Hello. I noticed test_mailbox (test_add) fails on my win2k machine.
It's something like this.

======================================================================
ERROR: test_add (__main__.TestMbox)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "test_mailbox.py", line 60, in tearDown
    self._box.close()
  File "e:\python-dev\py3k\lib\mailbox.py", line 642, in close
    self.flush()
  File "e:\python-dev\py3k\lib\mailbox.py", line 600, in flush
    stop - self._file.tell()))
  File "e:\python-dev\py3k\lib\io.py", line 1625, in tell
    chars_decoded += len(decoder.decode(next_byte))
  File "e:\python-dev\py3k\lib\io.py", line 1295, in decode
    output = self.decoder.decode(input, final=final)
TypeError: decode() argument 1 must be string or pinned buffer, not
bytearray

And this is simple reproducable code. ("a.txt" contains some text)

f = open("a.txt")
f.read(1)
f.tell() # boom

I searched the place where raises this error in C code,
and I found mbidecoder_decode() is.

    if (!PyArg_ParseTupleAndKeywords(args, kwargs, "t#|i:decode",
    		incrementalkwarglist, &data, &size, &final))
	return NULL;

This uses "t#", so cannot accept bytearray, probably.
I hope attached file solves this issue. Thank you.

Your patch looks right to me. It should also be backported to 2.6.

Well, MultibyteCodec_Decode() also uses Py_buffer in py3k but not in
trunk. Is this also backported?

Le dimanche 17 août 2008 à 11:44 +0000, Hirokazu Yamamoto a écrit :
> Hirokazu Yamamoto <ocean-city@m2.ccsnet.ne.jp> added the comment:
> 
> Well, MultibyteCodec_Decode() also uses Py_buffer in py3k but not in
> trunk. Is this also backported?

It should be :-)

OK, done.

Please state the revision numbers when closing an issue due to a commit;
this makes it easier to find the change later.

Sorry, fixed in r65760 and r65762.