[midnight adapter] Adapter - #786
Conversation
fa42572 to
eaab822
Compare
peyha
left a comment
peyha
left a comment
There was a problem hiding this comment.
Nice 🔥 🔥 using a linked list to manage maturities is smart
If I understand correctly, the allocation workflow should be
An allocator creates a make offer for the vault
A taker takes this offer onchain
| morphoV2 calls onBuy on the adapter
| | the adapter updates its state
| | the adapter calls allocate on its parent vault
| | | the vault V2 calls allocate on the market V2 adapter
| | | | the adapter returns the corresponding ids and allocated assets
|
|
||
| // Do not cleanup the linked list if we end up at 0 growth | ||
| function withdraw(Obligation memory obligation, uint256 obligationUnits, uint256 shares) external { | ||
| require(IVaultV2(parentVault).isAllocator(msg.sender), NotAuthorized()); |
There was a problem hiding this comment.
shouldn't lenders also be able to trigger a withdraw on morpho V2 ?
There was a problem hiding this comment.
(note: it's been renamed withdrawToVault for clarity)
- It's safe to make it permissionless but I don't know if we want that
- Even if permissioned the user can buy at 1 through forceDeallocate then withdraw, it costs him the settlement fee + the penalty.
| maturityData.netCredit += boughtNetCredit.toUint128(); | ||
| marketData.netCredit += boughtNetCredit.toUint128(); | ||
|
|
||
| // Insert the maturity in the list if needed |
There was a problem hiding this comment.
would a dedicated lib for this kind of logic make sense ?
There was a problem hiding this comment.
I don't think we would gain anything in this case, unlike the DurationsLib's get it's not reused. Btw DurationLib's pack is not reused either in deployed code, it could be inlined, but it's useful to have it in tests.
| uint256 netCreditLoss = uint256(marketData.netCredit) + boughtNetCredit - currentNetCredit(marketId); | ||
| decreaseNetCredit(marketId, market.maturity, netCreditLoss); | ||
|
|
||
| IVaultV2(parentVault) |
There was a problem hiding this comment.
the vault v2 doc specifies here that "[adapters] must not re-enter (directly or indirectly) the vault" and this line is breaking this property, a few possibilities to move forward
- Adapt the vault v2 doc to take this specific case into account
- Use a new intermediary contract "Relayer" to handle the callback
onBuy/onSellbefore calling the vault (but it adds operational complexity) - Refactor this function to make it clear that
onBuyis not interacting dangerously with the vault
There was a problem hiding this comment.
Good point. I think adapting the docs works in this case.
| IVaultV2(parentVault) | ||
| .deallocate(address(this), abi.encode(zeroedDurationsIds, -int256(uint256(maturityData.netCredit))), 0); | ||
| } |
There was a problem hiding this comment.
another way to update these caps would be to return these ids/change in allocate/deallocate to let the vault update it automatically
There was a problem hiding this comment.
I didn't get this.
| /// @dev Returns the number of durations in packedDurations that are at most the time to maturity. | ||
| function durationCount(uint256 maturity) internal view returns (uint256 count) { | ||
| uint256 timeToMaturity = maturity.zeroFloorSub(block.timestamp); | ||
| while (count < durationsLength && timeToMaturity >= packedDurations.get(count)) count++; |
There was a problem hiding this comment.
could speed this up with fenwick trees or segment tree + lazy propagation. not worth it for small duration array though
There was a problem hiding this comment.
yes very likely too complex for 8 durations
# Conflicts: # foundry.toml
…ory under EIP-170
…multiple fitting EIP-170
2 participants
TODO
Design notes notion page
Notes:
Things to decide:
forceDeallocateby selling vsforceDeallocateby minting sharesgrowth = interest / ttmso any take whereinterest < ttmthe remainder goes to current assets. In the WBTC example6.3e8 * 5/100is ~ 1 year in seconds. But for others it's not as bad (for usdc 1 year 5%, everything goes to assets below $630).State creation costs increases in glamsterdam that can bias more than before towards creating fewer slots.