Michel Hidalgo
Michel Hidalgo
Additional considerations: - Define a QA procedure based on autonomously traversing a fixed path. - Record bag file throughout the entire procedure for post-mortem analysis. - (stretch) Use InOrbit for...
We are closer to automated performance regression tests than we are from HIL tests. I'll close this for now, we can resurrect when applicable.
May want to give this another try once #266 and #267 land.
> Evaluate the performance of Bresenham's algorithm for the Bean Sensor Model to determine if it's worth optimizing Note benchmarks are included in #160. Integer arithmetic-based Bresenham algorithm benchmark ```...
FWIW I don't think we can improve the computation itself much further. We need to reduce the number of computations.
> Which theme did you test it with? I'm using the `sphinx-book-theme` but without many of the [customizations](https://github.com/boschglobal/doxysphinx/blob/main/docs/_static/sphinx-book-theme-custom.css) you have in this repo.
AMD docs look pretty too. Don't mind which path we take so long as we can get the dark mode to work. Great tool BTW.
I always thought this was MBARI's code doing some form of self-check, but I guess it could also be a very underdamped pitch controller.
https://github.com/protocolbuffers/protobuf/issues/4126 suggests loading/unloading protobuf message shared libraries leaves bad global state behind. Perhaps we can preload them?
We know why this happens and it's very much core to how `protobuf` works. I'm inclined to close this as `wont fix`.