Re: [gpsd-users] Segmentation fault on Qualcomm MDM9615

[Eric S. Raymond]

Jordi Augé <address@hidden>:
> Hello all, 
> 
> I am trying to use gpsd on a Qualcomm MDM9615 device, which is an Embedded 
> ARM device, with an embedded modem and GPS, and running Linux. 
> 
> I've cross compiled the latest version without dbus, python or bluetooth 
> support, as I'm attaching to a serial device already present on the target 
> device. 
> 
> The serial port is identified by gpsd as being NMEA0183, but after a few 
> messages, gpsd throws a segmentation fault. This happens at every run, after 
> about 8-10 messages are processed. 
> 
> I've posted the full session log here: http://pastebin.com/fxTS3nLY 
> In the paste, there is: 
> - gpsd -h (to see which drivers are compiled in) 
> - gpsd -V 
> - A run of gpsd with debugging enabled, ending on a segmentation fault 
> - Some messages extracted directly from /dev/nmea, for reference 
> 
> As this test is being run indoors, there's no GPS fix, I'll start testing 
> outdoors as soon as I get rid of the segmentation faults here. 
> 
> Can anyone shed any light on why this is happening? 
> 
> Thanks, 
> Jordi 

Not yet.  If you can give us an NMEA log that triggers a core dump when run
through gpsfake, though, we can probably find and fix this very fast.
Best if it has an explanatory comment header, like this:

# Name: Qualcomm MDM9615
# Description: Embedded ARM device, with modem and GPS
# Submitter: Jordi Augé <address@hidden>
# Date: 2016-09-01
# Location: <fill in your location, including longitude/latitude>

That is, what we're looking for is a file foo.log such that

gpsfake -p -1 foo.log 

reproduces the core dump.  Since you say this always happens within 8 or 10
lines, any capture of 20 or 30 NMEA lines from the device should do it.

Once we figure out what the problem is and fix it, we'll add your log to our
regression tests. 
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>