On Fri, Jun 23, 2006 at 12:44:17PM +0900, Junichi Uekawa wrote: [...] > I don't have thetime to track it down right now, but I think you can > reproduce the problem by running iconv with a sequence of characters ' > 漢字\'.
Right; I did (under a UTF-8 locale) $ echo '漢字\' | iconv -f utf8 -t iso-2022-jp | iconv -f iso-2022-jp -t utf8 漢字\ > Converting from iso-2022-jp to euc-jp and converting it back > to iso-2022-jp will result in backslash becpoming a Zenkaku > backslash. This is a critical bug in that this kind of code sequence > happens often in C programs and TeX sources. My currenyt workaround is > to not allow people to use VIM to come anywhere near my sources, and > force them to use emacs. > > I don't know if this is a related problem with vim, or it's a > different problem with glibc. As shown above, it seems that there is a bug in glibc. I cannot reproduce it with libc6 from experimental, so it looks like it has been fixed in 2.4. My current bet is that getting libc/iconvdata/jis0208.[ch] from CVS HEAD could do the trick with glibc 2.3.6, will test it soon. If you have better ideas, please let me know ;) Denis