Re: Move pending_event to remote_notif_state ([PATCH 1/6] Move notif_queue to remote_state)

[Pedro Alves <palves@redhat.com>]

Eh, sorry, somehow I only noticed this patch now.

(Still, I'd have preferred if the previous patch didn't
create an intermediate state that leaks.)

Anyway...

On 09/30/2013 08:56 AM, Yao Qi wrote:

>  /* Return an allocated remote_notif_state.  */
>  
>  struct remote_notif_state *
> @@ -238,6 +246,8 @@ remote_notif_state_allocate (void)
>      = create_async_event_handler (remote_async_get_pending_events_handler,
>  				  notif_state);
>  
> +  memset (notif_state->pending_event, 0, sizeof (notif_state->pending_event));

Could use xzalloc to begin with.  Might be more future proof.

> @@ -5414,16 +5409,22 @@ discard_pending_stop_replies (struct inferior *inf)
>  {
>    int i;
>    struct queue_iter_param param;
> -  struct stop_reply *reply
> -    = (struct stop_reply *) notif_client_stop.pending_event;
> +  struct stop_reply *reply;
> +  struct remote_state *rs = get_remote_state ();
> +  struct remote_notif_state *rns = rs->notif_state;
> +
> +  /* This function can be notified when an inferior exists.  When the
> +     target is not remote, the notification state is NULL.  */
> +  if (rns == NULL)
> +    return;

I'd rather we use the same "target is remote" predicate we use everywhere else:

  if (rs->remote_desc == NULL)
    return;

Otherwise this looks fine.

Thanks,
-- 
Pedro Alves