Re: [gpsd-dev] PPS offset fields and regression testing

[Eric S. Raymond]

Gary E. Miller <address@hidden>:
> > 1. You have a serial GPS that emits Garmin binary packets, with
> > DLE-stuffing and a checksum.
> 
> Serial is not the problem, USB Garmin is.

There are two different problems I have in mind. One is verifying that the
driver's packet-inspection logic doesn't regress between releases.  For 
that purpose, serial packets are sufficient and no handshaking are 
sufficient.

Though it would be nice to be able to test-jig the USB Garmin behavior
more completely, that's a separate problem.
 
> > 2. You have gpscat, gpspipe -r, and gpsmon -l.
> > 
> > What's so hard about capturing 10 or 15 seconds of packets?
> 
> I can capture it, but the key part, the part that keeps breaking, is
> the ACK/NACK, and no way to capture that so gpsfake can duplicate it.

The duplication does not have to occur for the test to be useful.  The
Trimbles like to do elaborate handshakes too, but a Trimble packet log
is still a useful regression-test load without them. Verifying that
it still extracts the same navigation-solution bits to the same cooked
data that it did N releases ago is the point.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>

signature.asc
Description: Digital signature