Difference between "Pass/Fail Testing" "Test To Failure" "Degradation Test"

This is a prototype functional test and can be verified without thorough testing. If the prototype passes the test, it fails if it fails, it fails.

The test is very weak and it is possible to detect defects related to "wear" problems, material over stress problems, deterioration, etc.

Test the prototype multiple times and stop the test when the prototype fails. Then investigate the root cause of the failure and improve the design to eliminate the cause, if necessary. Examples of testing are ALT (accelerated life test) and HALT (high acceleration life test). ALT can detect wear mechanism. HALT may detect design problems where the "margin" is too small and the component is weak.

Test the prototype for a certain period (or cycle), measure the specified key characteristics, then continue the test, measure the same characteristics again after a certain period (or test period), and so on. It allows you to see how important functions change over time / cycle - this is "degradation". Because the prototype function detects a change in function before it fails, the detection method is superior to the first two.

If no failure is detected after a specific round pass / fail test, the estimated failure rate will be zero if the individual tests are segregated. Pass / fail tests are often misunderstood that the zero defect rate of a particular sample tested does not guarantee that the defect rate of the whole product or part will be zero. If failures were not reported during the sample tests, the researchers naturally tended to ignore the maximum failure rate possible for the entire population. It is necessary to understand the maximum failure rate of the population, not the sample. It also needs to be part of risk assessment related to decision making.

It is an abstraction. The program failed due to some test cases. What condition of the test case leads to a specific failure? The Delta Debugging algorithm summarizes several failed test cases and reduces it to the smallest failed test case; it also separates the difference between working test cases and failed test cases. In the case study, the Mozilla Web browser crashed after 95 user actions. Our prototype implementation automatically reduces the input to three related user actions. Again, it simplifies 896 lines of HTML to one line, causing a failure. In case studies, 139 automated tests or 35 minutes are required on a 500 MHz PC.