DbgVis: The Debugger Visualizer

Introduction

DbgVis is a library that helps in visualizing data from within a debugger. Most debuggers are very powerful and allow developers to see every single bit of code and data associated to it, but, unfortunately, nothing more than the bits. However, there are times in debugging where visualizing the data could be extremely useful, let it be a graph or a image.

The purpose of DbgVis is to allow debuggers to display data in a more useful format to the developer than as a string of bytes.

DbgVis' design aims to be extensible in 3 aspects:

• Data parsing
• Visualization backend
• Debugger backend

Data parsing

DbgVis allows new data types support to be added.

Currently supported types are:

• OpenCV's Mat

Planned types:

• OpenCV's UMat (OpenCL)
• OpenCV's GpuMat (CUDA)

Visualization backend

Currently supported visualizers are:

• OpenCV's built-in GUI (imshow)

Planned visualizers:

• Matplotlib
• Pillow

Debugger backend

Currently supported debuggers are:

• GDB

Planned debuggers:

• LLDB

Requirements

• GDB with Python 3 support
• Python 3
• OpenCV 3.0+ (compiled with Python support)
• Numpy

Usage

Clone DbgVis repository:

git clone https://github.com/pentschev/DbgVis.git

Add DbgVis repository to Python module search path:

export PYTHONPATH=$(pwd)/DbgVis:${PYTHONPATH}

Now GDB is ready to display images. From the GDB terminal you can now import the DbgVis interface for GDB:

source /path/to/DbgVis/gdb.py

There are currently two commands available:

• show_cv_mat
• show_cv_mat_ptr

The first one can be used to display a cv::Mat from a value (e.g., a local scope variable), whereas the second allows displaying a cv::Mat from a pointer (or a memory address) from the inferior (inferior is the GDB terminology for a processing running within GDB or attached to it).

Both show_cv_mat and show_cv_mat_ptr take exactly one argument: the value and memory address to a cv::Mat, respectively.

WARNING: API is experimental and subject to change.