Trait gimli::read::cfi::UnwindContextStorage

source ·

pub trait UnwindContextStorage<T: ReaderOffset>: Sized {
    type Rules: ArrayLike<Item = (Register, RegisterRule<T>)>;
    type Stack: ArrayLike<Item = UnwindTableRow<T, Self>>;
}

Expand description

Specification of what storage should be used for UnwindContext.

Normally you would only need to use StoreOnHeap, which places the stack on the heap using Box. This is the default storage type parameter for UnwindContext.

You may want to supply your own storage type for one of the following reasons:

In rare cases you may run into failed unwinds due to the fixed stack size used by StoreOnHeap, so you may want to try a larger Box. If denial of service is not a concern, then you could also try a Vec-based stack which can grow as needed.
You may want to avoid heap allocations entirely. You can use a fixed-size stack with in-line arrays, which will place the entire storage in-line into UnwindContext.

Here’s an implementation which uses a fixed-size stack and allocates everything in-line, which will cause UnwindContext to be large:

struct StoreOnStack;

impl<T: ReaderOffset> UnwindContextStorage<T> for StoreOnStack {
    type Rules = [(Register, RegisterRule<T>); 192];
    type Stack = [UnwindTableRow<T, Self>; 4];
}

let mut ctx = UnwindContext::<_, StoreOnStack>::new_in();

// Initialize the context by evaluating the CIE's initial instruction program,
// and generate the unwind table.
let mut table = some_fde.rows(&eh_frame, &bases, &mut ctx)?;
while let Some(row) = table.next_row()? {
    // Do stuff with each row...
}