Fix: avoid mmap on empty ripgrep cache file and normalize input to rg

seagoat/sources/ripgrep.py

@@ -86,8 +86,20 @@ def rebuild(self):
             with open(self.file_path, encoding="utf-8") as cache_file:
                 self._data = cache_file.read()
         else:
+            # On POSIX systems, mmap() cannot map a zero-length file.
+            # If the cache file is empty, avoid mmap and use an empty bytes buffer instead.
+            # See Issue #945 for details.
             with open(self.file_path, "r+b") as cache_file:
-                self._data = mmap.mmap(cache_file.fileno(), 0)
+                try:
+                    cache_file.seek(0, 2)  # move to end to determine size
+                    size = cache_file.tell()
+                    cache_file.seek(0)
+                except Exception:
+                    size = 0
+                if size == 0:
+                    self._data = b""
+                else:
+                    self._data = mmap.mmap(cache_file.fileno(), 0)
 
     def encode(self, *args, **kwargs):  # type: ignore
         return self._data
@@ -128,8 +140,28 @@ def _fetch(query_text: str, path: str, limit: int, cache: RipGrepCache):
         ]
 
         try:
+            # Determine safe text input for subprocess
+            try:
+                raw = cache.encode("utf-8")  # RipGrepCache overrides encode()
+            except Exception:
+                raw = cache.as_input()
+
+            if isinstance(raw, (bytes, bytearray)):
+                cache_input_text = raw.decode("utf-8", errors="ignore")
+            elif hasattr(raw, "read"):
+                # Likely an mmap object
+                try:
+                    b = bytes(raw)
+                except Exception:
+                    b = b""
+                cache_input_text = b.decode("utf-8", errors="ignore")
+            elif isinstance(raw, str):
+                cache_input_text = raw
+            else:
+                cache_input_text = str(raw)
+
             rg_output = subprocess.check_output(
-                cmd, encoding="utf-8", input=cache.as_input()
+                cmd, encoding="utf-8", input=cache_input_text
             )
         except subprocess.CalledProcessError as exception:
             rg_output = exception.output

tests/test_issue_945.py

@@ -0,0 +1,65 @@
+from pathlib import Path
+
+import pytest
+
+from seagoat.repository import Repository
+from seagoat.sources.ripgrep import initialize
+
+
+@pytest.fixture(name="initialize_empty_source")
+def _initialize_empty_source(repo):
+    # Remove any pre-populated supported files so the ripgrep cache ends up empty
+    for filename in ["file1.md", "file2.py", "file3.py", "file4.js", "file4.md"]:
+        p = Path(repo.working_dir) / filename
+        if p.exists():
+            try:
+                p.unlink()
+            except FileNotFoundError:
+                pass
+
+    # Optionally add an unsupported file to ensure repository contains files but none supported
+    unsupported = Path(repo.working_dir) / "rock.mp3"
+    unsupported.write_text("12345", encoding="utf-8")
+
+    def _init():
+        my_repo = Repository(repo.working_dir)
+        my_repo.analyze_files()
+        source = initialize(my_repo)
+        # This should build an empty cache file and must not raise due to mmap on empty files
+        source["cache_repo"]()
+        return source["fetch"]
+
+    return _init
+
+
+def test_mmap_empty_file_issue_945(initialize_empty_source):
+    fetch = initialize_empty_source()
+    # Should not raise and should return no results for empty input cache
+    results = list(fetch("anything", limit=10))
+    assert results == []
+
+
+def test_non_empty_mapping_unchanged(repo):
+    contents = """
+hello foo bar baz
+hello foo bar baz 23
+"""
+    repo.add_file_change_commit(
+        file_name="sample.txt",
+        contents=contents,
+        author=repo.actors["John Doe"],
+        commit_message="Add sample text",
+    )
+
+    my_repo = Repository(repo.working_dir)
+    my_repo.analyze_files()
+    source = initialize(my_repo)
+    source["cache_repo"]()
+    fetch = source["fetch"]
+
+    fetched_results = list(fetch("baz|23", limit=100))
+    # Behavior for non-empty files should remain intact
+    assert len(fetched_results) == 1
+    file = fetched_results[0]
+    assert file.gitfile.path == "sample.txt"
+    assert set(file.lines) != set()  # we should have some matching lines