[vlc-devel] [RFC PATCH] playlist/item: redirect request following deletion of node

[Thomas Guillem]

On Thu, Sep 1, 2016, at 05:15, Filip Roséen wrote:
> :: Introduction
> ----------------------------------------------------------------------
> 
> When we are dealing with a flat playlist, a request to add children to
> an item results in that item being deleted, and its children being
> inserted into the parent of said item. This is done in order to not
> introduce any nested levels in the tree.
> 
> What this further means is that /if/ the item that is effectivelly
> removed is
> scheduled for playback prior to the removal, the playlist might try to
> play an
> entity which does not exist anymore; which is problematic for several
> reasons.
> 
> Note: The problem does not exist when --playlist-tree is enabled since
>       the item on which children should be attached is then preserved.
> 
> :: Solution
> ----------------------------------------------------------------------
> 
> The solution to the previously described problem is to "redirect" any
> pending request that refers to the deleted entity, so that it instead
> refers to its first child (if any).
> 
> closes #17183
> 
> --
> 
> vlc-devel:
> 
>     This is the least intrusive change to the spaghetti code which is
>     the playlist that I could think of; anyone got some other viable
>     idea on how to solve the issue?
> 
> ---
>  src/playlist/item.c | 30 ++++++++++++++++++++++++++++++
>  1 file changed, 30 insertions(+)
> 
> diff --git a/src/playlist/item.c b/src/playlist/item.c
> index 9250284..d6966f5 100644
> --- a/src/playlist/item.c
> +++ b/src/playlist/item.c
> @@ -54,6 +54,7 @@ static void input_item_add_subitem_tree ( const
> vlc_event_t * p_event,
>  {
>      input_item_t *p_input = p_event->p_obj;
>      playlist_t *p_playlist = (( playlist_item_t* )
>      user_data)->p_playlist;
> +    playlist_private_t *p_sys = pl_priv( p_playlist );
>      input_item_node_t *p_new_root =
>      p_event->u.input_item_subitem_tree_added.p_root;
>  
>      PL_LOCK;
> @@ -91,6 +92,8 @@ static void input_item_add_subitem_tree ( const
> vlc_event_t * p_event,
>      /* If we have to flatten out, then take the item's position in the
>      parent as
>      insertion point and delete the item */
>  
> +    bool b_redirect_request = false;
> +
>      if( b_flat )
>      {
>          playlist_item_t *p_parent = p_item->p_parent;
> @@ -109,6 +112,17 @@ static void input_item_add_subitem_tree ( const
> vlc_event_t * p_event,
>  
>          playlist_DeleteItem( p_playlist, p_item, true );
>  
> +        /* If there is a pending request referring to the item we just
> deleted
> +         * it needs to be updated so that we do not try to play an
> entity that
> +         * is no longer part of the playlist. */
> +
> +        if( p_sys->request.b_request &&
> +            ( p_sys->request.p_item == p_item ||
> +              p_sys->request.p_node == p_item ) )
> +        {
> +            b_redirect_request = true;
> +        }
> +
>          p_item = p_parent;
>      }
>      else
> @@ -125,6 +139,22 @@ static void input_item_add_subitem_tree ( const
> vlc_event_t * p_event,
>                                                   p_new_root,
>                                                   pos,
>                                                   b_flat );
> +    if( b_redirect_request )
> +    {
> +        /* a redirect of the pending request is required, as such we
> update the
> +         * request to refer to the item that would have been the next in
> line
> +         * (if any). */
> +
> +        assert( b_flat );
> +
> +        playlist_item_t* p_redirect = NULL;
> +
> +        if( p_item->i_children > pos )
> +            p_redirect = p_item->pp_children[pos];
> +
> +        p_sys->request.p_item = p_redirect;
> +        p_sys->request.p_node = NULL;
> +    }
>  
>      if( !b_flat ) var_SetInteger( p_playlist, "leaf-to-parent",
>      p_item->i_id );
>  
> -- 
> 2.9.3

OK for me

> 
> _______________________________________________
> vlc-devel mailing list
> To unsubscribe or modify your subscription options:
> https://mailman.videolan.org/listinfo/vlc-devel