| OLD | NEW |
|---|
| (Empty) |
| 1 # Copyright 2017 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 """Classes that comprise the data model for binary size analysis. | |
| 5 | |
| 6 The primary classes are Symbol, and SymbolGroup. | |
| 7 | |
| 8 Description of common properties: | |
| 9 * address: The start address of the symbol. | |
| 10 May be 0 (e.g. for .bss or for SymbolGroups). | |
| 11 * size: The number of bytes this symbol takes up, including padding that comes | |
| 12 before |address|. | |
| 13 * padding: The number of bytes of padding before |address| due to this symbol. | |
| 14 * name: Symbol names with parameter list removed. | |
| 15 Never None, but will be '' for anonymous symbols. | |
| 16 * full_name: Symbols names with parameter list left in. | |
| 17 Never None, but will be '' for anonymous symbols, and for symbols that do | |
| 18 not contain a parameter list. | |
| 19 * is_anonymous: True when the symbol exists in an anonymous namespace (which | |
| 20 are removed from both full_name and name during normalization). | |
| 21 * section_name: E.g. ".text", ".rodata", ".data.rel.local" | |
| 22 * section: The second character of |section_name|. E.g. "t", "r", "d". | |
| 23 """ | |
| 24 | |
| 25 import collections | |
| 26 import copy | |
| 27 import os | |
| 28 import re | |
| 29 | |
| 30 import match_util | |
| 31 | |
| 32 | |
| 33 METADATA_GIT_REVISION = 'git_revision' | |
| 34 METADATA_MAP_FILENAME = 'map_file_name' # Path relative to output_directory. | |
| 35 METADATA_ELF_FILENAME = 'elf_file_name' # Path relative to output_directory. | |
| 36 METADATA_ELF_MTIME = 'elf_mtime' # int timestamp in utc. | |
| 37 METADATA_ELF_BUILD_ID = 'elf_build_id' | |
| 38 METADATA_GN_ARGS = 'gn_args' | |
| 39 | |
| 40 | |
| 41 SECTION_TO_SECTION_NAME = { | |
| 42 'b': '.bss', | |
| 43 'd': '.data', | |
| 44 'r': '.rodata', | |
| 45 't': '.text', | |
| 46 } | |
| 47 | |
| 48 | |
| 49 class SizeInfo(object): | |
| 50 """Represents all size information for a single binary. | |
| 51 | |
| 52 Fields: | |
| 53 section_sizes: A dict of section_name -> size. | |
| 54 raw_symbols: A flat list of all symbols. | |
| 55 symbols: A SymbolGroup containing raw_symbols, but with some Symbols grouped | |
| 56 into sub-SymbolGroups. | |
| 57 metadata: A dict. | |
| 58 """ | |
| 59 __slots__ = ( | |
| 60 'section_sizes', | |
| 61 'raw_symbols', | |
| 62 'symbols', | |
| 63 'metadata', | |
| 64 ) | |
| 65 | |
| 66 """Root size information.""" | |
| 67 def __init__(self, section_sizes, raw_symbols, grouped_symbols=None, | |
| 68 metadata=None): | |
| 69 self.section_sizes = section_sizes # E.g. {'.text': 0} | |
| 70 # List of symbols sorted by address per-section. | |
| 71 self.raw_symbols = raw_symbols | |
| 72 # Root SymbolGroup. Cloned symbols grouped together within sub-SymbolGroups. | |
| 73 self.symbols = grouped_symbols | |
| 74 self.metadata = metadata or {} | |
| 75 | |
| 76 | |
| 77 class SizeInfoDiff(object): | |
| 78 """What you get when you Diff() two SizeInfo objects. | |
| 79 | |
| 80 Fields: | |
| 81 section_sizes: A dict of section_name -> size delta. | |
| 82 symbols: A SymbolDiff with all symbols in it. | |
| 83 old_metadata: metadata of the "old" SizeInfo. | |
| 84 new_metadata: metadata of the "new" SizeInfo. | |
| 85 """ | |
| 86 __slots__ = ( | |
| 87 'section_sizes', | |
| 88 'symbols', | |
| 89 'old_metadata', | |
| 90 'new_metadata', | |
| 91 ) | |
| 92 | |
| 93 def __init__(self, section_sizes, symbols, old_metadata, new_metadata): | |
| 94 self.section_sizes = section_sizes | |
| 95 self.symbols = symbols | |
| 96 self.old_metadata = old_metadata | |
| 97 self.new_metadata = new_metadata | |
| 98 | |
| 99 | |
| 100 class BaseSymbol(object): | |
| 101 """Base class for Symbol and SymbolGroup. | |
| 102 | |
| 103 Refer to module docs for field descriptions. | |
| 104 """ | |
| 105 __slots__ = () | |
| 106 | |
| 107 @property | |
| 108 def section(self): | |
| 109 """Returns the one-letter section. | |
| 110 | |
| 111 E.g. If section_name == '.rodata', then section == 'r'. | |
| 112 """ | |
| 113 return self.section_name[1] | |
| 114 | |
| 115 @property | |
| 116 def size_without_padding(self): | |
| 117 return self.size - self.padding | |
| 118 | |
| 119 @property | |
| 120 def end_address(self): | |
| 121 return self.address + self.size_without_padding | |
| 122 | |
| 123 def IsBss(self): | |
| 124 return self.section_name == '.bss' | |
| 125 | |
| 126 def IsGroup(self): | |
| 127 return False | |
| 128 | |
| 129 def IsGenerated(self): | |
| 130 # TODO(agrieve): Also match generated functions such as: | |
| 131 # startup._GLOBAL__sub_I_page_allocator.cc | |
| 132 return self.name.endswith(']') and not self.name.endswith('[]') | |
| 133 | |
| 134 def _Key(self): | |
| 135 """Returns a tuple that can be used to see if two Symbol are the same. | |
| 136 | |
| 137 Keys are not guaranteed to be unique within a SymbolGroup. For example, it | |
| 138 is common to have multiple "** merge strings" symbols, which will have a | |
| 139 common key.""" | |
| 140 stripped_full_name = self.full_name | |
| 141 if stripped_full_name: | |
| 142 clone_idx = stripped_full_name.find(' [clone ') | |
| 143 if clone_idx != -1: | |
| 144 stripped_full_name = stripped_full_name[:clone_idx] | |
| 145 return (self.section_name, stripped_full_name or self.name) | |
| 146 | |
| 147 | |
| 148 class Symbol(BaseSymbol): | |
| 149 """Represents a single symbol within a binary. | |
| 150 | |
| 151 Refer to module docs for field descriptions. | |
| 152 """ | |
| 153 | |
| 154 __slots__ = ( | |
| 155 'address', | |
| 156 'full_name', | |
| 157 'is_anonymous', | |
| 158 'object_path', | |
| 159 'name', | |
| 160 'padding', | |
| 161 'section_name', | |
| 162 'source_path', | |
| 163 'size', | |
| 164 ) | |
| 165 | |
| 166 def __init__(self, section_name, size_without_padding, address=None, | |
| 167 name=None, source_path=None, object_path=None, | |
| 168 full_name=None, is_anonymous=False): | |
| 169 self.section_name = section_name | |
| 170 self.address = address or 0 | |
| 171 self.name = name or '' | |
| 172 self.full_name = full_name or '' | |
| 173 self.source_path = source_path or '' | |
| 174 self.object_path = object_path or '' | |
| 175 self.size = size_without_padding | |
| 176 # Change this to be a bitfield of flags if ever there is a need to add | |
| 177 # another similar thing. | |
| 178 self.is_anonymous = is_anonymous | |
| 179 self.padding = 0 | |
| 180 | |
| 181 def __repr__(self): | |
| 182 return ('%s@%x(size_without_padding=%d,padding=%d,name=%s,path=%s,anon=%d)' | |
| 183 % (self.section_name, self.address, self.size_without_padding, | |
| 184 self.padding, self.name, self.source_path or self.object_path, | |
| 185 int(self.is_anonymous))) | |
| 186 | |
| 187 | |
| 188 class SymbolGroup(BaseSymbol): | |
| 189 """Represents a group of symbols using the same interface as Symbol. | |
| 190 | |
| 191 SymbolGroups are immutable. All filtering / sorting will return new | |
| 192 SymbolGroups objects. | |
| 193 | |
| 194 Overrides many __functions__. E.g. the following are all valid: | |
| 195 * len(group) | |
| 196 * iter(group) | |
| 197 * group[0] | |
| 198 * group['0x1234'] # By symbol address | |
| 199 * without_group2 = group1 - group2 | |
| 200 * unioned = group1 + group2 | |
| 201 """ | |
| 202 | |
| 203 __slots__ = ( | |
| 204 '_padding', | |
| 205 '_size', | |
| 206 '_symbols', | |
| 207 '_filtered_symbols', | |
| 208 'full_name', | |
| 209 'name', | |
| 210 'section_name', | |
| 211 'is_sorted', | |
| 212 ) | |
| 213 | |
| 214 def __init__(self, symbols, filtered_symbols=None, name=None, | |
| 215 full_name=None, section_name=None, is_sorted=False): | |
| 216 self._padding = None | |
| 217 self._size = None | |
| 218 self._symbols = symbols | |
| 219 self._filtered_symbols = filtered_symbols or [] | |
| 220 self.name = name or '' | |
| 221 self.full_name = full_name | |
| 222 self.section_name = section_name or '.*' | |
| 223 self.is_sorted = is_sorted | |
| 224 | |
| 225 def __repr__(self): | |
| 226 return 'Group(name=%s,count=%d,size=%d)' % ( | |
| 227 self.name, len(self), self.size) | |
| 228 | |
| 229 def __iter__(self): | |
| 230 return iter(self._symbols) | |
| 231 | |
| 232 def __len__(self): | |
| 233 return len(self._symbols) | |
| 234 | |
| 235 def __eq__(self, other): | |
| 236 return self._symbols == other._symbols | |
| 237 | |
| 238 def __getitem__(self, key): | |
| 239 """|key| can be an index or an address. | |
| 240 | |
| 241 Raises if multiple symbols map to the address. | |
| 242 """ | |
| 243 if isinstance(key, slice): | |
| 244 return self._symbols.__getitem__(key) | |
| 245 if isinstance(key, basestring) or key > len(self._symbols): | |
| 246 found = self.WhereAddressInRange(key) | |
| 247 if len(found) != 1: | |
| 248 raise KeyError('%d symbols found at address %s.' % (len(found), key)) | |
| 249 return found[0] | |
| 250 return self._symbols[key] | |
| 251 | |
| 252 def __sub__(self, other): | |
| 253 other_ids = set(id(s) for s in other) | |
| 254 new_symbols = [s for s in self if id(s) not in other_ids] | |
| 255 return self._CreateTransformed(new_symbols, section_name=self.section_name) | |
| 256 | |
| 257 def __add__(self, other): | |
| 258 self_ids = set(id(s) for s in self) | |
| 259 new_symbols = self._symbols + [s for s in other if id(s) not in self_ids] | |
| 260 return self._CreateTransformed(new_symbols, section_name=self.section_name, | |
| 261 is_sorted=False) | |
| 262 | |
| 263 @property | |
| 264 def address(self): | |
| 265 first = self._symbols[0].address | |
| 266 return first if all(s.address == first for s in self._symbols) else 0 | |
| 267 | |
| 268 @property | |
| 269 def is_anonymous(self): | |
| 270 first = self._symbols[0].is_anonymous | |
| 271 return first if all( | |
| 272 s.is_anonymous == first for s in self._symbols) else False | |
| 273 | |
| 274 @property | |
| 275 def object_path(self): | |
| 276 first = self._symbols[0].object_path | |
| 277 return first if all(s.object_path == first for s in self._symbols) else None | |
| 278 | |
| 279 @property | |
| 280 def source_path(self): | |
| 281 first = self._symbols[0].source_path | |
| 282 return first if all(s.source_path == first for s in self._symbols) else None | |
| 283 | |
| 284 @property | |
| 285 def size(self): | |
| 286 if self._size is None: | |
| 287 if self.IsBss(): | |
| 288 self._size = sum(s.size for s in self) | |
| 289 self._size = sum(s.size for s in self if not s.IsBss()) | |
| 290 return self._size | |
| 291 | |
| 292 @property | |
| 293 def padding(self): | |
| 294 if self._padding is None: | |
| 295 self._padding = sum(s.padding for s in self) | |
| 296 return self._padding | |
| 297 | |
| 298 def IsGroup(self): | |
| 299 return True | |
| 300 | |
| 301 def _CreateTransformed(self, symbols, filtered_symbols=None, name=None, | |
| 302 section_name=None, is_sorted=None): | |
| 303 if is_sorted is None: | |
| 304 is_sorted = self.is_sorted | |
| 305 return SymbolGroup(symbols, filtered_symbols=filtered_symbols, name=name, | |
| 306 section_name=section_name, is_sorted=is_sorted) | |
| 307 | |
| 308 def Sorted(self, cmp_func=None, key=None, reverse=False): | |
| 309 # Default to sorting by abs(size) then name. | |
| 310 if cmp_func is None and key is None: | |
| 311 cmp_func = lambda a, b: cmp((a.IsBss(), abs(b.size), a.name), | |
| 312 (b.IsBss(), abs(a.size), b.name)) | |
| 313 | |
| 314 new_symbols = sorted(self._symbols, cmp_func, key, reverse) | |
| 315 return self._CreateTransformed( | |
| 316 new_symbols, filtered_symbols=self._filtered_symbols, | |
| 317 section_name=self.section_name, is_sorted=True) | |
| 318 | |
| 319 def SortedByName(self, reverse=False): | |
| 320 return self.Sorted(key=(lambda s:s.name), reverse=reverse) | |
| 321 | |
| 322 def SortedByAddress(self, reverse=False): | |
| 323 return self.Sorted(key=(lambda s:s.address), reverse=reverse) | |
| 324 | |
| 325 def SortedByCount(self, reverse=False): | |
| 326 return self.Sorted(key=(lambda s:len(s) if s.IsGroup() else 1), | |
| 327 reverse=not reverse) | |
| 328 | |
| 329 def Filter(self, func): | |
| 330 filtered_and_kept = ([], []) | |
| 331 for symbol in self: | |
| 332 filtered_and_kept[int(bool(func(symbol)))].append(symbol) | |
| 333 return self._CreateTransformed(filtered_and_kept[1], | |
| 334 filtered_symbols=filtered_and_kept[0], | |
| 335 section_name=self.section_name) | |
| 336 | |
| 337 def WhereBiggerThan(self, min_size): | |
| 338 return self.Filter(lambda s: s.size >= min_size) | |
| 339 | |
| 340 def WhereInSection(self, section): | |
| 341 if len(section) == 1: | |
| 342 ret = self.Filter(lambda s: s.section == section) | |
| 343 ret.section_name = SECTION_TO_SECTION_NAME[section] | |
| 344 else: | |
| 345 ret = self.Filter(lambda s: s.section_name == section) | |
| 346 ret.section_name = section | |
| 347 return ret | |
| 348 | |
| 349 def WhereIsGenerated(self): | |
| 350 return self.Filter(lambda s: s.IsGenerated()) | |
| 351 | |
| 352 def WhereNameMatches(self, pattern): | |
| 353 regex = re.compile(match_util.ExpandRegexIdentifierPlaceholder(pattern)) | |
| 354 return self.Filter(lambda s: regex.search(s.name)) | |
| 355 | |
| 356 def WhereObjectPathMatches(self, pattern): | |
| 357 regex = re.compile(match_util.ExpandRegexIdentifierPlaceholder(pattern)) | |
| 358 return self.Filter(lambda s: regex.search(s.object_path)) | |
| 359 | |
| 360 def WhereSourcePathMatches(self, pattern): | |
| 361 regex = re.compile(match_util.ExpandRegexIdentifierPlaceholder(pattern)) | |
| 362 return self.Filter(lambda s: regex.search(s.source_path)) | |
| 363 | |
| 364 def WherePathMatches(self, pattern): | |
| 365 regex = re.compile(match_util.ExpandRegexIdentifierPlaceholder(pattern)) | |
| 366 return self.Filter(lambda s: (regex.search(s.source_path) or | |
| 367 regex.search(s.object_path))) | |
| 368 | |
| 369 def WhereMatches(self, pattern): | |
| 370 """Looks for |pattern| within all paths & names.""" | |
| 371 regex = re.compile(match_util.ExpandRegexIdentifierPlaceholder(pattern)) | |
| 372 return self.Filter(lambda s: (regex.search(s.source_path) or | |
| 373 regex.search(s.object_path) or | |
| 374 regex.search(s.full_name or '') or | |
| 375 regex.search(s.name))) | |
| 376 | |
| 377 def WhereAddressInRange(self, start, end=None): | |
| 378 """Searches for addesses within [start, end). | |
| 379 | |
| 380 Args may be ints or hex strings. Default value for |end| is |start| + 1. | |
| 381 """ | |
| 382 if isinstance(start, basestring): | |
| 383 start = int(start, 16) | |
| 384 if end is None: | |
| 385 end = start + 1 | |
| 386 return self.Filter(lambda s: s.address >= start and s.address < end) | |
| 387 | |
| 388 def WhereHasAnyAttribution(self): | |
| 389 return self.Filter(lambda s: s.name or s.source_path or s.object_path) | |
| 390 | |
| 391 def Inverted(self): | |
| 392 """Returns the symbols that were filtered out by the previous filter. | |
| 393 | |
| 394 Applies only when the previous call was a filter. | |
| 395 | |
| 396 Example: | |
| 397 # Symbols that do not have "third_party" in their path. | |
| 398 symbols.WherePathMatches(r'third_party').Inverted() | |
| 399 # Symbols within third_party that do not contain the string "foo". | |
| 400 symbols.WherePathMatches(r'third_party').WhereMatches('foo').Inverted() | |
| 401 """ | |
| 402 return self._CreateTransformed( | |
| 403 self._filtered_symbols, filtered_symbols=self._symbols, is_sorted=False) | |
| 404 | |
| 405 def GroupBy(self, func, min_count=0): | |
| 406 """Returns a SymbolGroup of SymbolGroups, indexed by |func|. | |
| 407 | |
| 408 Args: | |
| 409 func: Grouping function. Passed a symbol and returns a string for the | |
| 410 name of the subgroup to put the symbol in. If None is returned, the | |
| 411 symbol is omitted. | |
| 412 min_count: Miniumum number of symbols for a group. If fewer than this many | |
| 413 symbols end up in a group, they will not be put within a group. | |
| 414 Use a negative value to omit symbols entirely rather than | |
| 415 include them outside of a group. | |
| 416 """ | |
| 417 new_syms = [] | |
| 418 filtered_symbols = [] | |
| 419 symbols_by_token = collections.defaultdict(list) | |
| 420 # Index symbols by |func|. | |
| 421 for symbol in self: | |
| 422 token = func(symbol) | |
| 423 if token is None: | |
| 424 filtered_symbols.append(symbol) | |
| 425 symbols_by_token[token].append(symbol) | |
| 426 # Create the subgroups. | |
| 427 include_singles = min_count >= 0 | |
| 428 min_count = abs(min_count) | |
| 429 for token, symbols in symbols_by_token.iteritems(): | |
| 430 if len(symbols) >= min_count: | |
| 431 new_syms.append(self._CreateTransformed( | |
| 432 symbols, name=token, section_name=self.section_name, | |
| 433 is_sorted=False)) | |
| 434 elif include_singles: | |
| 435 new_syms.extend(symbols) | |
| 436 else: | |
| 437 filtered_symbols.extend(symbols) | |
| 438 return self._CreateTransformed( | |
| 439 new_syms, filtered_symbols=filtered_symbols, | |
| 440 section_name=self.section_name, is_sorted=False) | |
| 441 | |
| 442 def GroupBySectionName(self): | |
| 443 return self.GroupBy(lambda s: s.section_name) | |
| 444 | |
| 445 def GroupByNamespace(self, depth=0, fallback='{global}', min_count=0): | |
| 446 """Groups by symbol namespace (as denoted by ::s). | |
| 447 | |
| 448 Does not differentiate between C++ namespaces and C++ classes. | |
| 449 | |
| 450 Args: | |
| 451 depth: When 0 (default), groups by entire namespace. When 1, groups by | |
| 452 top-level name, when 2, groups by top 2 names, etc. | |
| 453 fallback: Use this value when no namespace exists. | |
| 454 min_count: Miniumum number of symbols for a group. If fewer than this many | |
| 455 symbols end up in a group, they will not be put within a group. | |
| 456 Use a negative value to omit symbols entirely rather than | |
| 457 include them outside of a group. | |
| 458 """ | |
| 459 def extract_namespace(symbol): | |
| 460 # Remove template params. | |
| 461 name = symbol.name | |
| 462 template_idx = name.find('<') | |
| 463 if template_idx: | |
| 464 name = name[:template_idx] | |
| 465 | |
| 466 # Remove after the final :: (not part of the namespace). | |
| 467 colon_idx = name.rfind('::') | |
| 468 if colon_idx == -1: | |
| 469 return fallback | |
| 470 name = name[:colon_idx] | |
| 471 | |
| 472 return _ExtractPrefixBeforeSeparator(name, '::', depth) | |
| 473 return self.GroupBy(extract_namespace, min_count=min_count) | |
| 474 | |
| 475 def GroupBySourcePath(self, depth=0, fallback='{no path}', | |
| 476 fallback_to_object_path=True, min_count=0): | |
| 477 """Groups by source_path. | |
| 478 | |
| 479 Args: | |
| 480 depth: When 0 (default), groups by entire path. When 1, groups by | |
| 481 top-level directory, when 2, groups by top 2 directories, etc. | |
| 482 fallback: Use this value when no namespace exists. | |
| 483 fallback_to_object_path: When True (default), uses object_path when | |
| 484 source_path is missing. | |
| 485 min_count: Miniumum number of symbols for a group. If fewer than this many | |
| 486 symbols end up in a group, they will not be put within a group. | |
| 487 Use a negative value to omit symbols entirely rather than | |
| 488 include them outside of a group. | |
| 489 """ | |
| 490 def extract_path(symbol): | |
| 491 path = symbol.source_path | |
| 492 if fallback_to_object_path and not path: | |
| 493 path = symbol.object_path | |
| 494 path = path or fallback | |
| 495 return _ExtractPrefixBeforeSeparator(path, os.path.sep, depth) | |
| 496 return self.GroupBy(extract_path, min_count=min_count) | |
| 497 | |
| 498 def GroupByObjectPath(self, depth=0, fallback='{no path}', min_count=0): | |
| 499 """Groups by object_path. | |
| 500 | |
| 501 Args: | |
| 502 depth: When 0 (default), groups by entire path. When 1, groups by | |
| 503 top-level directory, when 2, groups by top 2 directories, etc. | |
| 504 fallback: Use this value when no namespace exists. | |
| 505 min_count: Miniumum number of symbols for a group. If fewer than this many | |
| 506 symbols end up in a group, they will not be put within a group. | |
| 507 Use a negative value to omit symbols entirely rather than | |
| 508 include them outside of a group. | |
| 509 """ | |
| 510 def extract_path(symbol): | |
| 511 path = symbol.object_path or fallback | |
| 512 return _ExtractPrefixBeforeSeparator(path, os.path.sep, depth) | |
| 513 return self.GroupBy(extract_path, min_count=min_count) | |
| 514 | |
| 515 | |
| 516 class SymbolDiff(SymbolGroup): | |
| 517 """A SymbolGroup subclass representing a diff of two other SymbolGroups. | |
| 518 | |
| 519 All Symbols contained within have a |size| which is actually the size delta. | |
| 520 Additionally, metadata is kept about which symbols were added / removed / | |
| 521 changed. | |
| 522 """ | |
| 523 __slots__ = ( | |
| 524 '_added_ids', | |
| 525 '_removed_ids', | |
| 526 ) | |
| 527 | |
| 528 def __init__(self, added, removed, similar, name=None, full_name=None, | |
| 529 section_name=None): | |
| 530 self._added_ids = set(id(s) for s in added) | |
| 531 self._removed_ids = set(id(s) for s in removed) | |
| 532 symbols = [] | |
| 533 symbols.extend(added) | |
| 534 symbols.extend(removed) | |
| 535 symbols.extend(similar) | |
| 536 super(SymbolDiff, self).__init__(symbols, name=name, full_name=full_name, | |
| 537 section_name=section_name) | |
| 538 | |
| 539 def __repr__(self): | |
| 540 return '%s(%d added, %d removed, %d changed, %d unchanged, size=%d)' % ( | |
| 541 'SymbolGroup', self.added_count, self.removed_count, self.changed_count, | |
| 542 self.unchanged_count, self.size) | |
| 543 | |
| 544 def _CreateTransformed(self, symbols, filtered_symbols=None, name=None, | |
| 545 section_name=None, is_sorted=None): | |
| 546 ret = SymbolDiff.__new__(SymbolDiff) | |
| 547 # Printing sorts, so fast-path the same symbols case. | |
| 548 if len(symbols) == len(self._symbols): | |
| 549 ret._added_ids = self._added_ids | |
| 550 ret._removed_ids = self._removed_ids | |
| 551 else: | |
| 552 ret._added_ids = set(id(s) for s in symbols if self.IsAdded(s)) | |
| 553 ret._removed_ids = set(id(s) for s in symbols if self.IsRemoved(s)) | |
| 554 super(SymbolDiff, ret).__init__( | |
| 555 symbols, filtered_symbols=filtered_symbols, name=name, | |
| 556 section_name=section_name, is_sorted=is_sorted) | |
| 557 return ret | |
| 558 | |
| 559 @property | |
| 560 def added_count(self): | |
| 561 return len(self._added_ids) | |
| 562 | |
| 563 @property | |
| 564 def removed_count(self): | |
| 565 return len(self._removed_ids) | |
| 566 | |
| 567 @property | |
| 568 def changed_count(self): | |
| 569 not_changed = self.unchanged_count + self.added_count + self.removed_count | |
| 570 return len(self) - not_changed | |
| 571 | |
| 572 @property | |
| 573 def unchanged_count(self): | |
| 574 return sum(1 for s in self if self.IsSimilar(s) and s.size == 0) | |
| 575 | |
| 576 def IsAdded(self, sym): | |
| 577 return id(sym) in self._added_ids | |
| 578 | |
| 579 def IsSimilar(self, sym): | |
| 580 key = id(sym) | |
| 581 return key not in self._added_ids and key not in self._removed_ids | |
| 582 | |
| 583 def IsRemoved(self, sym): | |
| 584 return id(sym) in self._removed_ids | |
| 585 | |
| 586 def WhereNotUnchanged(self): | |
| 587 return self.Filter(lambda s: not self.IsSimilar(s) or s.size) | |
| 588 | |
| 589 | |
| 590 def Diff(new, old): | |
| 591 """Diffs two SizeInfo or SymbolGroup objects. | |
| 592 | |
| 593 When diffing SizeInfos, a SizeInfoDiff is returned. | |
| 594 When diffing SymbolGroups, a SymbolDiff is returned. | |
| 595 | |
| 596 Returns: | |
| 597 Returns a SizeInfo when args are of type SizeInfo. | |
| 598 Returns a SymbolDiff when args are of type SymbolGroup. | |
| 599 """ | |
| 600 if isinstance(new, SizeInfo): | |
| 601 assert isinstance(old, SizeInfo) | |
| 602 section_sizes = { | |
| 603 k:new.section_sizes[k] - v for k, v in old.section_sizes.iteritems()} | |
| 604 symbol_diff = Diff(new.symbols, old.symbols) | |
| 605 return SizeInfoDiff(section_sizes, symbol_diff, old.metadata, new.metadata) | |
| 606 | |
| 607 assert isinstance(new, SymbolGroup) and isinstance(old, SymbolGroup) | |
| 608 return _DiffSymbols(new, old) | |
| 609 | |
| 610 | |
| 611 def _NegateAll(symbols): | |
| 612 ret = [] | |
| 613 for symbol in symbols: | |
| 614 if symbol.IsGroup(): | |
| 615 duped = SymbolDiff([], _NegateAll(symbol), [], name=symbol.name, | |
| 616 full_name=symbol.full_name, | |
| 617 section_name=symbol.section_name) | |
| 618 else: | |
| 619 duped = copy.copy(symbol) | |
| 620 duped.size = -duped.size | |
| 621 duped.padding = -duped.padding | |
| 622 ret.append(duped) | |
| 623 return ret | |
| 624 | |
| 625 | |
| 626 def _DiffSymbols(new_group, old_group): | |
| 627 symbols_by_key = collections.defaultdict(list) | |
| 628 for s in old_group: | |
| 629 symbols_by_key[s._Key()].append(s) | |
| 630 | |
| 631 added = [] | |
| 632 similar = [] | |
| 633 # For similar symbols, padding is zeroed out. In order to not lose the | |
| 634 # information entirely, store it in aggregate. | |
| 635 padding_by_section_name = collections.defaultdict(int) | |
| 636 for new_sym in new_group: | |
| 637 matching_syms = symbols_by_key.get(new_sym._Key()) | |
| 638 if matching_syms: | |
| 639 old_sym = matching_syms.pop(0) | |
| 640 if old_sym.IsGroup() and new_sym.IsGroup(): | |
| 641 merged_sym = _DiffSymbols(new_sym, old_sym) | |
| 642 else: | |
| 643 size_diff = new_sym.size_without_padding - old_sym.size_without_padding | |
| 644 merged_sym = Symbol(new_sym.section_name, size_diff, | |
| 645 address=new_sym.address, name=new_sym.name, | |
| 646 source_path=new_sym.source_path, | |
| 647 object_path=new_sym.object_path, | |
| 648 full_name=new_sym.full_name, | |
| 649 is_anonymous=new_sym.is_anonymous) | |
| 650 | |
| 651 # Diffs are more stable when comparing size without padding, except when | |
| 652 # the symbol is a padding-only symbol. | |
| 653 if new_sym.size_without_padding == 0 and size_diff == 0: | |
| 654 merged_sym.padding = new_sym.padding - old_sym.padding | |
| 655 else: | |
| 656 padding_by_section_name[new_sym.section_name] += ( | |
| 657 new_sym.padding - old_sym.padding) | |
| 658 | |
| 659 similar.append(merged_sym) | |
| 660 else: | |
| 661 added.append(new_sym) | |
| 662 | |
| 663 removed = [] | |
| 664 for remaining_syms in symbols_by_key.itervalues(): | |
| 665 if remaining_syms: | |
| 666 removed.extend(_NegateAll(remaining_syms)) | |
| 667 | |
| 668 for section_name, padding in padding_by_section_name.iteritems(): | |
| 669 if padding != 0: | |
| 670 similar.append(Symbol(section_name, padding, | |
| 671 name="** aggregate padding of diff'ed symbols")) | |
| 672 return SymbolDiff(added, removed, similar, name=new_group.name, | |
| 673 full_name=new_group.full_name, | |
| 674 section_name=new_group.section_name) | |
| 675 | |
| 676 | |
| 677 def _ExtractPrefixBeforeSeparator(string, separator, count=1): | |
| 678 idx = -len(separator) | |
| 679 prev_idx = None | |
| 680 for _ in xrange(count): | |
| 681 idx = string.find(separator, idx + len(separator)) | |
| 682 if idx < 0: | |
| 683 break | |
| 684 prev_idx = idx | |
| 685 return string[:prev_idx] | |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 2813963002:
//tools/binary_size: Consolidate most tools into "supersize" command (Closed)
