liblo-devel Mailing List for LibLO

Hi,

Thanks for the bug report. I have added your patch to CVS.
I also added two new methods: lo_server_get_protocol() and =20
lo_address_get_protocol()

nick.


On 26 Mar 2006, at 13:43, Sz=E9kelyi Szabolcs wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> [please Cc: me, i'm not on liblo-devel]
>
> Hi!
>
> Tiny but ugly bug...
>
> lo_address_get_url() returns an osc.udp:// URL regardless of the
> protocol for the address specified in its argument.
>
> A bug report was also submitted to SF.net BTS.
>
> I've attached a test case and the patch. The test case should print =20=

> the
> same URL twice, but the second time prints a UDP URL instead of TCP.
>
> Bye,
> - --
> Sze'kelyi Szabolcs
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.2.2 (GNU/Linux)
>
> iD8DBQFEJoxwGJRwVVqzMkMRAuZqAJ941rIl0IcAQKgaBw6VJkOUie2yeACcCCZi
> 61jKMQgVLGY5iIp4NIXWD0A=3D
> =3DRUre
> -----END PGP SIGNATURE-----
> --- liblo-0.23.orig/src/address.c	2006-03-26 13:23:02.000000000 =
+0200
> +++ liblo-0.23/src/address.c	2006-03-26 14:01:31.000000000 +0200
> @@ -108,6 +108,21 @@
>      return a->port;
>  }
>
> +static const char* lo_address_get_protocol_name(int proto)
> +{
> +    switch(proto) {
> +    case LO_UDP:
> +        return "udp";
> +    case LO_TCP:
> +        return "tcp";
> +#ifndef WIN32
> +    case LO_UNIX:
> +        return "unix";
> +#endif
> +    }
> +    return NULL;
> +}
> +
>  char *lo_address_get_url(lo_address a)
>  {
>      char *buf;
> @@ -120,13 +135,15 @@
>      } else {
>  	fmt =3D "osc.%s://%s:%s/";
>      }
> -    ret =3D snprintf(NULL, 0, fmt, "udp", a->host, a->port);
> +    ret =3D snprintf(NULL, 0, fmt,
> +		   lo_address_get_protocol_name(a->proto), a->host, =
a->port);
>      if (ret <=3D 0) {
>  	/* this libc is not C99 compliant, guess a size */
>  	ret =3D 1023;
>      }
>      buf =3D malloc((ret + 2) * sizeof(char));
> -    snprintf(buf, ret+1, fmt, "udp", a->host, a->port);
> +    snprintf(buf, ret+1, fmt,
> +	     lo_address_get_protocol_name(a->proto), a->host, a->port);
>
>      return buf;
>  }
> #include <stdio.h>
> #include <lo/lo.h>
>
> int main(int argc, char** argv)
> {
>   lo_server s =3D lo_server_new_with_proto("1025", LO_TCP, NULL);
>   char* url1 =3D lo_server_get_url(s);
>   char* url2 =3D lo_address_get_url(lo_address_new_from_url(url1));
>
>   fprintf(stderr, "%s\n", url1);
>   fprintf(stderr, "%s\n", url2);
>
>   return 0;
> }
>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

[please Cc: me, i'm not on liblo-devel]

Hi!

Tiny but ugly bug...

lo_address_get_url() returns an osc.udp:// URL regardless of the
protocol for the address specified in its argument.

A bug report was also submitted to SF.net BTS.

I've attached a test case and the patch. The test case should print the
same URL twice, but the second time prints a UDP URL instead of TCP.

Bye,
- --
Sze'kelyi Szabolcs

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)

iD8DBQFEJoxwGJRwVVqzMkMRAuZqAJ941rIl0IcAQKgaBw6VJkOUie2yeACcCCZi
61jKMQgVLGY5iIp4NIXWD0A=
=RUre
-----END PGP SIGNATURE-----

Hi,

I've been attempting to use a liblo server in TCP mode to send multiple
messages on a connection that stays open.  I ended up looking through the
source code and it doesn't seem possible to do this.

In the function,
void *lo_server_recv_raw_stream(lo_server s, size_t *size),

the socket is accepted:
sock = accept(s->socket, (struct sockaddr *)&addr, &addr_len);

then it's used to receive a single message, but there's no way to get to
the connected socket after this. It's not stored anywhere, and apparently
not closed either:

    //close(sock);

I've been thinking about how this might be fixed.  What do you think about
master/slave servers?  The master would listen on the socket, and when a
connection is accept()ed, a new slave lo_server would be created that uses the
new socket.  lo_server would need a list of slaves (add a next pointer so they
can be linked together?) and there would probably need to be some functions to
get access to the them: lo_server_get_first_slave(), lo_server_get_next_slave()
or something like that.

Also, as Dave Robillard points out in the bug tracker:
"lo_server_thread_new_with_proto missing
The corresponding function to lo_server_new_with_proto
for server threads is missing, which means you can't
make a server thread which listens to TCP (not good)"

I'm thinking that TCP server threads could work in two ways:
1) New thread for each connection
2) Do the select() thing on all of the slave connections

I may try some of these ideas out and get back with some patches, but wanted to
see what developers thought.

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com