SPF WCA

SPF-WCA is a tool for algorithmic complexity analysis based on symbolic execution. It uses history-based guidance policies that enable exploration at large input sizes (e.g., number of elements to be sorted) of the target program.

To cite SPF-WCA, please use the most recent paper that was accepted at ICST 2017 (won the Best Paper Award):

Kasper Luckow, Rody Kersten, Corina Pasareanu, Symbolic Complexity Analysis using Context-preserving Histories, Proceedings of the 10th IEEE International Conference on Software Testing, Verification and Validation (ICST 2017), [pdf] [bibtex].

Installation

We provide two ways of installing SPF-WCA:

• Installation on local machine
• Virtual machine with Docker

Local Machine

Before continuing, make sure that jpf-core and jpf-symbc are installed.

To install SPF-WCA, update your site.properties file (usually ~/.jpf/site.properties) and set the spf-wca variable to point to the directory of your SPF-WCA installation.

spf-wca=/path/to/spf-wca

Don't add spf-wca to the extensions variable.

Make sure you have Ivy installed on your system. To bootstrap the Ivy ant task, you can run:

$ ant bootstrap

Then, obtain all the dependencies by running:

$ ant resolve

The dependencies will be downloaded to lib/.

Now SPF-WCA can be built by simply running:

$ ant build

Docker

Assuming you have Docker installed, simply run:

$ docker build -t spf-wca .
# Will take some time to build the image...
$ docker run -it spf-wca

Note that, because there is no X11 available, the fitting functions cannot be displayed. However, the results can be checked in the CSV file under verbose/heuristics in the results folder set by JPF option symbolic.worstcase.outputpath (see below for more details). To copy the results to the host machine from the Docker container, use:

# Note the container id (CID) from the following command (column CONTAINER ID)
$ docker ps
$ docker cp <CID>:/path/to/csv/file/in/container /destination/on/hostmachine

Usage

The Java PathFinder shell wcanalysis.WorstCaseAnalyzer can be used to set up Phase 1 (policy generation) and Phase 2 (guided search) of the analysis. The analysis can be performed by executing the JPF config file that specifies the parameters of the analysis, the constraint solver, the entry point of the system under analysis etc:

$ ./jpf-core/bin/jpf <path-to-jpf-file>

This will generate a results folder according to the option symbolic.worstcase.outputpath (see below).

In the results folder, a subfolder verbose/heuristic/ will be generated that contains the constraints found for the analyzed input size. In addition a csv file is produced which summarizes all the results. Note that consecutive runs of the analysis will produced and new data set that is appended to this file.

The csv file includes various statistics:

• inputSize Current input size to the SUT. It assumes this information is available in the jpf file as the value of the target.args property (it is for the examples in benchmarks.heuristic)
• wcInstrExec Shows how many instructions were executed on the worst case path
• cfgInputSize Specifies which input size was used for generating the annotated CFG used for the heuristic. This will by default not be added to the .csv file
• analysisTime The analysis time when using the heuristic
• mem The peak memory consumption during heuristic-based exploration
• depth The maximum depth recorded (which constitutes the worst case path)
• paths Number of paths explored with the search heuristic
• resolvedChoicesNum How many decisions that were resolved as a consequence of using the heuristic
• unresolvedChoicesNum How many decisions that could not be resolved by the heuristic
• newChoicesNum New decisions encountered during exploration for which the heuristic has no information
• wcConstraint The constraint recorded for the worst case path. Any solution to this provides test inputs that are guaranteed to exercise worst case behavior of the SUT

In addition, data points <inputSize, depth> (here depth is the notion of worst case for a path) are generated, and the shell will automatically generate a plot showing the raw data, and the various fitted functions based on regression analysis. The user can zoom in on the graph by highlighting a region.

Configuration

The following must be supplied in jpf file (or imported from another jpf file using the @include directive):

@using spf-wca

shell=wcanalysis.WorstCaseAnalyzer

classpath=${spf-wca}/build/examples
target=fully.qualified.name.of.target.class

symbolic.worstcase.policy.inputsize=XXX
symbolic.worstcase.input.max=YYY

symbolic.wc.policy.history.size=ZZZ

The classpath variable should be updated according to the system under test. target denotes the entry point of the system under test.

Replace XXX with the input size at which the policy should be obtained. This corresponds to phase 1. Replace YYY with the maximum input size at which the heuristic search (phase 2) should be run. The heuristic search will run from input size 1-YYY.

symbolic.wc.policy.history.size is important because it controls whether the guidance policy produced in phase 1 is memoryless or history-based. By setting this variable to 0, a memoryless policy is used; otherwise, a history with the specified size ZZZ will be used.

Optional Configuration

In addition, the user can optionally use the following for the WorstCaseAnalyzer shell:

• symbolic.heuristic.measuredmethods=<method desc(s)> A list (separated by semicolons) that specifies from which method(s), the value for the worst case path should start counting. Default is the value of symbolic.method i.e. the "symbolic target method".

• symbolic.worstcase.verbose=<true | false> This will generate verbose output e.g. analysis statistics.

• symbolic.worstcase.outputpath=<path> This will output the constraints for each worst case path and in addition summarize analysis statistics in a csv file. See above.

• symbolic.worstcase.req.maxinputsize=<Number> Plot the budget input size from the requirement

• symbolic.worstcase.req.maxres=<Number> Plot the budget max resource size from the requirement

• symbolic.worstcase.predictionmodel.size=<Number> The maximum plotted domain of the fitting functions.

• symbolic.worstcase.reusepolicy=<true | false> By setting this to true, a computed policy will be reused if it has been previously computed.

LICENSE

SPF-WCA is Copyright (c) 2017, The ISSTAC Authors and is released under the MIT License. See the LICENSE file in the root of this project and the headers of the individual files in the src/ folder for the details.

SPF-WCA uses benchmarks from the WISE project by Jacob Burnim, Sudeep Juvekar, Koushik Sen. The benchmarks are available here WISE-1.0.tar.gz.

WISE is Copyright (c) 2011, Regents of the University of California, and is released under an open-source BSD-style license. See the individual source files under src/examples/wise for details. A copy of the README file of WISE that includes license details can be found in the file licenses/README.WISE.

Benchmark code in src/examples/wise/*/*.java is based on the code obtained from the WISE project. It is Copyright (c) 2011, Regents of the University of California, and is released under an open-source BSD-style license.

We repeat here the license details from the README file (with file paths adjusted) in the WISE distribution from above:

The code in src/examples/wise/rbtree/ for red-black trees is by Tuomo Saarni, obtained from:

http://users.utu.fi/~tuiisa/Java/index.html

under the following license:

Here's some java sources I've made. Most codes are free to download. If you use some of my sources just remember give me the credits.

The code in src/examples/wise/java15/{util,lang}/ is originally from the Oracle Java (TM) 2 Platform Standard Edition Development Kit 5.0 Update 22, obtained and redistributed under the Java Research License v1.5 -- please see licenses/JavaResearchLicense.txt for details. Use and distribution of this technology is subject to the Java Research License included herein.

In addition, SPF-WCA relies on several other libraries:

• Google Guava: Copyright (c) The Guava Authors and is distributed under the Apache License, Version 2.0. The license for Google Guava can be found in the file licenses/COPYING.GUAVA.
• Apache Commons Math3, which is distributed under the Apache License, Version 2.0. A copy of the license can be found in the file licenses/LICENSE.COMMONS_MATH3. The NOTICE file of Apache Commons Math3 can be found in the file licenses/NOTICE.COMMONS_MATH3.
• Apache Commons CSV, which is distributed under the Apache License, Version 2.0. A copy of the license can be found in the file licenses/LICENSE.COMMONS_CSV. The NOTICE file of Apache Commons CSV can be found in the file licenses/NOTICE.COMMONS_CSV.
• Apache Commons CLI, which is distributed under the Apache License, Version 2.0. A copy of the license can be found in the file licenses/LICENSE.COMMONS_CLI. The NOTICE file of Apache Commons CLI can be found in the file licenses/NOTICE.COMMONS_CLI.
• Apache Commons Lang, which is distributed under the Apache License, Version 2.0. A copy of the license can be found in the file licenses/LICENSE.COMMONS_LANG. The NOTICE file of Apache Commons LANG can be found in the file licenses/NOTICE.COMMONS_LANG.
• JFreeChart, which is distributed under the GNU Lesser General Public License (LGPL) Version 2.1. A copy of the license can be found in the file licenses/LICENSE.JFREECHART.