
Instead of opening the file in unicode mode, we need only encode data which potentially has non-ASCII characters as UTF8 before writing. This should work across both Python versions, and is also far simpler than anything else discussed. llvm-svn: 258969