Handle shared memory symbols better in the binarysize tool.

tools/binary_size/explain_binary_size_delta.py

Index: tools/binary_size/explain_binary_size_delta.py
diff --git a/tools/binary_size/explain_binary_size_delta.py b/tools/binary_size/explain_binary_size_delta.py
index cb99fe6cc93981b3bec3ecff66e8ea0cb1a9fa07..75b6aea17f3bf4758a563be01d152c0ddbb216e6 100755
--- a/tools/binary_size/explain_binary_size_delta.py
+++ b/tools/binary_size/explain_binary_size_delta.py
@@ -38,6 +38,8 @@ dumps. Example:
 """
 
 import collections
+from collections import Counter
+from math import ceil
 import operator
 import optparse
 import os
@@ -46,6 +48,19 @@ import sys
 import binary_size_utils
 
 
+def CalculateSharedAddresses(symbols):
+  """Checks how many symbols share the same memory space."""

[Andrew Hayden (chromium.org), 2015/02/16 12:47:38]

Please document the return type.

[Daniel Bratell, 2015/02/18 13:31:53]

Done.

+  count = Counter()
+  for _, _, _, _, address in symbols:
+    count[address] += 1
+
+  # metacount = Counter()

[Andrew Hayden (chromium.org), 2015/02/16 12:47:38]

Please remove the commented lines 57-61 or uncomment and hide them behind a
debug/verbose flag.

[Daniel Bratell, 2015/02/18 13:31:53]

Done.

+  # for share_count in count.itervalues():
+  #   metacount[share_count] += 1
+  # for (key, value) in metacount.items():
+  #   print("%d symbols were shared %d times" % ((value * key), key))
+  return count
+
 def Compare(symbols1, symbols2):
   """Executes a comparison of the symbols in symbols1 and symbols2.
 
@@ -59,19 +74,28 @@ def Compare(symbols1, symbols2):
 
   cache1 = {}
   cache2 = {}
-  # Make a map of (file, symbol_type) : (symbol_name, symbol_size)
-  for cache, symbols in ((cache1, symbols1), (cache2, symbols2)):
-    for symbol_name, symbol_type, symbol_size, file_path in symbols:
+  # Make a map of (file, symbol_type) : (symbol_name, effective_symbol_size)
+  share_count1 = CalculateSharedAddresses(symbols1)
+  share_count2 = CalculateSharedAddresses(symbols2)
+  for cache, symbols, share_count in ((cache1, symbols1, share_count1),
+                                      (cache2, symbols2, share_count2)):
+    for symbol_name, symbol_type, symbol_size, file_path, address in symbols:
       if 'vtable for ' in symbol_name:
         symbol_type = '@' # hack to categorize these separately
       if file_path:
         file_path = os.path.normpath(file_path)
       else:
         file_path = '(No Path)'
+      shared_count = share_count[address]
+      if shared_count == 1:
+        effective_symbol_size = symbol_size
+      else:
+        assert shared_count > 1
+        effective_symbol_size = int(ceil(symbol_size / float(shared_count)))

[Andrew Hayden (chromium.org), 2015/02/16 12:47:38]

For clarity, I'd prefer that we turn this line into a function like
"CalculateEffectiveSize" or something that we can potentially tweak (or at least
serves as a good way to document what we're trying to do here - there could be
other, better strategies that aren't just an equal distribution).

[Daniel Bratell, 2015/02/18 13:31:53]

Done.

       key = (file_path, symbol_type)
       bucket = cache.setdefault(key, {})
       size_list = bucket.setdefault(symbol_name, [])
-      size_list.append(symbol_size)
+      size_list.append(effective_symbol_size)
 
   # Now diff them. We iterate over the elements in cache1. For each symbol
   # that we find in cache2, we record whether it was deleted, changed, or
@@ -177,6 +201,7 @@ def CrunchStats(added, removed, changed, unchanged, showsources, showsymbols):
   sections = [new_symbols, removed_symbols, grown_symbols, shrunk_symbols]
   for section in sections:
     for file_path, symbol_type, symbol_name, size1, size2 in section.symbols:
+      assert not '&' in file_path, "File path: " + file_path

[Andrew Hayden (chromium.org), 2015/02/16 12:47:38]

Is there some context for this being added?

[Daniel Bratell, 2015/02/18 13:31:53]

I really can't remember. Probably remains from some debugging. I will remove it.

       section.sources.add(file_path)
       if size1 is not None:
         section.before_size += size1