>> Perhaps, though, I should use svn_subst_translate_string2() with the >> encoding set to "UTF-8". It would greatly simplify normalize_string(), > > +1. Just call it with encoding=NULL; using the same API in both > branches makes life easier for the next person to revv that API.
One thing that worries me about using NULL for ENCODING is that svn_subst_translate_string2() calls svn_utf_cstring_to_utf8() on the string in that case, implying some sort of re-encoding from the native narrow string encoding to UTF-8. Couldn't this be problematic?
