TODO list for launch

[j-berman]

Wallet features

• Restore
• Sync
• Transfers
• Sweeps
• Support >8 input txs
  • FCMP++: support up to 128 inputs per tx #57
• Payments to integrated address
  • carrot+fcmp: send to integrated address and fix transfer tests #61
• scan_tx
  • fcmp++: scan_tx & RPC for path by global output id #49
  • fcmp++: Consolidate paths in RPC to get path by output ID #89
• Background sync
• Hot/cold wallet integration
  • Carrot tx proposal overhaul, wallet2_basic lib, sweep fixes, and preliminary hot/cold types  #51
  • wallet: complete hot-cold implementation for Carrot/FCMP++ #52
• HW wallet interface changes
  • Guide: How to implement HW devices for Carrot/FCMP++ [WIP] #63
• Tx proofs
• Multisig
  • Get multisig working pre-FCMP++/Carrot fork #291
• 3rd party HW wallet integrations
  • Helpful link: Guide: How to implement HW devices for Carrot/FCMP++ [WIP] #63
  • Trezor
  • Ledger
• Clean up CLI references to rings in UI
• Clean up GUI references to rings
• Wallet RPC describe_transfer command for Carrot transaction proposals
  • Resolved by: wallet: complete hot-cold implementation for Carrot/FCMP++ #52
• Don't remove existing tx dests when opening an old wallet with the new software
  • [untested] wallet: persist dest addrs when updating to FCMP++ wallet #106
• Don't need simple_wallet::prompt_if_old after fork
  • Don't need Transaction spends more than one very old outputs. Privacy would be better if they were sent separately. after FCMP++.
• Remove block hashes download on restore
  • FCMP++: get rid of init hash download on restore + fix deep reorg handling + more #81
• Speed up prove()
  • Non-linear prove times when n inputs >16 kayabaNerve/fcmp-plus-plus#34
  • Integrate faster prove() into fcmp++-stage
  • Solved by Move to monero-oxide/monero-oxide #93
• When the CLI identifies a receives and requires password and/or user inputs the wrong password, make block popping instant
  • See this comment
  • fcmp++: fast pop blocks in wallet + fix tree cache ref counter + better logging #96

Daemon features

• Build FCMP++ tree
• DB migration for FCMP++ tree
• Validate FCMP++ txs
• Validate Carrot txs
• Include init sync data needed to restore a wallet from arbitrary restore height
• Include FCMP++ tree root and n tree layers in PoW hash
• Prohibit torsioned outputs at consensus
• Endpoint to get path by output id
  • fcmp++: scan_tx & RPC for path by global output id #49
  • fcmp++: Consolidate paths in RPC to get path by output ID #89
• Hardened tx weight calculations (Revisit FCMP++ transaction weight and scaling parameters #44)
  • fcmp++: use weight approx equal to byte size #232
  • Revisit FCMP++ transaction weight limit since a 128-in tx can be bigger than 1/2 the penalty free block size
• Dynamic block weight changes
  • Proposed: https://github.com/ArticMine/Monero-Documents/blob/master/MoneroScaling2025-07.pdf
• PoW-enabled relay (PoWER)
  • PoWER #230
  • Proof-of-Work-Enabled Relay ("PoWER") monero-project/research-lab#133
• RandomX commitments w/double hashing
  • cryptonote_core: support RandomX V2 and commitments in v17 monero-project/monero#10038
• monerod revalidates the whole tx-pool on HFs monero-project/monero#10142
  • Resolved by cache input verification results

Known bugs

• [BUG] error: m_blockchain <> m_tree_cache top block idx mismatch #45
  • FCMP++: get rid of init hash download on restore + fix deep reorg handling + more #81 may help
• [BUG] node failed to verify FCMP++ proof #46
  • FCMP++: get rid of init hash download on restore + fix deep reorg handling + more #81 may help
• @nahuhh reported a segfault in wallet, broken wallet state
  • Should be fixed by Destroy Rust FFI types correctly #39 and fcmp++: fix path_for_proof when a non-root layer has single elem #59
• Sweep when some txs don't cover fees.
  • Mitigated by Carrot tx proposal overhaul, wallet2_basic lib, sweep fixes, and preliminary hot/cold types  #51.
  • Should be fixed by carrot_impl: sweep_all filters dust automatically #54.
• Handle uncaught errors in tx construction (e.g. no unlocked outputs triggering "no input candidates provided" currently breaks the RPC wallet)
  • Possibly carrot+fcmp: send to integrated address and fix transfer tests #61
  • Add a functional test for error handling in the RPC wallet
• Sporadic failure in k_anonymity functional test (link)
  • Maybe this one is ok to leave for now since allowing >8 inputs per tx should solve it.
• Handle tx dests taking Carrot changes into account.
  • carrot+fcmp: send to integrated address and fix transfer tests #61
• Handle tx priv keys taking Carrot changes into account.
  • carrot+fcmp: send to integrated address and fix transfer tests #61
• Fee-included transfers.
  • carrot+fcmp: send to integrated address and fix transfer tests #61
• wallet2::get_payment_id updated for Carrot.
  • carrot+fcmp: send to integrated address and fix transfer tests #61
• Require new wallets point to updated daemon (required for building the tree from an arbitrary restore height)
  • Require updated daemon on init wallet connection #58
• TODO: Revert breaking changes to wallet2::create_transactions_*() methods regarding payment IDs #70
• Creating watch-only wallet from full wallet, wallet2::import_blockchain has a FIXME for tree
• Blockchain import should add locked outputs and update FCMP++ tree
• Fix CLI/RPC displaying locked status of time-based locked outputs after FCMP++ fork
• Fix monero-blockchain-prune utility
  • blockchain_prune: add FCMP tables and check DB version [STRESSNET] #196

Misc.

• Use new unbiased hash to point for Carrot outputs

  • monero/src/crypto/crypto.cpp

    Lines 650 to 653 in 64190fe

    	// TODO @jeffro256
    	static void hash_to_ec(const public_key &key, ge_p3 &res) {
    	return biased_hash_to_ec(key, res);
    	}

  • Set the tree up here: https://github.com/j-berman/monero/commits/unbiased-hash-curve-trees/
• FFI cleanup
  • Destroy Rust FFI types correctly #39
  • FFI: remove all unwraps and asserts, use err ints #90
  • FFI: clippy+fmt #91
• fcmp++: faster & cleaner transparent amount commitment handling, and detangle deps #43
• fcmp++: avoid expensive inner proof_size calls #40
• Don't store output pubkeys and commitments in db twice
  • fcmp++: store indexes for outputs in tree instead of complete OutputContext, saving 65 bytes/output #62
• Test pre-RCT to FCMP++
  • FCMP++/Carrot: core test creating FCMP++/Carrot tx from {Coinbase,Regular} Pre-RCT tx and BP+ tx #67
• Clean up compile warnings
  • Fix FCMP++/Carrot compile warnings #66
• xmrig
• p2pool
• Block explorer
  • Discussion: Changes for Carrot/FCMP++ moneroexamples/onion-monero-blockchain-explorer#335
• LWS
• Updated documentation for the FCMP++ integration
  • [WIP] FCMP++ integration documentation #110
• Carrot integration documentation
  • Heuristics wallet implementers must keep in mind (e.g. zero-change dummies back to sender, input selection prefers powers of 2, 1/2-input probability)
  • New things users should understand (new priv keys, tx keys, tx proofs, view balance key)
• Document general wallet changes users should keep in mind
  • Updated wallets must point to updated daemons, old wallets can point to updated daemons
  • Deprecated scan_tx scanning future txs relative to wallet sync height, faster wallet restore at high height
  • Updated wallets re-sync from restore height (in order to sync tree)
• Document general daemon changes users should keep in mind
  • Long migration
  • Deprecated building on top of alt chains
• Make sure functional tests pass excluding the HF from hardforks.cpp too
  • [DO NOT MERGE] Test pre-fork functional tests #111
• Use optimized libs (https://github.com/j-berman/fcmp-plus-plus-optimization-competition)
• Move to monero-oxide/monero-oxide #93
• Remove static proof len table and use the latest fast proof len size calc from the FCMP++ lib instead
  • Move the static proof len table into a test just to sanity check the proof len size calc

On hold

• Blinds cache (fcmp++: cache that blinds branches and outputs in the background #25) (no longer necessary thanks to contest success)
• Update references to "weight" instead of "bytes" as necessary
  • Not necessary for launch once fcmp++: use weight approx equal to byte size #232 is in, since weight will be much closer to bytes with that change.

[jeffro256]

Stressnet Blockers

The following tasks are blockers before a useful stressnet network can be launched.

• Support >8 input txs (including fixing quadratic prove bug)
• Divisors re-impl from SLVer bullet (edit: see this MRL discussion)
• Pulling in @fabrizio-m's ec-divisors submission
• Pulling in @Lederstrumpf's helioselene submission, plus kayaba's inversion changes
• src: dynamic block sync size #78
• FCMP++: get rid of init hash download on restore + fix deep reorg handling + more #81
• fcmp++: fix stop/start curve tree migration #87
• Remove broken get_transaction_hash in scanning tool #88

These tasks may significantly impact the performance characteristics of the nodes and should be merged before undergoing stressnet testing.

monero-project#9473 is a nice to have as well.

[spackle-xmr]

Unresolved from the 2024 stressnet

• Serialization limits: Daemons processing big blocks may bump against serializer sanity checks and fail to sync #9388 and contrib: change default value of epee::serialization to fix limitation #9433

• Dynamic block sync size: src: dynamic block sync size #9494

• Dynamic span: src: dynamic span, to calculate span limit dynamically #9495

[nahuhh]

Unresolved from the 2024 stressnet

• Serialization limits: Daemons processing big blocks may bump against serializer sanity checks and fail to sync #9388 and contrib: change default value of epee::serialization to fix limitation #9433

serialization limits might best be resolved by 8867

• Dynamic block sync size: src: dynamic block sync size #9494

this should be ready for review, and had been tested many times. This works around the serialization limits (by keeping batches, by default, below them), while also syncing batches based on size instead of an arbitrary 100 or 20 blocks.

• Dynamic span: src: dynamic span, to calculate span limit dynamically #9495

This is unrelated to stressnet / hopes to fix an unrelated issue

---

So unless we plan on going above ~30mb blocks, 9494 (dynamic block sync size) is the only one of these 4 prs that need to make it into the stressnet

edit: also need #9473. When txpool > 100mb, you cannot connect a wallet to the unrestricted RPC. Restricted does not have the problem because it batches into 100tx.