test: add tests for force_full parameter and fix Windows path normalization

- Add API endpoint tests for force_full query parameter
- Add sync service test validating force_full bypasses watermark optimization
- Fix normalize_project_path() to handle Windows absolute paths correctly
- Windows paths (e.g., C:\Users\...) now returned unchanged instead of prepending slash

Fixes #407 - ensures force_full parameter is properly tested

Signed-off-by: phernandez <paul@basicmachines.co>

Author: phernandez

src/basic_memory/utils.py

@@ -20,11 +20,13 @@ def normalize_project_path(path: str) -> str:
     prefix from project paths to avoid leaking implementation details and to
     ensure paths match the actual S3 bucket structure.
 
+    For local paths (including Windows paths), returns the path unchanged.
+
     Args:
-        path: Project path (e.g., "/app/data/basic-memory-llc")
+        path: Project path (e.g., "/app/data/basic-memory-llc" or "C:\\Users\\...")
 
     Returns:
-        Normalized path (e.g., "/basic-memory-llc")
+        Normalized path (e.g., "/basic-memory-llc" or "C:\\Users\\...")
 
     Examples:
         >>> normalize_project_path("/app/data/my-project")
@@ -33,13 +35,21 @@ def normalize_project_path(path: str) -> str:
         '/my-project'
         >>> normalize_project_path("app/data/my-project")
         '/my-project'
+        >>> normalize_project_path("C:\\\\Users\\\\project")
+        'C:\\\\Users\\\\project'
     """
-    # Handle both absolute and relative paths
+    # Check if this is a Windows absolute path (e.g., C:\Users\...)
+    # Windows paths have a drive letter followed by a colon
+    if len(path) >= 2 and path[1] == ":":
+        # Windows absolute path - return unchanged
+        return path
+
+    # Handle both absolute and relative Unix paths
     normalized = path.lstrip("/")
     if normalized.startswith("app/data/"):
         normalized = normalized.removeprefix("app/data/")
 
-    # Ensure leading slash for absolute paths
+    # Ensure leading slash for Unix absolute paths
     if not normalized.startswith("/"):
         normalized = "/" + normalized
 

test2/foo.md

@@ -466,6 +466,40 @@ async def test_sync_project_endpoint(test_graph, client, project_url):
     assert "Filesystem sync initiated" in data["message"]
 
 
+@pytest.mark.asyncio
+async def test_sync_project_endpoint_with_force_full(test_graph, client, project_url):
+    """Test the project sync endpoint with force_full parameter."""
+    # Call the sync endpoint with force_full=true
+    response = await client.post(f"{project_url}/project/sync?force_full=true")
+
+    # Verify response
+    assert response.status_code == 200
+    data = response.json()
+
+    # Check response structure
+    assert "status" in data
+    assert "message" in data
+    assert data["status"] == "sync_started"
+    assert "Filesystem sync initiated" in data["message"]
+
+
+@pytest.mark.asyncio
+async def test_sync_project_endpoint_with_force_full_false(test_graph, client, project_url):
+    """Test the project sync endpoint with force_full=false."""
+    # Call the sync endpoint with force_full=false
+    response = await client.post(f"{project_url}/project/sync?force_full=false")
+
+    # Verify response
+    assert response.status_code == 200
+    data = response.json()
+
+    # Check response structure
+    assert "status" in data
+    assert "message" in data
+    assert data["status"] == "sync_started"
+    assert "Filesystem sync initiated" in data["message"]
+
+
 @pytest.mark.asyncio
 async def test_sync_project_endpoint_not_found(client):
     """Test the project sync endpoint with nonexistent project."""

test2/foot.md

@@ -147,6 +147,64 @@ async def test_file_count_increased_uses_incremental_scan(
     assert "file3.md" in report.new
 
 
+@pytest.mark.asyncio
+async def test_force_full_bypasses_watermark_optimization(
+    sync_service: SyncService, project_config: ProjectConfig
+):
+    """Test that force_full=True bypasses watermark optimization and scans all files.
+
+    This is critical for detecting changes made by external tools like rclone bisync
+    that don't update mtimes detectably. See issue #407.
+    """
+    project_dir = project_config.home
+
+    # Create initial files
+    await create_test_file(project_dir / "file1.md", "# File 1\nOriginal")
+    await create_test_file(project_dir / "file2.md", "# File 2\nOriginal")
+
+    # First sync - establishes watermark
+    report = await sync_service.sync(project_dir)
+    assert len(report.new) == 2
+
+    # Verify watermark was set
+    project = await sync_service.project_repository.find_by_id(
+        sync_service.entity_repository.project_id
+    )
+    assert project.last_scan_timestamp is not None
+    initial_timestamp = project.last_scan_timestamp
+
+    # Sleep to ensure time passes
+    await sleep_past_watermark()
+
+    # Modify a file WITHOUT updating mtime (simulates external tool like rclone)
+    # We do this by reading the current mtime, modifying the file, then restoring the mtime
+    file_path = project_dir / "file1.md"
+    original_stat = file_path.stat()
+    await create_test_file(file_path, "# File 1\nModified by external tool")
+    # Restore original mtime to simulate external tool behavior
+    import os
+
+    os.utime(file_path, (original_stat.st_atime, original_stat.st_mtime))
+
+    # Normal incremental sync should NOT detect the change (mtime unchanged)
+    report = await sync_service.sync(project_dir)
+    assert len(report.modified) == 0, (
+        "Incremental scan should not detect changes with unchanged mtime"
+    )
+
+    # Force full scan should detect the change via checksum comparison
+    report = await sync_service.sync(project_dir, force_full=True)
+    assert len(report.modified) == 1, "Force full scan should detect changes via checksum"
+    assert "file1.md" in report.modified
+
+    # Verify watermark was still updated after force_full
+    project = await sync_service.project_repository.find_by_id(
+        sync_service.entity_repository.project_id
+    )
+    assert project.last_scan_timestamp is not None
+    assert project.last_scan_timestamp > initial_timestamp
+
+
 # ==============================================================================
 # Incremental Scan Base Cases
 # ==============================================================================