[Discuss-gnuradio] bug on get_mboard_sensor("gps_time") : may return old value

[Bastien Auneau]

Hi

Using USRP N210, UHD 3.2.1, windows7 (VC++ 9), GPSDO (Jackson Labs)

I first use :

  get_mboard_sensor("gps_time")

  set_time_next_pps(0, 1, 54000000)

Since gps time returns soon after pps edge occurs, I know that time is set to 0 at a known GPS time called gps_t_0 (which is gps time returned by sensor + 1)

(this is based on previous comments of Nick http://lists.gnu.org/archive/html/discuss-gnuradio/2011-08/msg00097.html)

Then, I want to check that gps_t_0 is really my starting point, then I discovered the bug

Here is pseudo code (in real code I use get_full_secs())

    loop(){

        //Sleep(3000)

        gps_now = get_mboard_sensor("gps_time")

        time_now = get_time_now()

        if ((gps_now - gps_t_0) != time_now){

            //NOT OK

            std::cout << (gps_now - gps_t_0) - time_now

        }

    }

If there is no Sleep(3000), results are OK

If I enable the Sleep(3000), it prints at every loop iteration either -2 or -3

If I change the sleep into Sleep(6000), it prints at every loop iteration -5 or -6

This means that query of gps_time is proportionally to Sleep(x) in the past

So I guess there is a problem of data flushing, the sensor query returning first gps phrase generated by GPSDO since last gps_time query

A way to solve this bug may be to flush previous gps_time phrase generated by GPSDO prior to get current time ??

Trying to dig into code (so far searching around get_nmea(...) in gps_ctrl.cpp l.148), but not sure this can be solved in host code ??

Regards

Bastien