A tool allowing the quantification of software properties through Model Counting: We use CBMC to transform C-code into SAT instances and then run (approximate) model counting (e.g. ApproxMC) on those instances.
This work was publised by Samuel Teuber and Alexander Weigl at QEST 2021 (BibTeX).
The experimental setup and results can be found in this repository or this evaluated artifact.
- Tested with Python 3.8 and CBMC 5.11
- Expects CBMC and cpp (C preprocessing) to be in PATH
- Expects
requirements.txtto be installed through pip (specifically we are using pycparser)
Run counterSharp as:
python -m counterSharp -amm <assume miss dimacs> --amh <assume hit dimacs> --asm <assert miss dimacs> --ash <assert hit dimacs> <inputfiles...>
counterSharp will read and transform the input files and use CBMC to produce 4 dimacs files on which model counting can be applied by tools like ApproxMC.
Frequently helpful:
| Parameter | Description |
|---|---|
-h, --help |
Show help message and exit |
-d |
Debug Mode (more output) |
--function <fun> |
Name of the function to be analyze |
--cbmcArg <args> |
Additional input paramters for CBMC |
--unwindDepth <depth> |
The unwind depth of CBMC (otherwise found through iteration over depths) |
The default for --function is main. Splicing must explictly be activated for CBMC.
Less frequently helpful:
| Parameter | Description |
|---|---|
--assertMissVar <name> |
Name of the assert miss state variable |
--assumeMissVar <name> |
Name of the assume miss state variable |
--assertFunction <name> |
Name of the assert function |
--assumeFunction <name> |
Name of the assume function |
--returnLabel <name> |
Name of the return label |
All these are set to default values which are unlikely to clash with your code unless you use variables/functions/labels which begin with __counterSharp_
counterSharp counts (thus sharp/#) counterexamples (thus counter) for a given specification.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent