Support manually selecting inputs consuming their entire value#4575
Support manually selecting inputs consuming their entire value#4575wpaulino wants to merge 2 commits intolightningdevkit:mainfrom
Conversation
|
👋 Thanks for assigning @TheBlueMatt as a reviewer! |
| ); | ||
|
|
||
| if !self.inputs.is_empty() { | ||
| if !self.inputs.is_empty() && self.input_mode == Some(FundingInputMode::CoinSelected) { |
There was a problem hiding this comment.
Bug: Backwards compatibility issue with persisted FundingContribution objects.
FundingContribution is persisted in PendingFunding.contributions (channel.rs:2917). When deserializing contributions created before this PR, input_mode will be None (it's a new option TLV field). Old coin-selected contributions with inputs will have input_mode == None, causing this condition to be false even though they should take the coin-selected branch.
This causes two problems for old persisted coin-selected contributions:
- Wrong fee buffer calculation: Uses
holder_balance + net_value_without_feeinstead ofestimated_fee + change_value, potentially allowing or rejecting feerate adjustments incorrectly. - Change output silently dropped:
compute_feerate_adjustmentreturnsNonefor change, andat_feeratesetschange_output = None, losing the change value.
This is reachable via for_acceptor_at_feerate / for_initiator_at_feerate called on contributions loaded from pending_splice.contributions (channel.rs lines 12504, 12944, 13127, 13145).
Fix: use self.input_mode != Some(FundingInputMode::Manual) instead of self.input_mode == Some(FundingInputMode::CoinSelected) to preserve old behavior for contributions where input_mode is None:
| if !self.inputs.is_empty() && self.input_mode == Some(FundingInputMode::CoinSelected) { | |
| if !self.inputs.is_empty() && self.input_mode != Some(FundingInputMode::Manual) { |
There was a problem hiding this comment.
There is no backwards compatibility concern because the serialized object has not been included in a release yet.
| if let Some(PriorContribution { contribution: prior_contribution, .. }) = | ||
| self.prior_contribution.as_ref() | ||
| { | ||
| if prior_contribution.input_mode == Some(FundingInputMode::CoinSelected) |
There was a problem hiding this comment.
Minor: Same backwards-compat pattern as line 880. Old persisted coin-selected contributions will have input_mode == None, so this guard won't fire for them. In practice this is mostly mitigated by the check at line 1309 (value_added > 0 && manually_selected_inputs non-empty), but it could miss edge cases where the old prior had value_added() == 0 (inputs exactly covered outputs + fees).
Consider using != Some(FundingInputMode::Manual) combined with a non-empty inputs check:
| if prior_contribution.input_mode == Some(FundingInputMode::CoinSelected) | |
| if prior_contribution.input_mode != Some(FundingInputMode::Manual) | |
| && !prior_contribution.inputs.is_empty() |
|
I've thoroughly re-reviewed the entire PR diff. The two previously flagged backwards-compatibility issues (lines 866 and 1310) remain the only actionable issues. No new bugs, security problems, or logic errors found. Review SummaryPreviously Flagged Issues (still open)Both backwards-compatibility issues from the prior review are still present and unresolved:
New IssuesNo new issues found. The rest of the implementation is correct:
|
12e5994 to
412ec3d
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #4575 +/- ##
==========================================
+ Coverage 86.22% 86.31% +0.09%
==========================================
Files 159 159
Lines 109170 109681 +511
Branches 109170 109681 +511
==========================================
+ Hits 94136 94676 +540
+ Misses 12424 12385 -39
- Partials 2610 2620 +10
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
TheBlueMatt
left a comment
There was a problem hiding this comment.
All LGTM. One question.
|
🔔 1st Reminder Hey @jkczyz! This PR has been waiting for your review. |
412ec3d to
228cae8
Compare
|
🔔 2nd Reminder Hey @jkczyz! This PR has been waiting for your review. |
| value_added: Amount, | ||
| manually_selected_inputs: Vec<FundingTxInput>, |
There was a problem hiding this comment.
Seems making these an enum would simplify some of the checks? Or is there a good argument for keeping these separate fields? It would essentially be changing FundingInputs to own the manual inputs and using that here.
There was a problem hiding this comment.
I kept them separate so we don't expose errors on the add/remove_value/input builder methods. Instead, we detect when the user is attempting both at build time, since technically they could have provided a coin selection source to the builder but still opted to manually select inputs.
| (Some(value_added), contribution.inputs, Some(FundingInputMode::CoinSelected)) | ||
| }, | ||
| FundingInputs::ManuallySelected { inputs } => { | ||
| (None, inputs.to_vec(), Some(FundingInputMode::Manual)) |
There was a problem hiding this comment.
FWIW, with the enum approach we could avoid this allocation.
elnosh
left a comment
There was a problem hiding this comment.
Not that familiar with splicing-related code in the codebase but looking at this for learning purposes. After much staring, changes make sense although agree with @jkczyz comment on making an enum since those fields (value_added and manually_selected_inputs) are mutually exclusive in the ways inputs can be provided.
| /// used if the request cannot be satisfied by reusing a prior contribution, by using only | ||
| /// manually selected inputs, or by building a pure splice-out directly. |
There was a problem hiding this comment.
I'm not sure I follow this comment. If we have a builder with either sync or async CoinSelectionSource then manually added inputs are not allowed.
There was a problem hiding this comment.
Manually selected inputs are still allowed as long as the added_value is 0.
| return Err(FundingContributionError::InvalidSpliceValue); | ||
| } | ||
|
|
||
| validate_inputs(&self.manually_selected_inputs)?; |
There was a problem hiding this comment.
shouldn't happen but wouldn't hurt for this to check for duplicate inputs provided?
There was a problem hiding this comment.
We already fail this when we get to the negotiation stage. Not sure it's worth failing early given how unlikely this would be, but happy to add it if others agree.
228cae8 to
17c864b
Compare
This commit introduces an alternative way of splicing in funds without coin selection by requiring the full UTXO to be provided. Each UTXO's entire value (minus fees) is allocated towards the channel, which provides unified balance wallets a more intuitive API when splicing funds into the channel, as they don't particularly care about maintaining a portion of their balance onchain. To simplify the implementation, we require that contributions are not allowed to mix coin-selected inputs with manually-selected ones. Users will need to start a fresh contribution if they want to change the funding input mode.
There's no reason not to do so, and it allows us to fail earlier when the user's net contribution exceeds their spliceable balance.
17c864b to
c3da7fe
Compare
This commit introduces an alternative way of splicing in funds without coin selection by requiring the full UTXO to be provided. Each UTXO's entire value (minus fees) is allocated towards the channel, which provides unified balance wallets a more intuitive API when splicing funds into the channel, as they don't particularly care about maintaining a portion of their balance onchain.
To simplify the implementation, we require that contributions are not allowed to mix coin-selected inputs with manually-selected ones. Users will need to start a fresh contribution if they want to change the funding input mode.