DeVIL

Debugger Validation via Cross-Level Differential Debugging

Table of Contents

• Background
• Install
• Usage
• Result

Background

DeVIL employs Cross-Level Differential Debugging (CLDD) to validate debugger toolchains. CLDD involves comparing traces of dynamic program states at different levels of debugging strategies, specifically source-level debugging and instruction-level debugging.

Install

This project uses gcc, gdb, llvm and lldb. Go check them out if you don't have them locally installed:

apt install python3 gcc gdb llvm lldb

Usage

C Source File

• For a given c source program test.c, the command used for DeVIL over the GDB toolchain is:

  python3 main.py -s test.c --compiler='gcc' --debugger='gdb'

• For a given c source program test.c, the command used for DeVIL over the LLDB toolchain is:

  python3 main.py -s test.c --compiler='clang' --debugger='lldb'

C Source Folder

• For a given folder FilesDir with multiple C source files, the command used for DeVIL over the GDB toolchain is:

  python3 main.py -s FilesDir --compiler='gcc' --debugger='gdb'

• For a given folder FilesDir with multiple C source files, the command used for DeVIL over the LLDB toolchain is:

  python3 main.py -s FilesDir --compiler='clang' --debugger='lldb'

• Note that, we recommend to use the parallel option when using our tool if your computer has multi-cores:

  $ python3 main.py --parallel -s FilesDir --compiler='gcc' --debugger='gdb'
  $ python3 main.py --parallel -s FilesDir --compiler='clang' --debugger='lldb'

• Let's take the gcc 12.1.0 testsuite for example

  $ wget https://ftp.gnu.org/gnu/gcc/gcc-12.1.0/gcc-12.1.0.tar.gz
  $ tar -zxvf gcc-12.1.10.tar.gz
  $ python3 main.py -s gcc-12.1.1/gcc/testsuite --parallel --compiler='gcc' --debugger='gdb'
  $ python3 main.py -s gcc-12.1.1/gcc/testsuite --parallel --compiler='clang' --debugger='lldb'

Result

The results will be save into the Expr folder

• step folder saves the results for cross-level differential debugging
• optimization folder saves the results for differential optimizations (SOTA)