bug: cycles detection fixes in discovery

internal/discovery/discovery.go

@@ -1256,21 +1256,61 @@ func RemoveCycles(c component.Components) (component.Components, error) {
 		cfg *component.Component
 	)
 
+	// Repeatedly break cycles by removing outgoing dependency edges on the offending node,
+	// rather than removing the node entirely. This preserves all units for downstream queueing.
 	for range maxCycleRemovalAttempts {
 		if cfg, err = c.CycleCheck(); err == nil {
-			break
+			// no cycles
+			return c, nil
 		}
 
-		// Cfg should never be nil if err is not nil,
-		// but we do this check to avoid a nil pointer dereference
-		// if our assumptions change in the future.
 		if cfg == nil {
-			break
+			// unexpected: cycle reported but no component; return existing error
+			return c, err
 		}
 
-		c = c.RemoveByPath(cfg.Path)
+		// Rebuild the graph, clearing dependencies originating from the offending node.
+		// 1) Create new nodes indexed by path.
+		nodes := make(map[string]*component.Component, len(c))
+		for _, n := range c {
+			nodes[n.Path] = &component.Component{
+				Parsed:           n.Parsed,
+				DiscoveryContext: n.DiscoveryContext,
+				Kind:             n.Kind,
+				Path:             n.Path,
+				Reading:          append([]string(nil), n.Reading...),
+				External:         n.External,
+			}
+		}
+
+		// 2) Reconnect dependencies, skipping edges that originate from the offending node.
+		for _, n := range c {
+			if n.Path == cfg.Path {
+				// Drop all outgoing edges from this node to break the cycle
+				continue
+			}
+
+			if deps := n.Dependencies(); len(deps) > 0 {
+				src := nodes[n.Path]
+				for _, dep := range deps {
+					if dst, ok := nodes[dep.Path]; ok {
+						src.AddDependency(dst)
+					}
+				}
+			}
+		}
+
+		// 3) Replace components slice with rebuilt nodes (sorted for determinism)
+		rebuilt := make(component.Components, 0, len(nodes))
+		for _, nn := range nodes {
+			rebuilt = append(rebuilt, nn)
+		}
+
+		c = rebuilt.Sort()
+		// loop again to ensure no remaining cycles
 	}
 
+	// If we exit the loop with an error, return the best-effort graph and the error
 	return c, err
 }
 

internal/queue/queue.go

@@ -23,7 +23,6 @@
 package queue
 
 import (
-	"errors"
 	"slices"
 	"sort"
 	"sync"
@@ -287,7 +286,17 @@ func NewQueue(discovered component.Components) (*Queue, error) {
 		}
 	}
 
-	return q, errors.New("cycle detected during queue construction")
+	// Cycle fallback: include all entries in a dependency-agnostic order (by path) and mark them ready.
+	sort.SliceStable(entries, func(i, j int) bool { return entries[i].Component.Path < entries[j].Component.Path })
+
+	for _, e := range entries {
+		e.Status = StatusReady
+	}
+
+	q.Entries = entries
+	q.IgnoreDependencyOrder = true
+
+	return q, nil
 }
 
 // GetReadyWithDependencies returns all entries that are ready to run and have all dependencies completed (or no dependencies).

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/ecr-cache/main.tf

@@ -0,0 +1,3 @@
+output "cache_url" {
+  value = "ecr-cache-url"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/ecr-cache/terragrunt.hcl

@@ -0,0 +1,3 @@
+terraform {
+  source = "."
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/eks-baseline/main.tf

@@ -0,0 +1,8 @@
+variable "cluster_id" {
+  type    = string
+  default = ""
+}
+
+output "baseline_status" {
+  value = "eks-baseline-${var.cluster_id}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/eks-baseline/terragrunt.hcl

@@ -0,0 +1,29 @@
+terraform {
+  source = "."
+}
+
+# Dependencies from which we may or may not consume outputs
+dependencies {
+  paths = try(values.dependencies, [])
+}
+
+# CRITICAL: This dependency uses values.dependency_path which points to the GENERATED unit path
+# This creates a cycle because:
+# - eks-baseline (in stacks/k8s) depends on eks-cluster via values.dependency_path
+# - values.dependency_path.eks-cluster = "./.terragrunt-stack/k8s/.terragrunt-stack/eks-cluster"
+# - But eks-baseline itself gets generated to "./.terragrunt-stack/k8s/.terragrunt-stack/eks-baseline"
+# This inter-stack dependency via generated paths triggers the cycle detection bug
+dependency "eks-cluster" {
+  config_path = values.dependency_path.eks-cluster
+
+  mock_outputs = {
+    cluster_id = "mock-cluster-id"
+  }
+
+  mock_outputs_merge_strategy_with_state  = "shallow"
+  mock_outputs_allowed_terraform_commands = ["init", "validate", "destroy", "plan"]
+}
+
+inputs = {
+  cluster_id = dependency.eks-cluster.outputs.cluster_id
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/eks-cluster/main.tf

@@ -0,0 +1,3 @@
+output "cluster_id" {
+  value = "eks-cluster-123"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/eks-cluster/terragrunt.hcl

@@ -0,0 +1,3 @@
+terraform {
+  source = "."
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/grafana-baseline/main.tf

@@ -0,0 +1,8 @@
+variable "cluster_id" {
+  type    = string
+  default = ""
+}
+
+output "grafana_status" {
+  value = "grafana-baseline-${var.cluster_id}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/grafana-baseline/terragrunt.hcl

@@ -0,0 +1,23 @@
+terraform {
+  source = "."
+}
+
+# Dependencies from which we may or may not consume outputs
+dependencies {
+  paths = try(values.dependencies, [])
+}
+
+dependency "eks-cluster" {
+  config_path = values.dependency_path.eks-cluster
+
+  mock_outputs = {
+    cluster_id = "mock-cluster-id"
+  }
+
+  mock_outputs_merge_strategy_with_state  = "shallow"
+  mock_outputs_allowed_terraform_commands = ["init", "validate", "destroy", "plan"]
+}
+
+inputs = {
+  cluster_id = dependency.eks-cluster.outputs.cluster_id
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/id/main.tf

@@ -0,0 +1,3 @@
+output "id" {
+  value = "id-123"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/id/terragrunt.hcl

@@ -0,0 +1,3 @@
+terraform {
+  source = "."
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/rancher-baseline/main.tf

@@ -0,0 +1,8 @@
+variable "bootstrap_status" {
+  type    = string
+  default = ""
+}
+
+output "baseline_status" {
+  value = "rancher-baseline-${var.bootstrap_status}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/rancher-baseline/terragrunt.hcl

@@ -0,0 +1,23 @@
+terraform {
+  source = "."
+}
+
+# Dependencies from which we may or may not consume outputs
+dependencies {
+  paths = try(values.dependencies, [])
+}
+
+dependency "rancher-bootstrap" {
+  config_path = values.dependency_path.rancher-baseline
+
+  mock_outputs = {
+    bootstrap_status = "mock-bootstrap-status"
+  }
+
+  mock_outputs_merge_strategy_with_state  = "shallow"
+  mock_outputs_allowed_terraform_commands = ["init", "validate", "destroy", "plan"]
+}
+
+inputs = {
+  bootstrap_status = dependency.rancher-bootstrap.outputs.bootstrap_status
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/rancher-bootstrap/main.tf

@@ -0,0 +1,8 @@
+variable "cluster_id" {
+  type    = string
+  default = ""
+}
+
+output "bootstrap_status" {
+  value = "rancher-bootstrap-${var.cluster_id}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/rancher-bootstrap/terragrunt.hcl

@@ -0,0 +1,23 @@
+terraform {
+  source = "."
+}
+
+# Dependencies from which we may or may not consume outputs
+dependencies {
+  paths = try(values.dependencies, [])
+}
+
+dependency "eks-cluster" {
+  config_path = values.dependency_path.eks-cluster
+
+  mock_outputs = {
+    cluster_id = "mock-cluster-id"
+  }
+
+  mock_outputs_merge_strategy_with_state  = "shallow"
+  mock_outputs_allowed_terraform_commands = ["init", "validate", "destroy", "plan"]
+}
+
+inputs = {
+  cluster_id = dependency.eks-cluster.outputs.cluster_id
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/tailscale-router/main.tf

@@ -0,0 +1,8 @@
+variable "vpc_id" {
+  type    = string
+  default = ""
+}
+
+output "router_id" {
+  value = "tailscale-router-${var.vpc_id}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/tailscale-router/terragrunt.hcl

@@ -0,0 +1,15 @@
+terraform {
+  source = "."
+}
+
+dependency "vpc" {
+  config_path = "../vpc"
+
+  mock_outputs = {
+    vpc_id = "mock-vpc-id"
+  }
+}
+
+inputs = {
+  vpc_id = dependency.vpc.outputs.vpc_id
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/vpc-endpoints/main.tf

@@ -0,0 +1,8 @@
+variable "vpc_id" {
+  type    = string
+  default = ""
+}
+
+output "endpoint_id" {
+  value = "vpc-endpoint-${var.vpc_id}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/vpc-endpoints/terragrunt.hcl

@@ -0,0 +1,15 @@
+terraform {
+  source = "."
+}
+
+dependency "vpc" {
+  config_path = "../vpc"
+
+  mock_outputs = {
+    vpc_id = "mock-vpc-id"
+  }
+}
+
+inputs = {
+  vpc_id = dependency.vpc.outputs.vpc_id
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/vpc-nat/main.tf

@@ -0,0 +1,8 @@
+variable "vpc_id" {
+  type    = string
+  default = ""
+}
+
+output "nat_id" {
+  value = "vpc-nat-${var.vpc_id}"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/vpc-nat/terragrunt.hcl

@@ -0,0 +1,15 @@
+terraform {
+  source = "."
+}
+
+dependency "vpc" {
+  config_path = "../vpc"
+
+  mock_outputs = {
+    vpc_id = "mock-vpc-id"
+  }
+}
+
+inputs = {
+  vpc_id = dependency.vpc.outputs.vpc_id
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/vpc/main.tf

@@ -0,0 +1,3 @@
+output "vpc_id" {
+  value = "vpc-123"
+}

test/fixtures/regressions/nested-stacks-runner-pool/_source/units/vpc/terragrunt.hcl

@@ -0,0 +1,3 @@
+terraform {
+  source = "."
+}

test/fixtures/regressions/nested-stacks-runner-pool/live/terragrunt.stack.hcl

@@ -0,0 +1,59 @@
+
+locals {
+  stack_root = "${get_terragrunt_dir()}/.terragrunt-stack"
+
+  dependency_path = {
+    id        = "${local.stack_root}/id"
+    ecr-cache = "${local.stack_root}/ecr-cache"
+  }
+}
+
+unit "id" {
+  path   = "id"
+  source = "${get_repo_root()}/_source/units/id"
+
+  values = {
+    prefix = "aio"
+  }
+}
+
+unit "ecr-cache" {
+  path   = "ecr-cache"
+  source = "${get_repo_root()}/_source/units/ecr-cache"
+
+  values = {
+    dependency_path = local.dependency_path
+  }
+}
+
+stack "network" {
+  path   = "network"
+  source = "../stacks/network"
+
+  values = {
+    dependency_path = local.dependency_path
+  }
+}
+
+stack "k8s" {
+  path   = "k8s"
+  source = "../stacks/k8s"
+
+  values = {
+    dependencies = [
+      "${local.stack_root}/network/.terragrunt-stack/vpc-nat",
+      "${local.stack_root}/network/.terragrunt-stack/tailscale-router",
+    ]
+
+    dependency_path = merge(
+      local.dependency_path,
+      {
+        vpc = "${local.stack_root}/network/.terragrunt-stack/vpc",
+        eks-cluster = "${local.stack_root}/k8s/.terragrunt-stack/eks-cluster",
+        rancher-baseline = "${local.stack_root}/k8s/.terragrunt-stack/rancher-baseline",
+      }
+    )
+  }
+}
+
+