I've studied numerous articles on software regression testing optimization techniques so far, but I've rarely seen any of them notice whether or not it takes less time to carry out the proposed technique plus the time spent on executing the output test cases of the technique than the time spent on executing the test suite without any optimization(retest-all).
Why do you think the articles have not mentioned this point?
We ran millions and millions of automated tests on our software 24x7 on 20 platforms, for years. It seems to me that there are tests to prove that particular features or changes work and tests that prove the product still works after the change. Tests for change are small, whereas tests for the whole product are usually big. Optimization is complex to design (which tests to run, and why), even if it is easy to execute, so optimization takes labor and can be expensive. The only optimization that matters is which tests to run - tests to check a new feature or tests to prove the product right. I totally agree with Peter - hopefully tests can be run in parallel, so they are probably inexpensive compared to human labor.
- Badges
- Science topic
- Similar topics
- Engineering
- Computer Engineering