On Thu, Mar 14, 2002 at 05:22:24PM -0600, Dave Kleikamp wrote: > JFS stores file names in unicode (16-bit) and converts the names into > and out of the system codepage. By default, the codepage used is > whatever CONFIG_NLS_DEFAULT is set to in the kernel. This can be > overridden with the mount option "iocharset=<charset>". > > It appears that the characters in the name ?????? don't map to the > current codepage. A possible, though involved, solution may be to make > sure that nls_utf8.o is either built in the kernel or as a module and > mount the file system with "-o iocharset=utf8". > > In OS/2, JFS was able to get the codepage from the current process and > use that to convert from unicode. Linux processes don't appear to have > that information available (at least in the kernel).
Most other linux file systems just have the policy of leaving such conversion to the user space by not interpreting the filenames. Userspace can store UTF-8 if it wants and convert it to any other set for display. How about supporting that with a mount option in JFS too? I guess that would fix the problem. -Andi _______________________________________________ Jfs-discussion mailing list [EMAIL PROTECTED] http://www-124.ibm.com/developerworks/oss/mailman/listinfo/jfs-discussion
