[Spice-devel] [PATCH spice-server 3/3] red-channel: Allows to retrieve client callback saves pointer

[Christophe Fergeau]

On Fri, Aug 25, 2017 at 11:24:41AM +0100, Frediano Ziglio wrote:
> The data pointer for client callbacks was not used.
> The code was saving this information using callback registration
> and g_object_set_data. Remove the g_object_set_data using
> the data registered.

Ah, hmm, I actually had different plans for this 'data' stuff, which is
to just kill it ;) (and do some more reworking of the client_cbs code).
Yet another patch series I have lying around but never got around to
testing/sending. Yes this means to keep living with the
g_object_set_data() use for 'dispatcher', I did not really pay attention
to it until now, so don't know if there are alternative ways of getting
rid of it.

Some comments below if we go this way:

> 
> Signed-off-by: Frediano Ziglio <fziglio at redhat.com>
> ---
>  server/red-channel.c |  5 +++++
>  server/red-channel.h |  2 ++
>  server/red-qxl.c     | 12 ++++++------
>  server/red-worker.c  |  2 --
>  4 files changed, 13 insertions(+), 8 deletions(-)
> 
> diff --git a/server/red-channel.c b/server/red-channel.c
> index 9ff3474a..3c4d236b 100644
> --- a/server/red-channel.c
> +++ b/server/red-channel.c
> @@ -372,6 +372,11 @@ void red_channel_register_client_cbs(RedChannel *channel, const ClientCbs *clien
>      channel->priv->data = cbs_data;
>  }
>  
> +void *red_channel_get_registered_client_cbs_data(RedChannel *channel)

I'd just name this "red_channel_get_client_cbs_data()

> +{
> +    return channel->priv->data;

This could be channel->priv->client_cbs.data I think? I really prefer
when the user data is where it belongs, rather than a generic 'data'
member in the bigger Channel class.

> +}
> +
>  static void add_capability(uint32_t **caps, int *num_caps, uint32_t cap)
>  {
>      int nbefore, n;
> diff --git a/server/red-channel.h b/server/red-channel.h
> index e65eea1e..0068294a 100644
> --- a/server/red-channel.h
> +++ b/server/red-channel.h
> @@ -136,6 +136,8 @@ void red_channel_remove_client(RedChannel *channel, RedChannelClient *rcc);
>  void red_channel_init_stat_node(RedChannel *channel, const RedStatNode *parent, const char *name);
>  
>  void red_channel_register_client_cbs(RedChannel *channel, const ClientCbs *client_cbs, gpointer cbs_data);
> +// retrieve data pointer stored with red_channel_register_client_cbs
> +void *red_channel_get_registered_client_cbs_data(RedChannel *channel);
>  // caps are freed when the channel is destroyed
>  void red_channel_set_common_cap(RedChannel *channel, uint32_t cap);
>  void red_channel_set_cap(RedChannel *channel, uint32_t cap);
> diff --git a/server/red-qxl.c b/server/red-qxl.c
> index ba869e54..0eaf0e83 100644
> --- a/server/red-qxl.c
> +++ b/server/red-qxl.c
> @@ -82,7 +82,7 @@ static void red_qxl_set_display_peer(RedChannel *channel, RedClient *client,
>      Dispatcher *dispatcher;
>  
>      spice_debug("%s", "");
> -    dispatcher = (Dispatcher *)g_object_get_data(G_OBJECT(channel), "dispatcher");
> +    dispatcher = (Dispatcher *)red_channel_get_registered_client_cbs_data(channel);

It might be more consistent to do
 typedef void (*channel_client_connect_proc)(RedChannel *channel, RedClient *client, RedsStream *stream,
-                                              int migration, RedChannelCapabilities *caps);
+                                              int migration, RedChannelCapabilities *caps, gpointer user_data);
instead, and to not have a get_client_cbs_data() method at all.

Maybe wait a bit until I take a closer look at the work I had done
before reworking this series?

Christophe