This seems like something that a semantic diff could aid in tremendously. Determining which code to test via coverage or attempting some text parsing based on

This seems like something that a semantic diff could aid in tremendously. Determining which code to test via coverage or attempting some text parsing based on commits is likely be very error-prone. Determining code to test based on differences in a syntax tree seems more concrete to me.