Everett Wang wrote:
> Thanks. I looked my code again, which is modified from example
> cross-link.c. The section looks like:
> 
> ...
>                  irq_time = rx_event.rxpend_timestamp;
>                 read = rt_dev_read(read_fd, &write_time, sz);
>                 if (read == sz) {
>                         read_time = rt_timer_read();
>                         printf("%3d |%16llu |%16llu |%16llu\n", nr,
>                                irq_time  - write_time,
>                                read_time - irq_time,
>                                read_time - write_time);
>                         nr++;
>                 } else if (read < 0 ) {
>                         printf(RTASK_PREFIX "error on rt_dev_read, code %s\n",
>                                strerror(-err)); <<<<<<<===========
>                         break;
>                 } else {
>                         printf(RTASK_PREFIX "only %d / %d byte received \n",
>                                read, sz);
>                         break;
>                 }
> ...
> I guess I should change the strerror(-err) to strerror(-read).

Indeed, the example suffers from some copy&paste mistakes. Same for the
write path. And closing in a loop is no longer required. A cleanup is
needed.

> If I do
> that, I some
> times get a " code input/output error". That is better. What can cause
> input/output error?

Try RTSER_RTIOC_GET_STATUS to learn more.

> I get my serial port in that state consistently
> when I
> finish running my GPS program. I can reset it to a usable serial port
> again by remove and re-load xeno-16550A module.

Re-opening the rtser port, ie. restarting your program doesn't help?
Does re-plugging the GPS mouse change the picture? If yes, you may have
to reset its hardware somehow. If no, there might be a sticky state in
our serial driver, ie. a bug.

> But setserial -G can't
> see anything unusual. Could it be some unclosed serial port?
> 
> My simple code is attached.
> 
> Thanks,
> 
> Everett
> 

Jan