Navit goes into tight loop if gpsd socket is closed

[mattcallow]

when using gpsd (for example with gpsfake, but possibly with a real gps as well), if gpsd is stopped (and the socket closed) Navit does into an endless loop. This is beacuse it does not detect that the socket has closed.

Possible patch:

Index: vehicle/gpsd/vehicle_gpsd.c
===================================================================
RCS file: /cvsroot/navit/navit/src/vehicle/gpsd/vehicle_gpsd.c,v
retrieving revision 1.4
diff -u -r1.4 vehicle_gpsd.c
--- vehicle/gpsd/vehicle_gpsd.c 10 Jan 2008 17:56:33 -0000      1.4
+++ vehicle/gpsd/vehicle_gpsd.c 11 Jan 2008 04:26:19 -0000
@@ -85,8 +85,10 @@
        } else
                priv->gps = gps_open(source + 7, NULL);
        g_free(source);
-       if (!priv->gps)
+       if (!priv->gps) {
+               dbg(0, "gps_open failed\n");
                return 0;
+       }
        gps_query(priv->gps, "w+x\n");
        gps_set_raw_hook(priv->gps, vehicle_gpsd_callback);
        priv->iochan = g_io_channel_unix_new(priv->gps->gps_fd);
@@ -124,7 +126,12 @@
        if (condition == G_IO_IN) {
                if (priv->gps) {
                        vehicle_last = priv;
-                       gps_poll(priv->gps);
+                       if (gps_poll(priv->gps)) {
+                               dbg(0, "gps_poll failed\n");
+                               vehicle_gpsd_close(priv);
+                               vehicle_gpsd_open(priv);
+                       }
+
                }
                return TRUE;
        }

Would a retry mechanism be appropriate if gps_open fails ?

[KaZeR]

Patch has been applied. Thank you. A retry mechanism would be useful yes, so that if gpsd is started too late, navit can retry the connection. I let the ticket open for now until it is implemented.

[mattcallow]

Retry has been added to cvs.