I'm using rsync to copy a directory from a windows server. This directory is very big and has thousands of files. Some of the files have some non-common ASCII characters in their filenames. I'm copying these files to a JFS filesystem based external storage. Whenever kernel encounters such characters in the filenames, it throws up messages like this in /var/log/messages.
jfs_strtoUCS: char2uni returned -22.
charset = iso8859-1, char = 0x84
So, there are thousands of such messages in the log file. Apparently, these files are written by Mac computers to the windows server. Anyways, Windows was able to recognize the characters and store them properly. I believe its filesystem limitation, am I not correct? Is it just a parameter tweak in JFS or is it an inherent limitation in JFS and can't be changed. Does this limitation exist in ext3 too? 'cos if doesn't then I will change the filesystem to ext3.
Or am I in the wrong impression that this unicode conversion is a limitation of filesystem? Any ideas ???
My last option is to rename these files on windows server. Any ideas how to write such a script that parses all the filenames in a directory (recursively) and substitute few characters with some other common characters? I'm sure it will be painful for me to find out all such characters (and thier hex codes) from the log file.
Sanjay
http://www.techiesabode.com/
(text/plain)