Add `Node::splice_out_max` by tankyleo · Pull Request #1181 · lightningdevkit/ldk-node

tankyleo · 2026-10-20T05:09:29Z

    Add `Node::splice_out_max`

    In case a user wants to withdraw on-chain the maximum amount possible
    from a channel, the sum of the value of the spliced output and the fees
    they contribute to the splice has to be equal to
    `ChannelDetails::next_splice_out_maximum_sat`.

    It is currently not ergonomic to express this via the current
    `Node::splice_out` API, so we add `Node::splice_out_max` to satisfy
    this gap.

    This API remains unfinished; we would like any fee bumps of such splices
    to decrease the value of the spliced output instead of increasing the
    total value of the inputs of the splice transaction.

    Co-Authored-By: HAL 9000

ldk-reviews-bot · 2026-10-20T05:09:32Z

👋 Thanks for assigning @jkczyz as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.

tankyleo · 2026-10-20T05:21:44Z

A heads up this currently conflicts with #1176

shubhamkmr04 · 2026-10-20T13:46:43Z

 				return Err(Error::ChannelSplicingFailed);
 			};

+			// TODO: Retain maximum-withdrawal intent across restarts so its output can shrink


The docs cover that a bump here may be funded by the on-chain wallet, but that fallback doesn't hold back the anchor reserve: select_confirmed_utxos has no reserve logic, and bump_channel_funding_fee doesn't check funds.

Since a max withdrawal leaves no channel balance for a higher fee, almost any bump (unless the channel balance has grown since) pays the whole output plus fees from the wallet, and can spend the reserve. A plain splice_out can hit this too, but here it's the expected path.

Maybe refuse the wallet fallback for these until the TODO is done, or check the reserve first? Related: #1166.

I suggest we fix issue 1166 for both splice_out and splice_out_with_all in a separate PR.

Maybe refuse the wallet fallback for these until the TODO is done

This would require persisting the intention to withdraw the maximum from the channel instead of funding a destination address with a fixed amount. I think this should be done as part of a PR to shrink the size of the output instead of increasing the value of the inputs.

splice_out refused any amount above the channel's outbound capacity. That capacity is computed against the current channel reserve, but a splice-out shrinks the channel and with it the reserve, so LDK can accept a larger splice-out than the guard allowed. Compare against LDK's next_splice_out_maximum_sat instead, which accounts for the reduced reserve, so the guard only rejects amounts LDK would refuse. Suggested by tankyleo in lightningdevkit#1103. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

jkczyz · 2026-10-20T18:29:52Z

+	/// This API is experimental. Currently, a splice-out will be marked as an inbound payment if
+	/// paid to an address associated with the on-chain wallet, but this classification may change
+	/// in the future.
+	pub fn splice_out_with_all(


Should we rename this splice_out_max? The _with_all suffix refers to using all inputs when splicing in. But here we are withdrawing the maximum amount from the channel, which isn't the entire balance.

jkczyz · 2026-10-20T18:30:18Z

+			// TODO: Retain maximum-withdrawal intent across restarts so its output can shrink
+			// to cover higher fees while preserving the channel debit.


IIUC, this should be done upstream instead of having an intent. If so, could you open an issue upstream with the approach mentioned offline?

Issue filed here: https://git.rust-bitcoin.org/lightningdevkit/rust-lightning/issues/5305

jkczyz · 2026-10-20T20:01:52Z

+	do_splice_out_with_all(true, Some(FeeRate::from_sat_per_kwu(2_500))).await;
+}
+
+async fn do_splice_out_with_all(is_funder: bool, fee_rate: Option<FeeRate>) {


Should we test RBF behavior here?

I had codex do a concise expansion to test RBF behavior below. As described it currently is not what the user expects: we keep the destination output the same value, and shift to using wallet inputs to fund the higher fee.

jkczyz

LGTM. Please squash fixups.

In case a user wants to withdraw on-chain the maximum amount possible from a channel, the sum of the value of the spliced output and the fees they contribute to the splice has to be equal to `ChannelDetails::next_splice_out_maximum_sat`. It is currently not ergonomic to express this via the current `Node::splice_out` API, so we add `Node::splice_out_max` to satisfy this gap. This API remains unfinished; we would like any fee bumps of such splices to decrease the value of the spliced output instead of increasing the total value of the inputs of the splice transaction. Co-Authored-By: HAL 9000

tankyleo · 2026-10-21T21:40:21Z

squashed with edits to the tests, documentation, and changelog.

tankyleo · 2026-10-21T21:46:08Z

Use git diff e7957ef7 af0f5e12 to view the overall diff.

tankyleo requested review from jkczyz and tnull October 20, 2026 05:09

shubhamkmr04 reviewed Oct 20, 2026

View reviewed changes

tankyleo force-pushed the 2026-10-20-splice-out-with-all branch from cb33776 to 2a7d714 Compare October 20, 2026 18:00

tankyleo force-pushed the 2026-10-20-splice-out-with-all branch from 2a7d714 to f9ef993 Compare October 20, 2026 18:02

jkczyz mentioned this pull request Oct 20, 2026

Build splice_in_with_all from manually selected inputs #1176

Open

jkczyz reviewed Oct 20, 2026

View reviewed changes

tankyleo requested a review from jkczyz October 20, 2026 23:45

jkczyz reviewed Oct 21, 2026

View reviewed changes

tankyleo force-pushed the 2026-10-20-splice-out-with-all branch 2 times, most recently from 05b0ce8 to 431a376 Compare October 21, 2026 21:38

tankyleo force-pushed the 2026-10-20-splice-out-with-all branch from 431a376 to af0f5e1 Compare October 21, 2026 21:39

tankyleo requested a review from jkczyz October 21, 2026 21:48

jkczyz changed the title ~~Add Node::splice_out_with_all~~ Add Node::splice_out_max Oct 21, 2026

jkczyz approved these changes Oct 21, 2026

View reviewed changes

jkczyz merged commit 7b47a0b into lightningdevkit:main Oct 21, 2026
17 of 27 checks passed

tankyleo deleted the 2026-10-20-splice-out-with-all branch October 21, 2026 22:46

shubhamkmr04 mentioned this pull request Oct 23, 2026

Count the funding fee in anchor reserve checks for channel opens #1204

Open

		// TODO: Retain maximum-withdrawal intent across restarts so its output can shrink
		// to cover higher fees while preserving the channel debit.

Conversation

tankyleo commented Oct 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ldk-reviews-bot commented Oct 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tankyleo commented Oct 20, 2026

Uh oh!

shubhamkmr04 Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

tankyleo Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

jkczyz Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

tankyleo Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

jkczyz Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

tankyleo Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

jkczyz Oct 20, 2026

Choose a reason for hiding this comment

Uh oh!

tankyleo Oct 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jkczyz left a comment

Choose a reason for hiding this comment

Uh oh!

tankyleo commented Oct 21, 2026

Uh oh!

tankyleo commented Oct 21, 2026

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

tankyleo commented Oct 20, 2026 •

edited

Loading

ldk-reviews-bot commented Oct 20, 2026 •

edited

Loading

tankyleo Oct 20, 2026 •

edited

Loading