Re: [gpsd-dev] RTCM relaying problem

[Eric S. Raymond]

José Miguel Gonçalves <address@hidden>:
> Working with a RTCM caster I've noticed that not all RTCM packets
> that gpsd received were relayed to my GPS receiver, what caused DGPS
> corrections to be intermittently applied or not applied at all. To
> fix this I needed to apply the following patch to avoid RTCM packet
> overlapping:
> 
> 
> diff --git a/gpsd.c b/gpsd.c
> index d8f229b..abdaca1 100644
> --- a/gpsd.c
> +++ b/gpsd.c
> @@ -1422,7 +1422,7 @@ static void consume_packets(struct
> gps_device_t *device)
>      }
>  #endif /* NETFEED_ENABLE */
> 
> -    for (fragments = 0; ; fragments++) {
> +    for (fragments = 0; (changed & (RTCM2_SET | RTCM3_SET)) == 0;
> fragments++) {
>         changed = gpsd_poll(device);
> 
>         if (changed == ERROR_SET) {

This code is not correct.  The variable 'changed' is auto and not initialized,
which means random garbage on the stack could cause the loop to fail before
it is ever set.

But simply patching in an initialization wouldn't be the right thing here.
We need to figure out what is actually going on, because if you look at line
1519 and following you'll see exactly where the RTCM2/RTCM3 case is handled
*inside the loop this test is terminating*.  Your patch should, accordingly
have made things worse.  We need to understand why it didn't and incorporate
that understanding in the eventual fix. 
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>