From 2b9f79aced964dfe01f096a71c53222b89af0d94 Mon Sep 17 00:00:00 2001
From: abdul-khaliq-khalid <abdul@bugqore.com>
Date: Sat, 18 Jul 2026 11:44:39 +0530
Subject: [PATCH] detect old (v7) tar headers when routing archives

---
 lib/utils.py   | 41 +++++++++++++++++++++++++++++++++++++++--
 saferpickle.py | 14 +++++++-------
 2 files changed, 46 insertions(+), 9 deletions(-)

diff --git a/lib/utils.py b/lib/utils.py
index fffe129..3044bb9 100644
--- a/lib/utils.py
+++ b/lib/utils.py
@@ -357,10 +357,47 @@ def extract_gzip_contents(file_bytes: bytes | IO[bytes]) -> IO[bytes]:
   return cast(IO[bytes], gzip.open(stream))
 
 
+def has_tar_header(header: bytes) -> bool:
+  """Checks if a leading byte prefix looks like a tar header block.
+
+  ustar, GNU and pax archives all carry the "ustar" magic at offset 257, but
+  old (v7) archives carry no magic at all. Those are still readable by
+  `tarfile.open(mode="r:*")`, so they are identified here by verifying the
+  header checksum stored at offset 148 instead.
+
+  Args:
+    header: The leading bytes of the file. At least 512 bytes are needed to
+      recognise a v7 header.
+
+  Returns:
+    True if the prefix looks like a tar header, False otherwise.
+  """
+  if len(header) >= 262 and header[257:262] == b"ustar":
+    return True
+  if len(header) < 512:
+    return False
+
+  stored = header[148:156].split(b"\0")[0].strip()
+  if not stored:
+    return False
+  try:
+    expected = int(stored, 8)
+  except ValueError:
+    return False
+
+  # The checksum is computed with the checksum field itself read as spaces.
+  unsigned = sum(header[:148]) + (ord(" ") * 8) + sum(header[156:512])
+  signed = (
+      sum(b - 256 if b > 127 else b for b in header[:148])
+      + (ord(" ") * 8)
+      + sum(b - 256 if b > 127 else b for b in header[156:512])
+  )
+  return expected in (unsigned, signed)
+
+
 def is_tar_bytes(file_bytes: bytes | BinaryIO) -> bool:
   """Checks if the provided bytes represent a tar file."""
-  peeked = _peek_bytes(file_bytes, 262)
-  return len(peeked) >= 262 and peeked[257:262] == b"ustar"
+  return has_tar_header(_peek_bytes(file_bytes, 512))
 
 
 def extract_tar_contents(
diff --git a/saferpickle.py b/saferpickle.py
index ef29938..1755397 100644
--- a/saferpickle.py
+++ b/saferpickle.py
@@ -973,13 +973,13 @@ def security_scan(
   try:
     is_archive = False
     if not isinstance(pickle_bytes, bytes):
-      # Peek first 262 bytes to identify archive streams
+      # Peek first 512 bytes to identify archive streams
       current_pos = pickle_bytes.tell()
-      header = pickle_bytes.read(262)
+      header = pickle_bytes.read(512)
       pickle_bytes.seek(current_pos)
       if header.startswith(
           (b"PK\x03\x04", b"BZh", b"\xfd7zXZ\x00", b"\x1f\x8b")
-      ) or (len(header) >= 262 and header[257:262] == b"ustar"):
+      ) or utils.has_tar_header(header):
         is_archive = True
 
     if is_archive:
@@ -1042,7 +1042,7 @@ def security_scan(
             fail_fast=fail_fast,
             check_magic_bytes=check_magic_bytes,
         )
-      elif len(pickle_bytes) >= 262 and pickle_bytes[257:262] == b"ustar":
+      elif utils.has_tar_header(pickle_bytes):
         return _extract_and_scan_archive(
             pickle_bytes,
             "tar",
@@ -1251,7 +1251,7 @@ def _security_scan_internal(
     stream.seek(current_pos)
     is_archive = header_bytes.startswith(
         (b"PK\x03\x04", b"BZh", b"\xfd7zXZ\x00", b"\x1f\x8b")
-    ) or (len(header_bytes) >= 262 and header_bytes[257:262] == b"ustar")
+    ) or utils.has_tar_header(header_bytes)
     if not is_archive:
       start_offset = utils.find_pickle_start_offset(stream)
     else:
@@ -1415,7 +1415,7 @@ def _scan_and_load(
       header_bytes = pickle_file.read(1024)
       is_archive = header_bytes.startswith(
           (b"PK\x03\x04", b"BZh", b"\xfd7zXZ\x00", b"\x1f\x8b")
-      ) or (len(header_bytes) >= 262 and header_bytes[257:262] == b"ustar")
+      ) or utils.has_tar_header(header_bytes)
       if not is_archive:
         start_offset = utils.find_pickle_start_offset(header_bytes)
       else:
@@ -1435,7 +1435,7 @@ def _scan_and_load(
     data_bytes = pickle_file_or_bytes
     is_archive = data_bytes.startswith(
         (b"PK\x03\x04", b"BZh", b"\xfd7zXZ\x00", b"\x1f\x8b")
-    ) or (len(data_bytes) >= 262 and data_bytes[257:262] == b"ustar")
+    ) or utils.has_tar_header(data_bytes)
     if not is_archive:
       start_offset = utils.find_pickle_start_offset(data_bytes)
     else: