Suggestion: comparative fuzzing

[nlyan]

This is just an idea.

I've had some success in the past setting up an AFL job which runs 2 different parsers in tandem and deliberately fails when the results differ. This can be useful in finding out if they interpret the official spec differently.

[fitzgen]

I've also had this in the back of my mind, would definitely be a cool thing to do after passing all the libiberty test suite!