From 13960104371019b87781f1ea68b5de90946298c9 Mon Sep 17 00:00:00 2001
From: catlog22 <catlog22@github.com>
Date: Mon, 29 Dec 2025 12:33:23 +0800
Subject: [PATCH] fix(embedder): add lock protection for cache read operations

Protect fast path cache read in get_embedder() to prevent KeyError

during concurrent access and cache clearing operations.

Solution-ID: SOL-1735392000001

Issue-ID: ISS-1766921318981-2

Task-ID: T1
---
 codex-lens/src/codexlens/semantic/embedder.py | 15 ++--
 codex-lens/tests/test_embedder.py             | 85 +++++++++++++++++++
 2 files changed, 91 insertions(+), 9 deletions(-)
 create mode 100644 codex-lens/tests/test_embedder.py

diff --git a/codex-lens/src/codexlens/semantic/embedder.py b/codex-lens/src/codexlens/semantic/embedder.py
index bb1bc856..e2d21717 100644
--- a/codex-lens/src/codexlens/semantic/embedder.py
+++ b/codex-lens/src/codexlens/semantic/embedder.py
@@ -21,7 +21,7 @@ logger = logging.getLogger(__name__)
 
 # Global embedder cache for singleton pattern
 _embedder_cache: Dict[str, "Embedder"] = {}
-_cache_lock = threading.Lock()
+_cache_lock = threading.RLock()
 
 
 def get_embedder(profile: str = "code", use_gpu: bool = True) -> "Embedder":
@@ -43,15 +43,12 @@ def get_embedder(profile: str = "code", use_gpu: bool = True) -> "Embedder":
     # Cache key includes GPU preference to support mixed configurations
     cache_key = f"{profile}:{'gpu' if use_gpu else 'cpu'}"
 
-    # Fast path: check cache without lock
-    if cache_key in _embedder_cache:
-        return _embedder_cache[cache_key]
-
-    # Slow path: acquire lock for initialization
+    # All cache access is protected by _cache_lock to avoid races with
+    # clear_embedder_cache() during concurrent access.
     with _cache_lock:
-        # Double-check after acquiring lock
-        if cache_key in _embedder_cache:
-            return _embedder_cache[cache_key]
+        embedder = _embedder_cache.get(cache_key)
+        if embedder is not None:
+            return embedder
 
         # Create new embedder and cache it
         embedder = Embedder(profile=profile, use_gpu=use_gpu)
diff --git a/codex-lens/tests/test_embedder.py b/codex-lens/tests/test_embedder.py
new file mode 100644
index 00000000..3d6850a1
--- /dev/null
+++ b/codex-lens/tests/test_embedder.py
@@ -0,0 +1,85 @@
+"""Tests for embedder cache concurrency."""
+
+from __future__ import annotations
+
+import threading
+import time
+
+import pytest
+
+import codexlens.semantic.embedder as embedder_module
+
+
+def _patch_embedder_for_unit_tests(monkeypatch: pytest.MonkeyPatch) -> None:
+    """Make get_embedder() tests deterministic and fast (no model downloads)."""
+
+    monkeypatch.setattr(embedder_module, "SEMANTIC_AVAILABLE", True)
+    monkeypatch.setattr(embedder_module, "get_optimal_providers", lambda *args, **kwargs: [])
+    monkeypatch.setattr(embedder_module, "is_gpu_available", lambda: False)
+    monkeypatch.setattr(embedder_module.Embedder, "_load_model", lambda self: None)
+
+
+def test_embedder_instances_are_cached_and_reused(monkeypatch: pytest.MonkeyPatch) -> None:
+    _patch_embedder_for_unit_tests(monkeypatch)
+    embedder_module.clear_embedder_cache()
+
+    first = embedder_module.get_embedder(profile="code", use_gpu=False)
+    second = embedder_module.get_embedder(profile="code", use_gpu=False)
+
+    assert first is second
+
+
+def test_concurrent_cache_access(monkeypatch: pytest.MonkeyPatch) -> None:
+    _patch_embedder_for_unit_tests(monkeypatch)
+    embedder_module.clear_embedder_cache()
+
+    profiles = ["fast", "code", "balanced", "multilingual"]
+    for profile in profiles:
+        embedder_module.get_embedder(profile=profile, use_gpu=False)
+
+    errors: list[BaseException] = []
+    errors_lock = threading.Lock()
+
+    def record_error(err: BaseException) -> None:
+        with errors_lock:
+            errors.append(err)
+
+    worker_count = 20
+    start_barrier = threading.Barrier(worker_count + 1)
+    stop_at = time.monotonic() + 1.0
+
+    def clear_worker() -> None:
+        try:
+            start_barrier.wait()
+            while time.monotonic() < stop_at:
+                embedder_module.clear_embedder_cache()
+                time.sleep(0)
+        except BaseException as err:
+            record_error(err)
+
+    def access_worker(profile: str) -> None:
+        try:
+            start_barrier.wait()
+            while time.monotonic() < stop_at:
+                embedder_module.get_embedder(profile=profile, use_gpu=False)
+        except BaseException as err:
+            record_error(err)
+
+    threads: list[threading.Thread] = [
+        threading.Thread(target=clear_worker, name="clear-embedder-cache"),
+    ]
+    for idx in range(worker_count):
+        threads.append(
+            threading.Thread(
+                target=access_worker,
+                name=f"get-embedder-{idx}",
+                args=(profiles[idx % len(profiles)],),
+            )
+        )
+
+    for thread in threads:
+        thread.start()
+    for thread in threads:
+        thread.join(timeout=10)
+
+    assert not errors, f"Unexpected errors during concurrent access: {errors!r}"