Zero next_commitment_number does not immediately fail channel

[tankyleo]

Core Lightning does not immediately fail an established channel when receiving channel_reestablish with next_commitment_number = 0. On master at f80895bac, CLN instead sends a warning for the stale next_revocation_number, records a transient failure, and does not broadcast its commitment.

BOLT #2 states:

if next_commitment_number is zero: MUST immediately fail the channel and broadcast any relevant latest commitment transaction.

BOLT #2 requirements

In channeld.c, next_revocation_number is evaluated first. A sufficiently stale value calls peer_failed_warn_nodisconnect, so the commitment-number validation is never reached.

The existing zero check is nested under:

next_commitment_number == peer->next_index[REMOTE] - 1

which does not cover an advanced channel receiving zero.

Moving that check immediately after parsing and validating channel_reestablish, before other reconnect processing, was validated end to end on regtest with CLN master and unmodified LDK Server main: CLN immediately failed the channel, broadcast its latest commitment, and the recovering peer swept its output:

diff --git a/channeld/channeld.c b/channeld/channeld.c
index 3848f0d2..6b64f807 100644
--- a/channeld/channeld.c
+++ b/channeld/channeld.c
@@ -5890,6 +5890,12 @@ static void peer_reconnect(struct peer *peer,
                     next_revocation_number,
                     tal_count(peer->splice_state->inflights),
                     peer->splice_state->count);
+       if (next_commitment_number == 0)
+               peer_failed_err(peer->pps,
+                                &peer->channel_id,
+                                "bad reestablish commitment_number: %"
+                                PRIu64,
+                                next_commitment_number);

        local_next_funding = (send_tlvs ? send_tlvs->next_funding : NULL);
        remote_next_funding = (recv_tlvs ? recv_tlvs->next_funding : NULL);
@@ -6102,14 +6108,6 @@ static void peer_reconnect(struct peer *peer,
         *       `commitment_signed`.
         */
        if (next_commitment_number == peer->next_index[REMOTE] - 1) {
-               /* We completed opening, we don't re-transmit that one! */
-               if (next_commitment_number == 0)
-                       peer_failed_err(peer->pps,
-                                        &peer->channel_id,
-                                        "bad reestablish commitment_number: %"
-                                        PRIu64,
-                                        next_commitment_number);
-
                if (!recv_tlvs || !recv_tlvs->next_funding)
                        retransmit_commitment_signed = true;
                else