Utf-8 and diverging filenames

We noticed a problem with unicode characters, namely “u+0308 → COMBINING DIAERESIS”. If the client encounters a file containing this character in combination with any of ‘a’, ‘o’, ‘u’ it will treat them as ‘ä’, ‘ö’, ‘ü’. Synching usually works, although the filename on local fs and on the server differ.

This got to our attention, because after one person deleted the file, a sync-client on a macOS System (client 8.0.10, macOS 13.2.1) recreated it. The system, on which the file was deleted was a windows system, as far as we know.

I guess this tells us, that unicode is not treated equally by all of seafile-components. Maybe it would be cleaner to sync all utf-8 characters similarly or alternatively display a warning on the sync-client if the filename contains characters that will be “sanitized”, when synched to the server.

By design all filenames are stored internally in UTF8 NFC format. All filenames on all systems will be converted into this format before uploading. Can you reproduce the problem? If so please provide us with the logs from the clients.