Re: [gpsd-dev] stuck in consume_packets/multipoll loop

[Eric S. Raymond]

Bernd Ocklin <address@hidden>:
> Hope this looks okay.
> 
> 
> diff --git a/libgpsd_core.c b/libgpsd_core.c
> index a8687df..68f36c1 100644
> --- a/libgpsd_core.c
> +++ b/libgpsd_core.c
> @@ -1548,6 +1548,18 @@ int gpsd_multipoll(const bool data_ready,
>             /* handle data contained in this packet */
>             if (device->packet.type != BAD_PACKET)
>                 /address@hidden@*/handler(device, changed);
> +
> +#ifdef __future__
> +           /* Exit when a full packet was received and parsed.
> +            * This allows other devices to be serviced even if
> +            * this device delivers a full packet at every single
> +            * read.
> +            * Otherwise we can sit here for a long time without
> +            * any for-loop exit condition being met.
> +            */
> +           if (changed & PACKET_SET)
> +               break;
> +#endif // __future__
>         }
>      }
>      else if (device->reawake>0 && timestamp()>device->reawake) {
> 
> 
> I think its better to exit the poll loop after each fully read and
> parsed package. Allowing to poll devices in a more fair round-robin
> fashion.

The patch is in good form.  I question whether it is actually the
right place to do this thing, though.  We'll consider it more
carefully after 3.10 has shipped.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>