Question: fuzzing testing integration

[tyler92]

Hi,

I recently created a PR for Boost.Beast fuzzing in the Google OSS-Fuzz repository. This PR includes three fuzzing targets: the HTTP request parser, the HTTP response parser, and the WebSocket server. I find these targets quite useful, e.g. they found #2881 and #2861 (reported not by me) using AddressSanitizer.

As far as I understood, it’s better to have fuzzing targets in the upstream repo rather than just in OSS-Fuzz. It keeps everything in one place and makes it easier to keep the fuzzing targets up to date with any changes.

Would you be open to integrating these fuzzing targets into the Beast repo? I’m happy to help with the initial setup. This would mean:

• Adding the *.cc files for the fuzzing targets
• Including a "corpus" of sample input data for the fuzzers
• Optionally setting up an automatic CI job to run the fuzzers on each PR

Let me know what you think.

[ashtum]

It would be great if we could add Boost.Beast to OSS-Fuzz. As you mentioned, #2881 and #2861 showed there is always a chance of finding new bugs. You can use Boost.URL as a reference for where to add the required files: https://github.com/boostorg/url/tree/develop/test/fuzz

Optionally setting up an automatic CI job to run the fuzzers on each PR

Currently, we have an overloaded GitHub Action CI. However, I believe we can add that to Drone. I'll take care of this part.

[tyler92]

Great! I will open a PR and I hope we will do it step by step.

[ashtum]

Addressed in #2888 and #2891.