It would be great if the documentation of c_char_p
(http://docs.python.org/library/ctypes.html#ctypes.c_char_p) could be
reformulated as follows (would have saved me quite some time):

class ctypes.c_char_p¶
    Represents the C char * datatype when it points a zero-terminated
string. For a general character pointer that may also point to binary
data, POINTER(c_char) must be used. The constructor accepts an integer
address, or a string.

Wouldn't a "general" pointer better be of type c_void_p?

I don't want to judge if the best way to represent binary data in C is a
void* or char*, but there is a lot of C code out there that uses char*,
and if we want to interface with such a library we need to use
POINTER(c_char).

Note that my docpatch doesn't say anything about using void_p or not:
"For a general *character* pointer that may point to binary data,
POINTER(c_char) must be used". This refers only to char*, so if the C
code uses void* instead, we can of course also use c_void_p on the
Python side.

The suggestion looks good to me.

OK, applied in r77108.