Document the data-management problem underlying ckuethe's POLL bug.

All regression tests pass.

2 files changed, 14 insertions, 0 deletions

diff --git a/gpsd.c b/gpsd.c
index 0220333e..81025a70 100644
--- a/gpsd.c
+++ b/gpsd.c
@@ -1104,6 +1104,13 @@ static void handle_request(struct subscriber_t *sub,
 	for (devp = devices; devp < devices + MAXDEVICES; devp++) {
 	    if (allocated_device(devp) && subscribed(sub, devp)) {
 		if ((devp->observed & GPS_TYPEMASK)!=0) {
+		    /*
+		     * Architectural problem: This is going to see just the
+		     * data dropped in during the last gps_poll() call.
+		     * What we really need to do is have the drivers put
+		     * new data someplace other than devp->gpsdata.fix, so
+		     * that it can be returned to use for accumulated fix data. 
+		     */
 		    json_tpv_dump(&devp->gpsdata, 
 				  reply + strlen(reply),
 				  replylen - strlen(reply));
diff --git a/libgpsd_core.c b/libgpsd_core.c
index abb7215f..f38f14c2 100644
--- a/libgpsd_core.c
+++ b/libgpsd_core.c
@@ -580,6 +580,13 @@ gps_mask_t gpsd_poll(struct gps_device_t *session)
     ssize_t newlen;
     bool first_sync = false;
 
+    /*
+     * This looks strange, but it works because the
+     * gpsdata.fix buffer has been pressed into service
+     * as a place for the drivers to drop new data. 
+     * The last accumulated fix actually lives in
+     * fixbuffer now.  This should probably be fixed.
+     */
     gps_clear_fix(&session->gpsdata.fix);
 
 #ifdef TIMING_ENABLE