aljaz/lightning
1
0
Fork 0
mirror of https://github.com/aljazceru/lightning.git synced 2026-01-05 07:14:36 +01:00
Code Issues Packages Projects Releases Wiki Activity

tests: on reconnect, we need to send any fulfills before adds

Browse Source 
The peer will fail the channel on reconnect if we send an 'add' we
don't have balance for yet; we can avoid this issue by always sending
fulfills (+) before sending adds (-)
This commit is contained in:
lisa neigut
2020-04-17 15:23:23 -05:00
committed by Rusty Russell
parent 8189e9a64d
commit 1b2f4990ab
1 changed files with 21 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
channeld/channeld.c
View File

@@ -2046,6 +2046,27 @@ static void resend_commitment(struct peer *peer, const struct changed_htlc *last
*/
/* In our case, we consider ourselves already committed to this, so
* retransmission is simplest. */
/* We need to send fulfills/failures before adds, so we split them
* up into two loops -- this is the 'fulfill/fail' loop */
for (i = 0; i < tal_count(last); i++) {
const struct htlc *h;
h = channel_get_htlc(peer->channel,
htlc_state_owner(last[i].newstate),
last[i].id);
/* I think this can happen if we actually received revoke_and_ack
* then they asked for a retransmit */
if (!h)
peer_failed(peer->pps,
&peer->channel_id,
"Can't find HTLC %"PRIu64" to resend",
last[i].id);
if (h->state == SENT_REMOVE_COMMIT)
send_fail_or_fulfill(peer, h);
}
/* We need to send fulfills/failures before adds, so we split them
* up into two loops -- this is the 'add' loop */
for (i = 0; i < tal_count(last); i++) {
const struct htlc *h;
@@ -2082,8 +2103,6 @@ static void resend_commitment(struct peer *peer, const struct changed_htlc *last
#endif
);
sync_crypto_write(peer->pps, take(msg));
} else if (h->state == SENT_REMOVE_COMMIT) {
send_fail_or_fulfill(peer, h);
}
}