put output errors behind a feature flag

We're going to start merging breaking functgionality behind feature
flags, to reduce the need for long-lived feature branches.

terraform/eval_output.go

@@ -68,18 +68,22 @@ func (n *EvalWriteOutput) Eval(ctx EvalContext) (interface{}, error) {
 
 	// handling the interpolation error
 	if err != nil {
-		if n.ContinueOnErr {
+		switch {
-			log.Printf("[ERROR] Output interpolation %q failed: %s", n.Name, err)
+		case featureOutputErrors:
-			// if we're continueing, make sure the output is included, and
+			if n.ContinueOnErr {
-			// marked as unknown
+				log.Printf("[ERROR] Output interpolation %q failed: %s", n.Name, err)
-			mod.Outputs[n.Name] = &OutputState{
+				// if we're continueing, make sure the output is included, and
-				Type:  "string",
+				// marked as unknown
-				Value: config.UnknownVariableValue,
+				mod.Outputs[n.Name] = &OutputState{
+					Type:  "string",
+					Value: config.UnknownVariableValue,
+				}
+				return nil, EvalEarlyExitError{}
 			}
-			return nil, EvalEarlyExitError{}
+			return nil, err
+		default:
+			log.Printf("[WARN] Output interpolation %q failed: %s", n.Name, err)
 		}
-
-		return nil, err
 	}
 
 	// Get the value from the config

terraform/features.go

@@ -0,0 +1,9 @@
+package terraform
+
+import (
+	"os"
+)
+
+// This file holds feature flags for the next release
+
+var featureOutputErrors = os.Getenv("TF_OUTPUT_ERRORS") != ""