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   p = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
  19   out, err = p.communicate()
  20   if p.returncode != 0:
  21     raise SystemExit(out)
  22
  23
  24 def FindNode(node, component):
  25   for child in node['children']:
  26     if child['name'] == component:
  27       return child
  28   return None
  29
  30
  31 def InsertIntoTree(tree, source_name, size):
  32   components = source_name[3:].split('\\')
  33   node = tree
  34   for index, component in enumerate(components):
  35     data = FindNode(node, component)
  36     if not data:
  37       data = { 'name': source_name, 'name': component }
  38       if index == len(components) - 1:
  39         data['size'] = size
  40       else:
  41         data['children'] = []
  42       node['children'].append(data)
  43     node = data
  44
  45
  46 def FlattenTree(tree):
  47   result = [['Path', 'Parent', 'Size', 'Value']]
  48   def Flatten(node, parent):
  49     name = node['name']
  50     if parent and parent != '/':
  51       name = parent + '/' + name
  52     if 'children' in node:
  53       result.append([name, parent, -1, -1])
  54       for c in node['children']:
  55         Flatten(c, name)
  56     else:
  57       result.append([name, parent, node['size'], node['size']])
  58   Flatten(tree, '')
  59   return result
  60
  61
  62 def GetAsset(filename):
  63   with open(os.path.join(BASE_DIR, filename), 'rb') as f:
  64     return f.read()
  65
  66
  67 def AppendAsScriptBlock(f, value, var=None):
  68   f.write('<script type="text/javascript">\n')
  69   if var:
  70     f.write('var ' + var + ' = ')
  71   f.write(value)
  72   if var:
  73     f.write(';\n')
  74   f.write('</script>\n')
  75
  76
  77 def main():
  78   jsons = []
  79   if len(sys.argv) > 1:
  80     dlls = sys.argv[1:]
  81   else:
  82     out_dir = os.path.join(BASE_DIR, '..', '..', '..', 'out', 'Release')
  83     dlls = [os.path.normpath(os.path.join(out_dir, dll))
  84             for dll in ('chrome.dll', 'chrome_child.dll')]
  85   for dll_path in dlls:
  86     if os.path.exists(dll_path):
  87       print 'Tallying %s...' % dll_path
  88       json_path = dll_path + '.json'
  89       Run(os.path.join(BASE_DIR, '..', '..', '..', 'third_party', 'syzygy',
  90                        'binaries', 'exe', 'experimental', 'code_tally.exe'),
  91           '--input-image=' + dll_path,
  92           '--input-pdb=' + dll_path + '.pdb',
  93           '--output-file=' + json_path)
  94       jsons.append(json_path)
  95   if not jsons:
  96     print 'Couldn\'t find dlls.'
  97     print 'Pass fully qualified dll name(s) if you want to use something other '
  98     print 'than out\\Release\\chrome.dll and chrome_child.dll.'
  99     return 1
 100
 101   # Munge the code_tally json format into an easier-to-view format.
 102   for json_name in jsons:
 103     with open(json_name, 'r') as jsonf:
 104       all_data = json.load(jsonf)
 105     html_path = os.path.splitext(json_name)[0] + '.html'
 106     print 'Generating %s... (standlone)' % html_path
 107     by_source = {}
 108     symbols_index = {}
 109     symbols = []
 110     for obj_name, obj_data in all_data['objects'].iteritems():
 111       for symbol, symbol_data in obj_data.iteritems():
 112         size = int(symbol_data['size'])
 113         # Sometimes there's symbols with no source file, we just ignore those.
 114         if 'contribs' in symbol_data:
 115           i = 0
 116           while i < len(symbol_data['contribs']):
 117             src_index = symbol_data['contribs'][i]
 118             i += 1
 119             per_line = symbol_data['contribs'][i]
 120             i += 1
 121             source = all_data['sources'][int(src_index)]
 122             if source not in by_source:
 123               by_source[source] = {'lines': {}, 'total_size': 0}
 124             size = 0
 125             # per_line is [line, size, line, size, line, size, ...]
 126             for j in range(0, len(per_line), 2):
 127               line_number = per_line[j]
 128               size += per_line[j + 1]
 129               # Save some time/space in JS by using an array here. 0 == size,
 130               # 1 == symbol list.
 131               by_source[source]['lines'].setdefault(line_number, [0, []])
 132               by_source[source]['lines'][line_number][0] += per_line[j + 1]
 133               if symbol in symbols_index:
 134                 symindex = symbols_index[symbol]
 135               else:
 136                 symbols.append(symbol)
 137                 symbols_index[symbol] = symindex = len(symbols) - 1
 138               by_source[source]['lines'][line_number][1].append(
 139                   symindex)
 140             by_source[source]['total_size'] += size
 141     binary_name = all_data['executable']['name']
 142     data = {}
 143     data['name'] = '/'
 144     data['children'] = []
 145     file_contents = {}
 146     line_data = {}
 147     for source, file_data in by_source.iteritems():
 148       InsertIntoTree(data, source, file_data['total_size'])
 149
 150       store_as = source[3:].replace('\\', '/')
 151       try:
 152         with codecs.open(source, 'rb', encoding='latin1') as f:
 153           file_contents[store_as] = f.read()
 154       except IOError:
 155         file_contents[store_as] = '// Unable to load source.'
 156
 157       line_data[store_as] = file_data['lines']
 158       # code_tally attempts to assign fractional bytes when code is shared
 159       # across multiple symbols. Round off here for display after summing above.
 160       for per_line in line_data[store_as].values():
 161         per_line[0] = round(per_line[0])
 162
 163     flattened = FlattenTree(data)
 164     maxval = 0
 165     for i in flattened[1:]:
 166       maxval = max(i[2], maxval)
 167     flattened_str = json.dumps(flattened)
 168
 169     to_write = GetAsset('template.html')
 170     # Save all data and what would normally be external resources into the
 171     # one html so that it's a standalone report.
 172     with open(html_path, 'w') as f:
 173       f.write(to_write)
 174       # These aren't subbed in as a silly workaround for 32-bit python.
 175       # The end result is only ~100M, but while substituting these into a
 176       # template, it otherwise raises a MemoryError, I guess due to
 177       # fragmentation. So instead, we just append them as variables to the file
 178       # and then refer to the variables in the main script.
 179       filedata_str = json.dumps(file_contents).replace(
 180           '</script>', '</scr"+"ipt>')
 181       AppendAsScriptBlock(f, filedata_str, var='g_file_contents')
 182       AppendAsScriptBlock(f, json.dumps(line_data), var='g_line_data')
 183       AppendAsScriptBlock(f, json.dumps(symbols), var='g_symbol_list')
 184       favicon_str = json.dumps(base64.b64encode(GetAsset('favicon.png')))
 185       AppendAsScriptBlock(f, favicon_str, var='g_favicon')
 186       AppendAsScriptBlock(f, flattened_str, var='g_raw_data')
 187       AppendAsScriptBlock(f, str(maxval), var='g_maxval')
 188       dllname_str = binary_name + ' ' + all_data['executable']['version']
 189       AppendAsScriptBlock(f, json.dumps(dllname_str), var='g_dllname')
 190       AppendAsScriptBlock(f, GetAsset('codemirror.js'))
 191       AppendAsScriptBlock(f, GetAsset('clike.js'))
 192       AppendAsScriptBlock(f, GetAsset('main.js'))
 193       f.write('</html>')
 194
 195   return 0
 196
 197
 198 if __name__ == '__main__':
 199   sys.exit(main())