Data-race Freedom and Memory Safety for Untyped Languages

The goal of this research is to strengthen data-race freedom and memory safety in software development.We focus on untyped programming languages such as Python and JavaScript, which are widely used across a wide range of domains. The increased use of these languages forces their evolution to embrace concurrency while staying safe.

This is challenging for the implementation of these languages and adds new classes of possible bugs to programs written in these languages.Our approach is to make a technique from the world of statically typed languages available in the untyped world: ownership.

Static ownership annotations have successfully been used in the past to avoid data races and ensure memory safety, using types to propagate ownership information and rule out violating programs at compile-time.

However, this entirely static approach excludes the class of untyped programming languages for which there is no compile-time: Worse: these programs often use reflective mechanisms to construct themselves incrementally during execution.

Our main challenges are: dealing with such dynamism without losing neither a strong notion of ownership nor destroying the advantage of this class of programming languages; tracking and enforcing ownership efficiently at run-time; provide alternatives to the static means for supporting programming with ownership so that programmers can take advantage of our results.

We evaluate our work through integration into the Python programming language.