- From: taka oshiyama <oshiyama@est.co.jp>
- Date: Tue, 05 Apr 2011 13:59:12 +0900
- To: fantasai <fantasai.lists@inkedblade.net>
- Cc: public-css-testsuite@w3.org
(2011/03/29 14:38), fantasai wrote: > On 02/21/2011 03:43 AM, taka oshiyama wrote: >> Hi >> >> Please find a subject testcase I submitted here >> http://test.csswg.org/source/contributors/east-tokyo/submitted/css3-text/text-emphasis-color-001.xht >> >> Please kindly review it and give me feedback if any. > > Hello Taka, > > Because the positioning of the emphasis marks is not > precisely defined by the spec (and also because the > shape and size of the marks are suggested rather than > required), these tests cannot be written with exact > references as you are doing here. Reftests must match > their references pixel-perfectly. The abpos technique > is very clever, and is close enough to give the tester a > clear idea of what's expected from a manual comparison, > but they won't work for an automated comparison. > > What you should do is: > > - Write instructions that the test text should have marks > "similar to" (rather than identical to) the one in the > reference over each character. This will make all these > files function accurately as self-describing tests. I agree to change the instructions with "similar to" rather than "identical to". > > - Make the reference file for text-emphasis-style using > text-emphasis: <string> syntax on the test part. For most > implementations that support text-emphasis, this should > provide an accurate reference. To create a reference > for text-emphasis: <string>, probably the best you can > do is create a negative reference using != instead of > == in the reftest.list file. Then compare the test to > a file that doesn't have text-emphasis: this will make > implementations that don't support text-empahasis: > <string> at all fail during an automated run. (Because > we can't make a true positive reference here, a manual > comparison will be necessary to ensure correctness for > this test.) As I will use text-emphasis: <string> to draw a reference, so that text-emphasis: <string> will be tested while text-emphasis-style property is tested at the same time. To avoid duplicated tests, I will do not run text-emphasis: <string> by stand alone. It will make sense. > > I'm not sure yet what how to make a reference for the > colors. We can leave it as a self-describing test only > for now. Make sure that the test instructions are asking > about matching the color, rather than paying attention > to shapes. I will revise the instruction to focus on colors rather than shape. > > - For the abspos references, it will probably give better > results to set both 'left' and 'right' and use > 'text-align: center' to center the mark over the character. Ok, I will use abspos technique. > > > - Lastly, use different kinds of characters here as the > base text, e.g. > ABC早午晚いろはابتdཀཁག > to test that the emphasis marks work correctly with > different scripts. Ok. > > > It would be a good idea to have a separate test that makes > sure text-emphasis marks work correctly with combining > characters, e.g. throw in some decomposed Korean and accented > Latin, and maybe a few Indic examples from UAX29. g̈각நிเก etc. > http://unicode.org/reports/tr29/ At this moment, I would like to defer this as our focus is Japanese text layouts. > > > ~fantasai > Hi fantasai I put my replies in-line, and please wait for a moment until revised ones are submitted. regards, -- 押山 隆 taka oshiyama oshiyama@est.co.jp (or takaoshiyama@gmail.com) mobile: 080-4148-7654 イースト株式会社 〒151-0053 渋谷区代々木2-22-8
Received on Tuesday, 5 April 2011 04:59:53 UTC