Skip to content
APALACHE: symbolic model checker for TLA+
Scala TLA Python Shell Java SMT
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
3rdparty
bin
docs using Apache commons-configuration2, which allows variable substitution Mar 30, 2019
mod-distribution version change to 0.5.1-SNAPSHOT Jun 25, 2019
mod-infra version change to 0.5.1-SNAPSHOT Jun 25, 2019
mod-tool version change to 0.5.1-SNAPSHOT Jun 25, 2019
profiling fixed the message Jul 5, 2019
script
src/assembly dependency problems resolved plus better packaging Aug 7, 2018
test
tla-assignments
tla-bmcmt
tla-import version change to 0.5.1-SNAPSHOT Jun 25, 2019
tlair
.gitignore uncovered the docs that were prevously hidden by accident Feb 9, 2019
.travis.yml
CHANGES.md changes for the release 0.5.0-RELEASE May 14, 2019
Dockerfile added mvn clean before mvn package Mar 2, 2019
INSTALL.md
LICENSE
README.md
TODOs_and_misc.txt comments by Jure Dec 22, 2017
apalache.ipr
apalache.iws
pom.xml version change to 0.5.1-SNAPSHOT Jun 25, 2019

README.md

APALACHE project

A symbolic model checker for TLA+

master: Build Status     unstable: Build Status

To get an intuition about how APALACHE is working, check our paper at OOPSLA19.

The current version of the tool is neither parameterized, nor it is using abstractions. As a first step, we are working on a symbolic bounded model checker that runs under the same assumptions as TLC. To see the list of supported TLA+ constructs, check the supported features page. Our tool runs simple type inference, in order to encode TLA+ expressions in Microsoft Z3. Check the note on type inference and annotations to get an idea of the supported TLA+ expressions and hints provided by the user.

Related reports and publications can be found at the apalache page.

Releases

Check the releases page. As the tool is in the early development stage, there are a few releases.

To try the latest features, you can download the latest unstable build.

Preparing the spec

Similar to TLC, before running the tool, you have to specify the input parameters. Since we do not have integration with the TLA+ Toolbox yet, we require all constants to be initialized in a spec. There are two ways to do that.

Option 1: Replace constants with operators

For instance, assume your spec contains a constant N for the number of processes and a constant Values for the set of possible values:

CONSTANT N, Values

Replace them with the operators that define concrete values:

\* CONSTANT N, Values
N == 4
Values == {0, 1}

Option 2: Initialize constants with ConstInit

This approach is similar to the Init operator, but applied to the constants. We define a special operator, e.g., called ConstInit:

CONSTANT N, Values

ConstInit ==
  /\ N \in {4}
  /\ Values \in {{0, 1}}

Note that in the current version, an assignments to a constants c should be written as c \in S, where S is a set (you can use SUBSET S and [S -> T as well]). Moreover, ConstInit should be a conjunction of such assignments and possibly of additional constraints on the constants.

As a bonus of this approach, you can have parameterized though bounded constraints on the constants. For example:

CONSTANT N, Values

ConstInit ==
  /\ N \in 3..10
  /\ Values \in SUBSET 0..4
  /\ Values /= {}

The model checker will try the instances for all the combinations of the parameters specified in ConstInit, that is, in our example, it will consider N \in 3..10 and all non-empty value sets that are subsets of 0..4.

Type annotations

Additionally, whenever you are using an empty set {}, an empty sequence <<>>, or a set of records with different sets of fields, you have to annotate these expressions with types, see type inference and annotations.

Running the binaries

First, make sure that you have installed Oracle JRE 8 (or JDK8) and Microsoft Z3 4.7.1.

The model checker can be run as follows:

$ bin/apalache-mc check [--init=Init] [--cinit=ConstInit] [--next=Next] [--inv=Inv] [--length=10] [--tuning=filename] myspec.tla

The arguments are as follows:

  • --init specifies the initialization predicate, the default name is Init
  • --next specifies the transition predicate, the default name is Next
  • --cinit specifies the constant initialization predicate, optional
  • --inv specifies the invariant to check, optional
  • --length specifies the upper bound on the length of the finite executions to explore
  • --tuning specifies the properties file that stores the options for fine tuning

If you like to check an inductive invariant Inv, you can do it as follows:

$ bin/apalache-mc check --init=Inv --inv=Inv --length=1 myspec.tla

Make sure that Inv contains necessary constraints on the shape of the variables. Usually, TypeOK as the first line of the invariant is exactly what is needed.

The tool will display only the important messages. A detailed log can be found in detailed.log.

Building from sources

To build the tool, you will need the following standard packages: Java 8 SDK, Scala, and Maven. You will also need Z3 Prover 4.7.1 and TLA tools. Both Z3 and TLA+ tools can be automatically installed in your local Maven repository by running the script ./3rdparty/install-local.sh. WARNING: Z3 API has been changed in version 4.8.0, so make sure that you install version 4.7.1.

To build the complete package, including the dependencies, type:

$ mvn package
You can’t perform that action at this time.