Use StringIO to write out requirements. Use more common convention of adding newline to each line of the file, not just intervening lines.

--HG--
extra : rebase_source : c22b55cde69bbf7fc6a075bcd8797017797c6225

est_manifest_is_written_with_utf8_encoding should use the filelist's append, NOT filelist.files.append

--HG--
extra : rebase_source : cdf0df0a11b96a60c29c10483579e6e043a7fcd1

Convert quiet to a context manager, so its comment is no longer complaining and to simplify the code.

--HG--
extra : source : 199c917b8a0be209144878872269c3bd08936d6a

--HG--
extra : source : 2e47fa11a272ed61f7c1bbf88aae27e81040fe93

*Since py2.5 has been dropped, we can use future imports to
make use of with statements.

*End goal was to always decode to utf-8 in write_file on 307

--HG--
extra : rebase_source : 502ea7128f4e3b843b16c6d64d6d0b2ac56ce87d

--HG--
extra : rebase_source : 0dba3308549833dc7fc5b242e8ae3a4ec9f3c119

FileList.append already excludes files that are not UTF-8 encodable, so rely on it when building the manifest. Ref #193

Test should use the same high-level interface for appending filenames as the object itself will use when building the filelist. Ideally, FileList wouldn't expose this attribute at all.

This code path doesn't make sense. If the UnicodeEncodeError occurred above, it will occur here too.

--HG--
extra : rebase_source : 5ca0514bbe3ec025094ea8f39921722d0ab3dd05