Merge pull request #1922 from DaveLak/fix-fuzz-submodule-expected-exceptions-handling

Add graceful handling of expected exceptions in fuzz_submodule.py

Author: Byron

fuzzing/fuzz-targets/fuzz_submodule.py

@@ -3,14 +3,23 @@
 import os
 import tempfile
 from configparser import ParsingError
-from utils import is_expected_exception_message
+from utils import is_expected_exception_message, get_max_filename_length
+from git import Repo, GitCommandError, InvalidGitRepositoryError
 
-if getattr(sys, "frozen", False) and hasattr(sys, "_MEIPASS"):
+if getattr(sys, "frozen", False) and hasattr(sys, "_MEIPASS"):  # pragma: no cover
     path_to_bundled_git_binary = os.path.abspath(os.path.join(os.path.dirname(__file__), "git"))
     os.environ["GIT_PYTHON_GIT_EXECUTABLE"] = path_to_bundled_git_binary
 
-with atheris.instrument_imports():
-    from git import Repo, GitCommandError, InvalidGitRepositoryError
+if not sys.warnoptions:  # pragma: no cover
+    # The warnings filter below can be overridden by passing the -W option
+    # to the Python interpreter command line or setting the `PYTHONWARNINGS` environment variable.
+    import warnings
+    import logging
+
+    # Fuzzing data causes some modules to generate a large number of warnings
+    # which are not usually interesting and make the test output hard to read, so we ignore them.
+    warnings.simplefilter("ignore")
+    logging.getLogger().setLevel(logging.ERROR)
 
 
 def TestOneInput(data):
@@ -42,12 +51,12 @@ def TestOneInput(data):
                     writer.release()
 
                 submodule.update(init=fdp.ConsumeBool(), dry_run=fdp.ConsumeBool(), force=fdp.ConsumeBool())
-
                 submodule_repo = submodule.module()
-                new_file_path = os.path.join(
-                    submodule_repo.working_tree_dir,
-                    f"new_file_{fdp.ConsumeUnicodeNoSurrogates(fdp.ConsumeIntInRange(1, 512))}",
+
+                new_file_name = fdp.ConsumeUnicodeNoSurrogates(
+                    fdp.ConsumeIntInRange(1, max(1, get_max_filename_length(submodule_repo.working_tree_dir)))
                 )
+                new_file_path = os.path.join(submodule_repo.working_tree_dir, new_file_name)
                 with open(new_file_path, "wb") as new_file:
                     new_file.write(fdp.ConsumeBytes(fdp.ConsumeIntInRange(1, 512)))
                 submodule_repo.index.add([new_file_path])
@@ -67,16 +76,24 @@ def TestOneInput(data):
                 )
                 repo.index.commit(f"Removed submodule {submodule_name}")
 
-        except (ParsingError, GitCommandError, InvalidGitRepositoryError, FileNotFoundError, BrokenPipeError):
+        except (
+            ParsingError,
+            GitCommandError,
+            InvalidGitRepositoryError,
+            FileNotFoundError,
+            FileExistsError,
+            IsADirectoryError,
+            NotADirectoryError,
+            BrokenPipeError,
+        ):
             return -1
-        except (ValueError, OSError) as e:
+        except ValueError as e:
             expected_messages = [
                 "SHA is empty",
                 "Reference at",
                 "embedded null byte",
                 "This submodule instance does not exist anymore",
                 "cmd stdin was empty",
-                "File name too long",
             ]
             if is_expected_exception_message(e, expected_messages):
                 return -1
@@ -85,6 +102,7 @@ def TestOneInput(data):
 
 
 def main():
+    atheris.instrument_all()
     atheris.Setup(sys.argv, TestOneInput)
     atheris.Fuzz()
 

fuzzing/fuzz-targets/utils.py

@@ -1,4 +1,5 @@
 import atheris  # pragma: no cover
+import os  # pragma: no cover
 from typing import List  # pragma: no cover
 
 
@@ -20,3 +21,17 @@ def is_expected_exception_message(exception: Exception, error_message_list: List
         if error.lower() in exception_message:
             return True
     return False
+
+
+@atheris.instrument_func
+def get_max_filename_length(path: str) -> int:  # pragma: no cover
+    """
+    Get the maximum filename length for the filesystem containing the given path.
+
+    Args:
+        path (str): The path to check the filesystem for.
+
+    Returns:
+        int: The maximum filename length.
+    """
+    return os.pathconf(path, "PC_NAME_MAX")