Not very scientific measures. :) I usually test out the code on real world geodata and verify results visually, then if everything looks right, incorporate some of the shapes as test fixtures. OSM vector tiles have a huge variety of data, so any edge case issues can be found pretty quickly. After that, I have to rely on other contributors to find edge cases that don't appear on the data I test.
Additionally, I rely on test coverage reports to make sure the test fixtures cover every branch of the code, so my small JS libraries usually have 100% test coverage, which also helps catch many issues.
Not very scientific measures. :) I usually test out the code on real world geodata and verify results visually, then if everything looks right, incorporate some of the shapes as test fixtures. OSM vector tiles have a huge variety of data, so any edge case issues can be found pretty quickly. After that, I have to rely on other contributors to find edge cases that don't appear on the data I test.
Additionally, I rely on test coverage reports to make sure the test fixtures cover every branch of the code, so my small JS libraries usually have 100% test coverage, which also helps catch many issues.