Automatic Testcase Generation for Flight Software

The TacSat3 project is applying Integrated Systems Health Management (ISHM) technologies to an Air Force spacecraft for operational evaluation in space. The experiment will demonstrate the effectiveness and cost of ISHM and vehicle systems management (VSM) technologies through onboard operation for extended periods. We present two approaches to automatic testcase generation for ISHM: 1) A blackbox approach that views the system as a blackbox, and uses a grammar-based specification of the system's inputs to automatically generate *all* inputs that satisfy the specifications (up to prespecified limits); these inputs are then used to exercise the system. 2) A whitebox approach that performs analysis and testcase generation directly on a representation of the internal behaviour of the system under test. The enabling technologies for both these approaches are model checking and symbolic execution, as implemented in the Ames' Java PathFinder (JPF) tool suite. Model checking is an automated technique for software verification. Unlike simulation and testing which check only some of the system executions and therefore may miss errors, model checking exhaustively explores all possible executions. Symbolic execution evaluates programs with symbolic rather than concrete values and represents variable values as symbolic expressions. We are applying the blackbox approach to generating input scripts for the Spacecraft Command Language (SCL) from Interface and Control Systems. SCL is an embedded interpreter for controlling spacecraft systems. TacSat3 will be using SCL as the controller for its ISHM systems. We translated the SCL grammar into a program that outputs scripts conforming to the grammars. Running JPF on this program generates all legal input scripts up to a prespecified size. Script generation can also be targeted to specific parts of the grammar of interest to the developers. These scripts are then fed to the SCL Executive. ICS's in-house coverage tools will be run to measure code coverage. Because the scripts exercise all parts of the grammar, we expect them to provide high code coverage. This blackbox approach is suitable for systems for which we do not have access to the source code. We are applying whitebox test generation to the Spacecraft Health INference Engine (SHINE) that is part of the ISHM system. In TacSat3, SHINE will execute an on-board knowledge base for fault detection and diagnosis. SHINE converts its knowledge base into optimized C code which runs onboard TacSat3. SHINE can translate its rules into an intermediate representation (Java) suitable for analysis with JPF. JPF will analyze SHINE's Java output using symbolic execution, producing testcases that can provide either complete or directed coverage of the code. Automatically generated test suites can provide full code coverage and be quickly regenerated when code changes. Because our tools analyze executable code, they fully cover the delivered code, not just models of the code. This approach also provides a way to generate tests that exercise specific sections of code under specific preconditions. This capability gives us more focused testing of specific sections of code.