This patch adds sethostname to the socket module (since socket has gethostname()).

I'm not sure why you use PyTuple_Size() and PyTuple_GetItem().
You should be able to do a first call to PyArg_ParseTuple() (using the "S" specifier to mandate a bytes object), and call PyErr_Clear() and fallback to the second PyArg_ParseTuple() if the first one fails.

Here is a simplified version.

New patch including a check for PyObject_GetBuffer()'s return value, the missing declaration of HAVE_SETHOSTNAME in pyconfig.h.in, and a test for giving a bytes values to the function.

The patch looks good.

Just to be clear, on my system running "autoreconf" adds the correct stuff to pyconfig.h.in

Isn't it best to leave it up to the committer to generate "configure" and "pyconfig.h.in", especially since different autoconf versions can create a lot of differences?

> The patch looks good.
> 
> Just to be clear, on my system running "autoreconf" adds the correct
> stuff to pyconfig.h.in

Ah, strange. I used "autoconf" and it didn't...

> Isn't it best to leave it up to the committer to generate "configure"
> and "pyconfig.h.in", especially since different autoconf versions can
> create a lot of differences?

Right, but in this case the version was the same and it was easier to
create the patch that way.

> Ah, strange. I used "autoconf" and it didn't...

From the man page of autoreconf:
 Run  `autoconf'  (and  `autoheader', `aclocal', `automake', `autopoint'    
 (formerly `gettextize'), and `libtoolize' where appropriate) repeatedly
 to  remake  the GNU Build System files in the DIRECTORIES or the direc‐
 tory trees driven by CONFIGURE-AC (defaulting to `.').

I guess it autoheader that regenerates pyconfig.h.in

Patch committed in r88685 (3.3), closing.