locking around source_funcs->finalize

[Tim Janik <timj imendio com>]

hi Owen.

reading up on g_source_unref() again, i notice that the 
GMainContext lock is removed and re-acquired around
callback_funcs->unref(), but not around source_funcs->finalize();
supposing you're unlocking around unref() so that the
handler may do things like adding a new idle handler
to the context, shouldnt the same semantics be provided
for source_funcs->finalize() as well?

> static void
> g_source_unref_internal (GSource      *source,
>                          GMainContext *context,
>                          gboolean      have_lock)
> {
>   gpointer old_cb_data = NULL;
>   GSourceCallbackFuncs *old_cb_funcs = NULL;
>
>   g_return_if_fail (source != NULL);
>
>   if (!have_lock && context)
>     LOCK_CONTEXT (context);
>
>   source->ref_count--;
>   if (source->ref_count == 0)
>     {
>       old_cb_data = source->callback_data;
>       old_cb_funcs = source->callback_funcs;
>
>       source->callback_data = NULL;
>       source->callback_funcs = NULL;
>
>       if (context && !SOURCE_DESTROYED (source))
>         {
>           g_warning (G_STRLOC ": ref_count == 0, but source is still attached to a context!");
>           source->ref_count++;
>         }
>       else if (context)
>         g_source_list_remove (source, context);
>
>       if (source->source_funcs->finalize)
>         source->source_funcs->finalize (source);

lock is being held if have_lock || context.

>       g_slist_free (source->poll_fds);
>       source->poll_fds = NULL;
>       g_free (source);
>     }
>
>   if (!have_lock && context)
>     UNLOCK_CONTEXT (context);
>
>   if (old_cb_funcs)
>     {
>       if (have_lock)
>         UNLOCK_CONTEXT (context);
>
>       old_cb_funcs->unref (old_cb_data);

no lock is being held.

>       if (have_lock)
>         LOCK_CONTEXT (context);
>     }
> }


---
ciaoTJ