Dashboard

This repository contains a CLI for comparing proof obligations in the context of formal verification. The dashboard lets you compare and filter proof obligations generated by formal methods tools.

Proof Obligation Format

A proof obligation is a JSON file with the following structure:

{
  "time": "The time it took to generate the proof obligation in seconds",
  "checks": [
    {
        "kind": "safe" | "warning" | "error",
        "title": "A category for the check, e.g. 'Invalid Memory Access'",
        "message": "A message describing the check",
        "range": {
            "start": {
                "line": "The line number where the check starts",
                "column": "The column number where the check starts",
                "file": "The file where the check starts"
            },
            "end": {
                "line": "The line number where the check ends",
                "column": "The column number where the check ends",
                "file": "The file where the check ends"
            }
        },
        "callstack": "An optional object containing a unique identifier for the current context; it can be any string that helps identify the check context",
    }
  ]
}

Check categories

The title field in the checks array can be one of the following categories:

• Assertion failure
• Invalid memory access
• Division by zero
• Signed integer overflow in arithmetic operator
• Unsigned integer overflow in arithmetic operator
• Signed integer overflow in explicit cast
• Unsigned integer overflow in explicit cast
• Signed integer overflow in implicit cast
• Unsigned integer overflow in implicit cast
• Invalid pointer comparison
• Invalid pointer subtraction
• Double free
• Negative array size
• Invalid floating point operation
• Stub condition
• Insufficient variadic arguments
• Insufficient format arguments
• Invalid type of format argument
• Floating-point division by zero
• Floating-point overflow
• Incorrect number of arguments
• Invalid shift

Any other value is treated as an invalid check.

Usage

The dashboard takes two required positional arguments for the proof obligations to compare. If one of them is stdin, the proof obligation is read as a single-line JSON value from standard input.

Argument	Description
--project <path>	Path to the project directory. Defaults to pwd.
--output <file>	File to write the JSON output to. Defaults to stdout, formatted for readability.
--analyze <file>	Filter checks to show only those related to the given file. Defaults to all files.
--exclude-not-found	Exclude files that are not found in the project directory from the output. Defaults to false.
--filter-error-category <category,...>	Filter conflicts by error category, as a comma-separated list. Defaults to all categories. See Check categories.

JSON Output

The JSON output format is still being finalized. See report.ml and conflict.ml for the current format.

Ranges

The error ranges are computed dynamically from the file contents. In the output, expect the ranges from the less precise tool.

Quickstart

To set up the dashboard, run opam install . --deps-only in the repository root, then build the project with dune build.

Run the examples on Mopsa and Goblint

Installing Mopsa

Follow the instructions on the installation page.

Installing Goblint

Dashboard output is not yet incorporated into the main Goblint repository. To use Goblint with the dashboard, clone this fork:

git clone -b checks https://github.com/Robotechnic/analyzer.git

Then choose one of the following options:

• Follow the installation instructions in the "Installing" section of the README. In this case, Goblint will be installed in a separate switch, which may be less convenient for development.
• Install Goblint in the current opam switch by running the following commands in the root of the cloned repository:

make deps
make install

Running the examples

mopsa-c -output mopsa.json -format=json -show-safe-checks examples/simples/overflow.c
goblint --outfile goblint.json --result dashboard examples/simples/overflow.c

Finally, run the dashboard to compare the two proof obligations: dune exec dashboard -- --exclude-not-found true mopsa.json goblint.json

Get the paper example

To get the example used in the paper, run screenshot_script/allexamples.sh. This creates a results directory containing the mopsa.json and goblint.json files needed for the GUI. To use the example in the GUI, compress examples/allErrorTest and upload it together with the two JSON files (mopsa.json and goblint.json).

If you only want to see the console output, run the dashboard with the following command:

dashboard -- --exclude-not-found true ./results/mopsa.json ./results/goblint.json

Docker Image

To build the Docker image, run the following command in the repository root:

docker build -t dashboard .

You can then start an environment with the dashboard CLI using:

docker run -it --rm -v $(pwd):/dashboard/app dashboard bash

This mounts the current directory so it is available inside the Docker container.

You can also run the dashboard CLI directly with:

docker run --rm -v $(pwd):/dashboard/app dashboard dune exec dashboard -- [args]

Replace [args] with the appropriate arguments.