[gpsd-dev] Tricky bug in the json parser

[Christian Gagneraud]

Hi all,

I just have one last bug to fix from my live test campaign. and it is a 
bit tricky to understand for me as I have to dive into the hairy json.c.

So basically, here are 2 AIS sentences which contain an array of 1 struct:
!AIVDM,1,1,,A,address@hidden,0*02
!AIVDM,1,1,,A,6h2E3N0rThqP0600J00000000000,0*51

The dumping doesn't seems to cause any trouble, but reading back the 
JSON crashes gpsdecode:
echo '!AIVDM,1,1,,A,address@hidden,0*02' | gpsdecode -ju
{"class":"AIS","device":"stdin","type":6,"repeat":3,"mmsi":2442101,"scaled":false,"seqno":2,"dest_mmsi":246351000,"retransmit":false,"dac":1,"fid":32,"month":0,"day":0,"tidals":[{"lon":-16777216,"lat":0,"from_hour":0,"from_min":0,"to_hour":0,"to_min":0,"cdir":0,"cspeed":0}]}
echo '!AIVDM,1,1,,A,address@hidden,0*02' | gpsdecode -ju | 
gpsdecode -jeu
Segmentation fault

I saw some other messages with arrays of 0 structures, they don't crash 
the json parser.

I have activated the debug messages in json.c, but it didn't help me 
that much (it seems that the parser doesn't "see" the end of struct or 
end of array).
Maybe someone familiar with the json parser code could help me to fix 
this nasty bug.

Chris

PS: I'm running an AIS test campaign on live data from aishub.net.
Input data is ca. 25 millions sentences (1.2GB of logs), and goes 
through these steps:
- AIVDM decoding
- JSON scaled dump
- JSON validity check
- JSON unscaled dump
- JSON validity check
- JSON idempotency check

I fixed a dozen of bugs so far, this one seems to be the last one!

--
Christian Gagneraud,
Embedded systems engineer.
Techworks Marine
1 Harbour road
Dun Laoghaire
Co. Dublin
Ireland
Tel: + 353 (0) 1 236 5990
Web: http://www.techworks.ie/