Previously we stopped the simulation to ensure parameter changes are
applied to the simulated vehicle. We can now reboot the simulated
autopilot, so these patches make use of those changes.
One big advantage is that the --gdb option is in play for the applying
of the parameters to the vehicle.
RTL may disarm the vehicle on completion. We RTL at several times in
the testing, and the subsequent tests were not rearming. This means we
had a race condition.
We now explicitly wait to be disarmed by the RTL mode, and rearm the
vehicle.
This is an interim patch until we decide whether to make each "test"
self-contained, and have a precondition of "on ground and disarmed".
If any of the tests throws an exception, it will be caught by the generic test function
that collects all errors and displays them before passing the result to the vehicle testcode.
Hence the name of the test and the exception that got raised get printed in a single
location. This nicely reduces code duplication.
Three of our models start a SITL instance to apply parameters
then restart it so that the parameter changes can take effect.
Factor that functionality out.
Autotest: common.py add new base class and test
Autotest : Common add heartbeat handling on wait_mode timeout
Autotest: common.py add reach_heading_manual and reach_distance_manual
Autotest: new autotest implementation for Rover
Autotest : rover add drive_square test
Autotest: new autotest implementation for Copter
Autotest: new autotest implementation for Plane
Autotest: new autotest implementation for Sub
Autotest: new autotest implementation
Autotest: new autotest implementation for QuadPlane
Autotest : Sub disable GCS_Failsafe on autotest to prevent timeout. The failsafe aren't currently tested on Sub
Tools: don't need to pass option on first mavproxy
Tools: reformat common.py and add commun function
Tools: use new common functions
Tools: move functions from rover to commun and reorder
Tools: add and use set_rc function with timeout
Tools: fix style for pep8
Added \r\n to the expect() string as recomended at:
http://pexpect.readthedocs.io/en/stable/overview.html#find-the-end-of-line-cr-lf-conventions
this should work on both windows and linux systems
pexpect says it will always do a minimal (non greedy) matching and docs explicitly say that a .+ expression will always return only one character. These lines in autotest are looking for \S+, which, believing the documentation, would only return one character of the log file path.
Now we know that's not true, neither for Linux or for Windows (given the logs from @karthikdesai), so I can only assume that it does a greedy match but only for the characters it has received at the time expect is called.
Apparently, in the machines we are using autotest, it isn't a problem since MAVProxy is likely fast to give its output to pexpect before the expect method is called. On @karthikdesai's machine that wasn't happening since his machine was more or less loaded.
Concluding, this looks like a correct fix in the sense that it extends the regex pattern to wait for the end of line (and probably other places could benefit from it too).
Based on work done by khancir
(https://github.com/ArduPilot/ardupilot/pull/6360)
Tools: arduplane.py change print to progress function
Tools: quadplane.py change print to progress function
Tools: ardusub.py change print to progress function
Try to make the various param files to follow similar namings:
- Lowercase the few ones using uppercase
- Use dash to separate name components
- Start with "copter", "plane", "rover", etc instead of the mix we
had
- Remove "params" from the filename since it's redundant with the
extension
Michael du Breuil
Peter Barker
Pierre Kancir
Guilherme Sousa
Karthik Desai
khancyr
Dr.-Ing. Amilcar Do Carmo Lucas
Pierre Kancir
Andrew Tridgell
Miguel Arroyo
Randy Mackay
Lucas De Marchi