Re: [gpsd-users] NMEA being discarded

[Eric S. Raymond]

Ed Simmons <address@hidden>:
> >Hm.  Can we have that code? I'm especially interested if the additional
> >sentences are stock NMEA.
> Sure, I'd be glad to have it in the codebase - I've added things to
> do with Raymarine sentences. For one reason or another GPSD was not
> accepting some of the sentence headers as being valid. Once I
> understood the output of the higher level of debug it became clear
> that the parser was not happy with some of the prefixes. A couple of
> additions to the packet sniffer cleared this up and I was happy to
> see my parser code working.
> 
> How do I go about submitting these changes?

Make diffs of whatever files you changed against the base release you
forked from and mail them to to me.  The NMEA code has been pretty
stable for the last couple years; I don't anticipate a lot of
difficulty with the merge.

I'll also want a capture log from the boat's NMEA bus.  It doesn't
have to be huge, just a representative sample of the traffic.

> It works flawlessly in other applications with less of the
> proprietary raymarine stuff flying about. My somewhat crude solution
> to this is to compile GPSD on the machine doing the logging with
> only the two possible baud rates in the switch in autobauding. This
> makes it rock solid, I didn't investigate further... also I think
> this problem was exacerbated by the high proportion of the received
> sentences it was discarding, that would probably not have helped
> autobauding.

Ah, that makes complete sense actually.  I think your diagnosis is
correct and you did the right thing to cope with it.  Now I'm
wondering if maybe I ought to hack the configuration machinery to give
people building from source finer-grained choices about which baud rates
to support.  Maybe max_bps and min_bps switches?

> >>                        for example I spent ages looking for
> >>how to get shared memory export working. If it is publicly
> >>documented, I couldn't find it.
> >Sorry, I didn't realize I hadn't done that yet.  I'll update the
> >manual page for the client-side libraries.

I looked and the shm stuff is actually pretty well documented in
recent releases. Look near the end of the Client HOWTO on the website.

> >>                                Also, drawing people's attention to
> >>the provided init scripts etc. would have saved me loads of time,
> >>but damn am I glad I found it in the end!
> >Where would that pointer have been more useful?  In the FAQ, the
> >source-build instructions, the INSTALL file, or somewhere else?
> Probably the source build instructions is the best place. It seems
> logical to draw the attention of the user to the existence of the
> scripts at the crucial moment just after you've run sudo scons
> install !!

I in fact pushed a description to that exact spot about two hours
ago, before I got your second mail. :-)
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>