Pointer parsing accepts non-canonical varuint encodings and normalizes on re-encode

[LorenzoRD2003]

Summary

pallas-addresses currently accepts non-canonical varuint encodings inside Pointer::parse, so two different byte sequences can parse to the same Pointer value.

That means the following converse property does not hold today:

• if Pointer::parse(bytes) == Ok(p), then p.to_vec() == bytes

The forward roundtrip still holds:

• Pointer::parse(p.to_vec()) == Ok(p)

Minimal repro

The varuint encoding [0x80, 0x00] is accepted as 0, even though the canonical encoding is [0x00].

So this parses successfully:

let bytes = [0x80, 0x00, 0x80, 0x00, 0x80, 0x00];
let parsed = Pointer::parse(&bytes).unwrap();

assert_eq!(parsed, Pointer::new(0, 0, 0));
assert_eq!(parsed.to_vec(), vec![0x00, 0x00, 0x00]);
assert_ne!(parsed.to_vec(), bytes);

Impact

Depending on how consumers use parsed values, this can affect:

• strict validation
• byte-level identity checks
• roundtrip assumptions
• normalization-sensitive caches or allowlists