cfg-gate async signing logic

We are intending to release without having completed our async
signing logic, which sadly means we need to cfg-gate it to ensure
we restore the previous state of panicking on signer errors, rather
than putting us in a stuck state with no way to recover.

Here we add a new `async_signing` cfg flag and use it to gate all
the new logic from #2558 effectively reverting commits
1da29290e7 through
014a336e59.

ci/check-cfg-flags.py

@@ -86,6 +86,8 @@ def check_cfg_tag(cfg):
         pass
     elif cfg == "taproot":
         pass
+    elif cfg == "async_signing":
+        pass
     elif cfg == "require_route_graph_test":
         pass
     else:

ci/ci-tests.sh

@@ -171,7 +171,6 @@ if [ -f "$(which arm-none-eabi-gcc)" ]; then
 	popd
 fi
 
-echo -e "\n\nTest Taproot builds"
-pushd lightning
+echo -e "\n\nTest cfg-flag builds"
 RUSTFLAGS="$RUSTFLAGS --cfg=taproot" cargo test --verbose --color always -p lightning
-popd
+RUSTFLAGS="$RUSTFLAGS --cfg=async_signing" cargo test --verbose --color always -p lightning

lightning/src/ln/channel.rs

@@ -2434,8 +2434,13 @@ impl<SP: Deref> ChannelContext<SP> where SP::Target: SignerProvider  {
 					.ok();
 
 				if funding_signed.is_none() {
-					log_trace!(logger, "Counterparty commitment signature not available for funding_signed message; setting signer_pending_funding");
-					self.signer_pending_funding = true;
+					#[cfg(not(async_signing))] {
+						panic!("Failed to get signature for funding_signed");
+					}
+					#[cfg(async_signing)] {
+						log_trace!(logger, "Counterparty commitment signature not available for funding_signed message; setting signer_pending_funding");
+						self.signer_pending_funding = true;
+					}
 				} else if self.signer_pending_funding {
 					log_trace!(logger, "Counterparty commitment signature available for funding_signed message; clearing signer_pending_funding");
 					self.signer_pending_funding = false;
@@ -4259,7 +4264,7 @@ impl<SP: Deref> Channel<SP> where
 
 	/// Indicates that the signer may have some signatures for us, so we should retry if we're
 	/// blocked.
-	#[allow(unused)]
+	#[cfg(async_signing)]
 	pub fn signer_maybe_unblocked<L: Deref>(&mut self, logger: &L) -> SignerResumeUpdates where L::Target: Logger {
 		let commitment_update = if self.context.signer_pending_commitment_update {
 			self.get_last_commitment_update_for_send(logger).ok()
@@ -4363,11 +4368,16 @@ impl<SP: Deref> Channel<SP> where
 			}
 			update
 		} else {
-			if !self.context.signer_pending_commitment_update {
-				log_trace!(logger, "Commitment update awaiting signer: setting signer_pending_commitment_update");
-				self.context.signer_pending_commitment_update = true;
+			#[cfg(not(async_signing))] {
+				panic!("Failed to get signature for new commitment state");
+			}
+			#[cfg(async_signing)] {
+				if !self.context.signer_pending_commitment_update {
+					log_trace!(logger, "Commitment update awaiting signer: setting signer_pending_commitment_update");
+					self.context.signer_pending_commitment_update = true;
+				}
+				return Err(());
 			}
-			return Err(());
 		};
 		Ok(msgs::CommitmentUpdate {
 			update_add_htlcs, update_fulfill_htlcs, update_fail_htlcs, update_fail_malformed_htlcs, update_fee,
@@ -6448,9 +6458,14 @@ impl<SP: Deref> OutboundV1Channel<SP> where SP::Target: SignerProvider {
 
 		let funding_created = self.get_funding_created_msg(logger);
 		if funding_created.is_none() {
-			if !self.context.signer_pending_funding {
-				log_trace!(logger, "funding_created awaiting signer; setting signer_pending_funding");
-				self.context.signer_pending_funding = true;
+			#[cfg(not(async_signing))] {
+				panic!("Failed to get signature for new funding creation");
+			}
+			#[cfg(async_signing)] {
+				if !self.context.signer_pending_funding {
+					log_trace!(logger, "funding_created awaiting signer; setting signer_pending_funding");
+					self.context.signer_pending_funding = true;
+				}
 			}
 		}
 
@@ -6796,7 +6811,7 @@ impl<SP: Deref> OutboundV1Channel<SP> where SP::Target: SignerProvider {
 
 	/// Indicates that the signer may have some signatures for us, so we should retry if we're
 	/// blocked.
-	#[allow(unused)]
+	#[cfg(async_signing)]
 	pub fn signer_maybe_unblocked<L: Deref>(&mut self, logger: &L) -> Option<msgs::FundingCreated> where L::Target: Logger {
 		if self.context.signer_pending_funding && self.context.is_outbound() {
 			log_trace!(logger, "Signer unblocked a funding_created");

lightning/src/ln/channelmanager.rs

@@ -7322,8 +7322,7 @@ where
 	/// attempted in every channel, or in the specifically provided channel.
 	///
 	/// [`ChannelSigner`]: crate::sign::ChannelSigner
-	#[cfg(test)] // This is only implemented for one signer method, and should be private until we
-	             // actually finish implementing it fully.
+	#[cfg(async_signing)]
 	pub fn signer_unblocked(&self, channel_opt: Option<(PublicKey, ChannelId)>) {
 		let _persistence_guard = PersistenceNotifierGuard::notify_on_drop(self);
 

lightning/src/ln/mod.rs

@@ -76,7 +76,7 @@ mod monitor_tests;
 #[cfg(test)]
 #[allow(unused_mut)]
 mod shutdown_tests;
-#[cfg(test)]
+#[cfg(all(test, async_signing))]
 #[allow(unused_mut)]
 mod async_signer_tests;