fix(testing): align CLI budget fixture paths across review runners

docs/development/testing-and-quality.md

@@ -683,6 +683,20 @@ override or merge bypass.
 断言，但必须保留解析、必需字段、锚点和语义差异检查；candidate 始终执行当前绝对
 预算。measurement-only 不能用于 candidate，也不是合并旁路。
 
+The real-CLI differential runner and pytest use the same fixed-width fixture
+alias **per default scenario**. Alias the scenario root, not just its parent:
+otherwise scenario-name suffixes change repeated absolute command paths and
+can create a size failure unrelated to output growth. Measure unmodified
+stdout, keep fixture populations and budgets unchanged, and retain the separate
+real-long-path command-integrity check. This aligns measurement layouts; it does
+not shorten production commands or qualify long-path output under short-path caps.
+
+独立 real-CLI 对照和 pytest 对每个默认场景使用相同的固定宽度 fixture 别名。
+别名应指向场景根目录，而非仅指向父目录；否则场景名会改变多处绝对命令路径，
+产生与输出增长无关的尺寸失败。仍测量未经改写的 stdout，保留原负载、预算及
+独立的真实长路径命令完整性检查。这仅统一测量布局，不缩短生产命令，也不将
+长路径输出冒充短路径预算已通过。
+
 The PR-review packet's `semantic_alignment` rule consumes this evidence through
 the existing `validation_matrix` and `observable_semantics` rows. It does not
 add a separate budget receipt or approval gate. The result checker verifies

examples/control_plane/cli-output-probe-runner.py

@@ -163,61 +163,64 @@ def _default_rows(
 ) -> list[dict]:
     rows: list[dict] = []
     for scenario in probe.SCENARIOS:
-        project, runtime, registry_path, state_file = probe._write_fixture(
-            fixture_root / scenario.name,
-            scenario,
-        )
-        for output_format in ("json", "markdown"):
-            commands = probe._surface_commands(
-                project=project,
-                runtime=runtime,
-                registry_path=registry_path,
-                state_file=state_file,
-                output_format=output_format,
+        # Match _measure_scenario: alias each scenario, not its parent. Otherwise
+        # emitted command paths include the scenario suffix only in this runner.
+        with probe._stable_budget_fixture_root(fixture_root / scenario.name) as root:
+            project, runtime, registry_path, state_file = probe._write_fixture(
+                root,
+                scenario,
             )
-            for surface_id, command in commands.items():
-                exit_code, text = probe._invoke_cli(command)
-                if exit_code != 0:
-                    raise AssertionError(f"{surface_id}/{output_format} failed")
-                measurement = probe.measure_cli_output(
-                    text, output_format=output_format
+            for output_format in ("json", "markdown"):
+                commands = probe._surface_commands(
+                    project=project,
+                    runtime=runtime,
+                    registry_path=registry_path,
+                    state_file=state_file,
+                    output_format=output_format,
                 )
-                surface = probe.CLI_OUTPUT_BUDGET_BY_ID[surface_id]
-                if enforce_budget:
-                    probe.assert_cli_output_baseline(
-                        surface,
-                        scenario=scenario.name,
-                        output_format=output_format,
-                        text=text,
-                        measurement=measurement,
+                for surface_id, command in commands.items():
+                    exit_code, text = probe._invoke_cli(command)
+                    if exit_code != 0:
+                        raise AssertionError(f"{surface_id}/{output_format} failed")
+                    measurement = probe.measure_cli_output(
+                        text, output_format=output_format
                     )
-                else:
-                    _assert_output_contract(
-                        output_format=output_format,
-                        text=text,
-                        measurement=measurement,
-                        semantic_json_keys=surface.semantic_json_keys,
-                        markdown_anchor=surface.markdown_anchor,
-                    )
-                rows.append(
-                    _receipt_row(
-                        semantics=semantics,
-                        row_id=f"surface/{surface_id}/{scenario.name}/{output_format}",
-                        surface_id=surface_id,
-                        scenario=scenario.name,
-                        output_format=output_format,
-                        qualification_policy=surface.qualification_policy,
-                        semantic_json_keys=surface.semantic_json_keys,
-                        markdown_anchor=surface.markdown_anchor,
-                        measurement=measurement,
-                        text=text,
-                        output_contract_version=(
-                            getattr(surface, "output_contract_version", None)
-                            if output_format == "json"
-                            else None
+                    surface = probe.CLI_OUTPUT_BUDGET_BY_ID[surface_id]
+                    if enforce_budget:
+                        probe.assert_cli_output_baseline(
+                            surface,
+                            scenario=scenario.name,
+                            output_format=output_format,
+                            text=text,
+                            measurement=measurement,
+                        )
+                    else:
+                        _assert_output_contract(
+                            output_format=output_format,
+                            text=text,
+                            measurement=measurement,
+                            semantic_json_keys=surface.semantic_json_keys,
+                            markdown_anchor=surface.markdown_anchor,
+                        )
+                    rows.append(
+                        _receipt_row(
+                            semantics=semantics,
+                            row_id=f"surface/{surface_id}/{scenario.name}/{output_format}",
+                            surface_id=surface_id,
+                            scenario=scenario.name,
+                            output_format=output_format,
+                            qualification_policy=surface.qualification_policy,
+                            semantic_json_keys=surface.semantic_json_keys,
+                            markdown_anchor=surface.markdown_anchor,
+                            measurement=measurement,
+                            text=text,
+                            output_contract_version=(
+                                getattr(surface, "output_contract_version", None)
+                                if output_format == "json"
+                                else None
+                            ),
                         ),
                     )
-                )
     return rows
 
 

tests/control_plane/test_cli_output_probe_runner.py

@@ -0,0 +1,90 @@
+from __future__ import annotations
+
+import runpy
+from pathlib import Path
+
+import pytest
+
+from loopx.control_plane.testing import cli_output_semantics
+from tests.control_plane import test_cli_output_budget as probe
+
+
+RUNNER = (
+    Path(__file__).resolve().parents[2]
+    / "examples/control_plane/cli-output-probe-runner.py"
+)
+
+
+@pytest.fixture
+def crowded_turn_probe(monkeypatch: pytest.MonkeyPatch):
+    commands = probe._surface_commands
+    monkeypatch.setattr(probe, "SCENARIOS", (probe.SCENARIOS[1],))
+
+    def turn_json_only(**kwargs):
+        if kwargs["output_format"] != "json":
+            return {}
+        return {"loopx_turn_plan": commands(**kwargs)["loopx_turn_plan"]}
+
+    monkeypatch.setattr(probe, "_surface_commands", turn_json_only)
+    return runpy.run_path(str(RUNNER))["_default_rows"]
+
+
+@pytest.mark.parametrize("layout", ["short", "long-runner-layout" * 8])
+def test_runner_uses_the_pytest_scenario_alias_without_rewriting_stdout(
+    tmp_path: Path,
+    monkeypatch: pytest.MonkeyPatch,
+    crowded_turn_probe,
+    layout: str,
+) -> None:
+    fixture_roots: list[Path] = []
+    write_fixture = probe._write_fixture
+    invoke_cli = probe._invoke_cli
+    emitted: list[str] = []
+
+    def capture_fixture(root, scenario):
+        fixture_roots.append(root)
+        assert (scenario.todo_count, scenario.agent_count, scenario.run_count) == (
+            36,
+            1,
+            12,
+        )
+        return write_fixture(root, scenario)
+
+    def capture_stdout(command):
+        rc, text = invoke_cli(command)
+        emitted.append(text)
+        return rc, text
+
+    monkeypatch.setattr(probe, "_write_fixture", capture_fixture)
+    monkeypatch.setattr(probe, "_invoke_cli", capture_stdout)
+    rows = crowded_turn_probe(probe, cli_output_semantics, tmp_path / layout)
+
+    assert len(rows) == len(emitted) == len(fixture_roots) == 1
+    root = fixture_roots[0]
+    assert root.parent == Path("/tmp")
+    assert root.name.startswith("loopx-cli-budget-")
+    assert len(root.name.removeprefix("loopx-cli-budget-")) == 12
+    assert not root.exists()  # The shared context cleans up its alias.
+    assert str(root) in emitted[0]  # Full CLI command paths are still emitted.
+    assert rows[0]["row_id"] == "surface/loopx_turn_plan/crowded/json"
+    assert rows[0]["chars"] == len(emitted[0])
+    assert (
+        "$.turn_envelope.replan_action_packet.writeback_contract.vision_authoring"
+        in (rows[0]["json_shape_paths"])
+    )
+
+
+def test_runner_still_rejects_actual_stdout_growth(
+    tmp_path: Path,
+    monkeypatch: pytest.MonkeyPatch,
+    crowded_turn_probe,
+) -> None:
+    invoke_cli = probe._invoke_cli
+
+    def oversized_stdout(command):
+        rc, text = invoke_cli(command)
+        return rc, text + " " * 15_000
+
+    monkeypatch.setattr(probe, "_invoke_cli", oversized_stdout)
+    with pytest.raises(AssertionError, match="baseline ceiling is 14500"):
+        crowded_turn_probe(probe, cli_output_semantics, tmp_path / "growth")