Fix slashing migration to v10

frame/staking/src/lib.rs

@@ -882,7 +882,7 @@ enum Releases {
 
 impl Default for Releases {
 	fn default() -> Self {
-		Releases::V8_0_0
+		Releases::V10_0_0
 	}
 }
 

frame/staking/src/migrations.rs

@@ -27,13 +27,29 @@ pub mod v10 {
 	#[storage_alias]
 	type EarliestUnappliedSlash<T: Config> = StorageValue<Pallet<T>, EraIndex>;
 
+	/// Apply any pending slashes that where queued.
+	///
+	/// That means we might slash someone a bit too early, but we will definitely
+	/// won't forget to slash them. The cap of 512 is somewhat randomly taken to
+	/// prevent us from iterating over an arbitrary large number of keys `on_runtime_upgrade`.
 	pub struct MigrateToV10<T>(sp_std::marker::PhantomData<T>);
 	impl<T: Config> OnRuntimeUpgrade for MigrateToV10<T> {
 		fn on_runtime_upgrade() -> frame_support::weights::Weight {
 			if StorageVersion::<T>::get() == Releases::V9_0_0 {
+				let pending_slashes = <Pallet<T> as Store>::UnappliedSlashes::iter().take(512);
+				for (era, slashes) in pending_slashes {
+					for slash in slashes {
+						// in the old slashing scheme, the slash era was the key at which we read
+						// from `UnappliedSlashes`.
+						log!(warn, "prematurely applying a slash ({:?}) for era {:?}", slash, era);
+						slashing::apply_slash::<T>(slash, era);
+					}
+				}
+
 				EarliestUnappliedSlash::<T>::kill();
 				StorageVersion::<T>::put(Releases::V10_0_0);
 
+				log!(info, "MigrateToV10 executed successfully");
 				T::DbWeight::get().reads_writes(1, 1)
 			} else {
 				log!(warn, "MigrateToV10 should be removed.");