[gpsd-dev] Telit HE910 GPS week rollover

[Titouan Christophe]

Dear GPSD developers,

I'm using gpsd on embedded mobile devices that use the Telit HE910-EUG
combined 3G+GPS receiver. I generate the system images for these
devices using Buildroot (https://buildroot.org/).

According to an application note from Telit (https://rutronik-tec.com/w
p-
content/uploads/2019/02/Telit_Cellular_WeekRollover_Application_Note_r0
.pdf), this HE910 chip will be subject to the GPS week rollover bug. As
I understand the application note, after the critical date, the year
reported in the date field of RMC sentences could be either "-1" or
"99".

I already discussed that particular issue with gemiller on the GPSD irc
channel on freenode. From our discussion, it seems that the application
note is ambiguous, because we had different interpretations for the
example RMC lines presented in section 3.2.

Because my use case is essentially syncing the system time with NTP,
and providing TPV vectors to business applications, I did write a patch
that corrects the date field in the NMEA0183 RMC sentences before using
it in the parser, such as to compensate for the week rollover. This is
the approach proposed in the Telit app note. However, in its current
form, this patch cannot be upstreamed in gpsd, because it is assuming
that the date should be at least in year 2019, and this would break at
least gpsfake, and probably other regression tests.

Thanks to Buildroot, it is very easy for me to keep this patch out of
the gpsd source code and apply it only for our devices, and I think
this is an acceptable process. There are two versions of the patch: 

- one that has been developed on top of the GPSD master branch: http://
paste.awesom.eu/5sVv
- one that has been backported to the 3.17 release (which I am
currently using in production): http://paste.awesom.eu/82ER

I'm sharing this patch here in the hope that it could be useful to
someone else. Also if you see something that could miserably fail, or
if you have suggestions to get this into GPSD mainline, I'd be glad to
help about it.


Thanks to all the GPSD contributors for providing such a useful piece
of software !

Best regards,

Titouan