Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.770274
Title: Revealing behaviours of concurrent functional programs by systematic testing
Author: Walker, M. S.
Awarding Body: University of York
Current Institution: University of York
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
We aim to make it easier for programmers to write correct concurrent programs and to demonstrate that concurrency testing techniques, typically described in the context of simple core languages, can be successfully applied to languages with more complex concurrency. In pursuit of these goals, we develop three lines of work: Testing concurrent Haskell We develop a library for testing concurrent Haskell programs using a typeclass abstraction of concurrency, which we give a formal semantics. Our tool implements systematic concurrency testing, a family of techniques for deterministically testing concurrent programs. Along the way we also tackle how to soundly handle daemon threads, and how to usefully present complex execution traces to a user. We not only obtain a useful tool for Haskell programs, but we also show that these techniques work well in languages with rich concurrency abstractions. Randomised concurrency testing We propose a new algorithm for randomly testing concurrent programs. This approach is fundamentally incomplete, but can be suitable in cases where systematic concurrency testing is not. We show that our algorithm performs as well as a pre-existing popular algorithm for a standard set of benchmarks. This pre-existing algorithm requires the use of program-specific parameters, but our algorithm does not. We argue that this makes use and implementation of our algorithm simpler. Finding properties of programs We develop a tool for finding properties of sets of concurrency functions operating on some shared state, such as the API for a concurrent data type. Our tool enumerates Haskell expressions and discovers properties by comparing execution results for a variety of inputs. Unlike other property discovery tools, we support side effects. We do so by building on our tool for testing concurrent Haskell programs. We argue that this approach can lead to greater understanding of concurrency functions.
Supervisor: Runciman, Colin Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.770274  DOI: Not available
Share: