Re: [gpsd-dev] GPSd with Broadcom bcm4752

[Eric S. Raymond]

amacias <address@hidden>:
> The GPS was a big problem because Broadcom's code is not free and for
> Android we only had the binary "gpsd" service.
> That "gpsd" service is NOT your GPSd, is a software like Gllin for OpenMoko
> http://wiki.openmoko.org/wiki/Gllin#Using_gpsd_with_gllin
> Is a software that reads binay data from /dev/ttyMFD1 and gives to you NMEA
> data in a socket (similar to /tmp/nmeaNP in Gllin).

That's interesting.  Full gpsd has been seen on Samsung Android phones.  This
is the first we've heard of a *different* gpsd being deployed.

> So my solution was to start the Android's "gpsd" in a "chrooted" Android and
> then start GPSd using that socket as the device.
> 
> BUT I had to change some code in GPSd because when you read NMEA data from
> the device you assume that the "read" call may return one (and only one)
> NMEA string. And in my case, the "read" call reads more than one NMEA
> string, so I had to change the code to split the strings each time a "\n" is
> detected.
> 
> Patch file attached. This patch applies on GPSd 3.9-3 (Ubuntu 14).
> 
> I've added a "-t" parameter to read the data as "text" instead of "binary".

Your changes mostly look good, except that adding configuration options
is forbidden.  The GPSD philosophy is to be self-configuring by watching
the environment.  Rather than having an option, you should change the logic
so that gpsd notices the protocol it has synced to is textual (see the
TEXTUAL_PACKET_TYPE macro in gpsd.h) and calls  packet_parse_as_text()
when this happens.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>