Review fixes: remove cast() calls, add debug_report to FakePlatform, remove dead code, fix stale doc paths

Author: rjpower

lib/iris/TESTING.md

@@ -118,7 +118,7 @@ types even in startup-polling loops.
   `@pytest.mark.e2e`.
 - Docker-dependent tests must also be marked `@pytest.mark.docker`.
 - E2E tests live in `tests/e2e/`.
-- Shared fakes live in `tests/cluster/platform/fakes.py` or
+- Shared fakes live in `tests/providers/fakes.py` or
   `src/iris/test_util.py`. Do not duplicate fakes across files.
 
 ## Protocols

lib/iris/docs/autoscaler-fix.md

@@ -712,7 +712,7 @@ elif availability.status == GroupAvailability.AT_MAX_SLICES:
 ```bash
 cd lib/iris
 uv run pytest tests/cluster/controller/test_autoscaler.py -x -v
-uv run pytest tests/cluster/platform/test_scaling_group.py -x -v
+uv run pytest tests/providers/test_scaling_group.py -x -v
 uv run pytest -m 'not slow' -x
 ```
 

lib/iris/docs/coreweave.md

@@ -362,7 +362,7 @@ The platform detects fatal errors before the full timeout expires:
 ### Cluster startup (`iris cluster start`)
 
 `CoreweavePlatform.start_controller()` orchestrates the full startup sequence.
-See `lib/iris/src/iris/cluster/platform/coreweave.py`.
+See `lib/iris/src/iris/providers/k8s/coreweave.py`.
 
 1. Apply RBAC prerequisites (Namespace, ServiceAccount, ClusterRole `iris-controller-{ns}`, ClusterRoleBinding `iris-controller-{ns}`)
 2. Create S3 credentials Secret (if S3 storage configured)
@@ -579,8 +579,8 @@ instantly. Fix in config and redeploy.
 
 | File | Description |
 |------|-------------|
-| `lib/iris/src/iris/cluster/platform/coreweave.py` | CoreWeave platform implementation (includes `ensure_rbac()`) |
+| `lib/iris/src/iris/providers/k8s/coreweave.py` | CoreWeave platform implementation (includes `ensure_rbac()`) |
 | `lib/iris/src/iris/cluster/runtime/kubernetes.py` | KubernetesRuntime (Pod-per-task) |
-| `lib/iris/src/iris/cluster/k8s/kubectl.py` | Kubectl CLI wrapper |
+| `lib/iris/src/iris/providers/k8s/service.py` | Kubectl CLI wrapper |
 | `lib/iris/examples/coreweave.yaml` | Example cluster config |
 | `lib/iris/AGENTS.md` | CoreWeave integration notes for agents |

lib/iris/docs/image-push.md

@@ -124,7 +124,7 @@ docker pull us-docker.pkg.dev/hai-gcp-models/ghcr-mirror/marin-community/iris-wo
 
 ## Code
 
-- **Rewrite logic**: `lib/iris/src/iris/cluster/platform/bootstrap.py`
+- **Rewrite logic**: `lib/iris/src/iris/providers/bootstrap.py`
   - `zone_to_multi_region()`: maps GCP zone → continent (`us`, `europe`, `asia`)
   - `rewrite_ghcr_to_ar_remote()`: rewrites `ghcr.io/...` → `{continent}-docker.pkg.dev/.../ghcr-mirror/...`
 - **Autoscaler**: `_per_group_bootstrap_config()` rewrites the worker image per scale group

lib/iris/src/iris/providers/base.py

@@ -527,10 +527,6 @@ def default_stop_all(
     for s in all_slices:
         targets.append((f"slice:{s.slice_id}", s.terminate))
 
-    if not targets:
-        logger.info("No resources to terminate")
-        return target_names
-
     logger.info("Terminating %d resource(s) in parallel", len(targets))
 
     errors: list[str] = []

lib/iris/src/iris/providers/factory.py

@@ -18,7 +18,6 @@
 """Factory for creating Platform and ControllerLifecycle instances from cluster configuration."""
 
 from dataclasses import dataclass
-from typing import cast
 
 from iris.providers.base import ControllerLifecycle, Platform
 from iris.providers.k8s.coreweave import CoreweaveControllerLifecycle, CoreweavePlatform
@@ -79,8 +78,8 @@ def create_platform_bundle(
             ssh_config=ssh_config,
         )
         return PlatformBundle(
-            platform=cast(Platform, gcp_platform),
-            lifecycle=cast(ControllerLifecycle, GcpControllerLifecycle(platform=gcp_platform)),
+            platform=gcp_platform,
+            lifecycle=GcpControllerLifecycle(platform=gcp_platform),
         )
 
     if which == "manual":
@@ -89,8 +88,8 @@ def create_platform_bundle(
             ssh_config=ssh_config,
         )
         return PlatformBundle(
-            platform=cast(Platform, manual_platform),
-            lifecycle=cast(ControllerLifecycle, ManualControllerLifecycle(platform=manual_platform)),
+            platform=manual_platform,
+            lifecycle=ManualControllerLifecycle(platform=manual_platform),
         )
 
     if which == "local":
@@ -102,8 +101,8 @@ def create_platform_bundle(
             gcp_service=gcp_service,
         )
         return PlatformBundle(
-            platform=cast(Platform, local_platform),
-            lifecycle=cast(ControllerLifecycle, GcpControllerLifecycle(platform=local_platform)),
+            platform=local_platform,
+            lifecycle=GcpControllerLifecycle(platform=local_platform),
         )
 
     if which == "coreweave":
@@ -112,8 +111,8 @@ def create_platform_bundle(
             label_prefix=label_prefix,
         )
         return PlatformBundle(
-            platform=cast(Platform, cw_platform),
-            lifecycle=cast(ControllerLifecycle, CoreweaveControllerLifecycle(platform=cw_platform)),
+            platform=cw_platform,
+            lifecycle=CoreweaveControllerLifecycle(platform=cw_platform),
         )
 
     raise ValueError(f"Unknown platform: {which}")

lib/iris/tests/providers/fakes.py

@@ -325,8 +325,11 @@ def list_vms(self, zones: list[str], labels: dict[str, str] | None = None) -> li
     def tunnel(self, address: str, local_port: int | None = None):
         return nullcontext(address)
 
+    def debug_report(self) -> None:
+        pass
+
     def shutdown(self) -> None:
-        """No-op: FakePlatform has no background threads to stop."""
+        pass
 
     def tick(self, ts: int | None = None) -> None:
         """Advance all slice state transitions.