feat: reconcile split-surface CORS

Authored-By: OpenAI (GPT-5) <noreply@openai.com>

README.md

@@ -52,6 +52,7 @@ artifacts:
   - name: distributions
     type: static
     content_dir: dist
+    cors_origins: ['*']
     publish:
       bucket: baseline-dist
       credentials:
@@ -141,6 +142,11 @@ protected input from entering a public artifact through dereference. Split
 storage endpoints are pinned to Garage, and each website
 authority is derived from its bucket; a site cannot expose an arbitrary backend.
 
+`cors_origins` is reconciled as a bucket policy on every split publication. Values are either `*`
+or HTTPS origins; browser access is limited to `GET` and `HEAD`. Omitting the field removes stale
+CORS from that bucket. Protected artifacts cannot allow wildcard CORS. All immutable objects and
+all bucket CORS policies complete before mutable channels change.
+
 Each split route gets a bucket-specific `<bucket>.web.sjc001.fritzlab.net`
 ExternalName Service annotated to disable pass-host-header and a separate Ingress. Route
 Ingresses share the hostname's certificate Secret. The access middleware and

scripts/deploy.py

@@ -8,6 +8,7 @@ import os
 import re
 import shutil
 import subprocess
+import tempfile
 from pathlib import Path
 from urllib.error import HTTPError, URLError
 from urllib.request import Request, urlopen
@@ -167,6 +168,41 @@ def publication_aws_env(artifact, credential_env_names=None):
     return aws_env
 
 
+def configure_cors(bucket, origins, endpoint, aws_env):
+    """Reconcile read-only browser access without exposing publication credentials."""
+    if origins is None:
+        return
+    if not origins:
+        run([
+            "aws", "--endpoint-url", endpoint, "s3api", "delete-bucket-cors",
+            "--bucket", bucket,
+        ], env=aws_env)
+        return
+    config = {
+        "CORSRules": [{
+            "AllowedOrigins": origins,
+            "AllowedMethods": ["GET", "HEAD"],
+            "AllowedHeaders": ["*"],
+            "ExposeHeaders": ["ETag"],
+            "MaxAgeSeconds": 3600,
+        }],
+    }
+    with tempfile.NamedTemporaryFile("w", suffix=".json", encoding="utf-8") as handle:
+        json.dump(config, handle)
+        handle.flush()
+        run([
+            "aws", "--endpoint-url", endpoint, "s3api", "put-bucket-cors",
+            "--bucket", bucket, "--cors-configuration", f"file://{handle.name}",
+        ], env=aws_env)
+
+
+def configure_artifact_cors(artifact, credential_env_names=None):
+    configure_cors(
+        artifact["bucket"], artifact["cors_origins"], artifact["s3_endpoint"],
+        publication_aws_env(artifact, credential_env_names),
+    )
+
+
 def publish_route_immutables(artifact, route, site_dir, credential_env_names=None):
     """Publish one route's immutable partitions during the global preflight."""
     html_dir = site_dir / artifact["build_dir"]
@@ -312,6 +348,11 @@ def deploy_static(site_name, site_dir, action_dir, token, cfg):
         publish_route_immutables(
             artifact_by_name[route["artifact"]], route, site_dir, credential_env_names,
         )
+    # Reconcile every browser-read policy before publishing mutable content.
+    # A CORS failure therefore cannot leave a new channel pointing at a release
+    # whose cross-origin assets browsers cannot consume.
+    for artifact in cfg["artifacts"]:
+        configure_artifact_cors(artifact, credential_env_names)
     for route in cfg["routes"]:
         s3_sync(artifact_by_name[route["artifact"]], route, site_dir, credential_env_names)
     if cfg["compatibility"]:

scripts/utils.py

@@ -104,6 +104,26 @@ def _strings(value, label):
     return values
 
 
+def _cors_origins(value, label):
+    origins = _strings(value or [], label)
+    if len(origins) != len(set(origins)):
+        raise ConfigError(f"{label} must not contain duplicates")
+    for origin in origins:
+        if origin == "*":
+            continue
+        parsed = urlparse(origin)
+        try:
+            port = parsed.port
+        except ValueError:
+            port = None
+        if parsed.scheme != "https" or not parsed.hostname or parsed.path or parsed.params or (
+            parsed.query or parsed.fragment or parsed.username or parsed.password
+            or (":" in parsed.netloc and port is None and not parsed.netloc.endswith("]"))
+        ):
+            raise ConfigError(f"{label} must contain '*' or HTTPS origins")
+    return origins
+
+
 def _hostname(value, label):
     if not isinstance(value, str) or len(value) > 253 or value.endswith("."):
         raise ConfigError(f"{label} must be a lowercase DNS hostname without a trailing dot")
@@ -215,6 +235,7 @@ def _legacy_config(raw, site_name):
         "website_authority": "garage-s3.storage.svc.k8s.sjc001.fritzlab.net",
         "credentials": {"access_key_env": "AWS_ACCESS_KEY_ID", "secret_key_env": "AWS_SECRET_ACCESS_KEY"},
         "cache_rules": [{"path": "", "cache_control": DEFAULT_CACHE_CONTROL}],
+        "cors_origins": None,
     }
     return {
         "version": 1,
@@ -234,7 +255,11 @@ def _legacy_config(raw, site_name):
 def _artifact(item, index):
     label = f"artifacts[{index}]"
     item = _mapping(item, label)
-    _known_keys(item, {"name", "type", "content_dir", "tidy", "excludes", "publish", "cache"}, label)
+    _known_keys(
+        item,
+        {"name", "type", "content_dir", "tidy", "excludes", "publish", "cache", "cors_origins"},
+        label,
+    )
     name = item.get("name")
     if not isinstance(name, str) or not NAME_RE.fullmatch(name):
         raise ConfigError(f"{label}.name must be a DNS label")
@@ -303,6 +328,7 @@ def _artifact(item, index):
         "website_authority": authority,
         "credentials": normalized_credentials,
         "cache_rules": cache_rules,
+        "cors_origins": _cors_origins(item.get("cors_origins"), f"{label}.cors_origins"),
     }
 
 
@@ -389,6 +415,10 @@ def _validate_multi(cfg):
                 raise ConfigError(f"protected route {route['name']} cannot use shared-cache max-age")
             if route["access"] == "public" and "private" in directives:
                 raise ConfigError(f"public route {route['name']} cannot use private cache policy")
+        if route["access"] == "protected" and "*" in artifact["cors_origins"]:
+            raise ConfigError(
+                f"protected route {route['name']} cannot allow wildcard CORS"
+            )
     root = next(route for route in routes if route["path"] == "/")
     if any(route["access"] == "protected" for route in routes) and root["access"] == "public":
         raise ConfigError("a public '/' catch-all would expose unmatched protected content")

tests/fixtures/split-site.yaml

@@ -15,6 +15,7 @@ artifacts:
   - name: distributions
     type: static
     content_dir: dist
+    cors_origins: ['*']
     publish:
       bucket: baseline-dist
       credentials:

tests/test_contract.py

@@ -113,6 +113,30 @@ class ConfigContractTests(unittest.TestCase):
         self.assertEqual(["/dist", "/"], [route["path"] for route in cfg["routes"]])
         credentials = {artifact["name"]: artifact["credentials"] for artifact in cfg["artifacts"]}
         self.assertNotEqual(credentials["distributions"], credentials["portal"])
+        distributions = next(
+            artifact for artifact in cfg["artifacts"] if artifact["name"] == "distributions"
+        )
+        self.assertEqual(["*"], distributions["cors_origins"])
+
+    def test_cors_origins_are_https_origins_or_wildcard(self):
+        self.assert_invalid(
+            lambda raw: raw["artifacts"][1].__setitem__(
+                "cors_origins", ["http://consumer.example"]
+            ),
+            "must contain '\\*' or HTTPS origins",
+        )
+        self.assert_invalid(
+            lambda raw: raw["artifacts"][1].__setitem__(
+                "cors_origins", ["https://consumer.example", "https://consumer.example"]
+            ),
+            "must not contain duplicates",
+        )
+
+    def test_protected_route_rejects_wildcard_cors(self):
+        self.assert_invalid(
+            lambda raw: raw["artifacts"][0].__setitem__("cors_origins", ["*"]),
+            "protected route portal cannot allow wildcard CORS",
+        )
 
     def test_equivalent_route_paths_are_ambiguous(self):
         self.assert_invalid(lambda raw: raw["routes"].append({
@@ -131,6 +155,7 @@ class ConfigContractTests(unittest.TestCase):
             raw["routes"][1]["access"] = {
                 "mode": "protected", "middleware": "authentik-forwardauth"
             }
+            raw["artifacts"][1]["cors_origins"] = []
             raw["artifacts"][0]["cache"]["rules"][0]["cache_control"] = (
                 "public, max-age=0, must-revalidate"
             )
@@ -317,6 +342,30 @@ class BuildTests(unittest.TestCase):
 
 
 class PublishingTests(unittest.TestCase):
+    def test_cors_policy_is_reconciled_as_read_only_browser_access(self):
+        captured = {}
+
+        def capture(command, **_kwargs):
+            config_path = command[command.index("--cors-configuration") + 1]
+            captured["command"] = command
+            captured["config"] = json.loads(Path(config_path.removeprefix("file://")).read_text())
+
+        with patch.object(deploy, "run", side_effect=capture):
+            deploy.configure_cors(
+                "baseline-dist", ["*"], "http://garage-s3.storage.svc:3900", {}
+            )
+
+        self.assertIn("put-bucket-cors", captured["command"])
+        self.assertEqual(["GET", "HEAD"], captured["config"]["CORSRules"][0]["AllowedMethods"])
+        self.assertEqual(["*"], captured["config"]["CORSRules"][0]["AllowedOrigins"])
+
+    def test_empty_cors_policy_removes_stale_bucket_cors(self):
+        with patch.object(deploy, "run") as request:
+            deploy.configure_cors(
+                "baseline-catalogue", [], "http://garage-s3.storage.svc:3900", {}
+            )
+        self.assertIn("delete-bucket-cors", request.call_args.args[0])
+
     def test_apps_clone_never_places_token_in_argv_or_log(self):
         calls = []
         secret = "clone-secret-must-not-appear"
@@ -437,15 +486,42 @@ class PublishingTests(unittest.TestCase):
                     patch.object(deploy, "validate_artifact_output"), patch.object(
                         deploy, "publish_route_immutables",
                         side_effect=[None, RuntimeError("immutable failed")],
-                    ) as immutable_publish, patch.object(deploy, "s3_sync") as mutable_sync, \
+                    ) as immutable_publish, patch.object(
+                        deploy, "configure_artifact_cors"
+                    ) as cors_configure, patch.object(deploy, "s3_sync") as mutable_sync, \
                     self.assertRaisesRegex(
                 RuntimeError, "immutable failed"
             ):
                 deploy.deploy_static("baseline", root, root, "token", cfg)
 
             self.assertEqual(2, immutable_publish.call_count)
+            cors_configure.assert_not_called()
             mutable_sync.assert_not_called()
 
+    def test_all_cors_policies_complete_before_mutable_publication(self):
+        cfg = normalize_site_config(fixture("split-site.yaml"), "baseline.fritzlab.net")
+        events = []
+        with tempfile.TemporaryDirectory() as tmp:
+            root = Path(tmp)
+            apps = root / "apps"
+            apps.mkdir()
+            with patch.object(deploy, "validate_publication_environment"), patch.object(
+                deploy, "validate_artifact_output"
+            ), patch.object(
+                deploy, "publish_route_immutables", side_effect=lambda *_args: events.append("immutable")
+            ), patch.object(
+                deploy, "configure_artifact_cors", side_effect=lambda *_args: events.append("cors")
+            ), patch.object(
+                deploy, "s3_sync", side_effect=lambda *_args: events.append("mutable")
+            ), patch.object(deploy, "clone_apps", return_value=apps), patch.object(
+                deploy, "render_site_manifests"
+            ), patch.object(deploy, "commit_and_push"):
+                deploy.deploy_static("baseline", root, root, "token", cfg)
+
+        self.assertEqual(
+            ["immutable", "immutable", "cors", "cors", "mutable", "mutable"], events
+        )
+
 
     def test_absent_artifact_is_detected_before_publish(self):
         cfg = normalize_site_config(fixture("split-site.yaml"), "baseline.fritzlab.net")