Re: windows vs linux behaviour: question

[jan v. k. <j.v...@gm...>]

Installing libusbx-1.012 seems to solve the problem.
Thanx for the excellent work in developing interfaces for true portability.
all the best

jan



2012/7/28 Xiaofan Chen <xia...@gm...>

> Please reply to the list. Thanks.
>
> On Sat, Jul 28, 2012 at 6:53 PM, jan van katwijk <j.v...@gm...>
> wrote:
> > Thanx
> >
> > The problem is that there is no error message, often just hanging of the
> > whole system.
> >
> > The libusb-1.-9 version I am using has as its last entry in the changelog
> > april 20.
> >
> > The main issue was whether behaviour under windows wrt simultaneously
> > receiving data using bulk transfer
> > (asynchronous API) while sending control bits, should be expected the
> same
> > as under Linux.
> > If so, I should instrument the code and see where I made an error, if
> > behaviour can be different I should
> > find a different programmatic solution.
> >
> > In detail: I have a Qthread for handling the asynchronous libusb
> interface
> > (i.e. catching the events), this thread
> > continues to deliver data on the required speed. It seems pretty standard
> > code, based on some examples
> > I found on the net and on the code in the osmocom library for the rtl,
> but
> > then adapted for use with Qt
> >
> > void    RTL2832U::run (void) {
> > int32_t i;
> > struct timeval tv       = {1, 0};
> >
> > rtl_is_running  = true;
> >         while (rtl_should_run) {
> >            while (TheDevice -> async_status == RTLSDR_INACTIVE)
> >               msleep (1);
> >
> > //         fprintf (stderr, "off we go\n");
> >            if (!resetBuffer ())
> >               fprintf (stderr, "failed to reset\n");
> > //
> > //      could have been done on init
> >            for (i = 0; i < (int)(TheDevice -> xfer_buf_num); i ++) {
> >               libusb_fill_bulk_transfer (TheDevice -> xfer [i],
> >                                          TheDevice -> devh,
> >                                          0x81,
> >                                          TheDevice      -> xfer_buf [i],
> >                                          TheDevice -> xfer_buf_len,
> >                                          _libusb_callback,
> >                                          (void *)TheDevice,
> >                                          BULK_TIMEOUT);
> >               libusb_submit_transfer (TheDevice -> xfer [i]);
> >     }
> >     while (TheDevice -> async_status != RTLSDR_INACTIVE) {
> >               libusb_handle_events_timeout (TheDevice -> ctx, &tv);
> >
> >               if (TheDevice -> async_status == RTLSDR_CANCELING) {
> >                  TheDevice -> async_status = RTLSDR_INACTIVE;
> >
> >                  if (TheDevice ->xfer == NULL)
> >                     break;
> >
> >                  for (i = 0; i < (int)(TheDevice -> xfer_buf_num); i ++)
> {
> >                     if (TheDevice -> xfer [i] == NULL)
> >                        continue;
> >
> >                     if (TheDevice -> xfer [i] -> status ==
> >                                          LIBUSB_TRANSFER_COMPLETED) {
> >                        libusb_cancel_transfer (TheDevice -> xfer [i]);
> >                        TheDevice -> async_status = RTLSDR_CANCELING;
> >                      }
> >                  }
> >
> >               }
> >            }
> > }
> >
> >
> > Another thread (the main thread in the Qt application) calls on the
> libusb
> > interface for setting and reading control data
> > all using constructs like
> > r       = libusb_control_transfer (TheDevice    -> devh,  CTRL_IN, 0,
> addr,
> > index, array,
> >                                            len, CTRL_TIMEOUT);
> >
> > I do have another radio through the usb gate connected, and control works
> > fine. The problem starts
> > as said above when sending control messages while the bulk transfers are
> on,
> > but again, only
> > windows. No problem whatsoever in Linux (Fedora 17 64 bits, fedora 16 64
> > bits, ubuntu 12.04 64 bits).
> >
> > If behaviour SHOULD be exactly the same under windows and linux, then
> > apparently I am making some programming
> > error, which can be as simple as an unitialized variable. Then I need to
> set
> > up a debugging environment for
> > use with Mingw64 and reinspect all elements of my code.
> >
> > anyway, thanx so far
>
> You may want to use libusb.git since libusb-1.0.9 has serious bugs
> under Windows. Or you can use libusbx-1.0.12.
> http://git.libusb.org/?p=libusb.git
> http://libusbx.org/
>
> --
> Xiaofan
>



-- 
Jan van Katwijk


+31 (0)15 3698980
+31 (0) 628260355