Tor Lillqvist <[EMAIL PROTECTED]> brought this up:
The technique of searching for directory separators in strings through the ISSLASH macro does, on Woe32, not support non-ASCII pathnames in most CJK locale encodings. Why? ISSLASH looks for a _byte_ with value 0x5C. However,
0x5C -or- 0x3F on Woe32.
in these locale encodings
Japanese: CP932 SHIFT-JIS Chinese: GBK GB18030 BIG5 BIG5-HKSCS CP950 Korean: JOHAB
the byte 0x5C occurs as second byte of some multibyte characters. If such a character is used inside a directory name, code that uses ISSLASH does not work correctly. All gnulib modules that use ISSLASH are affected.
Could this also be a problem on Unix systems using multibyte encoded (UTF-8) filesystems, if not now then in the future? Maybe some (future) Unix systems support multi-byte encoded filenames containing 0x3F in the second+ byte of a multi-byte character.
What can we do?
1) On Woe32, use 'wchar_t*' instead of 'char*' to denote pathnames. Use conditional macros like _TCHAR, _TEXT(), _tcslen() etc. that allow to accomodate these platform differences without too much #ifs.
2) On Woe32, expect UTF-8 encoded 'char*' strings to denote pathnames.
3) Use mbtowc() to step through pathnames while looking for a backslash.
4) Document this as a limitation. The workaround for the user is to switch to an UTF-8 locale.
It's probably best to choose one internal representation of pathnames and stick to it, but any representation other than single 'char' is a lot of work, as you say!
The drawbacks are:
1) Tons of code that deals with pathnames has to be changed to use typedef'ed types. Also, support for WindowsME and older is dropped.
Wouldn't it be possible to link against unicows.dll to support Win95/98/ME? Are there licensing problems with this?
2) Extra code must be added for every system call to convert pathname
arguments from UTF-8 to UTF-16, and pathname results from UTF-16
to UTF-8. Also, the user of the gnulib modules must be aware of the
semantic difference. Also, support for WindowsME and older is dropped.
It sounds like replacing the system calls with some wrapper functions with lots of conditional code. Maybe the wrapper functions could avoid converting to and from UTF-16 if they are running on WinME and earlier.
Strictly speaking, UTF-16 is a multi-16-bit-word encoding, but I don't know what support Woe32 systems have for characters outside the Unicode BMP (requiring more than one 16-bit word in their encoding).
I think (2) also implies (3). If you use UTF-8 internally, any parsing of pathnames need to change, e.g. the IS_PATH_WITH_DIR macro in pathname.h.
3) Tons of code that deals with pathnames has to be changed to use mbtowc(), _mbschr(), _mbsrchr() etc.
4) For users in CJK locales on Woe32, the contents of directories with some non-ASCII pathnames is inaccessible to GNU tools.
Microsoft recomments approach 1. GNOME has chosen approach 2. I would favour answer 4.
What do you think?
Bruno
My gut instinct would be to use UTF-8 internally, but I'm not doing the work! :-)
-- -=( Ian Abbott @ MEV Ltd. E-mail: <[EMAIL PROTECTED]> )=- -=( Tel: +44 (0)161 477 1898 FAX: +44 (0)161 718 3587 )=-
_______________________________________________ bug-gnulib mailing list bug-gnulib@gnu.org http://lists.gnu.org/mailman/listinfo/bug-gnulib
