test: Add fuzzing harnesses for various classes/functions in primitives/ #18744

 112 | @@ -111,4 +113,24 @@ void test_one_input(const std::vector<uint8_t>& buffer)
 113 |          static const uint256 u256_max(uint256S("ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"));
 114 |          TxToUniv(tx, u256_max, u);
 115 |      }
 116 | +
 117 | +    {
 118 | +        FuzzedDataProvider fuzzed_data_provider(buffer.data(), buffer.size());
 119 | +        const CScript script = ConsumeScript(fuzzed_data_provider);

MarcoFalke commented at 11:50 AM on April 23, 2020:

This code will never be executed when the buffer does not start with a valid transaction

practicalswift commented at 9:29 AM on April 24, 2020:

Oh, good catch! Thanks a lot for noticing! I missed the return:s above (L37/42/45/61). Now fixed. Please re-review updated version :)

  46 | +        if (mutable_tx_1 && mutable_tx_2) {
  47 | +            const CTransaction tx_1{*mutable_tx_1};
  48 | +            const CTransaction tx_2{*mutable_tx_2};
  49 | +            assert((tx_1 == tx_2) != (tx_1 != tx_2));
  50 | +        }
  51 | +    }

MarcoFalke commented at 11:32 AM on April 24, 2020:

It still seems odd. How do you know that the fuzzer won't favor inputs that start with a valid transaction and is more likely to discard inputs that do not?

practicalswift commented at 12:17 PM on April 24, 2020:

I see your point: yes, it could confuse fuzzers. To rule out that possibility I've now moved this to a new fuzzing harness file :)