TamiFlex facilitates static analysis of programs that use reflection and custom class loaders
We would appreciate if you could briefly let us know about how you use TamiFlex. To do so, click here. It will only take a minute. Thanks!
TamiFlex is a tool suite to facilitate static analyses of Java programs that use reflection and custom class loaders. The suite consists of two agents that use the java.lang.instrument API, one Play-out Agent and one Play-in Agent. Our Overview document gives more details.
TamiFlex consists of three components, the Play-out Agent, the Play-in Agent and the Booster.
The Play-out Agent allows you to:
The Booster takes as input a class-file folder and reflection log produced by the Play-Out Agent and produces as output an enriched version of the program that contains "materialized" versions of the original reflective calls in the form of standard Java method calls. Statically analyzing the "boosted" program instead of the original program allows static-analysis tools to treat the reflective calls just as standard Java method calls.
With the Play-in Agent you can cause the virtual machine to load classes from a specified directory instead of from they would normally be loaded from. This is useful for replacing classes by statically optimized classes irrespective of the program's class-loading setup.
TamiFlex is a joint effort of Eric Bodden, Andreas Sewe, Jan Sinschek and Mira Mezini, of the Software Technology Group and Secure Software Engineering Group at TU Darmstadt. This work is supported by CASED.
TamiFlex uses technology from ASM, which is under this OS license, and from Soot, which is under LGPL.
This document gives an Overview of TamiFlex.
Our Usage page tells you how to use TamiFlex.
TamiFlex enables static whole-program analysis of DaCapo benchmarks using Soot. Here you can read more about how this works.
If you are interested in the internal workings of TamiFlex, please consult our ICSE 2011 paper or our Technical Report.
We also have all benchmarking results from this Tech Report online, along with scripts to reproduce them. Please consult our Benchmarks page for details and also see this Github project.