503986 - Errors in Unicode upper/lower case conversion

Status: RESOLVED DUPLICATE of bug 394604

(Core :: JavaScript Engine, defect)

Description

[David Mandelin [:dmandelin]]

Attachment: Incorrect case conversions

I just compared the results of JS toUpperCase and toLowerCase against Unicode 5.1.0. The attached file shows the 638 code points for which JS does something wrong.

Comment 1

[David Mandelin [:dmandelin]]

Another thing I realized is that the char code compressed table format in jsstr.cpp cannot support correct Unicode case conversions. That format only allows a character to have an upper-case form or a lower-case form, but not both. But some characters (e.g., U+01C5 [1]) have both. I haven't analyzed those characters in detail but I think it might work to allow both bits set with a special interpretation of the offset area. I also notice 2 unused bits in the encoding that might be used to extend the offset area if needed.

[1] http://www.fileformat.info/info/unicode/char/01c5/index.htm)

Comment 2

[Brendan Eich [:brendan]]

The jsstr.cpp table comes from Java 1 in 1995 -- Unicode 2, IIRC. There is a bug on upgrading it: bug 394604 I think.

/be