[LinuxPPS] success

[James Boddington]

Rodolfo Giometti wrote:
> On Wed, Oct 31, 2007 at 09:56:09AM +1000, James Boddington wrote:
>> Rodolfo Giometti wrote:
>>>> It was the difference in sequence numbers that struck me as strange
>>>>
>>>> ntp ~ # cat /sys/class/pps/pps0/{assert,clear}
>>>> 1193700322.000030061#73747
>>>> 1193700322.100028874#73745
>>>>
>>>> The offset of 30us is correct and the pulse width of 100us is also 
>>>> correct. 2 more asserts than clears.
>>> The offset remains constant? Each time you load the PPS support you
>>> got the same result.
>> There is no problem with the offset. The only problem was the different 
>> sequence numbers with assert and clear and have not seen that since.
> 
> Can you replicate this situation?

I can now. It is a side effect of having to have the serial port open to get 
the  interrupts on the dcd line. If there is an assert event and the serial 
port is  closed before the clear event there will me more assert recorded than 
clear.

The day it happened I was using cat /dev/ttyS0 to open the serial port so I 
would get the interrupts. I am guessing I killed cat after an assert and before 
   a clear giving me more asserts than clears.

Have been playing. I got the following by timing when I opened and closed the 
serial port. I timed it so the serial port was opened after the start of the 
second. This meant the assert was missed and the port was open long enough to 
capture the clear event. End result is I have 10 more clear than assert recorded.

assert 1193993798.999997289#115096
clear  1193993810.099998731#115106

-- 
    James