Movement along non-cardinal directions is slighty off.

[stephank]

When moving along a direction one step away from a cardinal direction, the movement simulation is incorrect. These are the directions at 22.5°, 67.5°, 112.5°, etc.

This can be clearly reproduced when at a minimum velocity, by just tapping forward until movement starts, then turning to one of these directions. Movement appears 'rounded off' to the nearest cardinal direction.

The code to this is in src/tank.coffee. I tried to model it after WinBolo's movement code. Perhaps this is a floating point rounding thing?

[stephank]

Looks like winbolo actually does this too.