regex bug: can't match half-width kana in SJIS locale

[KIMURA Koichi]

Hello.

I found bug of regex in GNU sed 4.1.4.
I heard GNU sed's regex is drawn from glibc, so I send mail
to this mailing list.

Situation:
In Japanese ShiftJIS locale, half-witdth katakana in character class
with range does not match appropriately.

Reproduce:
set LANG=ja_JP.SJIS
export LANG
echo ABCDE | sed -ne '/[A-E]\+/p'

Actually, A B C D E is half-width katakana character.

Result:
nothig printed.

Please see data that appended this mail(uuencoded).

regards,

begin 644 testkana.sh
M<V5T($Q!3D<]:F%?2E`N4TI)4PIE>'!O<address@hidden<')I;G0*96-H;R"QX
MLK.TM2!\('-E9"`M;F4@)R];L;*SM+5=7"LO<"<*(VYO="!P<FEN=`IE8VAO4
>(+&RL[2U('P@<V5D("UN92`G+UNQ+;5=7"LO<"<*/
``
end
size 120


-- 
KIMURA Koichi