Add sanity check to ensure that all data decoded by parity_scale_codec are consumed

[appetrosyan]

I don't think this is an issue at all

Marin, [20.04.22 12:54]
the byte representations match for VersionedQueryResult and paginated version

Marin, [20.04.22 12:54]
to a point

Marin, [20.04.22 12:54]
if it was the other way around, it would fail. If you were encoding QueryResult and decoding PaginatedQueryResult

Marin, [20.04.22 12:55]
if there was something after if in the encoded form, it would fail as well

Aleksandr Petrosyan, [20.04.22 12:55]
[Forwarded from Aleksandr Petrosyan]
Thankfully it doesn't decode QueryResult, but the fact that it happily works with types containing VersionedQueryResult is a little worrying

Marin, [20.04.22 12:55]
but seems that scale happily disregards trailing bytes if it finished decoding an object

Aleksandr Petrosyan, [20.04.22 12:56]
[In reply to Marin]
It works more like a parser than a binary serialisation format.

Marin, [20.04.22 12:56]
we can inject a check I guess

Marin, [20.04.22 12:56]
to see if all bytes were consumed

Marin, [20.04.22 12:57]
check that parity_scale_codec::Input is depleted

Aleksandr Petrosyan, [20.04.22 12:58]
Letting unrecognised data into a blockchain is a ticking time-bomb.

[mversic]

I'll give an example of what's happening here. Let's say we have two structures:

struct Struct1 {
    a: u32,
    b: u16,
}
struct Struct2 {
    a: u32,
}

where Struct2 is a subtype of the Struct1 in the encoded format meaning that all valid encodings of Struct1 will successfully decode into Struct2 and there will always be 2 trailing bytes left at the end of the encoded array. I'm not sure this is a security issue or how could this be exploited, or if it's a feature or a bug.

If this is not a feature that we need/want then we have to check that the parity_scale_codec::Input is depleted after decoding. This can be done by wrapping the Decode::decode<I: Input>(input: &mut I) method with a wrapper and make sure that all the decode calls go through the wrapper

[Arjentix]

I think we can do it by introducing our own macro for (de-)coding and use it in derives