Introduce LLVMFuzzerInitialize support #128
Conversation
- Added `example_init` to demonstrate the use of initialization code with the `fuzz_target!` macro. - Updated `fuzz_target!` macro to support an `init` parameter for executing initialization code before fuzzing. - Updated CI script to build and run the new example.
| @@ -198,9 +185,31 @@ pub fn initialize(_argc: *const isize, _argv: *const *const *const u8) -> isize | |||
| /// `"arbitrary-derive"` cargo feature. | |||
| #[macro_export] | |||
| macro_rules! fuzz_target { | |||
| (|$bytes:ident| $body:expr) => { | |||
| (init: $init:expr, |$bytes:ident| $body:expr) => { | |||
There was a problem hiding this comment.
Using $init:expr matches the existing behavior for $body but I'm not sure if that's desirable. It permits all sorts of expressions, including those that return values, which may be misleading. This is code that's executed for side-effects, so $init:stmt seems more appropriate. However, non-block statements look strange to me in this position (fuzz_target!(init: let x = 0, |data| {}) without semicolon would be valid but meaningless). So maybe enforce that it's a block with $init:block and expand to let _: () = $init; so that it's required to not return anything meaningful?
There was a problem hiding this comment.
I'll wait and see what the maintainers have to say, but I get your point, still I don't have a preference.
example_initto demonstrate the use of initialization code with thefuzz_target!macro.fuzz_target!macro to support aninitparameter for executing initialization code before fuzzing.Should I update any doc with init usage example?
Is the example good enough? Maybe an example using
staticwould be more meaningful.Close after merge:
fuzz_target!code #110