From: Guenter M. <mi...@us...> - 2011-12-15 00:01:08
|
On 2011-12-12, Martin Bless wrote: > to cut a long story short: The answer is, well, not 42, but UTF-8! > I've done a lot of (re)search and testing now and I'm absolutely sure > that nothing else makes sense. You convinced me. (With a little help by David A. Wheeler http://www.dwheeler.com/essays/fixing-unix-linux-filenames.html). Encoding the dependency_record file in utf8 is * simple, save, portable * works out of the box in many cases. * not specific to Python, easy to recode (see example in config.txt) * backwards compatible with the "ascii" encoding used in releases <= 0.8. The alternative would be guesswork (locale_encoding or sys.filesystemencoding if not 'mcbs'). See Revision 7256 for the implementation, test, and documentation. I hope this fixes the remaining issues. Günter |