supersize: A few fixes for Android+clang builds

tools/binary_size/libsupersize/nm.py

 # Copyright 2017 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
 """Functions that rely on parsing output of "nm" tool."""
 
 import atexit
 import collections
 import errno
 import logging
 import os
 import subprocess
 import sys
 
 import concurrent
 
 _active_subprocesses = None
 
 
+def _IsRelevantNmName(name):
+  # Skip lines like:
+  # 00000000 t $t
+  # 00000000 r $d
+  # 0000041b r .L.str.38
+  # 00000344 N
+  return name and not name.startswith('.L.str.') and not (
+      len(name) == 2 and name.startswith('$'))
+
+
 def CollectAliasesByAddress(elf_path, tool_prefix):
   """Runs nm on |elf_path| and returns a dict of address->[names]"""
   names_by_address = collections.defaultdict(list)
 
   # About 60mb of output, but piping takes ~30s, and loading it into RAM
   # directly takes 3s.
   args = [tool_prefix + 'nm', '--no-sort', '--defined-only', '--demangle',
           elf_path]
   output = subprocess.check_output(args)
   for line in output.splitlines():
-    address_str, section, name = line.split(' ', 2)
+    space_idx = line.find(' ')
+    address_str = line[:space_idx]
+    section = line[space_idx + 1]
+    name = line[space_idx + 3:]
+
     # To verify that rodata does not have aliases:
     #   nm --no-sort --defined-only libchrome.so > nm.out
     #   grep -v '\$' nm.out | grep ' r ' | sort | cut -d' ' -f1 > addrs
     #   wc -l < addrs; uniq < addrs | wc -l
-    if section not in 'tT' or not name or name[0] == '$':
+    if section not in 'tT' or not _IsRelevantNmName(name):
       continue
 
     address = int(address_str, 16)
     if not address:
       continue
     # Constructors often show up twice.
     name_list = names_by_address[address]
     if name not in name_list:
       name_list.append(name)
 
@@ skipping 18 matching lines @@
   return concurrent.ForkAndCall(
       _CollectAliasesByAddressAsyncHelper, (elf_path, tool_prefix),
       decode_func=decode)
 
 
 def _ParseOneObjectFileOutput(lines):
   ret = []
   for line in lines:
     if not line:
       break
-    sep = line.find(' ')  # Skip over address.
-    sep = line.find(' ', sep + 1)  # Skip over symbol type.
-    name = line[sep + 1:]
-    # Skip lines like:
-    # 00000000 t $t
-    # 00000000 r $d
-    # 0000041b r .L.str.38
-    if name[0] not in '$.':
+    space_idx = line.find(' ')  # Skip over address.
+    name = line[space_idx + 3:]
+    if _IsRelevantNmName(name):
       ret.append(name)
   return ret
 
 
 def _BatchCollectNames(target, tool_prefix, output_directory):
   is_archive = isinstance(target, basestring)
   # Ensure tool_prefix is absolute so that CWD does not affect it
   if os.path.sep in tool_prefix:
     # Use abspath() on the dirname to avoid it stripping a trailing /.
     dirname = os.path.dirname(tool_prefix)
@@ skipping 157 matching lines @@
   except IOError, e:
     # Parent process exited.
     if e.errno == errno.EPIPE:
       sys.exit(1)
 
   logging.debug('nm bulk subprocess finished.')
 
 
 if __name__ == '__main__':
   _SubMain(*sys.argv[1:])