fix(config_file): refuse breaklines in config dict (jxmorris12#169)

Author: mdevolde

language_tool_python/config_file.py

@@ -12,6 +12,26 @@
 logger = logging.getLogger(__name__)
 
 
+def _reject_line_breaks(field_name: str, value: str) -> None:
+    """
+    Reject values that would break the one-option-per-line config format.
+
+    :param field_name: The name of the configuration field being validated.
+    :type field_name: str
+    :param value: The value of the configuration field to validate.
+    :type value: str
+    :raises ValueError: If the value contains line break characters or ends with an odd number of backslashes.
+    """
+    if "\n" in value or "\r" in value:
+        err = f"config {field_name} cannot contain line breaks"
+        raise ValueError(err)
+
+    trailing_backslashes = len(value) - len(value.rstrip("\\"))
+    if trailing_backslashes % 2 == 1:
+        err = f"config {field_name} cannot end with an odd number of backslashes"
+        raise ValueError(err)
+
+
 @dataclass(frozen=True)
 class OptionSpec:
     """
@@ -176,14 +196,17 @@ def _encode_config(config: Dict[str, Any]) -> Dict[str, str]:
     logger.debug("Encoding LanguageTool config with keys: %s", list(config.keys()))
     encoded: Dict[str, str] = {}
     for key, value in config.items():
+        _reject_line_breaks("key", key)
         if _is_lang_key(key) and key.count("-") == 1:  # lang-<code>
             logger.debug("Encoding language option %s=%r", key, value)
             encoded[key] = str(value)
+            _reject_line_breaks(key, encoded[key])
             continue
         if _is_lang_key(key) and key.count("-") == 2:  # lang-<code>-dictPath
             logger.debug("Encoding language dictPath %s=%r", key, value)
             _path_validator(value)
             encoded[key] = _path_encoder(value)
+            _reject_line_breaks(key, encoded[key])
             continue
 
         spec = CONFIG_SCHEMA.get(key)
@@ -197,6 +220,7 @@ def _encode_config(config: Dict[str, Any]) -> Dict[str, str]:
         if spec.validator is not None:
             spec.validator(value)
         encoded[key] = spec.encoder(value)
+        _reject_line_breaks(key, encoded[key])
     return encoded
 
 

tests/test_config.py

@@ -5,6 +5,7 @@
 
 import pytest
 
+from language_tool_python.config_file import LanguageToolConfig
 from language_tool_python.exceptions import LanguageToolError
 
 
@@ -175,3 +176,37 @@ def test_disabled_rule_in_config() -> None:
         text = "He realised that the organization was in jeopardy."
         matches = tool.check(text)
         assert len(matches) == 0
+
+
+@pytest.mark.parametrize(  # type: ignore[untyped-decorator]
+    "config",
+    [
+        {"blockedReferrers": "example.com\ntrustXForwardForHeader=true"},
+        {"disabledRuleIds": ["MORFOLOGIK_RULE_EN_US", "SAFE\rrequestLimit=0"]},
+        {"lang-en\ntrustXForwardForHeader": "true"},
+        {"lang-en": "custom-word\nrequestLimit=0"},
+    ],
+)
+def test_config_rejects_line_break_injection(config: dict[str, object]) -> None:
+    """
+    Test that config serialization cannot be escaped with CR/LF characters.
+    """
+    with pytest.raises(ValueError, match="cannot contain line breaks"):
+        LanguageToolConfig(config)
+
+
+@pytest.mark.parametrize(  # type: ignore[untyped-decorator]
+    "config",
+    [
+        {"blockedReferrers": "example.com\\"},
+        {"disabledRuleIds": ["MORFOLOGIK_RULE_EN_US", "SAFE\\"]},
+        {"lang-en\\": "true"},
+        {"lang-en": "custom-word\\"},
+    ],
+)
+def test_config_rejects_odd_trailing_backslashes(config: dict[str, object]) -> None:
+    """
+    Test that config serialization cannot escape the line ending with a backslash.
+    """
+    with pytest.raises(ValueError, match="odd number of backslashes"):
+        LanguageToolConfig(config)