dstockhouse / mr-fusion-capstone Goto Github PK

Right now, there are many places in the guidance code that may encounter an error and panic. During debugging this is acceptable, but for operation of the robot we would prefer to detect errors and tell the UI and/or exit gracefully and restart rather than just crash.

This is a possibility whenever an option type uses .unwrap()

Currently, our unit tests are mocking TCP connections, but some of these mocks are globally defined. This requries us to either

1. Use syncronization mechanisms in our unit tests to make sure the mocks return codes are not subject to races

2. Run our tests in one thread (which is currently what we are doing)

It would ideal if we could find a mocking library that does not require these mocks to be global or require explict condition compilation within our source code.

More test cases than the one "obvious" pass and fail cases.
Some test cases I want at a minimum:

• Not on graph if the point is on campus but outside the path (ex. library parking lot)
• Not on graph if the point is within the graph boundary but not close enough to a path (ex. the middle of the quad)
• Not on graph if the point is very close to any arbitrary segment, but not part of the segment itself (I want at least three of these on different segments of the path)
• Closest edge found if the robot is within the threshold of two different edges, but closer to one of them (ex. a sharp intersection, but still clearly on one edge)
• No path found for an unconnected graph (may have to create unconnected graph file like the triangle example)
• Find a path between some pairs of adjacent vertices
• Find a path between some mildly distant points towards the center of campus
• Find a path between a few pairs of points on opposite sides of campus

When you get to passing the output of closest_edge_to to shortest_path:

• All path cases above but mix starting right at vertex and starting in the middle of an edge, multiple "positive" cases to test for completeness here
• Single-edge paths found from closest edge to each adjacent vertex (so 2 times running shortest_path_found from the same starting point in the middle of an edge)

Every pass-by-pointer function needs to check if the input pointer is NULL before starting

Each error or success return code needs to have a defined constant explaining what the value represents, and so that the calling function can reference what error conditions there are

Header files should only have constants, type definitions, and public functions. Any functions not made available for use by external modules should not be in the header file