Explicitly bind lifetimes for NlaBuffer.value()

[vthriller]

Without this, the following results in an error:

fn subnlas_by_kind<'a>(nlas: &'a [u8]) -> HashMap<u16, Vec<&'a [u8]>> {
	let mut out = HashMap::new();
	for n in NlasIterator::new(nlas) {
		let n = n.unwrap();
		let entry: &mut Vec<_> = out.entry(n.kind()).or_default();
		(*entry).push(n.value())
	}
	out
}

error[E0515]: cannot return value referencing local variable `n`
    |
212 |         (*entry).push(n.value())
    |                       - `n` is borrowed here
213 |     }
214 |     out
    |     ^^^ returns a value referencing data owned by the current function

[cathay4t]

Please include unit test case proving your fix.

[vthriller]

There you go, also simplified example code a little bit.

[vthriller]

(and one more force-push to fix typo, sorry about that)

[cathay4t]

I was expecting something like:

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_foo() {

    }
}

[vthriller]

Sure, I can craft some inputs-outputs, but the whole point is that behavior doesn't change at all, it's all about lifetimes. OTOH there are no NlasIterator tests at all, so there's no redundant work to be done, might as well hit two birds with one stone.

[vthriller]

I don't like how this test looks, but at least there's no #[allow(dead_code)] anymore.

[vthriller]

I think I'm done for now.

While I don't like iter.next().unwrap().unwrap().value(), derive(PartialEq) for DecodeError would not make it easier to read this test in the future, so I stuck with that.