This repository contains the source code and experiment data for the paper GenTree: Using Decision Trees to Learn Interactions for Configurable Software.

• benchmarks: testsuites and configuration files for all benchmarks presented in the paper
• exp-data: raw benchmark resutls
• gentree-src: the GenTree source code (written in C++).

GenTree uses the CMake build system and requires the following external libraries:

• Boost 1.72 (https://www.boost.org/)
• fmt (https://github.com/fmtlib/fmt)
• GLog (https://github.com/google/glog)
• RocksDB (https://github.com/facebook/rocksdb)
• Z3 (https://github.com/Z3Prover/z3)
• Zstd (https://github.com/facebook/zstd)

To build GenTree, runs:

mkdir build && cd build
cmake ..
make

To run the example in the Illustration section, use

./gentree -J2 -cx -BF @ex_paper

To run the example with the full config space, use

./gentree -J2 -cx -BF @ex_paper --full

To run otter programs (vsftpd, ngircd), use

./gentree -J2 -cx -YF benchmarks/2/ngircd

(change benchmarks/2 to the correct folder containing benchmark configs)

To run gcov programs (e.g. id), use

./gentree -J2 -cx -GF benchmarks/2/id

(change benchmarks/2 to the correct folder containing benchmark configs; and change conf.gcov file appropriately for executable paths and testsuite directory)

For helps on GenTree CLI options, uses

./gentree --help

• exp-data/{ls, cat, ...}: GenTree's outputs (11 runs for each program), and optionally a full.txt file containing the results generated using all configuration space.
• exp-data/Analyze/{stat, mcc, cmin, stat_full}: aggregated experiment data that is used to fill in tables in the paper
  • stat: Table 3. GenTree’s results for the benchmark programs
  • mcc: Table 4. Comparing GenTree to exhaustive runs
  • cmin: Table 6. Minimal Configurations
  • stat_full: Table 7. Comparing GenTree to iGen.
• exp-data/Analyze_rand: Table 5. Comparing random search to exhaustive runs
• exp-data/{Analyze_progress, Analyze_progress_rand}: Fig. 10. Progress of GenTree on generating interactions for a subset of benchmark programs.