3.0.3 libusrp bug with usrp_basic_rx::start/stop


#1

The last thread I posted had indicated problems related to overruns.

The problem turns out that start/stop weren’t consistent in how they
started and stopped things in gnuradio-3.0.3. It worked out that
roughly every 800000 samples in my test, we got an overrun (except if
the total was before the first stop()).

The problem lies in the start order:
d_ephandle->start() followed by set_rx_enable(true)

and the stop order of:
d_ephandle->stop() followed by set_rx_enable(false).

To fix the problem, the calls in stop() had to be flipped.

Here’s the patch:
*** usrp_basic.cc.orig 2007-05-30 10:30:10.000000000 +0000
— usrp_basic.cc 2007-05-30 10:23:11.000000000 +0000


*** 541,556 ****
— 541,565 ----
{
bool ok = usrp_basic::stop();

  • #if 1

  • if (!set_rx_enable(false)){

  • fprintf (stderr, "usrp_basic_rx: set_rx_enable(false) failed\n");
    
  • usb_strerror ();
    
  • ok = false;
    
  • }

  • #endif
    if (!d_ephandle->stop()){
    fprintf (stderr, “usrp_basic_rx: failed to stop end point
    streaming”);
    usb_strerror ();
    ok = false;
    }

  • #if 0
    if (!set_rx_enable(false)){
    fprintf (stderr, “usrp_basic_rx: set_rx_enable(false) failed\n”);
    usb_strerror ();
    ok = false;
    }

  • #endif
    return false;
    }

    Dave


#2

On Wed, May 30, 2007 at 11:29:42AM -0700, Dave Gotwisner wrote:

and the stop order of:
{
fprintf (stderr, “usrp_basic_rx: failed to stop end point streaming”);
return false;
}

Dave

Hi Dave,

Thanks for your efforts on this. I’ll apply the patch to the trunk.

Glad to hear it’s working for you.

Eric