Build fails on macos

[jsbsim-team]

Build fails on macos

[agodemar]

Hi Ewald, thank you for bringing out this issue.

At the moment I do not have a MacOS available. Let's see if someone can shed some light on the compilation errors.

What I know is that since 10.9 one has to replace libstd++ with libc++

[WillemEerland]

Right you are!

I don't have a fix -- I'm going to leave that to the coding guru's. I did find the following article, which may be useful to that particular guru:

https://philbull.wordpress.com/2012/07/27/portability-issues-with-floating-point-exception-handling/
https://philbull.wordpress.com/2012/12/09/update-floating-point-exception-handling-on-mac-os-x/

What I do have, is a workaround, which you can find here:

https://github.com/WillemEerland/jsbsim/tree/macos-fix

It comments out two lines, and switched around a declaration of Aeromatic++ (Why does the compiler on MacOS complain, and not on Linux?)

It compiles on MacOS, and completes almost all tests, except "TestInputSocket".

97% tests passed, 1 tests failed out of 34

Total Test time (real) = 158.06 sec

The following tests FAILED:
10 - TestInputSocket (Failed)
Errors while running CTest

I believe it will only influence the Python library and only the input socket exception handling... so depending on what you're planning, this may work for you!

[ezietsman]

OK!

Aeromatic seems to not build using clang but it does with GCC. Here's what I did:

export CXX=g++-8
export CC=gcc-8
cmake ..
make -j 4

and this worked fine after I commented out those two lines in fpectmodule.cpp

[bcoconni]

@ezietsman and @WillemEerland the fpectl Python module is only used to make sure that none of the tests raise a floating point exception. JSBSim does not need it for a day to day usage. Even the tests can be run without fpectl, it is just that if a floating point exception is raised, it will be ignored and the test will continue.

I will investigate later how the fpectl code can be modified to run on MacOS X as well.

Regarding TestInputSocket it fails from time to time. This is most likely due to a race condition in it, you can run it again using the following command

> ctest -R TestInputSocket

If the failure was due to the race condition then it should succeed after a couple of try. If it does not, then you can get more information about the test failure with the -V (verbose) flag.

> ctest -R TestInputSocket -V

[WillemEerland]

I tried it five times, failed five times. Here's (part of) the more detailed report

10: End of vehicle configuration loading.
10: -------------------------------------------------------------------------------
10: ok
10:
10: ======================================================================
10: FAIL: test_no_input (main.TestInputSocket)
10: ----------------------------------------------------------------------
10: Traceback (most recent call last):
10: File "TestInputSocket.py", line 156, in test_no_input
10: TelnetInterface(fdm, 5., 1137)
10: AssertionError: error not raised
10:
10: ----------------------------------------------------------------------
10: Ran 3 tests in 2.856s
10:
10: FAILED (failures=1)
1/1 Test #10: TestInputSocket ..................***Failed 3.33 sec

[ezietsman]

OK so it seems fpectl isn't in python 3, which for various reasons, I need to use at the moment. I'm trying to integrate jsbsim into a game engine, so all I really want to do is to pass in (throttle, rudder, roll, pitch, delta_time) inputs and want back (X, Y, Z, roll, pitch, yaw, Vx, Vy, Vz) for each time step.

I'm happy to hack at JSBSim to allow it to build python 3 bindings but I may need some guidance.

[bcoconni]

@WillemEerland OK thanks for the report. It seems that TestInputSocket fails for a reason different from a race condition on your system. Could you open a separate issue to address this topic ?

And in the meantime could you please check that the following lines of Python code raise a socket.error exception on your platform as well ?

import telnetlib
telnetlib.Telnet("localhost", 1137, 2.0)

[bcoconni]

@ezietsman and @WillemEerland the PR #23 has been merged. I think this issue can now be closed ?