to make sure they are not affected, e.g. large kana remain
> large, and alphabetic characters do not get uppercased or lowercased.
> Yes, I updated.

Aren't there half-width kana characters as well? They should be
tested as well, otherwise the test is not truly comprehensive.

> 2. Since this is very easy and very possible, test should ask for a
> comparison of the test text to a reference text, such as in these
> tests:
> http://test.csswg.org/suites/css2.1/20110111/xhtml1/text-transform-bicameral-008.xht
> http://test.csswg.org/suites/css2.1/20110111/xhtml1/content-counters-009.xht
> (The reftest reference would then be an exact copy of the test, except
> with two control elements instead of one control and one test.

Very good. Some comments on this:

1. Use blue, not green.
    http://www.w3.org/Style/CSS/Test/guidelines.html#color

2. The reference character is also affected by the text-transform.
    If you look at the source code for the tests you are using as
    a model, e.g.
      http://test.csswg.org/suites/css2.1/20101210/html4/text-transform-bicameral-008.htm
    you will notice that they are written so that only the character
    being tested is affected by the text transform: the reference
    character is unaffected.

3. Large kana and small kana look alike; they differ only in size.
    Therefore the instructions should emphasizes that the match must
    be in both shape *and size*. Otherwise the tester will look and
    assume the test passes even if the kana are not transformed: the
    two shapes obviously match.

4. Use a larger font size, e.g. 200%. This will make it easier to
    see differences in size, which are particularly tricky to detect.

>
> 3. The text-transform does not actually require the use of a particular font.
> It only requires that the font used have a glyph for the characters being
> tested. (CSS's fallback algorithm should take care of that.) If a system
> completely lacks a Japanese font, then the IPA font can be used as a
> suggestion, but the test shouldn't be flagged as requiring a special font.
> Yes, I updated accordingly, namely, no flag of IPA, instead I inserted scripts in order to use IPA font only if the system
> lacks a Japanese font.

OK

Last comment: I would suggest naming the test text-transform-kana-001;
if other people find things to test for "text-transform: kana", this
will make it easier to keep related tests together.

~fantasai

Received on Tuesday, 8 February 2011 08:45:17 UTC