I'm 'bout off to work, but I remembered something that just needs a little help from extra eyes.
One of our problems with reading and writing files with names containing characters outside of the 7-bit ASCII range is with library and/or API calls that take filenames directly.
Our uses of fopen have been mainly fixed, as are some of the pass-it-a-filename calls. However... there might be a few places that still pass a filename on directly.
Could people take a quick peek to see if we have spots where that happens? For things to work well, we have to do one of two things
1) Convert filenames from UTF-8 to filename encoding before calling 2) open the file and pass the API an open file or a data source (depending on the needs of the API).
#1 will work for Linux, but *not* Windows XP and friends, since all characters possible on the filesystem can't be respresented in the current ANSI codepage.
If anyone has a chance and can find a few of the lingering places, I can go in more quickly tonight and fix them.
Thanks.