25266 2014-04-04 20:44:02 +0000 Consider adding 34 code points to the EUC-JP decoder present in Blink/ICU's euc-jp-2007. 2014-04-28 21:35:45 +0000 1 1 1 Unclassified WHATWG Encoding unspecified All All RESOLVED WONTFIX P2 normal Unsorted 1 jshin annevk adrianba hsivonen mike pub-w3 travil VYV03354 www-international sideshowbarker+encodingspec oldest_to_newest 103434 0 jshin 2014-04-04 20:44:02 +0000 While updating the EUC-JP mapping table per the encoding standard (mainly dropping most of JIS X 212 characters among other things) for Blink/Chromium, I found that there are about 20 code points that are missing in the EUC-JP decoder. They're listed below: # 1. 0x8E 0xE0 to 0x8E 0xE2 # 00A2 00A3 00AC # 2. JIS X 0212 extra (0x8F 0xF3 0xhh) # 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 # 2172 2173 2174 2175 2176 2177 2178 2179 221A 2220 2229 222A 222B 2235 2252 # 2261 22A5 3231 # 3. JIS X 0208 extra : 0xFC 0xFB => FFE2 103745 1 annevk 2014-04-11 13:16:09 +0000 This seems like a subset of bug 16941. 103748 2 annevk 2014-04-11 13:23:26 +0000 I do not mind adding these. Are other vendors likely to add them? 103749 3 1470 VYV03354 2014-04-11 13:45:54 +0000 Created attachment 1470 Testcase I'm for adding 0xFC 0xFB => FFE2. Gecko and Trident already support this. But I'm adding other code points for two reasons: 1. No official specs have those mappings. 2. They are not interoperable between browsers. Gecko will convert all of them to U+FFFD. Trident will convert 0x8E 0xE0 .. 0x8E 0xE2 to U+73EE U+7AE2 U+9D5D. Also Trident does not support triple-byte sequences at all (0x8F 0xF3 0xhh will be converted to U+5834 U+xxxx). 103751 4 VYV03354 2014-04-11 13:46:57 +0000 (In reply to Masatoshi Kimura from comment #3) > But I'm adding other code points for two reasons: But I'm against adding... 103880 5 annevk 2014-04-15 16:07:17 +0000 I don't understand the requested 0xFC 0xFB mapping. Per the euc-jp decoder algorithm that becomes 8644 as pointer, which maps to U+FFE2 in index jis0208. 104564 6 annevk 2014-04-28 15:15:03 +0000 *** Bug 16941 has been marked as a duplicate of this bug. *** 104582 7 jshin 2014-04-28 21:35:45 +0000 From bug 16141 : The ones that are still missing can be found in an IBM extension to JIS X 0212, Rows 83--84. Tables: <http://coq.no/character-tables/ibmjapan2.pdf> <http://coq.no/character-tables/ibmjapan2.js> However, if Firefox has never supported them [1], I'm not wed to them, either. And I'll just get rid of them from Chromium's new euc-jp table. Then, it'll be exactly the same as specified in the encoding spec. > I don't understand the requested 0xFC 0xFB mapping. Per the euc-jp decoder > algorithm that becomes 8644 as pointer, which maps to U+FFE2 in index jis0208. You're right. It's there. Sorry about the noise on this code point. I'm closing this as wontfix. [1] The Chrome's current table (not the one I'm adding based on the encoding spec but the one that is currently used in released versions) was made by comparing IE, Firefox and ICU tables, but lost is how exactly I 'curated' and 'merged' them. 1470 2014-04-11 13:45:54 +0000 2014-04-11 13:45:54 +0000 Testcase euctest.html text/html 423 VYV03354 PCFET0NUWVBFIGh0bWw+CjxzdHlsZT4KYm9keSB7IHdvcmQtd3JhcDogYnJlYWstd29yZCB9Cjwv c3R5bGU+CjxzY3JpcHQ+CnZhciBmciA9IG5ldyBGaWxlUmVhZGVyKCkKdmFyIGlucHV0ID0gWzB4 OEUsMHhFMCwweDhFLDB4RTEsMHg4RSwweEUyXTsKZm9yICh2YXIgYyA9IDB4QTE7IGMgPD0gMHhC NzsgKytjKSB7CiAgaW5wdXQgPSBpbnB1dC5jb25jYXQoWzB4OEYsMHhGMyxjXSk7Cn0KaW5wdXQg PSBpbnB1dC5jb25jYXQoWzB4RkMsMHhGQl0pOwpmci5yZWFkQXNUZXh0KG5ldyBCbG9iKFtuZXcg VWludDhBcnJheShpbnB1dCldKSwgImV1Yy1qcCIpOwpmci5vbmxvYWQgPSBmdW5jdGlvbigpIHsK ICBkb2N1bWVudC5ib2R5Lmluc2VydEFkamFjZW50SFRNTCgiQmVmb3JlRW5kIiwgZXNjYXBlKGZy LnJlc3VsdCkpOwp9Owo8L3NjcmlwdD4K