On Mon, 2018-07-16 at 20:25 -0400, Yclept Nemo wrote:
> If 'one_per_name_type' is true, only add one tunnel per avahi service
> name and type. In effect, do not add more than one sink or source per
> remote pulseaudio server.
Is avahi service name the address? And is service type sink or source?
What's the reason you want this feature?
> ---
> src/modules/module-zeroconf-discover.c | 46 +++++++++++++++++++++++++++++++---
> 1 file changed, 43 insertions(+), 3 deletions(-)
>
> diff --git a/src/modules/module-zeroconf-discover.c b/src/modules/module-zeroconf-discover.c
> index 2c49d13a..1c960983 100644
> --- a/src/modules/module-zeroconf-discover.c
> +++ b/src/modules/module-zeroconf-discover.c
> @@ -53,6 +53,7 @@ PA_MODULE_LOAD_ONCE(true);
> static const char* const valid_modargs[] = {
> "disable_ipv4",
> "disable_ipv6",
> + "one_per_name_type",
> NULL
> };
>
> @@ -74,6 +75,14 @@ struct userdata {
> pa_hashmap *tunnels;
> };
>
> +static unsigned tunnel_hash_simple(const void *p) {
> + const struct tunnel *t = p;
> +
> + return
> + pa_idxset_string_hash_func(t->name) +
> + pa_idxset_string_hash_func(t->type) +
> +}
This doesn't build.
> +
> static unsigned tunnel_hash(const void *p) {
> const struct tunnel *t = p;
>
> @@ -85,6 +94,18 @@ static unsigned tunnel_hash(const void *p) {
> pa_idxset_string_hash_func(t->domain);
> }
>
> +static int tunnel_compare_simple(const void *a, const void *b) {
> + const struct tunnel *ta = a, *tb = b;
> + int r;
> +
> + if ((r = strcmp(ta->name, tb->name)))
> + return r;
> + if ((r = strcmp(ta->type, tb->type)))
> + return r;
> +
> + return 0;
> +}
> +
> static int tunnel_compare(const void *a, const void *b) {
> const struct tunnel *ta = a, *tb = b;
> int r;
> @@ -148,6 +169,12 @@ static void resolver_cb(
>
> tnl = tunnel_new(interface, protocol, name, type, domain);
>
> + if (pa_hashmap_get(u->tunnels, tnl)) {
> + pa_log_debug("Tunnel [%i,%i,%s,%s,%s] already mapped, skipping.",
> + interface, protocol, name, type, domain);
> + goto finish;
> + }
> +
> if (event != AVAHI_RESOLVER_FOUND)
> pa_log("Resolving of '%s' failed: %s", name, avahi_strerror(avahi_client_errno(u->client)));
> else {
> @@ -295,6 +322,8 @@ static void browser_cb(
>
> if (event == AVAHI_BROWSER_NEW) {
>
> + /* Since the resolver is asynchronous and the hashmap may not yet be
> + * updated, this check must be duplicated in the resolver callback. */
> if (!pa_hashmap_get(u->tunnels, t))
> if (!(avahi_service_resolver_new(u->client, interface, protocol, name, type, domain, u->protocol, 0, resolver_cb, u)))
> pa_log("avahi_service_resolver_new() failed: %s", avahi_strerror(avahi_client_errno(u->client)));
> @@ -304,9 +333,11 @@ static void browser_cb(
> * it from the callback */
>
> } else if (event == AVAHI_BROWSER_REMOVE) {
> - struct tunnel *t2;
> + struct tunnel *t2 = pa_hashmap_get(u->tunnels, t);
>
> - if ((t2 = pa_hashmap_get(u->tunnels, t))) {
> + /* A full comparison is required even if 'one_per_name_type' is true.
> + * Yes, this is redundant if it's false. */
This comment is a bit weird. It would make more sense to me if you
dropped the words "even" and "yes". You could also add an explanation
why it's redundant when one_per_name_type is false.
> + if (t2 && !tunnel_compare(t2, t)) {
> pa_module_unload_request_by_index(u->core, t2->module_index, true);
> pa_hashmap_remove(u->tunnels, t2);
> tunnel_free(t2);
--
Tanu
https://www.patreon.com/tanuk
https://liberapay.com/tanuk
