Can't get file size, possibly because of charset problem

Question

I have an application that needs to recursively find directory sizes, I have a class extending java.io.File with an internal method to do this:
private static long getSize(File file) {
                long size = file.length();
                if (file.exists()) {
                        if (file.isDirectory()) {
                                File[] files = file.listFiles();
                                for (int i = 0; i < files.length; i++) {
                                        size += getSize(files);
} else {
System.out.println("Problem checking size on \"" + file + "\", reported file size is: " + file.length());
return size;
It works (matches du -b output) with the exception of some problematic files I have.
Output from ls:
$ ls -l
total 5461
-rw-r--r--  1 xyz users    1003 Feb 16 22:06 FileWRS.class
-rw-r--r--  1 xyz users     831 Feb 16 22:04 FileWRS.java
-rw-r--r--  1 xyz users     489 Feb 16 22:07 FolderSizes.class
-rw-r--r--  1 xyz users     198 Feb 16 21:23 FolderSizes.java
-rwxrwxrwx  1 xyz users 5568138 Apr 25  2004 test.?.abcOutput from du -b:
$ du -b
5583203 .Output from my code:
$ java FolderSizes .
Problem checking size on "./test.?.abc", reported file size is: 0
15065The character displaying as "?" is actually an accented e character ("�") and displays correctly in konqueror (my gui file browser).
I don't really know why it can't pick up the character, but I'm more interested in just getting the file size and ignoring the character problem, any solutions/suggestions for places to look? Do I need to supply more information?

Answer 1

Try:
LC_ALL=en_US.ISO8859-1 java FolderSizes .
Does this help?Yes, that does fix it for my terminal, but if possible I would like a solution that can determine this at run time and solve (so that when it's running on some other terminal, with some other charset, with someother file system charset it will still work).
Any ideas for doing this from withing the VM?
p.s. thanks for the fast reply.