Fix a race that could occur if somehow the P2P network got the depend…

…ency tx to us before the lh server responded to the pledge upload request. Never seen this before but today it happened!

common/src/main/java/lighthouse/wallet/PledgingWallet.java

@@ -216,8 +216,13 @@ public LHProtos.Pledge commit(boolean andBroadcastDependencies, LHProtos.Pledge
             log.info("Committing pledge for stub: {}", stub);
             committed = true;
             if (dependency != null) {
-                commitTx(dependency);
-                if (andBroadcastDependencies) {
+                // It's possible that by the time we arrive here, the dependency is already committed, thus we use the
+                // maybe variant. The reason is, for a server-assisted project the server will broadcast the dependency
+                // for us to avoid races where the server thinks a pledge is invalid because it can't see the stub
+                // output. If the server responds to our HTTP upload request and we get here *slower* that the p2p
+                // network manages to propagate the transaction back to us, we might have already processed the
+                // dependency tx!
+                if (maybeCommitTx(dependency) && andBroadcastDependencies) {
                     log.info("Broadcasting dependency");
                     vTransactionBroadcaster.broadcastTransaction(dependency);
                 }