1#!/usr/bin/env python 2# 3#===- exploded-graph-rewriter.py - ExplodedGraph dump tool -----*- python -*--# 4# 5# Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 6# See https://llvm.org/LICENSE.txt for license information. 7# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 8# 9#===-----------------------------------------------------------------------===# 10 11 12from __future__ import print_function 13 14import argparse 15import collections 16import difflib 17import json 18import logging 19import os 20import re 21 22 23#===-----------------------------------------------------------------------===# 24# These data structures represent a deserialized ExplodedGraph. 25#===-----------------------------------------------------------------------===# 26 27 28# A helper function for finding the difference between two dictionaries. 29def diff_dicts(curr, prev): 30 removed = [k for k in prev if k not in curr or curr[k] != prev[k]] 31 added = [k for k in curr if k not in prev or curr[k] != prev[k]] 32 return (removed, added) 33 34 35# Represents any program state trait that is a dictionary of key-value pairs. 36class GenericMap: 37 def __init__(self, items): 38 self.generic_map = collections.OrderedDict(items) 39 40 def diff(self, prev): 41 return diff_dicts(self.generic_map, prev.generic_map) 42 43 def is_different(self, prev): 44 removed, added = self.diff(prev) 45 return len(removed) != 0 or len(added) != 0 46 47 48# A deserialized source location. 49class SourceLocation: 50 def __init__(self, json_loc): 51 logging.debug('json: %s' % json_loc) 52 self.line = json_loc['line'] 53 self.col = json_loc['column'] 54 self.filename = os.path.basename(json_loc['file']) \ 55 if 'file' in json_loc else '(main file)' 56 self.spelling = SourceLocation(json_loc['spelling']) \ 57 if 'spelling' in json_loc else None 58 59 def is_macro(self): 60 return self.spelling is not None 61 62 63# A deserialized program point. 64class ProgramPoint: 65 def __init__(self, json_pp): 66 self.kind = json_pp['kind'] 67 self.tag = json_pp['tag'] 68 self.node_id = json_pp['node_id'] 69 self.is_sink = bool(json_pp['is_sink']) 70 self.has_report = bool(json_pp['has_report']) 71 if self.kind == 'Edge': 72 self.src_id = json_pp['src_id'] 73 self.dst_id = json_pp['dst_id'] 74 elif self.kind == 'Statement': 75 logging.debug(json_pp) 76 self.stmt_kind = json_pp['stmt_kind'] 77 self.cast_kind = json_pp['cast_kind'] \ 78 if 'cast_kind' in json_pp else None 79 self.stmt_point_kind = json_pp['stmt_point_kind'] 80 self.stmt_id = json_pp['stmt_id'] 81 self.pointer = json_pp['pointer'] 82 self.pretty = json_pp['pretty'] 83 self.loc = SourceLocation(json_pp['location']) \ 84 if json_pp['location'] is not None else None 85 elif self.kind == 'BlockEntrance': 86 self.block_id = json_pp['block_id'] 87 88 89# A single expression acting as a key in a deserialized Environment. 90class EnvironmentBindingKey: 91 def __init__(self, json_ek): 92 # CXXCtorInitializer is not a Stmt! 93 self.stmt_id = json_ek['stmt_id'] if 'stmt_id' in json_ek \ 94 else json_ek['init_id'] 95 self.pretty = json_ek['pretty'] 96 self.kind = json_ek['kind'] if 'kind' in json_ek else None 97 98 def _key(self): 99 return self.stmt_id 100 101 def __eq__(self, other): 102 return self._key() == other._key() 103 104 def __hash__(self): 105 return hash(self._key()) 106 107 108# Deserialized description of a location context. 109class LocationContext: 110 def __init__(self, json_frame): 111 self.lctx_id = json_frame['lctx_id'] 112 self.caption = json_frame['location_context'] 113 self.decl = json_frame['calling'] 114 self.loc = SourceLocation(json_frame['location']) \ 115 if json_frame['location'] is not None else None 116 117 def _key(self): 118 return self.lctx_id 119 120 def __eq__(self, other): 121 return self._key() == other._key() 122 123 def __hash__(self): 124 return hash(self._key()) 125 126 127# A group of deserialized Environment bindings that correspond to a specific 128# location context. 129class EnvironmentFrame: 130 def __init__(self, json_frame): 131 self.location_context = LocationContext(json_frame) 132 self.bindings = collections.OrderedDict( 133 [(EnvironmentBindingKey(b), 134 b['value']) for b in json_frame['items']] 135 if json_frame['items'] is not None else []) 136 137 def diff_bindings(self, prev): 138 return diff_dicts(self.bindings, prev.bindings) 139 140 def is_different(self, prev): 141 removed, added = self.diff_bindings(prev) 142 return len(removed) != 0 or len(added) != 0 143 144 145# A deserialized Environment. This class can also hold other entities that 146# are similar to Environment, such as Objects Under Construction or 147# Indices Of Elements Under Construction. 148class GenericEnvironment: 149 def __init__(self, json_e): 150 self.frames = [EnvironmentFrame(f) for f in json_e] 151 152 def diff_frames(self, prev): 153 # TODO: It's difficult to display a good diff when frame numbers shift. 154 if len(self.frames) != len(prev.frames): 155 return None 156 157 updated = [] 158 for i in range(len(self.frames)): 159 f = self.frames[i] 160 prev_f = prev.frames[i] 161 if f.location_context == prev_f.location_context: 162 if f.is_different(prev_f): 163 updated.append(i) 164 else: 165 # We have the whole frame replaced with another frame. 166 # TODO: Produce a nice diff. 167 return None 168 169 # TODO: Add support for added/removed. 170 return updated 171 172 def is_different(self, prev): 173 updated = self.diff_frames(prev) 174 return updated is None or len(updated) > 0 175 176 177# A single binding key in a deserialized RegionStore cluster. 178class StoreBindingKey: 179 def __init__(self, json_sk): 180 self.kind = json_sk['kind'] 181 self.offset = json_sk['offset'] 182 183 def _key(self): 184 return (self.kind, self.offset) 185 186 def __eq__(self, other): 187 return self._key() == other._key() 188 189 def __hash__(self): 190 return hash(self._key()) 191 192 193# A single cluster of the deserialized RegionStore. 194class StoreCluster: 195 def __init__(self, json_sc): 196 self.base_region = json_sc['cluster'] 197 self.bindings = collections.OrderedDict( 198 [(StoreBindingKey(b), b['value']) for b in json_sc['items']]) 199 200 def diff_bindings(self, prev): 201 return diff_dicts(self.bindings, prev.bindings) 202 203 def is_different(self, prev): 204 removed, added = self.diff_bindings(prev) 205 return len(removed) != 0 or len(added) != 0 206 207 208# A deserialized RegionStore. 209class Store: 210 def __init__(self, json_s): 211 self.ptr = json_s['pointer'] 212 self.clusters = collections.OrderedDict( 213 [(c['pointer'], StoreCluster(c)) for c in json_s['items']]) 214 215 def diff_clusters(self, prev): 216 removed = [k for k in prev.clusters if k not in self.clusters] 217 added = [k for k in self.clusters if k not in prev.clusters] 218 updated = [k for k in prev.clusters if k in self.clusters 219 and prev.clusters[k].is_different(self.clusters[k])] 220 return (removed, added, updated) 221 222 def is_different(self, prev): 223 removed, added, updated = self.diff_clusters(prev) 224 return len(removed) != 0 or len(added) != 0 or len(updated) != 0 225 226 227# Deserialized messages from a single checker in a single program state. 228# Basically a list of raw strings. 229class CheckerLines: 230 def __init__(self, json_lines): 231 self.lines = json_lines 232 233 def diff_lines(self, prev): 234 lines = difflib.ndiff(prev.lines, self.lines) 235 return [l.strip() for l in lines 236 if l.startswith('+') or l.startswith('-')] 237 238 def is_different(self, prev): 239 return len(self.diff_lines(prev)) > 0 240 241 242# Deserialized messages of all checkers, separated by checker. 243class CheckerMessages: 244 def __init__(self, json_m): 245 self.items = collections.OrderedDict( 246 [(m['checker'], CheckerLines(m['messages'])) for m in json_m]) 247 248 def diff_messages(self, prev): 249 removed = [k for k in prev.items if k not in self.items] 250 added = [k for k in self.items if k not in prev.items] 251 updated = [k for k in prev.items if k in self.items 252 and prev.items[k].is_different(self.items[k])] 253 return (removed, added, updated) 254 255 def is_different(self, prev): 256 removed, added, updated = self.diff_messages(prev) 257 return len(removed) != 0 or len(added) != 0 or len(updated) != 0 258 259 260# A deserialized program state. 261class ProgramState: 262 def __init__(self, state_id, json_ps): 263 logging.debug('Adding ProgramState ' + str(state_id)) 264 265 if json_ps is None: 266 json_ps = { 267 'store': None, 268 'environment': None, 269 'constraints': None, 270 'dynamic_types': None, 271 'constructing_objects': None, 272 'index_of_element': None, 273 'checker_messages': None 274 } 275 276 self.state_id = state_id 277 278 self.store = Store(json_ps['store']) \ 279 if json_ps['store'] is not None else None 280 281 self.environment = \ 282 GenericEnvironment(json_ps['environment']['items']) \ 283 if json_ps['environment'] is not None else None 284 285 self.constraints = GenericMap([ 286 (c['symbol'], c['range']) for c in json_ps['constraints'] 287 ]) if json_ps['constraints'] is not None else None 288 289 self.dynamic_types = GenericMap([ 290 (t['region'], '%s%s' % (t['dyn_type'], 291 ' (or a sub-class)' 292 if t['sub_classable'] else '')) 293 for t in json_ps['dynamic_types']]) \ 294 if json_ps['dynamic_types'] is not None else None 295 296 self.constructing_objects = \ 297 GenericEnvironment(json_ps['constructing_objects']) \ 298 if json_ps['constructing_objects'] is not None else None 299 300 self.index_of_element = \ 301 GenericEnvironment(json_ps['index_of_element']) \ 302 if json_ps['index_of_element'] is not None else None 303 304 self.checker_messages = CheckerMessages(json_ps['checker_messages']) \ 305 if json_ps['checker_messages'] is not None else None 306 307 308# A deserialized exploded graph node. Has a default constructor because it 309# may be referenced as part of an edge before its contents are deserialized, 310# and in this moment we already need a room for predecessors and successors. 311class ExplodedNode: 312 def __init__(self): 313 self.predecessors = [] 314 self.successors = [] 315 316 def construct(self, node_id, json_node): 317 logging.debug('Adding ' + node_id) 318 self.ptr = node_id[4:] 319 self.points = [ProgramPoint(p) for p in json_node['program_points']] 320 self.node_id = self.points[-1].node_id 321 self.state = ProgramState(json_node['state_id'], 322 json_node['program_state'] 323 if json_node['program_state'] is not None else None); 324 325 assert self.node_name() == node_id 326 327 def node_name(self): 328 return 'Node' + self.ptr 329 330 331# A deserialized ExplodedGraph. Constructed by consuming a .dot file 332# line-by-line. 333class ExplodedGraph: 334 # Parse .dot files with regular expressions. 335 node_re = re.compile( 336 '^(Node0x[0-9a-f]*) \\[shape=record,.*label="{(.*)\\\\l}"\\];$') 337 edge_re = re.compile( 338 '^(Node0x[0-9a-f]*) -> (Node0x[0-9a-f]*);$') 339 340 def __init__(self): 341 self.nodes = collections.defaultdict(ExplodedNode) 342 self.root_id = None 343 self.incomplete_line = '' 344 345 def add_raw_line(self, raw_line): 346 if raw_line.startswith('//'): 347 return 348 349 # Allow line breaks by waiting for ';'. This is not valid in 350 # a .dot file, but it is useful for writing tests. 351 if len(raw_line) > 0 and raw_line[-1] != ';': 352 self.incomplete_line += raw_line 353 return 354 raw_line = self.incomplete_line + raw_line 355 self.incomplete_line = '' 356 357 # Apply regexps one by one to see if it's a node or an edge 358 # and extract contents if necessary. 359 logging.debug('Line: ' + raw_line) 360 result = self.edge_re.match(raw_line) 361 if result is not None: 362 logging.debug('Classified as edge line.') 363 pred = result.group(1) 364 succ = result.group(2) 365 self.nodes[pred].successors.append(succ) 366 self.nodes[succ].predecessors.append(pred) 367 return 368 result = self.node_re.match(raw_line) 369 if result is not None: 370 logging.debug('Classified as node line.') 371 node_id = result.group(1) 372 if len(self.nodes) == 0: 373 self.root_id = node_id 374 # Note: when writing tests you don't need to escape everything, 375 # even though in a valid dot file everything is escaped. 376 node_label = result.group(2).replace(' ', '') \ 377 .replace('\\"', '"') \ 378 .replace('\\{', '{') \ 379 .replace('\\}', '}') \ 380 .replace('\\\\', '\\') \ 381 .replace('\\|', '|') \ 382 .replace('\\<', '\\\\<') \ 383 .replace('\\>', '\\\\>') \ 384 .rstrip(',') 385 # Handle `\l` separately because a string literal can be in code 386 # like "string\\literal" with the `\l` inside. 387 # Also on Windows macros __FILE__ produces specific delimiters `\` 388 # and a directory or file may starts with the letter `l`. 389 # Find all `\l` (like `,\l`, `}\l`, `[\l`) except `\\l`, 390 # because the literal as a rule contains multiple `\` before `\l`. 391 node_label = re.sub(r'(?<!\\)\\l', '', node_label) 392 logging.debug(node_label) 393 json_node = json.loads(node_label) 394 self.nodes[node_id].construct(node_id, json_node) 395 return 396 logging.debug('Skipping.') 397 398 399#===-----------------------------------------------------------------------===# 400# Visitors traverse a deserialized ExplodedGraph and do different things 401# with every node and edge. 402#===-----------------------------------------------------------------------===# 403 404 405# A visitor that dumps the ExplodedGraph into a DOT file with fancy HTML-based 406# syntax highlighing. 407class DotDumpVisitor: 408 def __init__(self, do_diffs, dark_mode, gray_mode, 409 topo_mode, dump_dot_only): 410 self._do_diffs = do_diffs 411 self._dark_mode = dark_mode 412 self._gray_mode = gray_mode 413 self._topo_mode = topo_mode 414 self._dump_dot_only = dump_dot_only 415 self._output = [] 416 417 def _dump_raw(self, s): 418 if self._dump_dot_only: 419 print(s, end='') 420 else: 421 self._output.append(s) 422 423 def output(self): 424 assert not self._dump_dot_only 425 return ''.join(self._output) 426 427 def _dump(self, s): 428 s = s.replace('&', '&') \ 429 .replace('{', '\\{') \ 430 .replace('}', '\\}') \ 431 .replace('\\<', '<') \ 432 .replace('\\>', '>') \ 433 .replace('|', '\\|') 434 s = re.sub(r'(?<!\\)\\l', '<br />', s) 435 if self._gray_mode: 436 s = re.sub(r'<font color="[a-z0-9]*">', '', s) 437 s = re.sub(r'</font>', '', s) 438 self._dump_raw(s) 439 440 @staticmethod 441 def _diff_plus_minus(is_added): 442 if is_added is None: 443 return '' 444 if is_added: 445 return '<font color="forestgreen">+</font>' 446 return '<font color="red">-</font>' 447 448 @staticmethod 449 def _short_pretty(s): 450 if s is None: 451 return None 452 if len(s) < 20: 453 return s 454 left = s.find('{') 455 right = s.rfind('}') 456 if left == -1 or right == -1 or left >= right: 457 return s 458 candidate = s[0:left + 1] + ' ... ' + s[right:] 459 if len(candidate) >= len(s): 460 return s 461 return candidate 462 463 @staticmethod 464 def _make_sloc(loc): 465 if loc is None: 466 return '<i>Invalid Source Location</i>' 467 468 def make_plain_loc(loc): 469 return '%s:<b>%s</b>:<b>%s</b>' \ 470 % (loc.filename, loc.line, loc.col) 471 472 if loc.is_macro(): 473 return '%s <font color="royalblue1">' \ 474 '(<i>spelling at </i> %s)</font>' \ 475 % (make_plain_loc(loc), make_plain_loc(loc.spelling)) 476 477 return make_plain_loc(loc) 478 479 def visit_begin_graph(self, graph): 480 self._graph = graph 481 self._dump_raw('digraph "ExplodedGraph" {\n') 482 if self._dark_mode: 483 self._dump_raw('bgcolor="gray10";\n') 484 self._dump_raw('label="";\n') 485 486 def visit_program_point(self, p): 487 if p.kind in ['Edge', 'BlockEntrance', 'BlockExit']: 488 color = 'gold3' 489 elif p.kind in ['PreStmtPurgeDeadSymbols', 490 'PostStmtPurgeDeadSymbols']: 491 color = 'red' 492 elif p.kind in ['CallEnter', 'CallExitBegin', 'CallExitEnd']: 493 color = 'dodgerblue' if self._dark_mode else 'blue' 494 elif p.kind in ['Statement']: 495 color = 'cyan4' 496 else: 497 color = 'forestgreen' 498 499 self._dump('<tr><td align="left">%s.</td>' % p.node_id) 500 501 if p.kind == 'Statement': 502 # This avoids pretty-printing huge statements such as CompoundStmt. 503 # Such statements show up only at [Pre|Post]StmtPurgeDeadSymbols 504 skip_pretty = 'PurgeDeadSymbols' in p.stmt_point_kind 505 stmt_color = 'cyan3' 506 self._dump('<td align="left" width="0">%s:</td>' 507 '<td align="left" width="0"><font color="%s">' 508 '%s</font> </td>' 509 '<td align="left"><i>S%s</i></td>' 510 '<td align="left"><font color="%s">%s</font></td>' 511 '<td align="left">%s</td></tr>' 512 % (self._make_sloc(p.loc), color, 513 '%s (%s)' % (p.stmt_kind, p.cast_kind) 514 if p.cast_kind is not None else p.stmt_kind, 515 p.stmt_id, stmt_color, p.stmt_point_kind, 516 self._short_pretty(p.pretty) 517 if not skip_pretty else '')) 518 elif p.kind == 'Edge': 519 self._dump('<td width="0"></td>' 520 '<td align="left" width="0">' 521 '<font color="%s">%s</font></td><td align="left">' 522 '[B%d] -\\> [B%d]</td></tr>' 523 % (color, 'BlockEdge', p.src_id, p.dst_id)) 524 elif p.kind == 'BlockEntrance': 525 self._dump('<td width="0"></td>' 526 '<td align="left" width="0">' 527 '<font color="%s">%s</font></td>' 528 '<td align="left">[B%d]</td></tr>' 529 % (color, p.kind, p.block_id)) 530 else: 531 # TODO: Print more stuff for other kinds of points. 532 self._dump('<td width="0"></td>' 533 '<td align="left" width="0" colspan="2">' 534 '<font color="%s">%s</font></td></tr>' 535 % (color, p.kind)) 536 537 if p.tag is not None: 538 self._dump('<tr><td width="0"></td><td width="0"></td>' 539 '<td colspan="3" align="left">' 540 '<b>Tag: </b> <font color="crimson">' 541 '%s</font></td></tr>' % p.tag) 542 543 if p.has_report: 544 self._dump('<tr><td width="0"></td><td width="0"></td>' 545 '<td colspan="3" align="left">' 546 '<font color="red"><b>Bug Report Attached' 547 '</b></font></td></tr>') 548 if p.is_sink: 549 self._dump('<tr><td width="0"></td><td width="0"></td>' 550 '<td colspan="3" align="left">' 551 '<font color="cornflowerblue"><b>Sink Node' 552 '</b></font></td></tr>') 553 554 def visit_environment(self, e, prev_e=None): 555 self._dump('<table border="0">') 556 557 def dump_location_context(lc, is_added=None): 558 self._dump('<tr><td>%s</td>' 559 '<td align="left"><b>%s</b></td>' 560 '<td align="left" colspan="2">' 561 '<font color="gray60">%s </font>' 562 '%s</td></tr>' 563 % (self._diff_plus_minus(is_added), 564 lc.caption, lc.decl, 565 ('(%s)' % self._make_sloc(lc.loc)) 566 if lc.loc is not None else '')) 567 568 def dump_binding(f, b, is_added=None): 569 self._dump('<tr><td>%s</td>' 570 '<td align="left"><i>S%s</i></td>' 571 '%s' 572 '<td align="left">%s</td>' 573 '<td align="left">%s</td></tr>' 574 % (self._diff_plus_minus(is_added), 575 b.stmt_id, 576 '<td align="left"><font color="%s"><i>' 577 '%s</i></font></td>' % ( 578 'lavender' if self._dark_mode else 'darkgreen', 579 ('(%s)' % b.kind) if b.kind is not None else ' ' 580 ), 581 self._short_pretty(b.pretty), f.bindings[b])) 582 583 frames_updated = e.diff_frames(prev_e) if prev_e is not None else None 584 if frames_updated: 585 for i in frames_updated: 586 f = e.frames[i] 587 prev_f = prev_e.frames[i] 588 dump_location_context(f.location_context) 589 bindings_removed, bindings_added = f.diff_bindings(prev_f) 590 for b in bindings_removed: 591 dump_binding(prev_f, b, False) 592 for b in bindings_added: 593 dump_binding(f, b, True) 594 else: 595 for f in e.frames: 596 dump_location_context(f.location_context) 597 for b in f.bindings: 598 dump_binding(f, b) 599 600 self._dump('</table>') 601 602 def visit_environment_in_state(self, selector, title, s, prev_s=None): 603 e = getattr(s, selector) 604 prev_e = getattr(prev_s, selector) if prev_s is not None else None 605 if e is None and prev_e is None: 606 return 607 608 self._dump('<hr /><tr><td align="left"><b>%s: </b>' % title) 609 if e is None: 610 self._dump('<i> Nothing!</i>') 611 else: 612 if prev_e is not None: 613 if e.is_different(prev_e): 614 self._dump('</td></tr><tr><td align="left">') 615 self.visit_environment(e, prev_e) 616 else: 617 self._dump('<i> No changes!</i>') 618 else: 619 self._dump('</td></tr><tr><td align="left">') 620 self.visit_environment(e) 621 622 self._dump('</td></tr>') 623 624 def visit_store(self, s, prev_s=None): 625 self._dump('<table border="0">') 626 627 def dump_binding(s, c, b, is_added=None): 628 self._dump('<tr><td>%s</td>' 629 '<td align="left">%s</td>' 630 '<td align="left">%s</td>' 631 '<td align="left">%s</td>' 632 '<td align="left">%s</td></tr>' 633 % (self._diff_plus_minus(is_added), 634 s.clusters[c].base_region, b.offset, 635 '(<i>Default</i>)' if b.kind == 'Default' 636 else '', 637 s.clusters[c].bindings[b])) 638 639 if prev_s is not None: 640 clusters_removed, clusters_added, clusters_updated = \ 641 s.diff_clusters(prev_s) 642 for c in clusters_removed: 643 for b in prev_s.clusters[c].bindings: 644 dump_binding(prev_s, c, b, False) 645 for c in clusters_updated: 646 bindings_removed, bindings_added = \ 647 s.clusters[c].diff_bindings(prev_s.clusters[c]) 648 for b in bindings_removed: 649 dump_binding(prev_s, c, b, False) 650 for b in bindings_added: 651 dump_binding(s, c, b, True) 652 for c in clusters_added: 653 for b in s.clusters[c].bindings: 654 dump_binding(s, c, b, True) 655 else: 656 for c in s.clusters: 657 for b in s.clusters[c].bindings: 658 dump_binding(s, c, b) 659 660 self._dump('</table>') 661 662 def visit_store_in_state(self, s, prev_s=None): 663 st = s.store 664 prev_st = prev_s.store if prev_s is not None else None 665 if st is None and prev_st is None: 666 return 667 668 self._dump('<hr /><tr><td align="left"><b>Store: </b>') 669 if st is None: 670 self._dump('<i> Nothing!</i>') 671 else: 672 if self._dark_mode: 673 self._dump(' <font color="gray30">(%s)</font>' % st.ptr) 674 else: 675 self._dump(' <font color="gray">(%s)</font>' % st.ptr) 676 if prev_st is not None: 677 if s.store.is_different(prev_st): 678 self._dump('</td></tr><tr><td align="left">') 679 self.visit_store(st, prev_st) 680 else: 681 self._dump('<i> No changes!</i>') 682 else: 683 self._dump('</td></tr><tr><td align="left">') 684 self.visit_store(st) 685 self._dump('</td></tr>') 686 687 def visit_generic_map(self, m, prev_m=None): 688 self._dump('<table border="0">') 689 690 def dump_pair(m, k, is_added=None): 691 self._dump('<tr><td>%s</td>' 692 '<td align="left">%s</td>' 693 '<td align="left">%s</td></tr>' 694 % (self._diff_plus_minus(is_added), 695 k, m.generic_map[k])) 696 697 if prev_m is not None: 698 removed, added = m.diff(prev_m) 699 for k in removed: 700 dump_pair(prev_m, k, False) 701 for k in added: 702 dump_pair(m, k, True) 703 else: 704 for k in m.generic_map: 705 dump_pair(m, k, None) 706 707 self._dump('</table>') 708 709 def visit_generic_map_in_state(self, selector, title, s, prev_s=None): 710 m = getattr(s, selector) 711 prev_m = getattr(prev_s, selector) if prev_s is not None else None 712 if m is None and prev_m is None: 713 return 714 715 self._dump('<hr />') 716 self._dump('<tr><td align="left">' 717 '<b>%s: </b>' % title) 718 if m is None: 719 self._dump('<i> Nothing!</i>') 720 else: 721 if prev_m is not None: 722 if m.is_different(prev_m): 723 self._dump('</td></tr><tr><td align="left">') 724 self.visit_generic_map(m, prev_m) 725 else: 726 self._dump('<i> No changes!</i>') 727 else: 728 self._dump('</td></tr><tr><td align="left">') 729 self.visit_generic_map(m) 730 731 self._dump('</td></tr>') 732 733 def visit_checker_messages(self, m, prev_m=None): 734 self._dump('<table border="0">') 735 736 def dump_line(l, is_added=None): 737 self._dump('<tr><td>%s</td>' 738 '<td align="left">%s</td></tr>' 739 % (self._diff_plus_minus(is_added), l)) 740 741 def dump_chk(chk, is_added=None): 742 dump_line('<i>%s</i>:' % chk, is_added) 743 744 if prev_m is not None: 745 removed, added, updated = m.diff_messages(prev_m) 746 for chk in removed: 747 dump_chk(chk, False) 748 for l in prev_m.items[chk].lines: 749 dump_line(l, False) 750 for chk in updated: 751 dump_chk(chk) 752 for l in m.items[chk].diff_lines(prev_m.items[chk]): 753 dump_line(l[1:], l.startswith('+')) 754 for chk in added: 755 dump_chk(chk, True) 756 for l in m.items[chk].lines: 757 dump_line(l, True) 758 else: 759 for chk in m.items: 760 dump_chk(chk) 761 for l in m.items[chk].lines: 762 dump_line(l) 763 764 self._dump('</table>') 765 766 def visit_checker_messages_in_state(self, s, prev_s=None): 767 m = s.checker_messages 768 prev_m = prev_s.checker_messages if prev_s is not None else None 769 if m is None and prev_m is None: 770 return 771 772 self._dump('<hr />') 773 self._dump('<tr><td align="left">' 774 '<b>Checker State: </b>') 775 if m is None: 776 self._dump('<i> Nothing!</i>') 777 else: 778 if prev_m is not None: 779 if m.is_different(prev_m): 780 self._dump('</td></tr><tr><td align="left">') 781 self.visit_checker_messages(m, prev_m) 782 else: 783 self._dump('<i> No changes!</i>') 784 else: 785 self._dump('</td></tr><tr><td align="left">') 786 self.visit_checker_messages(m) 787 788 self._dump('</td></tr>') 789 790 def visit_state(self, s, prev_s): 791 self.visit_store_in_state(s, prev_s) 792 self.visit_environment_in_state('environment', 'Expressions', 793 s, prev_s) 794 self.visit_generic_map_in_state('constraints', 'Ranges', 795 s, prev_s) 796 self.visit_generic_map_in_state('dynamic_types', 'Dynamic Types', 797 s, prev_s) 798 self.visit_environment_in_state('constructing_objects', 799 'Objects Under Construction', 800 s, prev_s) 801 self.visit_environment_in_state('index_of_element', 802 'Indices Of Elements Under Construction', 803 s, prev_s) 804 self.visit_checker_messages_in_state(s, prev_s) 805 806 def visit_node(self, node): 807 self._dump('%s [shape=record,' 808 % (node.node_name())) 809 if self._dark_mode: 810 self._dump('color="white",fontcolor="gray80",') 811 self._dump('label=<<table border="0">') 812 813 self._dump('<tr><td bgcolor="%s"><b>State %s</b></td></tr>' 814 % ("gray20" if self._dark_mode else "gray70", 815 node.state.state_id 816 if node.state is not None else 'Unspecified')) 817 if not self._topo_mode: 818 self._dump('<tr><td align="left" width="0">') 819 if len(node.points) > 1: 820 self._dump('<b>Program points:</b></td></tr>') 821 else: 822 self._dump('<b>Program point:</b></td></tr>') 823 self._dump('<tr><td align="left" width="0">' 824 '<table border="0" align="left" width="0">') 825 for p in node.points: 826 self.visit_program_point(p) 827 self._dump('</table></td></tr>') 828 829 if node.state is not None and not self._topo_mode: 830 prev_s = None 831 # Do diffs only when we have a unique predecessor. 832 # Don't do diffs on the leaf nodes because they're 833 # the important ones. 834 if self._do_diffs and len(node.predecessors) == 1 \ 835 and len(node.successors) > 0: 836 prev_s = self._graph.nodes[node.predecessors[0]].state 837 self.visit_state(node.state, prev_s) 838 self._dump_raw('</table>>];\n') 839 840 def visit_edge(self, pred, succ): 841 self._dump_raw('%s -> %s%s;\n' % ( 842 pred.node_name(), succ.node_name(), 843 ' [color="white"]' if self._dark_mode else '' 844 )) 845 846 def visit_end_of_graph(self): 847 self._dump_raw('}\n') 848 849 if not self._dump_dot_only: 850 import sys 851 import tempfile 852 853 def write_temp_file(suffix, prefix, data): 854 fd, filename = tempfile.mkstemp(suffix, prefix, '.', True) 855 print('Writing "%s"...' % filename) 856 with os.fdopen(fd, 'w') as fp: 857 fp.write(data) 858 print('Done! Please remember to remove the file.') 859 return filename 860 861 try: 862 import graphviz 863 except ImportError: 864 # The fallback behavior if graphviz is not installed! 865 print('Python graphviz not found. Please invoke') 866 print(' $ pip install graphviz') 867 print('in order to enable automatic conversion to HTML.') 868 print() 869 print('You may also convert DOT to SVG manually via') 870 print(' $ dot -Tsvg input.dot -o output.svg') 871 print() 872 write_temp_file('.dot', 'egraph-', self.output()) 873 return 874 875 svg = graphviz.pipe('dot', 'svg', self.output().encode()).decode() 876 877 filename = write_temp_file( 878 '.html', 'egraph-', '<html><body bgcolor="%s">%s</body></html>' % ( 879 '#1a1a1a' if self._dark_mode else 'white', svg)) 880 if sys.platform == 'win32': 881 os.startfile(filename) 882 elif sys.platform == 'darwin': 883 os.system('open "%s"' % filename) 884 else: 885 os.system('xdg-open "%s"' % filename) 886 887 888#===-----------------------------------------------------------------------===# 889# Explorers know how to traverse the ExplodedGraph in a certain order. 890# They would invoke a Visitor on every node or edge they encounter. 891#===-----------------------------------------------------------------------===# 892 893 894# BasicExplorer explores the whole graph in no particular order. 895class BasicExplorer: 896 def explore(self, graph, visitor): 897 visitor.visit_begin_graph(graph) 898 for node in sorted(graph.nodes): 899 logging.debug('Visiting ' + node) 900 visitor.visit_node(graph.nodes[node]) 901 for succ in sorted(graph.nodes[node].successors): 902 logging.debug('Visiting edge: %s -> %s ' % (node, succ)) 903 visitor.visit_edge(graph.nodes[node], graph.nodes[succ]) 904 visitor.visit_end_of_graph() 905 906 907#===-----------------------------------------------------------------------===# 908# Trimmers cut out parts of the ExplodedGraph so that to focus on other parts. 909# Trimmers can be combined together by applying them sequentially. 910#===-----------------------------------------------------------------------===# 911 912 913# SinglePathTrimmer keeps only a single path - the leftmost path from the root. 914# Useful when the trimmed graph is still too large. 915class SinglePathTrimmer: 916 def trim(self, graph): 917 visited_nodes = set() 918 node_id = graph.root_id 919 while True: 920 visited_nodes.add(node_id) 921 node = graph.nodes[node_id] 922 if len(node.successors) > 0: 923 succ_id = node.successors[0] 924 succ = graph.nodes[succ_id] 925 node.successors = [succ_id] 926 succ.predecessors = [node_id] 927 if succ_id in visited_nodes: 928 break 929 node_id = succ_id 930 else: 931 break 932 graph.nodes = {node_id: graph.nodes[node_id] 933 for node_id in visited_nodes} 934 935 936# TargetedTrimmer keeps paths that lead to specific nodes and discards all 937# other paths. Useful when you cannot use -trim-egraph (e.g. when debugging 938# a crash). 939class TargetedTrimmer: 940 def __init__(self, target_nodes): 941 self._target_nodes = target_nodes 942 943 @staticmethod 944 def parse_target_node(node, graph): 945 if node.startswith('0x'): 946 ret = 'Node' + node 947 assert ret in graph.nodes 948 return ret 949 else: 950 for other_id in graph.nodes: 951 other = graph.nodes[other_id] 952 if other.node_id == int(node): 953 return other_id 954 955 @staticmethod 956 def parse_target_nodes(target_nodes, graph): 957 return [TargetedTrimmer.parse_target_node(node, graph) 958 for node in target_nodes.split(',')] 959 960 def trim(self, graph): 961 queue = self._target_nodes 962 visited_nodes = set() 963 964 while len(queue) > 0: 965 node_id = queue.pop() 966 visited_nodes.add(node_id) 967 node = graph.nodes[node_id] 968 for pred_id in node.predecessors: 969 if pred_id not in visited_nodes: 970 queue.append(pred_id) 971 graph.nodes = {node_id: graph.nodes[node_id] 972 for node_id in visited_nodes} 973 for node_id in graph.nodes: 974 node = graph.nodes[node_id] 975 node.successors = [succ_id for succ_id in node.successors 976 if succ_id in visited_nodes] 977 node.predecessors = [succ_id for succ_id in node.predecessors 978 if succ_id in visited_nodes] 979 980 981#===-----------------------------------------------------------------------===# 982# The entry point to the script. 983#===-----------------------------------------------------------------------===# 984 985 986def main(): 987 parser = argparse.ArgumentParser( 988 description='Display and manipulate Exploded Graph dumps.') 989 parser.add_argument('filename', type=str, 990 help='the .dot file produced by the Static Analyzer') 991 parser.add_argument('-v', '--verbose', action='store_const', 992 dest='loglevel', const=logging.DEBUG, 993 default=logging.WARNING, 994 help='enable info prints') 995 parser.add_argument('-d', '--diff', action='store_const', dest='diff', 996 const=True, default=False, 997 help='display differences between states') 998 parser.add_argument('-t', '--topology', action='store_const', 999 dest='topology', const=True, default=False, 1000 help='only display program points, omit states') 1001 parser.add_argument('-s', '--single-path', action='store_const', 1002 dest='single_path', const=True, default=False, 1003 help='only display the leftmost path in the graph ' 1004 '(useful for trimmed graphs that still ' 1005 'branch too much)') 1006 parser.add_argument('--to', type=str, default=None, 1007 help='only display execution paths from the root ' 1008 'to the given comma-separated list of nodes ' 1009 'identified by a pointer or a stable ID; ' 1010 'compatible with --single-path') 1011 parser.add_argument('--dark', action='store_const', dest='dark', 1012 const=True, default=False, 1013 help='dark mode') 1014 parser.add_argument('--gray', action='store_const', dest='gray', 1015 const=True, default=False, 1016 help='black-and-white mode') 1017 parser.add_argument('--dump-dot-only', action='store_const', 1018 dest='dump_dot_only', const=True, default=False, 1019 help='instead of writing an HTML file and immediately ' 1020 'displaying it, dump the rewritten dot file ' 1021 'to stdout') 1022 args = parser.parse_args() 1023 logging.basicConfig(level=args.loglevel) 1024 1025 graph = ExplodedGraph() 1026 with open(args.filename) as fd: 1027 for raw_line in fd: 1028 raw_line = raw_line.strip() 1029 graph.add_raw_line(raw_line) 1030 1031 trimmers = [] 1032 if args.to is not None: 1033 trimmers.append(TargetedTrimmer( 1034 TargetedTrimmer.parse_target_nodes(args.to, graph))) 1035 if args.single_path: 1036 trimmers.append(SinglePathTrimmer()) 1037 1038 explorer = BasicExplorer() 1039 1040 visitor = DotDumpVisitor(args.diff, args.dark, args.gray, args.topology, 1041 args.dump_dot_only) 1042 1043 for trimmer in trimmers: 1044 trimmer.trim(graph) 1045 1046 explorer.explore(graph, visitor) 1047 1048 1049if __name__ == '__main__': 1050 main() 1051