Testing Components in the Context of a System

1. Testing Components in the Context of a System CMSC 737 Fall 2006 Sharath Srinivas

2. Outline • Introduction to COTS • Advantages of COTS • Challenges in testing Components • Techniques for building testable components • Regression Testing of Component Systems • Conclusions

3. Introduction • What is Component Based Software Engineering (CBSE)? • This approach is based on the idea to develop software systems by selecting appropriate off-the-shelf components and then assemble them with a well-defined software architecture. • CBSE has three major functions: • Developing software from prefabricated, reusable parts. • The ability to use those parts in other applications. • Easily maintaining and customizing those parts to produce new functions and features.

4. Component 1 Component repository Component 2 Software system select Component n assemble Introduction… • What is a Commercial Off the shelf (COTS) component: • A component is an independent and replaceable part of a system that fulfills a clear function. • A component works in the context of a well defined architecture. • A component communicates with other components by its interfaces. • Can be developed by different developers, using different languages and different platforms.

5. Advantages of COTS • Development cost is reduced • Development time is reduced • Complex systems can be built by reusing pre-existing components • Testing effort is reduced.

6. Developer Builds Component Developer Tests Component Component Repository with adequately tested components Final Software Product Component User uses the components Component Based Systems: Architecture • Layman’s interpretation of Component Based System:

7. Component Based Systems: A cautionary Tale-Ariane example • Ariane Flight 501 crashed 30 sec after take-0ff • One of the reasons for the failure of the first Ariane 5 was that a conversion to integer its software failed (overflow) • This code was taken from Ariane 4, and had been successfully tested for Ariane 4. • But Ariane 5 had a higher initial velocity and was heavier than Ariane 4. • The story learnt: Even adequately tested components can fail when the context in which it is used changes.

8. Testing Components: Challenges • Testing is done by the developer. • System Requirements not available. Instead the developer has to use Component specifications. • Unpredictable Environment. • Unpredictable interaction with other components.

9. Context Sensitive Component Testing: Challenges • Testing is done by the Component User. • Unavailability of source code. • Source code in various languages. • Dependencies and interactions among components is complex.

10. Coverage Notions for Context Sensitive Component Testing Component Services are accessible through interfaces.

11. Coverage Notions for Context Sensitive Component Testing… • Interface Coverage criterion: Requires that all interfaces for a component should be exercised at least once. • Similar to Black box testing. Test these 2 interfaces Interface 1 Interface 2

12. Component 1 Component 2 Interface 1 Interface 2 Interface 1 Interface 2 Component 3 Interface 1 Interface 2 Coverage Notions for Context Sensitive Component Testing… • Event Coverage Criterion: • Test an interface against all its possible invocations • Test Interface 1 of component 3 invocated by Interface1 of Component1. • Test Interface 1 of Component 3 invocated by Interface 2 of component 2

13. Coverage Notions for Context Sensitive Component Testing… • Context Dependency Test Coverage criterion: • Sometimes events can have sequential dependencies on each other and the order in which they are triggered can result in different program behavior. • Such dependencies are called context dependencies. Component 3 Component 2 Interface 2 Interface 1 Interface 2 Component 1 Interface 1 Interface 1 Interface 2 • Test all possible operational sequences in the in the component • Similar to path Coverage

14. Coverage Notions for Context Sensitive Component Testing… • Content Dependency test criterion • A content dependency exists between 2 interfaces I1 and I2 if an operation of I2 modifies the value of a variable used in an operation of I1. • In this case I2 is said to be content dependent on I1. • Test all the content dependencies in the software. • Similar to All Def-Use criterion

15. Coverage Notions for Context Sensitive Component Testing…

16. Testable COTS components • Techniques for building testable components • Built In tests • Interface Probing • Traceable Components • Self testing components • State Machine Models…. • All these techniques involve the efforts of both the Component developer and Component User

17. Built In tests • The developer Packages components with the test cases. • These test cases may be executed on the component by using a test executor. • Using built in test cases a component user can validate the behavior of the COTS functionalities.

18. Interface Probing • Component users might employ this technique to understand the component functionalities. • The functionality can then be used as the basis for creating test oracles. • The component developer can reduce the effort required to determine component properties by automating certain aspects of interface probing.

19. Traceable Components • Tractability is a facility that a component provides to track its internal and external behavior. • Tractability can be used to create test oracles.

20. Self Testing Components • Augment the component with functionality of analysis and testing tools. • A component augmented accordingly is capable of performing some or all activities of the component user’s testing process. • Component can run in 2 modes: • Normal mode • Maintenance mode

21. State Machine Models • Techniques discussed in the previous slides are techniques for better exchange of information between Component developer and user. • Based on the knowledge gained the component user has to write test cases to validate if the component will work in its context. • An automated technique to check if a component will work in its context is needed

22. State Machine Model • Test cases generated for testing components in their context. • This component considers each component as a FSM. • For representing a system and its context all the Finite State Machines are combined together to form a Component Based Flow Graph (CBFG)

23. State Machine Models…

24. State Machine Model • Based on the CBFG test cases that are context adequate and content adequate are generated. • Problem in State Machine Models: The testing effort of a component can be greater than the re-development effort of the same component!

25. Modified CBS Architecture Developer Builds Testable Components Developer Tests Components Component Repository with adequately tested components User tests the COTS in their Context Final Software Product Component User validates the components

26. Regression Testing of Component Systems • Software built using components have system components and COTS. • New challenges: • System components might evolve while COTS components remain static. • COTS components might evolve while the system components remain static. • Both COTS and system components can evolve.

27. Conclusions • Testing components in the context of their system is not trivial! • Current research is focusing on 'building testable components' than 'testing components in their context' • This gives an indication that if a component is not testable it is difficult/ impossible to test it in it in the context of its system. • Future research in this field might focus on making use of the ‘testable components’ to automate testing components in the context of their system.