fix(benchmarks): harden memory test and ruff test glob

Reset tracemalloc peak before measuring large-file parse, assert non-empty
message count, extend E402 per-file-ignores to tests/**, and clarify README
that benchmark tests live under tests/benchmarks/.

Author: clean6378-max-it

benchmarks/README.md

@@ -1,5 +1,7 @@
 # Performance benchmarks
 
+Test files live under `tests/benchmarks/`; this directory holds only documentation and the informational `baselines.json` snapshot.
+
 Repeatable local measurements for parse, bulk export, and search hot paths.
 
 ## Run locally
@@ -12,7 +14,7 @@ pytest tests/benchmarks/ --benchmark-only -o addopts= -v
 ## Memory check
 
 ```bash
-pytest tests/benchmarks/test_parse_memory.py -v
+pytest tests/benchmarks/test_parse_memory.py -v -o addopts=
 ```
 
 The memory test also runs as part of the normal `pytest` suite (timing benchmarks are skipped via `--benchmark-skip` in `pyproject.toml`).

pyproject.toml

@@ -34,4 +34,4 @@ combine-as-imports = true
 # CLI bootstrap: sys.path must be set before local imports.
 "scripts/export.py" = ["E402"]
 # Tests mirror the same path bootstrap before importing app/utils.
-"tests/*.py" = ["E402"]
+"tests/**/*.py" = ["E402"]

tests/benchmarks/test_parse_memory.py

@@ -14,8 +14,10 @@ def test_large_parse_peak_memory_under_ceiling(parse_large_file: Path) -> None:
     ceiling = file_bytes * 10
 
     tracemalloc.start()
+    tracemalloc.clear_traces()
     try:
-        parse_session(str(path))
+        result = parse_session(str(path))
+        assert len(result["messages"]) > 0, "parse_session returned no messages"
         _, peak = tracemalloc.get_traced_memory()
     finally:
         tracemalloc.stop()

tests/benchmarks/test_search_bench.py

@@ -16,3 +16,4 @@ def _run() -> object:
 
     resp = benchmark(_run)
     assert resp.status_code == 200
+    assert resp.get_json(), "expected search hits from synthetic searchable tokens"