NAME¶
gpsfake - test harness for gpsd, simulating a GPS
SYNOPSIS¶
gpsfake [-?] [--baton] [--cycle interval]
[--gdb] [--help] [--initcmd initcmd] [--linedump] [--lldb]
[--monitor monitor] [--nowait] [--options=options]
[--pipe] [--port port] [--predump] [--promptme] [--quiet]
[--singleshot] [--slow] [--speed speed] [--tcp]
[--timeout timeout] [--udp] [--verbose] [--version] [-1] [-b]
[-c interval] [-D debuglevel] [-g] [-G] [-h]
[-i] [-l] [-m monitor] [-n] [-o=options] [-p]
[-P port] [-q] [-r initcmd] [-S]
[-s speed] [-t] [-T] [-u] [-v] [-V] [-W timeout]
[logfile...]
DESCRIPTION¶
gpsfake is a test harness for gpsd and its clients. It opens a pty
(pseudo-TTY), launches a gpsd instance that thinks the slave side of the pty
is its GPS device, and repeatedly feeds the contents of one or more test
logfiles through the master side to the GPS. If there are multiple logfiles,
sentences from them are interleaved in the order the files are
specified.
gpsfake does not require root privileges, and can be run
concurrently with a production gpsd instance without causing problems.
The logfiles may contain packets in any supported format,
including in particular NMEA, SiRF, TSIP, or Zodiac. Leading lines beginning
with # will be treated as comments and ignored, except in the following
special cases:
•a comment of the form #Date: yyyy-mm-dd (ISO8601
date format) may be used to set the initial date for the log.
•a comment of the form #Serial: [0-9]*
[78][NOE][12] may be used to set serial parameters for the log - baud rate,
word length, stop bits.
•a comment of the form #Transport: UDP may be used
to fake a UDP source rather than the normal pty.
The gpsd instance is run in foreground. The thread sending fake
GPS data to the daemon is run in background.
OPTIONS¶
-?, -h, --help
Print a usage message and exit.
-1, --singleshot
The logfile is interpreted once only rather than
repeatedly. This option is intended to facilitate regression testing.
-b, --baton
Enable a twirling-baton progress indicator on standard
error. At termination, it reports elapsed time.
-c COUNT, --cycle COUNT
Sets the delay between sentences in seconds. Fractional
values of seconds are legal. The default is zero (no delay).
-d LVL, --debug LVL
Pass a -D option to the daemon: thus -D 4
is shorthand for -o="-D 4".
-g, -G, --gdb, --lldb
Use the monitor facility to run the gpsd instance within
gpsfake under control of gdb or lldb, respectively. They also disable the
timeout on daemon inactivity, to allow for breakpointing. If necessary, the
timeout can be reenabled by a subsequent -W or --wait . If xterm
and $DISPLAY are available, these options launch the debugger in a separate
xterm window, to separate the debugger dialog from the program output, but
otherwise run it directly. In the gdb case, -tui is used with xterm but
not otherwise, since curses and program output don't play nicely together.
Although lldb lacks an equivalent option, some versions have a 'gui'
command.
-i, --promptme
Single-step through logfiles. It dumps the line or packet
number (and the sentence if the protocol is textual) followed by "?
". Only when the user keys Enter is the line actually fed to gpsd.
-l, --linedump
Print a line or packet number just before each sentence
is fed to the daemon. If the sentence is textual (e.g. NMEA), the text is
printed as well. If not, the packet will be printed in hexadecimal (except for
RTCM packets, which aren't dumped at all). This option is useful for checking
that gpsfake is getting packet boundaries right.
-m PROG, --monitor PROG
Specify a monitor program (PROG) inside which the daemon
should be run. This option is intended to be used with
valgrind(1),
gdb(1) and similar programs.
-n, --nowait
Pass -n to the daemon to start the daemon reading
the GPS without waiting for a client (equivalent to
-o="-n").
-o="OPTS", --option="OPTS"
Specify options to pass to the daemon. The equal sign (=)
and quotes are required so that gpsd options are not confused with gpsfake
options. To start the daemon reading the GPS without waiting for a client use
-o="-n" (equivalent to the -n) which passes -n
to the gpsd daemon. The option -o="-D 4" passes a -D 4
to the daemon, equivalent to the using -D 4.
-p, --pipe
Sets watcher mode and dump the NMEA and GPSD
notifications generated by the log to standard output. This is useful for
regression-testing.
-p PORT, --port PORT
Sets the daemon's listening port to PORT.
-q, --quiet
Tell gpsfake to suppress normal progress output and thus
act in a quiet manner.
-r STR, --clientinit STR
Specify an initialization command to use in pipe mode.
The default is
?WATCH={"enable":true,"json":true}.
-s SPEED, --speed SPEED
Sets the baud rate for the slave tty. The default is
4800.
-S, --slow
Tells gpsfake to insert realistic delays in the test
input rather than trying to stuff it through the daemon as fast as possible.
This will make the test(s) run much slower, but avoids flaky failures due to
machine load and possible race conditions in the pty layer.
-t, --tcp
Forces the test framework to use TCP rather than pty
devices. Besides being a test of TCP source handling, this may be useful for
testing from within chroot jails where access to pty devices is locked
out.
-T, --sysinfo
Makes gpsfake print some system information and then
exit.
-u, --udp
Forces the test framework to use UDP rather than pty
devices. Besides being a test of UDP source handling, this may be useful for
testing from within chroot jails where access to pty devices is locked
out.
-v, --verbose
Enable verbose progress reports to stderr. Use multiple
times to increase verbosity. It is mainly useful for debugging gpsfake
itself.
-w SEC, --wait SEC
Set the timeout on daemon inactivity, in seconds. The
default timeout is 60 seconds, and a value of 0 suppresses the timeout
altogether. Note that the actual timeout is longer due to internal delays,
typically by about 20 seconds.
-x, --predump
Dump packets as gpsfake gathers them. It is mainly useful
for debugging gpsfake itself.
The last argument(s) must be the name of a file or files
containing the data to be cycled at the device. gpsfake will print a
notification each time it cycles.
Normally, gpsfake creates a pty for each logfile and passes the
slave side of the device to the daemon. If the header comment in the logfile
contains the string "UDP", packets are instead shipped via UDP
port 5000 to the address 192.168.0.1.255. You can monitor them with this:
tcpdump -s0 -n -A -i lo udp and port 5000.
Certain magic comments in test load headers can change the
conditions of the test. These are:
Serial:
May contain a serial-port setting such as 4800 7N2 - baud
rate followed by 7 or 8 for byte length, N or O or E for parity and 1 or 2 for
stop bits. The test is run with those settings on the slave port that the
daemon sees.
Transport:
Values 'TCP' and 'UDP' force the use of TCP and UDP feeds
respectively (the default is a pty).
Delay-Cookie:
Must be followed by two whitespace-separated fields, a
delimiter character and a numeric delay in seconds. Instead of being broken up
by packet boundaries, the test load is split on the delimiters. The delay is
performed after each feed. Can be useful for imposing write boundaries in the
middle of packets.
CUSTOM TESTS¶
gpsfake is a trivial wrapper around a Python module, also named
gpsfake, that can be used to fully script sessions involving a gpsd
instance, any number of client sessions, and any number of fake GPSes
feeding the daemon instance with data from specified sentence logs.
Source and embedded documentation for this module is shipped with
the gpsd development tools. You can use it to torture-test either gpsd
itself or any gpsd-aware client application.
Logfiles for the use with gpsfake can be retrieved using gpspipe,
gpscat, or gpsmon from the gpsd distribution, or any other application which
is able to create a compatible output.
If gpsfake exits with "Cannot execute gpsd: executable not
found." the environment variable GPSD_HOME can be set to the path where
gpsd can be found. (instead of adding that folder to the PATH environment
variable
ENVIRONMENT¶
For unknown reasons gpsfake may sometimes time out and fail. Set
the WRITE_PAD environment value to a larger value to avoid this issue. A
starting point might be "WRITE_PAD = 0.005". Values as large os
0.200 may be required.
AUTHOR¶
Eric S. Raymond <esr@thyrsus.com>.
