"""Tests for judge.py encoding fix (debug print in retry loop). Verifies that: 1. Non-ASCII content in system_prompt/user_message doesn't raise exceptions when DEBUG_PROMPTS is enabled. 2. The ascii-safe encoding correctly replaces non-ASCII characters. 3. When DEBUG_PROMPTS is False, no output is produced. """ import io import os import sys import unittest from unittest.mock import patch class TestDebugPrintEncoding(unittest.TestCase): """Test the safe encoding pattern used in the judge retry loop.""" def _safe_print(self, label: str, text: str, max_len: int = 500): """Replicate the fixed debug print pattern from judge.py line 950-953.""" try: print(f"{label}:", text[:max_len].encode('ascii', 'replace').decode()) except Exception: pass def test_ascii_content_prints_normally(self): """ASCII-only content should print without modification.""" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("System Prompt", "You are a helpful assistant.") self.assertIn("You are a helpful assistant.", buf.getvalue()) def test_unicode_arrows_replaced(self): """Unicode arrows (U+2192) that caused the original bug should be replaced.""" text = "Step 1 → Step 2 → Step 3" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("User Message", text) output = buf.getvalue() # Arrow chars replaced with '?' self.assertIn("Step 1 ? Step 2 ? Step 3", output) self.assertNotIn("→", output) def test_emoji_content_replaced(self): """Emoji characters should be safely replaced, not crash.""" text = "The agent failed ❌ at step 5 ✅" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("Test", text) output = buf.getvalue() self.assertIn("step 5", output) def test_cjk_content_replaced(self): """CJK characters should be replaced without exception.""" text = "用户请求:查找产品信息" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("System Prompt", text) # Should not raise self.assertIn("System Prompt:", buf.getvalue()) def test_mixed_content_partial_replacement(self): """Mixed ASCII/Unicode should preserve ASCII parts.""" text = "Error at line 42: 无效操作 → crashed" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("Msg", text) output = buf.getvalue() self.assertIn("Error at line 42", output) self.assertIn("crashed", output) def test_long_content_truncated(self): """Content longer than max_len should be truncated before encoding.""" text = "A" * 1000 buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("Long", text, max_len=500) output = buf.getvalue() # 500 A's plus label self.assertEqual(output.count("A"), 500) def test_empty_string(self): """Empty strings should not cause issues.""" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("Empty", "") self.assertIn("Empty:", buf.getvalue()) def test_null_bytes_handled(self): """Null bytes in content should not crash the encoding.""" text = "before\x00after" buf = io.StringIO() with patch('sys.stdout', buf): self._safe_print("Null", text) # Should complete without exception self.assertIn("Null:", buf.getvalue()) class TestDebugPromptsFlag(unittest.TestCase): """Test that DEBUG_PROMPTS flag controls output.""" def test_debug_prompts_disabled_no_output(self): """When DEBUG_PROMPTS is False, the guarded block should not execute.""" DEBUG_PROMPTS = False buf = io.StringIO() with patch('sys.stdout', buf): if DEBUG_PROMPTS: try: print("System Prompt:", "test".encode('ascii', 'replace').decode()) except Exception: pass self.assertEqual(buf.getvalue(), "") def test_debug_prompts_enabled_produces_output(self): """When DEBUG_PROMPTS is True, output should be produced.""" DEBUG_PROMPTS = True buf = io.StringIO() with patch('sys.stdout', buf): if DEBUG_PROMPTS: try: print("System Prompt:", "test".encode('ascii', 'replace').decode()) except Exception: pass self.assertIn("System Prompt: test", buf.getvalue()) class TestModelNameEnvVar(unittest.TestCase): """Test that AGENT_VERIFY_MODEL_NAME env var is respected.""" def test_model_name_defaults_to_gpt5(self): """Without env var, MODEL_NAME should default to gpt-5.""" with patch.dict(os.environ, {}, clear=False): os.environ.pop("AGENT_VERIFY_MODEL_NAME", None) model_name = os.environ.get("AGENT_VERIFY_MODEL_NAME", "gpt-5") self.assertEqual(model_name, "gpt-5") def test_model_name_reads_from_env(self): """With env var set, MODEL_NAME should use the provided value.""" with patch.dict(os.environ, {"AGENT_VERIFY_MODEL_NAME": "gpt-4.1"}): model_name = os.environ.get("AGENT_VERIFY_MODEL_NAME", "gpt-5") self.assertEqual(model_name, "gpt-4.1") def test_deployment_and_model_name_should_match(self): """Both DEPLOYMENT and MODEL_NAME should be set to the same value for Azure.""" with patch.dict(os.environ, { "AGENT_VERIFY_DEPLOYMENT": "gpt-4.1", "AGENT_VERIFY_MODEL_NAME": "gpt-4.1", }): deployment = os.environ.get("AGENT_VERIFY_DEPLOYMENT", "gpt-5") model_name = os.environ.get("AGENT_VERIFY_MODEL_NAME", "gpt-5") self.assertEqual(deployment, model_name, "DEPLOYMENT and MODEL_NAME must match for Azure OpenAI") if __name__ == "__main__": unittest.main()