Problems with umlauts in file names on Mac OS X¶

Description of the Problem

Generally, there are two problems. The first occurs when you add a file with an umlaut in its name (e.g. Änderungen.txt) using the command line client on Mac OS X and check it out on Windows (or Linux). Although the file looks correctly in the Windows Explorer, the Command Shell shows the file name like this A?nderungen.txt. Even worse, you can create another file Änderungen.txt on Windows and now have two similar looking files in the same directory when using the Explorer!

The second problem occurs when you open the correctly looking working copy on Mac OS X using SmartSVN. You will see the two entries for the same file, one with the missing state, one with the unversioned state.

The Reason

Both problems have the same reason. Subversion has standardized to store file names encoded as UTF-8 in the repository. Unfortunately, there are different types of UTF-8 and Subversion has not defined what type to use. Windows and Linux (usually) create files with the so-called composed UTF-8 variant (e.g. Ä is a single character) where as Mac OS X uses the decomposed UTF-8 variant (the Ä will be stored as two characters, an A and ¨). Even when creating a file on Mac OS X using the composed form, it will be stored and later reported by the file system in the decomposed form.

Subversion maintains a list of the version controlled files of the current directory in the .svn sub-directory. When using the svn status command (which is used by SmartSVN to list the files in the project window), it basically compares this list with the file names reported from the file system. Problems occur if this list of version controlled files is using a different UTF-8 variant than the file system reports its files. Until now the subversion command line client has no knowledge about the different UTF-8 variants and hence compares the file names incorrectly (see http://subversion.tigris.org/issues/show_bug.cgi?id=2464).

The Solution

The solution for the first problem is either to avoid files with umlauts in their names, to wait until the Subversion team fixes this bug or to use SmartSVN. SmartSVN has additional code which converts the decomposed file names reported by the Mac OS X file system to the composed form and stores this in the repository.

The second problem only occurs if you used the command line client to add/commit files with decomposed umlaut characters. You could solve that by renaming the files in SmartSVN's Repository Browser and performing an update on the working copy.