http://git.collabora.co.uk/?p=user/cassidy/telepathy-glib;a=shortlog;h=refs/heads/simple-observer
should be ready for review.

It includes a example call observer. I can split it to a different branch for bug #24214 if needed but I think that should be easy enough to review.

+ * |[
+ * client = tp_simple_observer_new (dbus, TRUE, "MyObserver", FALSE,
+ *    my_observe_channels, user_data);

I'd like this comment to include a stub implementation of my_observe_channels:

static void
my_observe_channels (TpSimpleObserver *self,
    TpAccount *account,
    TpConnection *connection,
    GList *channels,
    TpChannelDispatchOperation *dispatch_operation,
    GList *requests,
    TpObserveChannelsContext *context,
    gpointer user_data)
{
  /* ... do something useful with the channels here ... */

  tp_observe_channels_context_accept (context);
}

I'd also prefer a more realistic filter: matching every TEXT channel is likely to be a mistake (unless your observer is written very carefully), and matching CONTACT TEXT channels would be more normal.

> + * but are not garanteed to be prepared.

"guaranteed"

> + * @channels: a #GPtrArray of #TpChannel having all %TP_CHANNEL_FEATURE_CORE
> + * prepared

"#GList of #TpChannel, all having"

> + * @user_data: arbitrary user-supplied data passed to the callback

This *is* the callback. "passed to tp_simple_observer_new()"

> +      case PROP_OBSERV_IMPL:

I'd prefer this property to be called "callback".

> +tp_simple_observer_new (TpDBusDaemon *dbus,

Shouldn't this take a GDestroyNotify for the user_data, too?

> + * Returns: a new #TpSimpleObserver

"Returns: (type Tp.SimpleObserver): a new #TpSimpleObserver" would be better, I think; we're deliberately "getting the type wrong" for C's benefit, but g-i (and documentation) can benefit from knowing more specifically.

It would be good to update telepathy-doc's example to use this, once it's in.

(In reply to comment #2)
> + * |[
> + * client = tp_simple_observer_new (dbus, TRUE, "MyObserver", FALSE,
> + *    my_observe_channels, user_data);
> 
> I'd like this comment to include a stub implementation of my_observe_channels:
> 
> static void
> my_observe_channels (TpSimpleObserver *self,
>     TpAccount *account,
>     TpConnection *connection,
>     GList *channels,
>     TpChannelDispatchOperation *dispatch_operation,
>     GList *requests,
>     TpObserveChannelsContext *context,
>     gpointer user_data)
> {
>   /* ... do something useful with the channels here ... */
> 
>   tp_observe_channels_context_accept (context);
> }

added (without the /* */, C comments can't be nested).

> I'd also prefer a more realistic filter: matching every TEXT channel is likely
> to be a mistake (unless your observer is written very carefully), and matching
> CONTACT TEXT channels would be more normal.

fixed.

> > + * but are not garanteed to be prepared.
> 
> "guaranteed"

fixed (one day I'll be able to spell this word).

> > + * @channels: a #GPtrArray of #TpChannel having all %TP_CHANNEL_FEATURE_CORE
> > + * prepared
> 
> "#GList of #TpChannel, all having"
Fixed.

> > + * @user_data: arbitrary user-supplied data passed to the callback
> 
> This *is* the callback. "passed to tp_simple_observer_new()"

Fixed.

> > +      case PROP_OBSERV_IMPL:
> 
> I'd prefer this property to be called "callback".

done.

> > +tp_simple_observer_new (TpDBusDaemon *dbus,
> 
> Shouldn't this take a GDestroyNotify for the user_data, too?

added.

> > + * Returns: a new #TpSimpleObserver
> 
> "Returns: (type Tp.SimpleObserver): a new #TpSimpleObserver" would be better, I
> think; we're deliberately "getting the type wrong" for C's benefit, but g-i
> (and documentation) can benefit from knowing more specifically.

done.

> It would be good to update telepathy-doc's example to use this, once it's in.

I opened bug #27888

Looks good to me!

Merged to master! Will be in 0.11.5.