From e9a40a3af342ed8a53f4b010f6329bbecd06a527 Mon Sep 17 00:00:00 2001
From: imSp4rky <george+slight@slight.me>
Date: Thu, 30 Jul 2026 17:11:18 -0600
Subject: [PATCH] perf: speed up the code hash with sha1 and a plain file
 reader

---
 tests/core/test_code_hash.py |  2 +-
 unshackle/core/__init__.py   | 15 +++++++++------
 2 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/tests/core/test_code_hash.py b/tests/core/test_code_hash.py
index 9aa10aa3..68ee6d92 100644
--- a/tests/core/test_code_hash.py
+++ b/tests/core/test_code_hash.py
@@ -16,7 +16,7 @@ pytestmark = pytest.mark.unit
 
 def test_is_hex_and_stable() -> None:
     first = code_hash()
-    assert len(first) == 32
+    assert len(first) == 40
     assert int(first, 16) >= 0
     assert first == code_hash()
 
diff --git a/unshackle/core/__init__.py b/unshackle/core/__init__.py
index 5005e996..130bca20 100644
--- a/unshackle/core/__init__.py
+++ b/unshackle/core/__init__.py
@@ -16,12 +16,13 @@ def _raise(error: OSError) -> None:
 
 def code_files() -> list[str]:
     """Framework source paths relative to the package root, in a platform-stable order."""
+    pkg = str(_PKG)
     rels = ["__main__.py"]
     for name in _CODE_DIRS:
-        for root, dirs, files in os.walk(_PKG / name, onerror=_raise):
+        for root, dirs, files in os.walk(os.path.join(pkg, name), onerror=_raise):
             dirs[:] = [d for d in dirs if d != "__pycache__"]
-            rel = os.path.relpath(root, _PKG)
-            rels.extend(os.path.join(rel, f).replace(os.sep, "/") for f in files if f.endswith(".py"))
+            rel = os.path.relpath(root, pkg).replace(os.sep, "/")
+            rels.extend(f"{rel}/{f}" for f in files if f.endswith(".py"))
     return sorted(rels)
 
 
@@ -30,17 +31,19 @@ def code_hash(
     read: Optional[Callable[[str], bytes]] = None,
 ) -> str:
     """
-    md5 of the framework source. Identifies the running code, not the release it claims to be.
+    sha1 of the framework source. Two installs on the same version differ here if their code does.
 
     Pass `files` and `read` to digest a different byte source, such as blobs from git history
     (see tools/resolve_code_hash.py). Returns "" when the source cannot be read.
     """
     if read is None:
+        pkg = str(_PKG)
 
         def read(rel: str) -> bytes:
-            return (_PKG / rel).read_bytes()
+            with open(os.path.join(pkg, rel), "rb") as fh:
+                return fh.read()
 
-    digest = hashlib.md5(usedforsecurity=False)
+    digest = hashlib.sha1(usedforsecurity=False)
     try:
         for rel in code_files() if files is None else files:
             digest.update(rel.encode())