Run tests on M1 MacOS runners.

[bcoconni]

Until now we were only testing that JSBSim was successfully compiling on Apple M1 chips using cross-compilation on Intel MacOS GitHub runners. The resulting binaries were not tested on the actual hardware so we were happily assuming that if it compiles then it runs swiftly. And everyone knows how reliable that criterion is 😄

These dark ages are now over: GitHub just introduced M1 MacOS runners, so this PR implements the testing of JSBSim on that platform.

Note that the floating point precision requirement had to be slightly relaxed at some places in the unit tests ...despite using the flag -fno-fast-math. If someone who has access to an M1 Macbook or equivalent knows better then by all means submit fixes 👍

Note also that the issue that was fixed by PR #143 (random failure of TestActuator.py) is resurfacing with M1 MacOS runners. This has been fixed again by extending further the waiting time of test_regression_bug_1503:

jsbsim/tests/TestActuator.py

Lines 94 to 130 in 6f0c4d8

	def test_regression_bug_1503(self):
	# First, the execution time of the script c1724.xml is measured. It
	# will be used as a reference to check if JSBSim hangs or not.
	fdm = self.create_fdm()
	start_time = time.time()
	self.ScriptExecution(fdm)
	exec_time = time.time() - start_time
	# Delete the FDM instance to make sure that all files are closed and
	# released before running the same script in another process.
	del fdm
	# Now the copy of the aircraft definition file will be altered: the
	# <rate_limit> element is split in two: one with the 'decr' sense, the
	# other with 'incr' sense.
	actuator_element = self.tree.getroot().find('flight_control/channel/actuator//rate_limit/..')
	rate_element = actuator_element.find('rate_limit')
	rate_element.attrib['sense'] = 'decr'
	new_rate_element = et.SubElement(actuator_element, 'rate_limit')
	new_rate_element.attrib['sense'] = 'incr'
	new_rate_element.text = str(float(rate_element.text) * 0.5)
	self.tree.write(os.path.join('aircraft', self.aircraft_name,
	self.aircraft_name+'.xml'))
	# A new process is created that launches the script. We wait for 10
	# times the reference execution time for the script completion. Beyond
	# that time, if the process is not completed, it is terminated and the
	# test is failed.
	p = Process(target=SubProcessScriptExecution,
	args=(self.sandbox, self.script_path))
	p.start()
	p.join(exec_time * 20.0) # Wait 20 times the reference time
	alive = p.is_alive()
	if alive:
	p.terminate()
	self.assertFalse(alive, msg="The script has hung")

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (6f0c4d8) 24.87% compared to head (b46fd66) 24.87%.
Report is 2 commits behind head on master.

❗ Current head b46fd66 differs from pull request most recent head d63bf5f. Consider uploading reports for the commit d63bf5f to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1030   +/-   ##
=======================================
  Coverage   24.87%   24.87%           
=======================================
  Files         168      168           
  Lines       18908    18908           
=======================================
  Hits         4704     4704           
  Misses      14204    14204           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[bcoconni]

And in case someone asks: the test job on M1 chips uses Python 3.10 (instead of Python 3.8 for other platforms) because Python<=3.9 is not available on the M1 MacOS runners.