test(jsonl): add parser coverage; harden message/usage handling

Add tests/test_jsonl_parser.py for tool-result dispatch, helpers, parse_session,
quick_session_info, and malformed entries (null/non-dict message and usage).
Introduce _entry_message() so only dict messages are read; keep usage as a dict
before token accounting. Apply the same normalization in quick_session_info.
Use full-string == assertions in TestStripSystemTags; drop unused pytest import;
add regression for string-shaped assistant message.

Author: clean6378-max-it

tests/test_jsonl_parser.py

@@ -5,8 +5,6 @@
 import sys
 import tempfile
 
-import pytest
-
 sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
 
 from utils.jsonl_parser import (  # noqa: E402
@@ -338,8 +336,7 @@ def test_sidechain_only_returns_untitled(self):
 class TestStripSystemTags:
     def test_system_reminder_removed(self):
         t = "<system-reminder>in</system-reminder>keep"
-        assert "system-reminder" not in _strip_system_tags(t)
-        assert "keep" in _strip_system_tags(t)
+        assert _strip_system_tags(t) == "keep"
 
     def test_ide_opened_file_removed(self):
         t = "<ide_opened_file>x</ide_opened_file>y"
@@ -351,9 +348,7 @@ def test_user_prompt_submit_hook_removed(self):
 
     def test_remaining_known_opening_closing_tags_stripped(self):
         t = "</ide_selection><command-name>foo</command-name>bar"
-        out = _strip_system_tags(t)
-        assert "ide_selection" not in out
-        assert "bar" in out
+        assert _strip_system_tags(t) == "foobar"
 
     def test_clean_text_unchanged(self):
         assert _strip_system_tags("hello world") == "hello world"
@@ -745,6 +740,22 @@ def test_null_message_assistant_no_crash(self):
         finally:
             os.unlink(path)
 
+    def test_non_dict_message_assistant_no_crash(self):
+        path = _write_jsonl([
+            {
+                "type": "assistant",
+                "timestamp": "2026-01-01T00:00:00Z",
+                "message": "not-a-dict",
+            },
+        ])
+        try:
+            s = parse_session(path)
+            assert s["metadata"]["total_input_tokens"] == 0
+            assert len(s["messages"]) == 1
+            assert s["messages"][0]["role"] == "assistant"
+        finally:
+            os.unlink(path)
+
     def test_non_dict_usage_assistant_no_crash(self):
         path = _write_jsonl([
             {

utils/jsonl_parser.py

@@ -125,6 +125,11 @@ def parse_session(filepath: str) -> dict:
     }
 
 
+def _entry_message(entry: dict) -> dict:
+    m = entry.get("message")
+    return m if isinstance(m, dict) else {}
+
+
 def _process_user(entry: dict, messages: list, metadata: dict):
     """Pull out text, tool results, and session-level metadata (cwd, version, etc.)
     from a user entry."""
@@ -137,7 +142,7 @@ def _process_user(entry: dict, messages: list, metadata: dict):
     if metadata["permission_mode"] is None:
         metadata["permission_mode"] = entry.get("permissionMode")
 
-    msg = entry.get("message") or {}
+    msg = _entry_message(entry)
     content = msg.get("content", [])
     text = _extract_text(content)
     images = _extract_images(content)
@@ -170,7 +175,7 @@ def _process_user(entry: dict, messages: list, metadata: dict):
 def _process_assistant(entry: dict, messages: list, metadata: dict):
     """Handle assistant responses -- splits content into text, thinking blocks,
     and tool_use calls, and accumulates token/model/tool stats."""
-    msg = entry.get("message") or {}
+    msg = _entry_message(entry)
     model = msg.get("model", "")
     if model and model != "<synthetic>":
         metadata["models_used"].add(model)
@@ -503,7 +508,7 @@ def quick_session_info(filepath: str) -> dict:
                 last_ts = ts  # keep updating in case file is small
 
             if title is None and entry.get("type") == "user":
-                msg = entry.get("message") or {}
+                msg = _entry_message(entry)
                 text = _extract_text(msg.get("content", []))
                 if text:
                     clean = _strip_system_tags(text).strip()