[EMAIL PROTECTED] (Ludovic Courtès) writes:
> Following this discussion, I propose the following addition which
> exposes the wrapping/unwrapping functions of `sockaddr' objects.
Hmm, I think your patch mixes the two ways we have to express a socket
address: one way is an argument convention used by connect, bind, and
sendto; the other way is a vector with the relevant data inside, as
returned by accept, getsockname, getpeername, and recvfrom!.
The scm_from_ and scm_to_ functions are meant to convert between a C
representation and a Scheme representation of some value. Thus, one
should be able to do things like
scm_equal_p (OBJ, scm_from_sockaddr (scm_to_sockaddr (OBJ)))
for example, and ideally get #t.
Thus, I propose the following functions:
- SCM scm_from_sockaddr (const struct sockaddr *addr, size_t addr_size);
Return the SCM representation of ADDR, which is a vector ...
- struct sockaddr *scm_to_sockaddr (SCM addr);
Return a newly-allocated `sockaddr' structure that reflects ADDR,
which is a SCM vector as returned by scm_from_sockaddr, scm_accept,
etc. The returned structure may be freed using `free ()'.
- SCM scm_make_socket_address (SCM fam, SCM address, SCM args);
Construct a socket address like connect would given the three
arguments and return it.
- struct sockaddr *scm_c_make_socket_address (SCM fam, SCM address,
SCM args);
Equivalent to scm_to_sockaddr (scm_make_socket_address (...)).
Also, scm_connect, scm_bind, and scm_sendto should probably be changed
to accept a SCM representation of a socket address, so that you could
write, for example:
(connect sock (make-socket-address ...))
That change should be backwards compatible, of course, which is a
little bit of pain because of the C API...
Ludovic, could you update your patch if you agree?
--
GPG: D5D4E405 - 2F9B BCCC 8527 692A 04E3 331E FAF8 226A D5D4 E405
_______________________________________________
Guile-user mailing list
Guile-user@gnu.org
http://lists.gnu.org/mailman/listinfo/guile-user
