Exploiting Formal Methods in Software Development

1
Exploiting Formal Methods in Software Development

• Duncan Clarke
• 14 September 2001

2
Me
   
PWI
RPI
MSU
BTL
Penn
IRISA/INRIA
UK
USC
3
Research

• 1996
• Automated test generation from formal
  specifications for resource-bound real-time
  systems

• 2001
• Software Engineering

4
Software Engineering is

• Process
• Management problems in the software development
  process
• Estimation
• Allocation
• Quality Assurance
• Deployment
• Etc.

5
a lot of different things

• Requirements
• Specification
• Reverse Engineering
• Analysis (Static and Dynamic)
• Testing
• Architectures
• Etc.

6
My piece of the puzzle

• Formal Methods
• Mathematically sound
• Well-defined semantics
• Concise syntax
• Support automated proofs of partial program
  correctness
• E.g., CSP, CCS, StateChart, etc.

• Software Development
• Specify (natural language)
• E.g., Natural language, tabular methods, UML,
  etc.
• Code
• Test partially

7
Why not Formal Specification?

• Proof technique
• State space exploration
• Industrial application
• State space explosion
• Other concerns
• Mathematical sophistication of users
• Learning curve
• Etc.

8
Emphasis on Application

• Getting more value from formal specs
• Skirting state-space explosion
• Testing techniques
• Symbolic techniques
• Hiding complexity of analysis
• Automated test generation and application
• Automated
• New uses

9
So Far

• Automated test generation from formal
  specifications for resource-bound real-time
  system models (with Insup Lee at Penn)

• Extracting executable tests with oracles from
  symbolic specifications (with T. Jeron, V. Rusu
  and E. Zinovieva at IRISA)

10
Next

• Concentrate on existing tools
• CAESAR, Spin
• Parallel state space generation
• Specification, test and debug issues in parallel
  programs
• Attack a specific class of defects for testing
• Race conditions in concurrent systems
• Knowledge xfer, specifier to programmer
• Driving visualizations