Avoid nullptr dereference when constructing Blockchain and tx_memory_pool

[DarkWingMcQuack]

Until now Blockchain and tx_memory_pool is constructed like the following:

std::unique_ptr<Blockchain> core_storage;
tx_memory_pool m_mempool(*core_storage);
core_storage.reset(new Blockchain(m_mempool));

This contains a nullptr dereference.

This PR fixes this by introducing a new struct BlockchainAndPool

struct BlockchainAndPool {
  cryptonote::Blockchain blockchain;
  cryptonote::tx_memory_pool tx_pool;

  BlockchainAndPool() : blockchain(tx_pool), tx_pool(blockchain) {}
};

which is used to construct Blockchain and tx_memory_pool to avoid the nullptr dereferencation.

[UkoeHB]

It seems strange to initialize an object with an uninitialized other object. Would this pattern be more intuitive?

struct BlockchainAndPool {
  cryptonote::Blockchain blockchain;
  cryptonote::tx_memory_pool tx_pool;

  BlockchainAndPool() : blockchain(), tx_pool(blockchain) {
    blockchain.set_tx_pool(tx_pool);
  }
};

[DarkWingMcQuack]

I was following how cryptonode::core does it here:
https://github.com/monero-project/monero/blob/528cd729687fcc73e2c082225af203d5f67b3eb6/src/cryptonote_core/cryptonote_core.cpp#L228

@moneromooo-monero has proposed this pattern on IRC/Matrix

Maybe a problem with using a setter would be that if the setter is not called the object would be in an invalid state.
Another reason is that Blockchain and tx_memory_pool hold a reference to each other and those references need to be initialized in the constructor.

[UkoeHB]

I see. The original design is dubious but I guess yours is the best solution without more intrusive changes.

[selsta]

Please undo the unrelated indentation changes :)

[selsta]

Please squash.

[DarkWingMcQuack]

done

[UkoeHB]

This PR does not fix all cases of the problematic pattern.

• tests/core_tests/chaingen.cpp init_blockchain() (nice dangling reference there...)
• tests/block_weight/block_weight.cpp PREFIX_WINDOW()
• tests/unit_tests/output_distribution.cpp get_output_distribution()

[mj-xmr]

Thanks for handling this. It was bothersome.

[mj-xmr]

As @UkoeHB noted, in case the PR is supposed to be merged as it is now, let's not forget to address later:

• tests/core_tests/chaingen.cpp init_blockchain()
• tests/block_weight/block_weight.cpp PREFIX_WINDOW()
• tests/unit_tests/output_distribution.cpp get_output_distribution()

[selsta]

Continuing in #8924