Validate experimentation opt-out with shared assignment fixtures

manifests/cpp.yml

@@ -189,6 +189,7 @@ manifest:
   tests/parametric/test_extract_behavior.py::Test_ExtractBehavior_Restart_With_Extract_First::test_baggage: missing_feature (Automatic baggage propagation is not implemented)
   tests/parametric/test_ffe/test_configuration_sources.py: irrelevant (cpp is not in scope for v1 agentless server SDK configuration-source coverage)
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: missing_feature
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=2.1.1'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

manifests/dotnet.yml

@@ -1183,6 +1183,7 @@ manifest:
       component_version: <3.55.0
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: '>=3.36.0'  # Modified by easy win activation script
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation::test_ffe_flag_evaluation: missing_feature  # Created by easy win activation script
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=3.54.0'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

manifests/golang.yml

@@ -2170,6 +2170,7 @@ manifest:
   tests/parametric/test_extract_behavior.py::Test_ExtractBehavior_Restart_With_Extract_First: missing_feature (DD_TRACE_PROPAGATION_EXTRACT_FIRST=true is unsupported; dd-trace-go panics when extraction fails)
   tests/parametric/test_ffe/test_configuration_sources.py: v2.11.0-dev  # Easy win for all weblogs and version 2.10.1
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: v2.5.0-dev
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=2.10.0'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

manifests/java.yml

@@ -4242,6 +4242,7 @@ manifest:
   : v1.65.0
   tests/parametric/test_ffe/test_configuration_sources.py::Test_Feature_Flag_Configuration_Source_Warm_State_Preservation: v1.65.0
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: v1.56.0
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=1.64.2'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

manifests/nodejs.yml

@@ -2658,6 +2658,7 @@ manifest:
   tests/parametric/test_extract_behavior.py::Test_ExtractBehavior_Restart_With_Extract_First::test_baggage: *ref_5_56_0
   tests/parametric/test_ffe/test_configuration_sources.py: *ref_6_8_0
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: *ref_5_75_0
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py:  # TODO: a lower version might be supported
     - declaration: missing_feature (dd-trace-js#8343)

manifests/php.yml

@@ -1598,6 +1598,7 @@ manifest:
   : missing_feature (FFL-2705 tracks PHP agentless configuration-source implementation; FFL-2731 tracks system-tests configuration-source contract)
   tests/parametric/test_ffe/test_configuration_sources.py::Test_Feature_Flag_Configuration_Source_Warm_State_Preservation: missing_feature (FFL-2705 tracks PHP agentless configuration-source implementation; FFL-2731 tracks system-tests configuration-source contract)
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: v1.21.0-dev
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=1.23.3'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

manifests/python.yml

@@ -2257,6 +2257,7 @@ manifest:
   tests/parametric/test_extract_behavior.py::Test_ExtractBehavior_Restart_With_Extract_First: v2.20.0
   tests/parametric/test_ffe/test_configuration_sources.py: v4.14.0-dev
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: v4.0.0
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=4.9.0'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

manifests/ruby.yml

@@ -2679,6 +2679,7 @@ manifest:
   tests/parametric/test_extract_behavior.py::Test_ExtractBehavior_Restart_With_Extract_First: incomplete_test_app (The parametric test app does not emit restart span links or preserve baggage)
   tests/parametric/test_ffe/test_configuration_sources.py: v2.46.0-dev
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: v2.23.0
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: missing_feature
   tests/parametric/test_headers_b3.py::Test_Headers_B3::test_headers_b3_extract_invalid:  # Easy win for all weblogs and version 2.27.0

manifests/rust.yml

@@ -258,6 +258,7 @@ manifest:
   tests/parametric/test_ffe/test_configuration_sources.py: irrelevant (rust is not in scope for v1 agentless server SDK configuration-source coverage)
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation: '>=0.5.3-dev+fd1307f'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_dynamic_evaluation.py::Test_Feature_Flag_Dynamic_Evaluation::test_ffe_flag_evaluation: missing_feature
+  tests/parametric/test_ffe/test_experimentation_opt_out.py: missing_feature
   tests/parametric/test_ffe/test_serial_id_metadata.py: missing_feature
   tests/parametric/test_ffe/test_span_enrichment.py: '>=0.5.3-dev+fd1307f'  # TODO: a lower version might be supported
   tests/parametric/test_ffe/test_span_enrichment.py::Test_Span_Enrichment_Child_Span_Propagation: missing_feature

tests/parametric/test_ffe/EXPERIMENTATION_OPT_OUT.md

@@ -0,0 +1,57 @@
+# Experimentation opt-out runtime contract
+
+This companion suite tests the real Java provider through its public evaluation API.
+It checks values, allocation keys, variant keys, serial IDs, exposure classification, and hook metadata.
+It does not infer success from an application default with an equal value.
+
+## Shared inputs
+
+The nested fixture snapshot comes from [shared fixture PR #48](https://github.com/DataDog/ffe-system-test-data/pull/48).
+The exact commit is `5e2430d9ad36644e2cecb6a32ddb8d05d7c98ac6`.
+`fixtures/experimentation-opt-out/SOURCE.json` records every original Git blob ID.
+The loader verifies every blob and all 136 case IDs before using the snapshot.
+Do not edit copied configurations or expectations. Update the shared repository and refresh the snapshot together.
+
+The runtime suite executes 116 cases in 11 configuration/provider-option groups.
+It excludes 20 cases with fixed allocation time boundaries.
+The Java SDK fixture adapter covers those cases with its fixed-clock test seam.
+The dogfooding companion also runs them while their original time window is valid.
+Changing generated dates would invalidate that proof.
+
+## Local candidate run
+
+Copy the agent, provider, and API JARs from the same candidate build into `binaries/`.
+Do not combine a released agent with a candidate provider.
+The Java image installer already accepts all three local JARs.
+
+```sh
+PYTEST_XDIST_AUTO_NUM_WORKERS=1 TEST_LIBRARY=java ./run.sh PARAMETRIC \
+  tests/parametric/test_ffe/test_experimentation_opt_out.py --runxfail
+```
+
+`run.sh` appends `-n auto`. The environment variable bounds that setting to one worker.
+Use `--skip-parametric-build` only when the image and candidate artifacts did not change.
+The manifest marks this unreleased behavior as `missing_feature` without inventing a release version.
+The local command uses `--runxfail` so a candidate failure fails the run.
+The suite enables the existing span-enrichment setting so ordinary allocation serial metadata is observable.
+This does not enable experimentation opt-out. The public provider option controls that behavior separately.
+
+`ffe_start(provider_options=...)` activates the public provider options before evaluation.
+A missing option method is a setup failure, not a behavioral negative control.
+The baseline negative control uses the existing evaluation path without the new method.
+Contexts remain per evaluation. The test does not replace server-global subject state.
+
+The Java controllers retain JSON integers that Jackson parses as `Long`.
+They convert these values through OpenFeature's supported `Double` context type.
+The bounded regression in `test_dynamic_evaluation.py` checks an exactly representable integer and both adjacent values.
+This does not add exact 64-bit integer support to the SDK.
+
+## Evidence boundary
+
+These groups check assignment behavior only. The SDK shares exposure deduplication across providers.
+Repeated fixture subjects can therefore suppress duplicate exposures correctly.
+The dogfooding companion restarts its own app between the four independent telemetry cases.
+It retains exact per-subject exposure and evaluation receipts, metrics, and correlated assignment traces.
+It also checks overlapping evaluations and confirms that a skipped assignment does not reserve a deduplication key.
+
+The 20 Precompute protocol cases and nine client state sequences do not apply to this server runtime suite.

tests/parametric/test_ffe/experimentation_contract.py

@@ -0,0 +1,63 @@
+"""Read an immutable shared fixture snapshot and compare actual provider details."""
+
+import hashlib
+import json
+from pathlib import Path
+from typing import Any
+
+FIXTURES = Path(__file__).parent / "fixtures" / "experimentation-opt-out"
+
+
+def load_contract() -> tuple[dict[str, Any], dict[str, Any], list[dict[str, Any]]]:
+    """Verify every source blob before loading the shared contract."""
+    source = json.loads((FIXTURES / "SOURCE.json").read_text())
+    assert source["commit"] == "5e2430d9ad36644e2cecb6a32ddb8d05d7c98ac6"
+    for name, expected in source["blobs"].items():
+        contents = (FIXTURES / name).read_bytes()
+        actual = hashlib.sha1(b"blob " + str(len(contents)).encode() + b"\0" + contents).hexdigest()
+        assert actual == expected, f"Modified shared fixture: {name}"
+    manifest = json.loads((FIXTURES / "manifest.json").read_text())
+    configurations = {
+        item["id"]: json.loads((FIXTURES / item["path"]).read_text()) for item in manifest["configurations"]
+    }
+    cases: list[dict[str, Any]] = []
+    for group in manifest["evaluationCases"]:
+        group_cases = json.loads((FIXTURES / group["path"]).read_text())
+        assert [case["id"] for case in group_cases] == group["ids"]
+        cases.extend(group_cases)
+    assert len(cases) == manifest["expectedCounts"]["evaluations"] == 136
+    return manifest, configurations, cases
+
+
+def has_time_boundary(configuration: dict[str, Any], case: dict[str, Any]) -> bool:
+    """Time-gated cases require fixed-clock SDK tests or a guarded live runtime."""
+    flag = configuration["flags"][case["evaluation"]["flag"]]
+    return any(allocation.get("startAt") or allocation.get("endAt") for allocation in flag["allocations"])
+
+
+def assert_details(result: dict[str, Any], case: dict[str, Any]) -> None:
+    """Require assignment identity and metadata, not only an equal value."""
+    expected = case["expected"]
+    assert result["value"] == expected["value"], (case["id"], result)
+    assignment = expected["assignment"]
+    if assignment is None:
+        assert not result.get("allocationKey"), (case["id"], result)
+        assert not result.get("variant"), (case["id"], result)
+        assert result.get("serialId") is None, (case["id"], result)
+        if expected["outcome"] != "no-assignment":
+            assert result.get("errorCode"), (case["id"], result)
+        return
+    assert not result.get("errorCode"), (case["id"], result)
+    assert result.get("reason"), (case["id"], result)
+    assert result.get("allocationKey") == assignment["allocationKey"], (case["id"], result)
+    assert result.get("variant") == assignment["variationKey"], (case["id"], result)
+    assert result.get("serialId") == assignment["serialId"], (case["id"], result)
+    metadata = result.get("flagMetadata", {})
+    assert metadata.get("__dd_do_log") is assignment["doLog"], (case["id"], result)
+    expected_features = {
+        "__dd_feature." + feature["key"]: feature["value"]
+        for feature in assignment.get("features", [])
+        if "HOOK" in feature["destinations"]
+    }
+    actual_features = {key: value for key, value in metadata.items() if key.startswith("__dd_feature.")}
+    assert actual_features == expected_features, (case["id"], actual_features, expected_features)