Socket duplication for windows

[sbt]

http://bugs.python.org/review/14310/diff/4411/15802
File Modules/socketmodule.c (right):

http://bugs.python.org/review/14310/diff/4411/15802#newcode3783
Modules/socketmodule.c:3783: if (!PyArg_ParseTuple(arg, "i", &processId))
Shouldn't "i" be "I" since DWORD is an unsigned 32bit int?

http://bugs.python.org/review/14310/diff/4411/15802#newcode3799
Modules/socketmodule.c:3799: the socket.");
I think you should say something about the string being passed as the 4th
argument of the constructor.

Also changes to the docs are necessary.

http://bugs.python.org/review/14310/diff/4411/15802#newcode3965
Modules/socketmodule.c:3965: /* recreate a socket that was duplicated */
You seems to have tabs here instead of spaces.

[Jim.Jewett]

http://bugs.python.org/review/14310/diff/4411/15802
File Modules/socketmodule.c (right):

http://bugs.python.org/review/14310/diff/4411/15802#newcode3835
Modules/socketmodule.c:3835: sock_duplicate_doc},
Aren't new methods usually added to the end, to provide at least a shot at
binary compatibility?  Or does the above #ifdef HAVE_GETPEERNAME already blow
that out of the water?

http://bugs.python.org/review/14310/diff/4411/15802#newcode3970
Modules/socketmodule.c:3970: "socket descriptor string has wrong size");
Is it worth returning the actual value and/or a sample expected value (or at
least the expected size)?

[krisvale]

Thanks, I'll submit an updated version soon.

http://bugs.python.org/review/14310/diff/4411/15802
File Modules/socketmodule.c (right):

http://bugs.python.org/review/14310/diff/4411/15802#newcode3783
Modules/socketmodule.c:3783: if (!PyArg_ParseTuple(arg, "i", &processId))
On 2012/03/15 12:36:06, sbt wrote:
> Shouldn't "i" be "I" since DWORD is an unsigned 32bit int?
Quite, thanks.

http://bugs.python.org/review/14310/diff/4411/15802#newcode3799
Modules/socketmodule.c:3799: the socket.");
On 2012/03/15 12:36:06, sbt wrote:
> I think you should say something about the string being passed as the 4th
> argument of the constructor.
> 
> Also changes to the docs are necessary.
Right, the fourth argument.  But I'm in the habit of not editing docs until it
is likely that the functional changes of my patches are likely to be accepted.
Code first, then docs, my grandpa always said. :)

http://bugs.python.org/review/14310/diff/4411/15802#newcode3835
Modules/socketmodule.c:3835: sock_duplicate_doc},
On 2012/03/15 16:37:01, Jim.Jewett wrote:
> Aren't new methods usually added to the end, to provide at least a shot at
> binary compatibility?  Or does the above #ifdef HAVE_GETPEERNAME already blow
> that out of the water?
The methods get added to a dict of methods in the class.  There is no binary
compatibility.  Anyone trying to invoke methods from binary relying on some
private method definition table is in for trouble.

http://bugs.python.org/review/14310/diff/4411/15802#newcode3965
Modules/socketmodule.c:3965: /* recreate a socket that was duplicated */
On 2012/03/15 12:36:06, sbt wrote:
> You seems to have tabs here instead of spaces.
Indeed.  That would be fixed on submission, must fix my editor settings.

http://bugs.python.org/review/14310/diff/4411/15802#newcode3970
Modules/socketmodule.c:3970: "socket descriptor string has wrong size");
On 2012/03/15 16:37:01, Jim.Jewett wrote:
> Is it worth returning the actual value and/or a sample expected value (or at
> least the expected size)?
Sure, I can return the expected size.  But the user really should know what he
was doing.

[sbt]

http://bugs.python.org/review/14310/diff/4472/15947
File Modules/socketmodule.c (right):

http://bugs.python.org/review/14310/diff/4472/15947#newcode3794
Modules/socketmodule.c:3794: "duplicate(process_id) -> bytes\n\
Should be "fromshare(process_id) -> bytes"

http://bugs.python.org/review/14310/diff/4472/15947#newcode3799
Modules/socketmodule.c:3799: socket.fromshared().");
Should be "socket.fromshare()"

http://bugs.python.org/review/14310/diff/4472/15947#newcode3965
Modules/socketmodule.c:3965: /* recreate a socket that was duplicated */
You still seem to be using 4th argument of socket constructor instead of
implementing socket.fromshare().

[krisvale]

http://bugs.python.org/review/14310/diff/4472/15947
File Modules/socketmodule.c (right):

http://bugs.python.org/review/14310/diff/4472/15947#newcode3794
Modules/socketmodule.c:3794: "duplicate(process_id) -> bytes\n\
On 2012/03/19 19:55:20, sbt wrote:
> Should be "fromshare(process_id) -> bytes"

Done.

http://bugs.python.org/review/14310/diff/4472/15947#newcode3799
Modules/socketmodule.c:3799: socket.fromshared().");
On 2012/03/19 19:55:20, sbt wrote:
> Should be "socket.fromshare()"

Done.

http://bugs.python.org/review/14310/diff/4472/15947#newcode3965
Modules/socketmodule.c:3965: /* recreate a socket that was duplicated */
On 2012/03/19 19:55:21, sbt wrote:
> You still seem to be using 4th argument of socket constructor instead of
> implementing socket.fromshare().
see socket.py.  socket.fromshare() is implemented in the same way as
socket.fromfd(), by invoking the socket.socket() constructor with the
undocumented fourth argument.