- 25% speed increse in tell();
- f.seek(0, 1) now maps to f.seek(f.tell(), 0) instead of to f.tell().

call self.buffer.tell().

Restore complex pickling.
Use cPickle in io.py.

without touching io.py or test_io.py.  The cause of the failure was that
bytes objects didn't pickle right.  As a stop-gap measure, I'm providing
bytes pickling via copy_reg.  Eventually, we should use a more efficient
protocol, e.g. __reduce_ex__ or __getstate__/__setstate__.

There is somewhat working (but slow) code supporting seek/tell for text files,
but extensive testing exposes a bug I can't nail down.

write() returns the number of bytes/characters written/buffered.
FileIO.close() calls self.flush().
Implement readinto() for buffered readers.
Tests th check all these.
Test proper behavior of __enter__/__exit__.

Reduce large file size to 2**31 (and a bit).

Etc., etc.

Removed unused _PyFileIO class.
Changed inheritance structure.
TODO: do the same kinds of things to TextIO.

Add tests for str <cmpop> bytes.

bytes objects can now be compared to anything that supports the buffer API.

(The new I/O library revealed the problem.)

Add closed attribute.
Support int argument to open() -- wrapping a file descriptor.
For b/w compat, support readline(n).
Support readlines() and readlines(n).
Flush on __del__.
Added some XXX comments.

instead of a filename.
Add a 'closed' attribute.

"import exceptions" to behave bizarrely.

(Should be backported to 2.6, really.)

This now appears to work when io.open is substituted for the real open
in fileinput.py -- at least the latter's unit tests pass.

Some cleanup of test_io.py and io.py.
Added seeking to buffered reader and writer, but no tests yet.

This is essentially a checkpoint.

mystery italics

Make readonly members defined in C throw an AttributeError on modification. This brings them into sync with Python-level attributes. Fixes bug #1687163.

(There was also a segfault but it disappeared when the tests
stopped erroring out; I presume the segfault is a pre-existing
problem somewhere in a destructor.)

NotImplemented.

(Is this worth backporting to 2.6?  It seems so useful...!)

inappropriately from ``e[0]`` to ``e.message`` instead of ``e.args[0]``.  The
reason it needs to be the last option is the dichotomy of 'message' and 'args':
'message' can be the empty string but args[0] can have a value if more than one
argument was passed.

(A symbol table entry was leaked every time a class was compiled.)

 * make some module variables static to prevent name pollution
 * Add some comments to clarify what's going on and some XXXs to address
 * Add a space after "for" before (
 * exc_value and tb can be NULL in some cases
 * Get working on Windows (I think)