Transaction id hashing and var_ints

[TierNolan]

The transaction serialization contains various var_int fields.  This means that you can change those and keep the actual data in the transaction the same.

For example, a transaction with 3 inputs could encode the 3 as "03" to encode 3 (single byte version) or "FD0300" (two byte version).  

Is it defined as the most compact version?  This isn't transaction malleability exactly.

The process when a tx is received is

- tx received
- convert to CTransaction object
- store 3 in the size of the vector
- calls UpdateHash
-- re-serializes transaction
-- compute hash
- store hash

This means that no matter how the transaction receives the 3, the 3 gets converted back into the shortest form before hashing for the txid.

Is this correct?

[harding]

It looks to me like CompactSize requires the smallest possible encoding.  See: https://github.com/bitcoin/bitcoin/blob/master/src/serialize.h#L218

It also looks like there are some tests to ensure non-canonical compatSize uints are forbidden in the test cases.  See: https://github.com/bitcoin/bitcoin/blob/master/src/test/serialize_tests.cpp#L81

[TierNolan]

It looks to me like CompactSize requires the smallest possible encoding.  See: https://github.com/bitcoin/bitcoin/blob/master/src/serialize.h#L218

Interesting, I thought it was valid but discouraged.