[gpsd-dev] Bug: Devices don't always respond if added while WATCH active...

[Ed W]

I can't quite reproduce this, so it's kind of a heads up to see if
someone else can?  Background: I have a USB GPS (Bluenext) which spits
out NMEA at 9600.  Additionally I have some udev scripts which
add/remove it using gpsdctl when I plug and unplug it.

Now: *IF* I run a watch via the network interface and then deliberately
unplug the device, and replug it in such as way it gets a new ttyUSB
port number, then sometimes it gets added into gpsd, but no output comes
out of gpsd?  I can kill and restart my watch or switch to gpspipe, but
although there is output on the serial port (tested this) it's not
output from gpsd...

Now, I can then jiggle the ttys again and the same device will work ok
on different ttys, but once I break it on a given tty then it continues
to remain broken on that tty even if I jiggle the device ordering...

So I'm using say another serial to usb adaptor that I connect/disconnect
in order to force the kernel to renumber things.  Say I break ttyUSB1,
then replugging to ttyUSB0 and I get output, back to ttyUSB1, nothing,
back to ttyUSB0 output...

Further I have sometimes noticed that the RSS size of gpsd is climbing
quite high once I have fiddled around connecting and disconnecting to
the gpsd socket.  I again can't quite correlate it, but it might be the
case that regular connects/disconnects cause the RSS size of gpsd to
increase?

OK, I said this isn't cleanly reproducible, so drop this report in the
bit bucket, but perhaps you will think of it if someone reports
something half similar?

Oh, I *think* the behaviour is only reproduced if the watch is done
using my simple app, which is a small perl script that just connects to
the gpsd net socket and consumes all the output.  Running a watch using
gpsd doesn't seem to trigger it (or my limited tests haven't triggered
it).  killing my app and running up gpspipe does NOT show output from
gpsd, so I don't think that the app is broken, only that it might be a
function in the way it consumes/connects which is tickling something
that gpspipe doesn't?)

I don't have gpsd logs of the problem interval yet (long story, but..)
Don't expect anyone to care without them, but ....

Thanks for listening...

Ed W