Spencer is a web service that hosts trace data that you can analyse using a web browser or an API. The data stores executions of programs running on the JVM, including (almost) all field stores/reads, method calls/exits. Spencer is a tool you can use – for instance – to quickly, and without any local setup, find behaviours in programs that are common or to find those objects that don’t follow these behaviours.

Spencer focuses on analyses of the heap, and comes with a combinator style DSL that lets you express analyses succinctly and readably. The DSL has combinators to compute reachability and logical connectives permitting complex analyses being implemented from simple ones. Analysis results are cached on the server, which improves performance to a level that enables analyses that are more detailed than what would otherwise be possible on commodity hardware.

In this talk, I will present Spencer, and also give a demo. http://spencer-t.racing

PhD student at Uppsala University, working on Aliasing Control. See my web page for more info!