Michael Behrisch

maybe we can implement something along the lines of https://github.com/getsentry/raven-python/commit/4c3b7eee56246fb3bb5118879171a59806b730fb

The usage of print_stack was wrong. It should be `print_stack(file=t)`.

The answer is surprisingly simple. A vehicle needs to be able to switch off the engine at coasting. Otherwise the engine and the wheels would work against each other. That's...

If I understand it correctly a "Motorbremsung" means exactly that. The wheels are "moving" the engine so the engine does not burn any fuel otherwise it would not brake.

here are average microseconds per vehicle update in the acosta plain scenario: ``` v0_25_0 2.843 v0_26_0 3.016 v0_27_0 3.346 v0_28_0 3.663 v0_29_0 3.764 v0_30_0 4.160 v0_31_0 4.083 v0_32_0 4.989 v0_32_0-1253-g63c03b44b5...

64ce3d0 increases vehicle update time by about 10%, this might be the major increase between 0.31 and 0.32

Without collision.action warn I get the following results for the A9: ``` v0_26_0 3.608 v0_27_0 3.898 v0_28_0 4.507 v0_29_0 4.267 v0_30_0 4.666 v0_31_0 5.143 v0_32_0 5.620 v0_32_0-1253-g63c03b44b5 5.889 v0_32_0-2706-g2834c9eaed 6.185...

The numbers for brunswick: ``` v0_25_0 6.605 v0_26_0 7.287 v0_27_0 8.274 v0_28_0 8.795 v0_29_0 8.212 v0_30_0 9.166 v0_31_0 9.286 v0_32_0 10.848 v0_32_0-1253-g63c03b44b5 11.034 v0_32_0-2706-g2834c9eaed 11.572 v1_0_1 11.171 ```

The numbers for extended/stops/stress: ``` v0_25_0 0.799 v0_26_0 0.870 v0_27_0 1.019 v0_28_0 1.192 v0_29_0 1.191 v0_30_0 1.386 v0_31_0 1.313 v0_32_0 1.708 v0_32_0-1253-g63c03b44b5 1.697 v0_32_0-2706-g2834c9eaed 1.852 v1_0_1 2.038 ```

not so surprisingly, the commit which is responsible for the major slowdown between 0.26 and 0.27 (in the stress test) is f3b08a38a. I am not sure how to proceed here....