test: fix handling of start height in generated E2E testnets (#5563)

In #5488 the E2E testnet generator changed to setting explicit `StartAt` heights for initial nodes. This broke the runner, which expected all initial nodes to have `StartAt: 0`, as well as validator set scheduling in the generator. Testnet loading now normalizes initial nodes to have `StartAt: 0`.

This also tweaks waiting for misbehavior heights to only use an additional wait if there actually is any misbehavior in the testnet, and to output information when waiting.

test/e2e/generator/generate.go

@@ -104,7 +104,7 @@ func generateTestnet(r *rand.Rand, opt map[string]interface{}) (e2e.Manifest, er
 		name := fmt.Sprintf("validator%02d", i)
 		manifest.Nodes[name] = generateNode(r, e2e.ModeValidator, startAt, i <= 2)
 
-		if startAt == 0 {
+		if startAt == manifest.InitialHeight {
 			(*manifest.Validators)[name] = int64(30 + r.Intn(71))
 		} else {
 			manifest.ValidatorUpdates[fmt.Sprint(startAt+5)] = map[string]int64{

test/e2e/pkg/testnet.go

@@ -151,6 +151,9 @@ func LoadTestnet(file string) (*Testnet, error) {
 			Perturbations:    []Perturbation{},
 			Misbehaviors:     make(map[int64]string),
 		}
+		if node.StartAt == testnet.InitialHeight {
+			node.StartAt = 0 // normalize to 0 for initial nodes, since code expects this
+		}
 		if nodeManifest.Mode != "" {
 			node.Mode = Mode(nodeManifest.Mode)
 		}
@@ -341,7 +344,12 @@ func (n Node) Validate(testnet Testnet) error {
 
 	for height, misbehavior := range n.Misbehaviors {
 		if height < n.StartAt {
-			return fmt.Errorf("misbehavior height %d is before start height %d", height, n.StartAt)
+			return fmt.Errorf("misbehavior height %d is below node start height %d",
+				height, n.StartAt)
+		}
+		if height < testnet.InitialHeight {
+			return fmt.Errorf("misbehavior height %d is below network initial height %d",
+				height, testnet.InitialHeight)
 		}
 		exists := false
 		for possibleBehaviors := range mcs.MisbehaviorList {
@@ -395,6 +403,19 @@ func (t Testnet) IPv6() bool {
 	return t.IP.IP.To4() == nil
 }
 
+// LastMisbehaviorHeight returns the height of the last misbehavior.
+func (t Testnet) LastMisbehaviorHeight() int64 {
+	lastHeight := int64(0)
+	for _, node := range t.Nodes {
+		for height := range node.Misbehaviors {
+			if height > lastHeight {
+				lastHeight = height
+			}
+		}
+	}
+	return lastHeight
+}
+
 // Address returns a P2P endpoint address for the node.
 func (n Node) AddressP2P(withID bool) string {
 	ip := n.IP.String()

test/e2e/runner/main.go

@@ -69,13 +69,16 @@ func NewCLI() *CLI {
 			if err := Start(cli.testnet); err != nil {
 				return err
 			}
-			if err := waitForAllMisbehaviors(cli.testnet); err != nil {
-				return err
+			if lastMisbehavior := cli.testnet.LastMisbehaviorHeight(); lastMisbehavior > 0 {
+				// wait for misbehaviors before starting perturbations
+				if err := WaitUntil(cli.testnet, lastMisbehavior+5); err != nil {
+					return err
+				}
 			}
 			if err := Perturb(cli.testnet); err != nil {
 				return err
 			}
-			if err := Wait(cli.testnet, interphaseWaitPeriod); err != nil { // allow some txs to go through
+			if err := Wait(cli.testnet, 5); err != nil { // allow some txs to go through
 				return err
 			}
 
@@ -84,7 +87,7 @@ func NewCLI() *CLI {
 				return err
 			}
 			// wait for network to settle before tests
-			if err := Wait(cli.testnet, interphaseWaitPeriod); err != nil {
+			if err := Wait(cli.testnet, 5); err != nil {
 				return err
 			}
 			if err := Test(cli.testnet); err != nil {

test/e2e/runner/start.go

@@ -15,6 +15,12 @@ func Start(testnet *e2e.Testnet) error {
 	sort.SliceStable(nodeQueue, func(i, j int) bool {
 		return nodeQueue[i].StartAt < nodeQueue[j].StartAt
 	})
+	if len(nodeQueue) == 0 {
+		return fmt.Errorf("no nodes in testnet")
+	}
+	if nodeQueue[0].StartAt > 0 {
+		return fmt.Errorf("no initial nodes in testnet")
+	}
 
 	// Start initial nodes (StartAt: 0)
 	logger.Info("Starting initial network nodes...")

test/e2e/runner/wait.go

@@ -7,8 +7,6 @@ import (
 	e2e "github.com/tendermint/tendermint/test/e2e/pkg"
 )
 
-const interphaseWaitPeriod = 5
-
 // Wait waits for a number of blocks to be produced, and for all nodes to catch
 // up with it.
 func Wait(testnet *e2e.Testnet, blocks int64) error {
@@ -16,30 +14,15 @@ func Wait(testnet *e2e.Testnet, blocks int64) error {
 	if err != nil {
 		return err
 	}
-	waitFor := block.Height + blocks
-	logger.Info(fmt.Sprintf("Waiting for all nodes to reach height %v...", waitFor))
-	_, err = waitForAllNodes(testnet, waitFor, 20*time.Second)
+	return WaitUntil(testnet, block.Height+blocks)
+}
+
+// WaitUntil waits until a given height has been reached.
+func WaitUntil(testnet *e2e.Testnet, height int64) error {
+	logger.Info(fmt.Sprintf("Waiting for all nodes to reach height %v...", height))
+	_, err := waitForAllNodes(testnet, height, 20*time.Second)
 	if err != nil {
 		return err
 	}
 	return nil
 }
-
-// WaitForAllMisbehaviors calculates the height of the last misbehavior and ensures the entire
-// testnet has surpassed this height before moving on to the next phase
-func waitForAllMisbehaviors(testnet *e2e.Testnet) error {
-	_, _, err := waitForHeight(testnet, lastMisbehaviorHeight(testnet))
-	return err
-}
-
-func lastMisbehaviorHeight(testnet *e2e.Testnet) int64 {
-	lastHeight := testnet.InitialHeight
-	for _, n := range testnet.Nodes {
-		for height := range n.Misbehaviors {
-			if height > lastHeight {
-				lastHeight = height
-			}
-		}
-	}
-	return lastHeight + interphaseWaitPeriod
-}