Change ZipInfo encoding of local extra data

A previous change in the zipfile refactor changeset defaulted the extra data to be encoded in the local file header to be empty bytes. This was because different content may appear in the local file extra data compared to the central directory extra data (different zip64 fields for instance). If opening a file from a ZipInfo instance, the extra data is initialised with data read from the central directory. On reflection, the zip64 difference is the only difference between the two encodings I know of and we can account for that by stripping and rewritting the zip64 content. Prior to this changeset the zip64 section was not stripped in the local file header which may have led to multiple zip64 sections appearing in files written after being opened with a ZipInfo instance which had zip64 data in its extra data.
python · danifus · Jul 10, 2019 · Jul 10, 2019 · Jul 10, 2019 · Jul 10, 2019
commit 0a718f7a644f50500c8bcc8a75922b5cf4a20518
diff --git a/Lib/zipfile.py b/Lib/zipfile.py
@@ -565,14 +565,16 @@ def get_local_header_params(self, zip64=False):
             compress_size = self.compress_size
             file_size = self.file_size
 
+        extra = _strip_extra(self.extra, (EXTRA_ZIP64,))
         # There are reports that windows 7 can only read zip 64 archives if the
         # zip 64 extra block is the first extra block present.
         min_version = 0
-        (extra,
+        (zip64_extra,
          file_size,
          compress_size,
          zip64_min_version,
          ) = self.zip64_local_header(zip64, file_size, compress_size)
+        extra = zip64_extra + extra
         min_version = min(min_version, zip64_min_version)
 
         if self.compress_type == ZIP_BZIP2: