solidity/test/tools/ossfuzz
2019-03-06 11:35:22 +01:00
..
config Add fuzzer config files for oss-fuzz and a solidity fuzzing dictionary. 2019-02-04 15:58:41 +01:00
CMakeLists.txt LibProtobuf Mutator (LPM) based fuzzer for yul subset 2019-03-05 12:32:55 +01:00
const_opt_ossfuzz.cpp Reduce input file size for solc and constant optimizer fuzzers. 2019-03-06 11:35:22 +01:00
protoToYul.cpp LibProtobuf Mutator (LPM) based fuzzer for yul subset 2019-03-05 12:32:55 +01:00
protoToYul.h LibProtobuf Mutator (LPM) based fuzzer for yul subset 2019-03-05 12:32:55 +01:00
README.md Add fuzzer config files for oss-fuzz and a solidity fuzzing dictionary. 2019-02-04 15:58:41 +01:00
solc_noopt_ossfuzz.cpp Reduce input file size for solc and constant optimizer fuzzers. 2019-03-06 11:35:22 +01:00
solc_opt_ossfuzz.cpp Reduce input file size for solc and constant optimizer fuzzers. 2019-03-06 11:35:22 +01:00
strictasm_diff_ossfuzz.cpp Create a differential fuzzer using tracing facility from the yul interpreter. 2019-03-04 20:44:17 +01:00
yulFuzzerCommon.cpp Create a differential fuzzer using tracing facility from the yul interpreter. 2019-03-04 20:44:17 +01:00
yulFuzzerCommon.h Create a differential fuzzer using tracing facility from the yul interpreter. 2019-03-04 20:44:17 +01:00
yulProto_diff_ossfuzz.cpp ossfuzz: Reduce test input size to less than equal to 600 bytes. 2019-03-05 16:55:55 +01:00
yulProto.proto LibProtobuf Mutator (LPM) based fuzzer for yul subset 2019-03-05 12:32:55 +01:00
yulProtoFuzzer.cpp ossfuzz: Reduce test input size to less than equal to 600 bytes. 2019-03-05 16:55:55 +01:00

Intro

oss-fuzz is Google's fuzzing infrastructure that performs continuous fuzzing. What this means is that, each and every upstream commit is automatically fetched by the infrastructure and fuzzed.

What does this directory contain?

To help oss-fuzz do this, we (as project maintainers) need to provide the following:

  • test harnesses: C/C++ tests that define the LLVMFuzzerTestOneInput API. This determines what is to be fuzz tested.
  • build infrastructure: (c)make targets per fuzzing binary. Fuzzing requires coverage and memory instrumentation of the code to be fuzzed.
  • configuration files: These are files with the .options extension that are parsed by oss-fuzz. The only option that we use currently is the dictionary option that asks the fuzzing engines behind oss-fuzz to use the specified dictionary. The specified dictionary happens to be solidity.dict.

solidity.dict contains Solidity-specific syntactical tokens that are more likely to guide the fuzzer towards generating parseable and varied Solidity input.

To be consistent and aid better evaluation of the utility of the fuzzing dictionary, we stick to the following rules-of-thumb:

  • Full tokens such as block.number are preceded and followed by a whitespace
  • Incomplete tokens including function calls such as msg.sender.send() are abbreviated .send( to provide some leeway to the fuzzer to sythesize variants such as address(this).send()
  • Language keywords are suffixed by a whitespace with the exception of those that end a line of code such as break; and continue;

What is libFuzzingEngine.a?

libFuzzingEngine.a is an oss-fuzz-related dependency. It is present in the Dockerized environment in which Solidity's oss-fuzz code will be built.

Is this directory relevant for routine Solidity CI builds?

No. This is the reason why the add_subdirectory(ossfuzz) cmake directive is nested under the if (OSSFUZZ) predicate. OSSFUZZ is a solidity-wide cmake option that is invoked by the ossfuzz solidity-builder-bot in order to compile solidity fuzzer binaries.