Site Navigation

• About
• Research
• Publications
• Teaching
• Services
• Curriculum Vitae (PDF)
• Keynotes and other Invited Talks
• Recognitions and Awards
• Tools
• University and Institute

@MastersThesis{Wilms2013,
  Title                    = {Instant Incremental Class Invariant Checking for Java Virtual Machines},
  Author                   = {Sebastian Wilms},
  School                   = {Johannes Kepler University (JKU), Linz, Austria},
  Year                     = {2013},

  Abstract                 = {Software systems fail. Although software developers mostly try their very best to avoid errors
in software systems, failures are unavoidable. Sometimes the failures are catastrophic, resulting
in financial loss or even danger to human safety. Often the faulty behavior leads to the system
being in an invalid state, which may cause follow-up errors later on. Yet, invalid states of a
system can be detected using class invariants. Class invariants express the valid states of a
software system and must not be violated. Through exhaustive testing, these invariants can
then be tested to ensure that they hold. Invariant checking may thus avoid failure.
Unfortunately, existing approaches for invariant checking require extensive manual intervention
in where/how to place the invariants in the code. Invariants only need to be validated
in situations where its validity is affected. This requires careful placement of invariants checks
in the code – a placement that needs to be reconsidered every time the code changes or the
invariant changes. The only alternative is to validate the invariants continuously at every step
of the software’s execution. However, doing so is unscalable.
This thesis presents a novel, tool-supported approach for invariant checking that automatically
decides when to validate invariants in a manner that is significantly more scalable compared
to placing invariants everywhere and in a manner that is guaranteed correct.
Apart from defining the invariant, no additional user interaction is required. The approach
avoids the invariant placement problem by observing the software system’s execution and triggering
an invariant check only if the software system’s state changed in a relevant manner. By
observing the execution and recognizing state changes of a software system during runtime, we
are able to drastically reduce the number of performed invariant checks. The proposed approach
has been implemented for the Java programming language but is, in principle, applicable to
other languages also.
We evaluated its correctness, performance and scalability on several open source case studies.},
  File                     = {:MSc Theses\\2013 Sebastian Wilms\\Wilms - Instant Class Invariant Checking for Java.pdf:PDF},
  Owner                    = {AK117794},
  Timestamp                = {2015.09.22}
}