Attempted fix for a close/reopen bug involvinf NMEA2000.

This was reported on gosd-users by Vignesh Krishnan <Krishnan_Vignesh@cat.com>:

> For the nmea2000 bug . I connect using gpsd control socket
>
> ./gpsd -n -F /tmp/control_sock
>
> Adding source is possible
> +nmea2000://can1
>
> Removing sourc is also possible with
> -nmea2000://can1
>
> Readding the source is not possible with
> +nmea2000://can1
>
> This is because the there is an array which stores the list of nmea
> devices. That is not cleared on removed. So adding the second time results
> in device exist error when actually device does not exist.

To address this, the driver close() method tries to zero the proper
array location.

All regression tests pass.

1 files changed, 15 insertions, 0 deletions

diff --git a/driver_nmea2000.c b/driver_nmea2000.c
index 309708b3..5b1e207d 100644
--- a/driver_nmea2000.c
+++ b/driver_nmea2000.c
@@ -1677,6 +1677,21 @@ void nmea2000_close(struct gps_device_t *session)
 		 session->gpsdata.gps_fd, session->gpsdata.dev.path);
 	(void)close(session->gpsdata.gps_fd);
 	INVALIDATE_SOCKET(session->gpsdata.gps_fd);
+
+	if (session->driver.nmea2000.unit_valid != 0) {
+	    unsigned int l1, l2;
+	    
+	    for (l1=0;l1<NMEA2000_NETS;l1++) {
+	        for (l2=0;l2<NMEA2000_UNITS;l2++) {
+		    if (session == nmea2000_units[l1][l2]) {
+		        session->driver.nmea2000.unit_valid = 0;
+		        session->driver.nmea2000.unit = 0;
+			session->driver.nmea2000.can_net = 0;
+			nmea2000_units[l1][l2] = NULL;
+		    }
+		}
+	    }
+	}
     }
 }