"codecs" module on Windows writes incorrect end-of-line, making it
impossible to write Unicode files.

See below, how-to reproduce bug (Python 2.5.1 on Windows XP)
===================================================================

#buggy unicode support module:
import codecs
filewr=codecs.open('myfile.txt','w','utf-8')
filewr.write("abc"+"\n")
===================================================================
Now, try to open this 'myfile.txt' using Windows Notepad.
The bug is perfectly visible.

The code below, will give correct results however:
===================================================================

filewr=open('myfile.txt','w')

filewr.write("abc"+"\n")
===================================================================
Basically this bugs _prevents_ me from writing Unicode text files.

NOTE: I'm not sure, if this bug should relate to "Windows" or "Unicode"
component.
NOTE: This bug is reproducible, even without writing Unicode characters.

-Technologov, 16.02.2008.

Could explain what exactly is wrong with the end-of-line on Windows ?

Note that "Unicode text files" on Windows are generally interpreted as
UTF-16 encoded files. Perhaps that's what makes you think there's a bug.

OK: try
filewr.write("abc"+"\n"+"abc")

The file will be generated with 7 bytes in it (must be 8, because
Windows has two-byte line-end).

Without using "codecs" modules, everything works fine, and the file will
have 8-bytes in it. (see 2nd example)

Plus, the text will be corrupted when opened with Windows Notepad.

-Technologov

As stated in the codecs.open() docstring: """Files are always opened in
binary mode, even if no binary mode was specified. This is done to avoid
data loss due to encodings using 8-bit values""". This certainly means
you have to insert "\r\n" yourself (instead of just "\n") if you want
the file contents to respect the end-of-line convention under Windows...

As Antoine already pointed out: the codecs.open() function does not
support the C lib's text mode. As a result, no magical conversion of a
single newline to a CRLF takes place.

Closing as invalid.

The note in the docstring wasn't in the documentation. Fixed this in r60873.

Thanks, Georg.