Telemetry Payload errors

[ptheywood]

The telemertry payload has a number of (potential errors) from a quick look at some data:

• [ ] pyflamegpu is never reported - this is (probably) due to the #ifdef SWIG being in a source file not a header file, and SWIG is only defined when building the pyflamegpu interface.
• [x] buildNumber has been reported as buildNumber rather than an actual hash. This value is set at CMake configuration time from a call to git, which should be on the path at configure time otherwise an error should occur so not sure how this is happening.
  • This appears to be an issue with the manylinux CI jobs not getting the build number at cmake time, so manylinux workflows need adjusting to fix the git related warning.
  • Fixed by #1036

[Robadob]

To catch python, the most sensible option is probably to augment the constructor for ModelDescription or CUDASimulation/CUDAEnsemble, to catch SWIG (pass a bool to the normal constructor defined private). Lazy version would be to detect use of RTC.

Similar to what's being done here for the visualiser detecting it.

[Robadob]

Marked this planned as it would be useful to identify C/Python breakdown if we're going to be publishing Python specific paper.

[Robadob]

Bit more testing this morning/yesterday, can't do the same trick as the visualiser for CUDASimulation::simulate() as it's a virtual function it cannot be inlined. Needs some more thought.