tools/win/sizeviewer/sizeviewer.py

   1 # Copyright 2013 The Chromium Authors. All rights reserved.
   2 # Use of this source code is governed by a BSD-style license that can be
   3 # found in the LICENSE file.
   4
   5 import base64
   6 import codecs
   7 import json
   8 import os
   9 import string
  10 import subprocess
  11 import sys
  12
  13
  14 BASE_DIR = os.path.dirname(os.path.abspath(__file__))
  15
  16
  17 def Run(*args):
  18   with open(os.devnull, 'w') as null:
  19     subprocess.check_call(args, stdout=null, stderr=null)
  20
  21
  22 def FindNode(node, component):
  23   for child in node['children']:
  24     if child['name'] == component:
  25       return child
  26   return None
  27
  28
  29 def InsertIntoTree(tree, source_name, size):
  30   components = source_name[3:].split('\\')
  31   node = tree
  32   for index, component in enumerate(components):
  33     data = FindNode(node, component)
  34     if not data:
  35       data = { 'name': source_name, 'name': component }
  36       if index == len(components) - 1:
  37         data['size'] = size
  38       else:
  39         data['children'] = []
  40       node['children'].append(data)
  41     node = data
  42
  43
  44 def FlattenTree(tree):
  45   result = [['Path', 'Parent', 'Size', 'Value']]
  46   def Flatten(node, parent):
  47     name = node['name']
  48     if parent and parent != '/':
  49       name = parent + '/' + name
  50     if 'children' in node:
  51       result.append([name, parent, -1, -1])
  52       for c in node['children']:
  53         Flatten(c, name)
  54     else:
  55       result.append([name, parent, node['size'], node['size']])
  56   Flatten(tree, '')
  57   return result
  58
  59
  60 def GetAsset(filename):
  61   with open(os.path.join(BASE_DIR, filename), 'rb') as f:
  62     return f.read()
  63
  64
  65 def AppendAsScriptBlock(f, value, var=None):
  66   f.write('<script type="text/javascript">\n')
  67   if var:
  68     f.write('var ' + var + ' = ')
  69   f.write(value)
  70   if var:
  71     f.write(';\n')
  72   f.write('</script>\n')
  73
  74
  75 def main():
  76   out_dir = os.path.join(BASE_DIR, '..', '..', '..', 'out', 'Release')
  77   jsons = []
  78   for dll in ('chrome.dll', 'chrome_child.dll'):
  79     dll_path = os.path.normpath(os.path.join(out_dir, dll))
  80     if os.path.exists(dll_path):
  81       print 'Tallying %s...' % dll_path
  82       json_path = dll_path + '.json'
  83       Run(os.path.join(BASE_DIR, '..', '..', '..', 'third_party', 'syzygy',
  84                        'binaries', 'exe', 'experimental', 'code_tally.exe'),
  85           '--input-image=' + dll_path,
  86           '--input-pdb=' + dll_path + '.pdb',
  87           '--output-file=' + json_path)
  88       jsons.append(json_path)
  89   if not jsons:
  90     print 'Couldn\'t find binaries, looking in', out_dir
  91     return 1
  92
  93   # Munge the code_tally json format into an easier-to-view format.
  94   for json_name in jsons:
  95     with open(json_name, 'r') as jsonf:
  96       all_data = json.load(jsonf)
  97     html_path = os.path.splitext(json_name)[0] + '.html'
  98     print 'Generating %s... (standlone)' % html_path
  99     by_source = {}
 100     symbols_index = {}
 101     symbols = []
 102     for obj_name, obj_data in all_data['objects'].iteritems():
 103       for symbol, symbol_data in obj_data.iteritems():
 104         size = int(symbol_data['size'])
 105         # Sometimes there's symbols with no source file, we just ignore those.
 106         if 'contribs' in symbol_data:
 107           i = 0
 108           while i < len(symbol_data['contribs']):
 109             src_index = symbol_data['contribs'][i]
 110             i += 1
 111             per_line = symbol_data['contribs'][i]
 112             i += 1
 113             source = all_data['sources'][int(src_index)]
 114             if source not in by_source:
 115               by_source[source] = {'lines': {}, 'total_size': 0}
 116             size = 0
 117             # per_line is [line, size, line, size, line, size, ...]
 118             for j in range(0, len(per_line), 2):
 119               line_number = per_line[j]
 120               size += per_line[j + 1]
 121               # Save some time/space in JS by using an array here. 0 == size,
 122               # 1 == symbol list.
 123               by_source[source]['lines'].setdefault(line_number, [0, []])
 124               by_source[source]['lines'][line_number][0] += per_line[j + 1]
 125               if symbol in symbols_index:
 126                 symindex = symbols_index[symbol]
 127               else:
 128                 symbols.append(symbol)
 129                 symbols_index[symbol] = symindex = len(symbols) - 1
 130               by_source[source]['lines'][line_number][1].append(
 131                   symindex)
 132             by_source[source]['total_size'] += size
 133     binary_name = all_data['executable']['name']
 134     data = {}
 135     data['name'] = '/'
 136     data['children'] = []
 137     file_contents = {}
 138     line_data = {}
 139     for source, file_data in by_source.iteritems():
 140       InsertIntoTree(data, source, file_data['total_size'])
 141
 142       store_as = source[3:].replace('\\', '/')
 143       try:
 144         with codecs.open(source, 'rb', encoding='latin1') as f:
 145           file_contents[store_as] = f.read()
 146       except IOError:
 147         file_contents[store_as] = '// Unable to load source.'
 148
 149       line_data[store_as] = file_data['lines']
 150       # code_tally attempts to assign fractional bytes when code is shared
 151       # across multiple symbols. Round off here for display after summing above.
 152       for per_line in line_data[store_as].values():
 153         per_line[0] = round(per_line[0])
 154
 155     flattened = FlattenTree(data)
 156     maxval = 0
 157     for i in flattened[1:]:
 158       maxval = max(i[2], maxval)
 159     flattened_str = json.dumps(flattened)
 160
 161     to_write = GetAsset('template.html')
 162     # Save all data and what would normally be external resources into the
 163     # one html so that it's a standalone report.
 164     with open(html_path, 'w') as f:
 165       f.write(to_write)
 166       # These aren't subbed in as a silly workaround for 32-bit python.
 167       # The end result is only ~100M, but while substituting these into a
 168       # template, it otherwise raises a MemoryError, I guess due to
 169       # fragmentation. So instead, we just append them as variables to the file
 170       # and then refer to the variables in the main script.
 171       filedata_str = json.dumps(file_contents).replace(
 172           '</script>', '</scr"+"ipt>')
 173       AppendAsScriptBlock(f, filedata_str, var='g_file_contents')
 174       AppendAsScriptBlock(f, json.dumps(line_data), var='g_line_data')
 175       AppendAsScriptBlock(f, json.dumps(symbols), var='g_symbol_list')
 176       favicon_str = json.dumps(base64.b64encode(GetAsset('favicon.png')))
 177       AppendAsScriptBlock(f, favicon_str, var='g_favicon')
 178       AppendAsScriptBlock(f, flattened_str, var='g_raw_data')
 179       AppendAsScriptBlock(f, str(maxval), var='g_maxval')
 180       dllname_str = binary_name + ' ' + all_data['executable']['version']
 181       AppendAsScriptBlock(f, json.dumps(dllname_str), var='g_dllname')
 182       AppendAsScriptBlock(f, GetAsset('codemirror.js'))
 183       AppendAsScriptBlock(f, GetAsset('clike.js'))
 184       AppendAsScriptBlock(f, GetAsset('main.js'))
 185       f.write('</html>')
 186
 187   return 0
 188
 189
 190 if __name__ == '__main__':
 191   sys.exit(main())