This depends on #12934 and #13407: review those first
As a follow up to #12934 which introduces some armature code to make message passing based communication a bit easier and moves ProcessNewBlock into its own thread, this does similar for AcceptToMemoryPool() in p2p
If this approach seems reasonable, I think logical next steps are (in forthcoming prs) to try to reduce dependency on cs_main in p2p so that the benefits of this parallelism can be better realized
If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first.
Create separate functions to handle responses when a transaction
received from a peer is accepted or rejected.
This commit introduces one behavior change:
Currently, if -promiscuousmempool is set (only allowed on regtest
and testnet), it is possible for AcceptToMemoryPool() to return
true, but for the CValidationState() to contain DoS points if the
transaction failed validation with policy standard script flags
but then passed with promiscuous flags. This will cause you to ban
the peer who sent the transaction. This commit fixes this behavior
by reseting the validation state in that one situation.
61+ if (!m_threads_observed.count(id)) {
62+ m_threads_observed.insert(std::this_thread::get_id());
63+
64+ // resubmit it so that it gets a chance to get to the right thread
65+ // when resubmitting, do not block and do not care about failures
66+ // theres a potential deadlock where we try to push this to a queue thats
codespell: theres (should be “there is”)
123+ {
124+ static_assert(m_consumer_mode == WorkerMode::BLOCKING, "");
125+
126+ T ret;
127+
128+ // use a temporary so theres no side effecting code inside an assert which could be disabled
