Supersize: Fix archive producing different results for the same input

A) Removes gzip header
B) Fixes sort key for string literals

Bug: 779895
Change-Id: Ia1c9872c42c5564134215b0973d5d9e27dcf966d
Reviewed-on: https://chromium-review.googlesource.com/746377Reviewed-by: Samuel Huang <huangs@chromium.org>
Commit-Queue: Samuel Huang <huangs@chromium.org>
Commit-Queue: agrieve <agrieve@chromium.org>
Cr-Commit-Position: refs/heads/master@{#512854}

tools/binary_size/libsupersize/archive.py

@@ -42,13 +42,11 @@ import paths
 _MAX_SAME_NAME_ALIAS_COUNT = 40  # 50kb is basically negligable.
 
 
-def _OpenMaybeGz(path, mode=None):
+def _OpenMaybeGz(path):
   """Calls `gzip.open()` if |path| ends in ".gz", otherwise calls `open()`."""
   if path.endswith('.gz'):
-    if mode and 'w' in mode:
-      return gzip.GzipFile(path, mode, 1)
-    return gzip.open(path, mode)
-  return open(path, mode or 'r')
+    return gzip.open(path, 'rb')
+  return open(path, 'rb')
 
 
 def _StripLinkerAddedSymbolPrefixes(raw_symbols):
@@ -390,7 +388,7 @@ def _CreateMergeStringsReplacements(merge_string_syms,
   logging.debug('Created %d string literal symbols', sum(len(x) for x in ret))
   logging.debug('Sorting string literals')
   for symbols in ret:
-    symbols.sort(key=lambda x: x.address)
+    symbols.sort(key=lambda x: (x.address, -x.size))
 
   logging.debug('Deduping string literals')
   num_removed = 0

tools/binary_size/libsupersize/file_format.py

@@ -6,6 +6,7 @@
 
 import cStringIO
 import calendar
+import contextlib
 import datetime
 import gzip
 import json
@@ -187,10 +188,18 @@ def _LoadSizeInfoFromFile(file_obj, size_path):
                          size_path=size_path)
 
 
+@contextlib.contextmanager
+def _OpenGzipForWrite(path):
+  # Open in a way that doesn't set any gzip header fields.
+  with open(path, 'wb') as f:
+    with gzip.GzipFile(filename='', mode='wb', fileobj=f, mtime=0) as fz:
+      yield fz
+
+
 def SaveSizeInfo(size_info, path):
   """Saves |size_info| to |path}."""
   if os.environ.get('SUPERSIZE_MEASURE_GZIP') == '1':
-    with gzip.open(path, 'wb') as f:
+    with _OpenGzipForWrite(path) as f:
       _SaveSizeInfoToFile(size_info, f)
   else:
     # It is seconds faster to do gzip in a separate step. 6s -> 3.5s.
@@ -199,7 +208,7 @@ def SaveSizeInfo(size_info, path):
 
     logging.debug('Serialization complete. Gzipping...')
     stringio.seek(0)
-    with gzip.open(path, 'wb') as f:
+    with _OpenGzipForWrite(path) as f:
       shutil.copyfileobj(stringio, f)