Package: grep Version: 2.5.1.ds2-4 Severity: minor Seems bad: $ echo $'\x80'|LC_ALL=C grep [$'\x80']|wc -c 2 $ echo $'\x80'|grep [$'\x80']|wc -c 0 $ locale LANG=zh_TW.utf8 LC_CTYPE=zh_TW.utf8 LC_NUMERIC="zh_TW.utf8" LC_TIME="zh_TW.utf8" LC_COLLATE=C LC_MONETARY="zh_TW.utf8" LC_MESSAGES=C LC_PAPER="zh_TW.utf8" LC_NAME="zh_TW.utf8" LC_ADDRESS="zh_TW.utf8" LC_TELEPHONE="zh_TW.utf8" LC_MEASUREMENT="zh_TW.utf8" LC_IDENTIFICATION="zh_TW.utf8" LC_ALL=
Upstream says J> I don't think this is a bug in GNU Grep itself, but maybe is a bug J> introduced by one of the patches that Debian apply to their package of J> Grep. You might want to report it to Debian. J> It has just occurred to me that, because the byte 0x80 is not a valid J> UTF-8 J> string (as far as I know), when Grep is asked to search for a pattern J> consisting of only that byte, in a UTF-8 locale, the behaviour could J> be expected to be undefined. Therefore other people might not regard J> this as a bug. J> - Julian OK, not sure if a bug or which exact package. Without the [] of course all is well. Maybe no bug at all.
