CS362 Lesson 17, 18

Tarantula Approach

A statistical approach to fault localization. It was originally a visual representation of how suspicious a line was. After many runs and analyzing the code, green, yellow, and red highlighting of lines would be displayed, showing what the program thought the areas with bugs could be in. Can quantify statements by showing how many tests failed and passed that statement, and all the failed tests and all the passed tests. Suspiciousness is calculated from these values. Suspiciousness of 0 means not suspicious at all, and 1 means we think there is highly highly likely to have fault. It actually instead of just saying says "here's a suspicious statement," it provides a ranking of every single statement in the whole program instead of just a set of notes.

Fault Localization

An effort to automatically find (one of the ) causes of an observable failure It is inherently difficult because there are many causes of the failure that are not the fault.

Renieris and Reiss' Localization

Basic idea is that we have a lot of test cases: some fail, but a much larger number pass the tests. You pick a failure and then compare that to the most similar successful test and report those differences. What if there are no similar successful tests?

Causality

When a test case fails we start debugging We assume the fault causes the failure What does "A causes B" mean? We don't know. General agreement that cause is something that makes a difference. If A does not take place, B also does not take place Causality does not depend on B being impossible without A.