ChiselBench is a suite of 10 programs (and counting) from GNU packages including GNU Core Utilities. These programs were chosen since they are open-source and widely used. More interestingly, their manually-reduced implementations are available in BusyBox, a lightweight UNIX utility package for embedded Linux.

The main objective of designing ChiselBench is to pave the path for researchers to explore program debloating.

Program Characteristics

Program	LOC	#Functions	#Statements
bzip-1.05	18688	108	6532
chown-8.2	69894	757	24792
date-8.21	75898	878	26147
grep-2.19	49011	432	12138
gzip-1.2.4	13223	93	4118
mkdir-5.2.1	28202	263	10679
rm-8.4	89694	764	27695
sort-8.16	71315	753	24890
tar-1.14	45778	502	13227
uniq-8.16	64915	665	22086

Test Specifications

Test specifications are written as bash scripts that specify the sets of desired and undesired program functionalities. Each time the script is executed, all the desired and undesired functionalities are run with the following Clang Sanitizers.

• Control Flow Integrity: detector of specific forms of undefined behavior that might lead to hijacking the program’s control flow
• Address Sanitizer: fast memory error detector
• Memory Sanitizer: uninitialized reads detector
• Undefined Behavior Sanitizer: detector of various kinds of undefined behavior such as misaligned/null pointer and integer overflow
• Leak Sanitizer: memory-leak detector

The list of sanitizers that are used by test scripts can be modified in the test-base.

Besides sanitizers, corner-case environment conditions including memory exhaustion and stream errors are simulated via statically-linked libraries. The list of these libraries can be modified in the test-base.

ChiselBench Suite Structure

ChiselBench contains three main elements for each program:

• A *.origin.c which is the original program (before reduction).
• A test script (test.sh) which specifies the desired and undesired features as well as all the Clang Sanitizers and corner-case environment simulators against which the user might want to test the program.
• A Makefile for automating the process of reduction.