Merge changes from topic "cherrypicker-L80200000962011019:N58900001388333251" into udc-dev

* changes:
  Handle symlinks when extracting zipfiles
  Handle zip64 extra fields better
  Fix python3.11's support for zip64
  Use python based unzip function for portability

tools/releasetools/check_target_files_signatures.py

@@ -241,7 +241,8 @@ class APK(object):
     # Signer (minSdkVersion=24, maxSdkVersion=32) certificate SHA-1 digest: 19da94896ce4078c38ca695701f1dec741ec6d67
     # ...
     certs_info = {}
-    certificate_regex = re.compile(r"(Signer (?:#[0-9]+|\(.*\))) (certificate .*):(.*)")
+    certificate_regex = re.compile(
+        r"(Signer (?:#[0-9]+|\(.*\))) (certificate .*):(.*)")
     for line in output.splitlines():
       m = certificate_regex.match(line)
       if not m:
@@ -312,7 +313,7 @@ class TargetFiles(object):
     # This is the list of wildcards of files we extract from |filename|.
     apk_extensions = ['*.apk', '*.apex']
 
-    with zipfile.ZipFile(filename) as input_zip:
+    with zipfile.ZipFile(filename, "r") as input_zip:
       self.certmap, compressed_extension = common.ReadApkCerts(input_zip)
     if compressed_extension:
       apk_extensions.append('*.apk' + compressed_extension)

tools/releasetools/common.py

@@ -35,6 +35,7 @@ import shlex
 import shutil
 import subprocess
 import sys
+import stat
 import tempfile
 import threading
 import time
@@ -2102,6 +2103,26 @@ def Gunzip(in_filename, out_filename):
     shutil.copyfileobj(in_file, out_file)
 
 
+def UnzipSingleFile(input_zip: zipfile.ZipFile, info: zipfile.ZipInfo, dirname: str):
+  # According to https://stackoverflow.com/questions/434641/how-do-i-set-permissions-attributes-on-a-file-in-a-zip-file-using-pythons-zip/6297838#6297838
+  # higher bits of |external_attr| are unix file permission and types
+  unix_filetype = info.external_attr >> 16
+
+  def CheckMask(a, mask):
+    return (a & mask) == mask
+
+  def IsSymlink(a):
+    return CheckMask(a, stat.S_IFLNK)
+  # python3.11 zipfile implementation doesn't handle symlink correctly
+  if not IsSymlink(unix_filetype):
+    return input_zip.extract(info, dirname)
+  if dirname is None:
+    dirname = os.getcwd()
+  target = os.path.join(dirname, info.filename)
+  os.makedirs(os.path.dirname(target), exist_ok=True)
+  os.symlink(input_zip.read(info).decode(), target)
+
+
 def UnzipToDir(filename, dirname, patterns=None):
   """Unzips the archive to the given directory.
 
@@ -2112,20 +2133,46 @@ def UnzipToDir(filename, dirname, patterns=None):
         archvie. Non-matching patterns will be filtered out. If there's no match
         after the filtering, no file will be unzipped.
   """
-  cmd = ["unzip", "-o", "-q", filename, "-d", dirname]
-  if patterns is not None:
+  with zipfile.ZipFile(filename, allowZip64=True, mode="r") as input_zip:
     # Filter out non-matching patterns. unzip will complain otherwise.
-    with zipfile.ZipFile(filename, allowZip64=True) as input_zip:
-      names = input_zip.namelist()
-    filtered = [
-        pattern for pattern in patterns if fnmatch.filter(names, pattern)]
+    entries = input_zip.infolist()
+    # b/283033491
+    # Per https://en.wikipedia.org/wiki/ZIP_(file_format)#Central_directory_file_header
+    # In zip64 mode, central directory record's header_offset field might be
+    # set to 0xFFFFFFFF if header offset is > 2^32. In this case, the extra
+    # fields will contain an 8 byte little endian integer at offset 20
+    # to indicate the actual local header offset.
+    # As of python3.11, python does not handle zip64 central directories
+    # correctly, so we will manually do the parsing here.
 
-    # There isn't any matching files. Don't unzip anything.
-    if not filtered:
-      return
-    cmd.extend(filtered)
+    # ZIP64 central directory extra field has two required fields:
+    # 2 bytes header ID and 2 bytes size field. Thes two require fields have
+    # a total size of 4 bytes. Then it has three other 8 bytes field, followed
+    # by a 4 byte disk number field. The last disk number field is not required
+    # to be present, but if it is present, the total size of extra field will be
+    # divisible by 8(because 2+2+4+8*n is always going to be multiple of 8)
+    # Most extra fields are optional, but when they appear, their must appear
+    # in the order defined by zip64 spec. Since file header offset is the 2nd
+    # to last field in zip64 spec, it will only be at last 8 bytes or last 12-4
+    # bytes, depending on whether disk number is present.
+    for entry in entries:
+      if entry.header_offset == 0xFFFFFFFF:
+        if len(entry.extra) % 8 == 0:
+          entry.header_offset = int.from_bytes(entry.extra[-12:-4], "little")
+        else:
+          entry.header_offset = int.from_bytes(entry.extra[-8:], "little")
+    if patterns is not None:
+      filtered = [info for info in entries if any(
+          [fnmatch.fnmatch(info.filename, p) for p in patterns])]
 
-  RunAndCheckOutput(cmd)
+      # There isn't any matching files. Don't unzip anything.
+      if not filtered:
+        return
+      for info in filtered:
+        UnzipSingleFile(input_zip, info, dirname)
+    else:
+      for info in entries:
+        UnzipSingleFile(input_zip, info, dirname)
 
 
 def UnzipTemp(filename, patterns=None):