The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Conflicts

Reviewers, this pull request conflicts with the following ones:

• Pass mempool reference to chainstate constructor #19826 (Pass mempool reference to chainstate constructor by MarcoFalke)
• validation: UTXO snapshot activation #19806 (validation: UTXO snapshot activation by jamesob)
• [net processing] Move Misbehaving() to PeerManager #19791 ([net processing] Move Misbehaving() to PeerManager by jnewbery)
• p2p: don't add AlreadyHave transactions to recentRejects #19753 (p2p: don't add AlreadyHave transactions to recentRejects by troygiorshev)
• Avoid locking CTxMemPool::cs recursively in Mempool{Info}ToJSON() #19652 (Add thread safety annotations to Mempool{Info}ToJSON() by hebasto)
• Add thread safety annotations to CTxMemPool methods #19647 (Add thread safety annotations to CTxMemPool methods by hebasto)
• Allow updating mempool-txn with cheaper witnesses #19645 (Allow updating mempool-txn with cheaper witnesses by ariard)
• p2p: refactor AlreadyHave(), CInv::type, INV/TX processing #19610 (p2p: refactor AlreadyHave(), CInv::type, INV/TX processing by jonatack)
• ZMQ: Create "sequence" notifier, enabling client-side mempool tracking #19572 (ZMQ: Create "sequence" notifier, enabling client-side mempool tracking by instagibbs)
• refactor: Add ParseBool to rpc/util #19544 (refactor: Add ParseBool to rpc/util by fjahr)
• Tidy up ProcessOrphanTx #19498 (Tidy up ProcessOrphanTx by jnewbery)
• Refactor mempool.dat to be extensible, and store missing info #19488 (Refactor mempool.dat to be extensible, and store missing info by luke-jr)
• Remove CTxMempool::mapLinks data structure member #19478 (Remove CTxMempool::mapLinks data structure member by JeremyRubin)
• validation: re-delegate absurd fee checking from mempool to clients #19339 (validation: re-delegate absurd fee checking from mempool to clients by gzhao408)
• RPC: Return transaction fee from testmempoolaccept #19093 (RPC: Return transaction fee from testmempoolaccept by rajarshimaitra)
• Change UpdateForDescendants to use Epochs #18191 (Change UpdateForDescendants to use Epochs by JeremyRubin)
• txmempool: split epoch logic into class #18017 (txmempool: split epoch logic into class by ajtowns)
• Allow fee estimation to work with lower fees #13990 (Allow fee estimation to work with lower fees by ajtowns)
• RPC: Add ancestor{count,size,fees} to listunspent output #12677 (RPC: Add ancestor{count,size,fees} to listunspent output by luke-jr)

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.

From IRC:

<sipa> eh, i agree - that change isn't a step in the right direction

@sipa
Which direction is right?

From IRC:

<jeremyrubin> I think it's basically getting rid of a recursive mutex for code that's still designed to take a recursive mutex
<gwillen> it's better than a true recursive mutex because it's not possible to recurse by accident, you have to declare at call time which behavior you want (although better if you had to declare statically at compile time)
<sipa> it looks like that
<jeremyrubin> The correct refactor would be to make the code not do anything fancy with locks, or to just leave it
<jeremyrubin> gwillen: I think the chances of a bug or error in custom logic is higher than a recursive mutex
<jeremyrubin> accidental recursion seems unlikely...
<jeremyrubin> and accidental recursion would be a bug lock or no
<gwillen> (sorry I mean, accidental mutex recursion, that is, calling a function while holding a mutex, not expecting the callee to also lock it, resulting in the callee violating the caller's invariants)
<gwillen> (this is the fundamental problem of recursive mutexes)
<gwillen> (and I assume the motivation behind hebasto's refactor)

Correct, @gwillen :)

Rebased 1faf43a -> a887d73 (pr19306.10 -> pr19306.12) due to the merge conflicts.