Re: [gpsd-dev] SiRF gpsmon client mode

[Gary E. Miller]

Yo Eric!

On Wed, 11 Feb 2015 23:30:18 -0500
"Eric S. Raymond" <address@hidden> wrote:

> Gary E. Miller <address@hidden>:
> > Yeah, I made it wider, but not wide enough for bad cases.  Also not
> > sure why PPS is in the packet 7 box?  Clearly it is not in a packet.
> 
> Both problems have the same cause: we're badly short on screen space
> if we want the top panel and at least one line of log window to fit
> in a 80x25 terminal emulator.  I spent a lot of ingenuity packing
> stuff in tight, but I admit the results are not entirely satisfactory.

Some programs, like atop, show more as the screen size grows.  Prolly
a PITA to code, but a possible direction to go.  Or maybe some page
flipping?

> > Plus still the garbage on the right.
> 
> I think that garbage is remnants of what was onscreen before the panel
> was painted.  You can test this by doing a 'clear' before calling
> gpsmon; if I'm right, there will be no garbage.

Sort of.  I see the garbage is some JSON displayed after gpsmon
starts and before the curses painted.  Annoying, but not fatal.

> > Pull git head and check it out.  I think the PPS in gpsmon now 
> > looking almost good.  Still need to fix that ntpshm_latch() is
> > called more than once a second in gpsmon hardware mode.
> 
> It's on my list of things to look at.

I committed a small patch to fix that.  Better.  The PPS now looks good
as I stare at it, so no obvious glitches in the steady state.

Tight now my SiRF PPS test host is tracking under 200 nSec jitter
while driving ntpd.  And gpsmon PPS offset data correlating well to the
'ntpq -p' output.

And a GR-601W doing its usual +/- 500 mSec with chronyd, also tracking
nicely with gpsmon PPS display.

But not gonna declare victory until it gets beat on more.  It is hard to
test the PPS when it is slewing hard on startup.  And of course other
people need to test.

> > Then test the garmin PPS and we may be good.
> 
> Not quite, though it is encouraging that the BSD tests now seem to be
> working. Other known issues:

I was being selfish and only thinking of what I cared about.
 
> * gpsdecode crashes on Mac OS X (probably a linker-option problem).
> * shmget() creation failures.

A pretty short list.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97701
        address@hidden  Tel:+1(541)382-8588