SCSISoftUSB bug fix

in mixed.RiscOS.Sources.HWSupport.SCSI.SCSISoftUSB.c.glue you have

usbd_status usbd_open_pipe(usbd_interface_handle iface, u_int8_t address, u_int8_t flags, usbd_pipe_handle *pipe)
{
  my_usb_device_t *device = (my_usb_device_t *) iface;
  char filename[sizeof "USBxxxxxxxxxxxxxxxx#interfacexxx;alternatexxx;endpointxxx;bulk;sizexxxxxx:" + 1];
  if (pipe) *pipe = 0; /* in case of error */
  
  sprintf(filename, "%s#alternate%d;interface%d;endpoint%d;bulk;size%d:",
         device->devicefs_name,
#if 1
/*
 * OPEN PIPE FIX
 */
         device->alternate,
         device->interface,
#else
         device->interface,
         device->alternate,
#endif
         address & 0x7F,
         BUFFER_SIZE);
  DEBUGf("\n usbd_open_pipe %s\n",filename);

As you can see the values passed to the sprintf statement are the wrong way around for interface and alternate.

As you can see the values passed to the sprintf statement are the wrong way around for interface and alternate.

Well spotted! Interestingly it was right originally, but they were swapped for some (unknown) reason perhaps to workaround the DeviceFS parsing problem. Seems a bit silly to not track down the actual problem when what was written in the sprintf statement was clearly right before.

I changed the formatter so that the special string is in the same order as the USB module template, to sidestep the DeviceFS problem if loaded on earlier machines (the SCSISoftUSB module pops out in the disc image for pre 5.18).

The “usbdi.c” change from a few weeks back is on ice until the 5.20 freeze is lifted.

You’ve reverted to the original format string. The USB module template is

"endpoint/Ninterface/Nalternate/Nreport/Ncontrol,isochronous,bulk,interrupt/Susbtimeout/Nsize/Nnopad/Sshort/S"

so endpoint wants to be at the beginning (and the order of the args changed) for it to work with 5.18