Re: ✘gpsd release coming

[Fred Wright]

On Fri, 31 Jul 2020, Greg Troxel wrote:

> I have just tested livetested gpsd on NetBSD 8 amd64 as of the following
> commit:
>
>  commit d3800a04000c9e94f93262505083743b3280548a (HEAD -> master, 
> origin/master, origin/HEAD)
>  Author: Greg Troxel <gdt@lexort.com>
>  Date:   Fri Jul 31 08:53:43 2020 -0400
>
> via making a tarball with scons dist, using that to build a binary
> package with pkgsrc, installing that package, running the resulting gpsd
> with a ublox 8, verified that it gets position, shows satellite location
> and status with xgps, and passes (non-PPS) time to ntpd.  I also checked
> ubxtool querying things.

You must not have looked very closely at the xgps output if you think it's 
OK.  You should see several pulsating values in the data box due to its 
seeing mixed NMEA/binary in "binary" mode.

And if you have any multi-GNSS configuration other than GPS+GLONASS, you 
should see at least one entire constellation coming and going in both the 
list and the satview, as well as in the seen/used counts in the data box. 
This can be seen even with a GPS-only antenna, if you configure it for 
GPS+Galileo.  With a multi-system antenna, configuring for 
GPS+BeiDou+Galileo shows the largest effect.  Though there seems to be 
some interaction with other timing that makes the exact behavior variable. 
It also appears that the special hack for GPS+GLONASS has some lingering 
effect, so restarting gpsd after turning off GLONASS makes it more 
representative.

Similar behavior can be seen in cgps, and gpsmon is crazier.

Since all of these issues relate to NMEA and UBX "fighting each other", 
they may be less significant at higher baud rates.  But they're definitely 
visible at 9600 (the default for both gpsd and u-blox 8), and at 19200.

I already reported this a while back in the "NEO-M8N NMEA" thread.  I have 
a series of commits that fix it, but I need to do more testing.  It is, 
notably, a regression since 3.20.

The other oddity is that, even with the other issues fixed, the gpsmon fix 
pane doesn't work now with a u-blox 8 and did in 3.20.  I thought I 
understood the reason (lack of support for NAV-PVT), until I noticed that 
NAV-PVT was already being used on u-blox 8 in 3.20, so that aspect hasn't 
changed.  Hence, the real mystery is why it *wasn't* broken in 3.20. 
Another thing to investigate.

Unrelated to all that, with Python 2, the current code to find 
libgpsdpacket fails on almost all platforms (including Linux).

Fred Wright