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 json 17import logging 18import re 19 20 21# A helper function for finding the difference between two dictionaries. 22def diff_dicts(curr, prev): 23 removed = [k for k in prev if k not in curr or curr[k] != prev[k]] 24 added = [k for k in curr if k not in prev or curr[k] != prev[k]] 25 return (removed, added) 26 27 28# Represents any program state trait that is a dictionary of key-value pairs. 29class GenericMap(object): 30 def __init__(self, generic_map): 31 self.generic_map = generic_map 32 33 def diff(self, prev): 34 return diff_dicts(self.generic_map, prev.generic_map) 35 36 def is_different(self, prev): 37 removed, added = self.diff(prev) 38 return len(removed) != 0 or len(added) != 0 39 40 41# A deserialized source location. 42class SourceLocation(object): 43 def __init__(self, json_loc): 44 super(SourceLocation, self).__init__() 45 self.line = json_loc['line'] 46 self.col = json_loc['column'] 47 self.filename = json_loc['filename'] \ 48 if 'filename' in json_loc else '(main file)' 49 50 51# A deserialized program point. 52class ProgramPoint(object): 53 def __init__(self, json_pp): 54 super(ProgramPoint, self).__init__() 55 self.kind = json_pp['kind'] 56 self.tag = json_pp['tag'] 57 if self.kind == 'Edge': 58 self.src_id = json_pp['src_id'] 59 self.dst_id = json_pp['dst_id'] 60 elif self.kind == 'Statement': 61 self.stmt_kind = json_pp['stmt_kind'] 62 self.pointer = json_pp['pointer'] 63 self.pretty = json_pp['pretty'] 64 self.loc = SourceLocation(json_pp['location']) \ 65 if json_pp['location'] is not None else None 66 elif self.kind == 'BlockEntrance': 67 self.block_id = json_pp['block_id'] 68 69 70# A single expression acting as a key in a deserialized Environment. 71class EnvironmentBindingKey(object): 72 def __init__(self, json_ek): 73 super(EnvironmentBindingKey, self).__init__() 74 self.stmt_id = json_ek['stmt_id'] 75 self.pretty = json_ek['pretty'] 76 77 def _key(self): 78 return self.stmt_id 79 80 def __eq__(self, other): 81 return self._key() == other._key() 82 83 def __hash__(self): 84 return hash(self._key()) 85 86 87# Deserialized description of a location context. 88class LocationContext(object): 89 def __init__(self, json_frame): 90 super(LocationContext, self).__init__() 91 self.lctx_id = json_frame['lctx_id'] 92 self.caption = json_frame['location_context'] 93 self.decl = json_frame['calling'] 94 self.line = json_frame['call_line'] 95 96 def _key(self): 97 return self.lctx_id 98 99 def __eq__(self, other): 100 return self._key() == other._key() 101 102 def __hash__(self): 103 return hash(self._key()) 104 105 106# A group of deserialized Environment bindings that correspond to a specific 107# location context. 108class EnvironmentFrame(object): 109 def __init__(self, json_frame): 110 super(EnvironmentFrame, self).__init__() 111 self.location_context = LocationContext(json_frame) 112 self.bindings = collections.OrderedDict( 113 [(EnvironmentBindingKey(b), 114 b['value']) for b in json_frame['items']] 115 if json_frame['items'] is not None else []) 116 117 def diff_bindings(self, prev): 118 return diff_dicts(self.bindings, prev.bindings) 119 120 def is_different(self, prev): 121 removed, added = self.diff_bindings(prev) 122 return len(removed) != 0 or len(added) != 0 123 124 125# A deserialized Environment. 126class Environment(object): 127 def __init__(self, json_e): 128 super(Environment, self).__init__() 129 self.frames = [EnvironmentFrame(f) for f in json_e] 130 131 def diff_frames(self, prev): 132 # TODO: It's difficult to display a good diff when frame numbers shift. 133 if len(self.frames) != len(prev.frames): 134 return None 135 136 updated = [] 137 for i in range(len(self.frames)): 138 f = self.frames[i] 139 prev_f = prev.frames[i] 140 if f.location_context == prev_f.location_context: 141 if f.is_different(prev_f): 142 updated.append(i) 143 else: 144 # We have the whole frame replaced with another frame. 145 # TODO: Produce a nice diff. 146 return None 147 148 # TODO: Add support for added/removed. 149 return updated 150 151 def is_different(self, prev): 152 updated = self.diff_frames(prev) 153 return updated is None or len(updated) > 0 154 155 156# A single binding key in a deserialized RegionStore cluster. 157class StoreBindingKey(object): 158 def __init__(self, json_sk): 159 super(StoreBindingKey, self).__init__() 160 self.kind = json_sk['kind'] 161 self.offset = json_sk['offset'] 162 163 def _key(self): 164 return (self.kind, self.offset) 165 166 def __eq__(self, other): 167 return self._key() == other._key() 168 169 def __hash__(self): 170 return hash(self._key()) 171 172 173# A single cluster of the deserialized RegionStore. 174class StoreCluster(object): 175 def __init__(self, json_sc): 176 super(StoreCluster, self).__init__() 177 self.base_region = json_sc['cluster'] 178 self.bindings = collections.OrderedDict( 179 [(StoreBindingKey(b), b['value']) for b in json_sc['items']]) 180 181 def diff_bindings(self, prev): 182 return diff_dicts(self.bindings, prev.bindings) 183 184 def is_different(self, prev): 185 removed, added = self.diff_bindings(prev) 186 return len(removed) != 0 or len(added) != 0 187 188 189# A deserialized RegionStore. 190class Store(object): 191 def __init__(self, json_s): 192 super(Store, self).__init__() 193 self.clusters = collections.OrderedDict( 194 [(c['pointer'], StoreCluster(c)) for c in json_s]) 195 196 def diff_clusters(self, prev): 197 removed = [k for k in prev.clusters if k not in self.clusters] 198 added = [k for k in self.clusters if k not in prev.clusters] 199 updated = [k for k in prev.clusters if k in self.clusters 200 and prev.clusters[k].is_different(self.clusters[k])] 201 return (removed, added, updated) 202 203 def is_different(self, prev): 204 removed, added, updated = self.diff_clusters(prev) 205 return len(removed) != 0 or len(added) != 0 or len(updated) != 0 206 207 208# A deserialized program state. 209class ProgramState(object): 210 def __init__(self, state_id, json_ps): 211 super(ProgramState, self).__init__() 212 logging.debug('Adding ProgramState ' + str(state_id)) 213 214 self.state_id = state_id 215 self.store = Store(json_ps['store']) \ 216 if json_ps['store'] is not None else None 217 self.environment = Environment(json_ps['environment']) \ 218 if json_ps['environment'] is not None else None 219 self.constraints = GenericMap(collections.OrderedDict([ 220 (c['symbol'], c['range']) for c in json_ps['constraints'] 221 ])) if json_ps['constraints'] is not None else None 222 # TODO: Objects under construction. 223 # TODO: Dynamic types of objects. 224 # TODO: Checker messages. 225 226 227# A deserialized exploded graph node. Has a default constructor because it 228# may be referenced as part of an edge before its contents are deserialized, 229# and in this moment we already need a room for predecessors and successors. 230class ExplodedNode(object): 231 def __init__(self): 232 super(ExplodedNode, self).__init__() 233 self.predecessors = [] 234 self.successors = [] 235 236 def construct(self, node_id, json_node): 237 logging.debug('Adding ' + node_id) 238 self.node_id = json_node['node_id'] 239 self.ptr = json_node['pointer'] 240 self.points = [ProgramPoint(p) for p in json_node['program_points']] 241 self.state = ProgramState(json_node['state_id'], 242 json_node['program_state']) \ 243 if json_node['program_state'] is not None else None 244 245 assert self.node_name() == node_id 246 247 def node_name(self): 248 return 'Node' + self.ptr 249 250 251# A deserialized ExplodedGraph. Constructed by consuming a .dot file 252# line-by-line. 253class ExplodedGraph(object): 254 # Parse .dot files with regular expressions. 255 node_re = re.compile( 256 '^(Node0x[0-9a-f]*) \\[shape=record,.*label="{(.*)\\\\l}"\\];$') 257 edge_re = re.compile( 258 '^(Node0x[0-9a-f]*) -> (Node0x[0-9a-f]*);$') 259 260 def __init__(self): 261 super(ExplodedGraph, self).__init__() 262 self.nodes = collections.defaultdict(ExplodedNode) 263 self.root_id = None 264 self.incomplete_line = '' 265 266 def add_raw_line(self, raw_line): 267 if raw_line.startswith('//'): 268 return 269 270 # Allow line breaks by waiting for ';'. This is not valid in 271 # a .dot file, but it is useful for writing tests. 272 if len(raw_line) > 0 and raw_line[-1] != ';': 273 self.incomplete_line += raw_line 274 return 275 raw_line = self.incomplete_line + raw_line 276 self.incomplete_line = '' 277 278 # Apply regexps one by one to see if it's a node or an edge 279 # and extract contents if necessary. 280 logging.debug('Line: ' + raw_line) 281 result = self.edge_re.match(raw_line) 282 if result is not None: 283 logging.debug('Classified as edge line.') 284 pred = result.group(1) 285 succ = result.group(2) 286 self.nodes[pred].successors.append(succ) 287 self.nodes[succ].predecessors.append(pred) 288 return 289 result = self.node_re.match(raw_line) 290 if result is not None: 291 logging.debug('Classified as node line.') 292 node_id = result.group(1) 293 if len(self.nodes) == 0: 294 self.root_id = node_id 295 # Note: when writing tests you don't need to escape everything, 296 # even though in a valid dot file everything is escaped. 297 node_label = result.group(2).replace('\\l', '') \ 298 .replace(' ', '') \ 299 .replace('\\"', '"') \ 300 .replace('\\{', '{') \ 301 .replace('\\}', '}') \ 302 .replace('\\\\', '\\') \ 303 .replace('\\|', '|') \ 304 .replace('\\<', '\\\\<') \ 305 .replace('\\>', '\\\\>') \ 306 .rstrip(',') 307 logging.debug(node_label) 308 json_node = json.loads(node_label) 309 self.nodes[node_id].construct(node_id, json_node) 310 return 311 logging.debug('Skipping.') 312 313 314# A visitor that dumps the ExplodedGraph into a DOT file with fancy HTML-based 315# syntax highlighing. 316class DotDumpVisitor(object): 317 def __init__(self, do_diffs): 318 super(DotDumpVisitor, self).__init__() 319 self._do_diffs = do_diffs 320 321 @staticmethod 322 def _dump_raw(s): 323 print(s, end='') 324 325 @staticmethod 326 def _dump(s): 327 print(s.replace('&', '&') 328 .replace('{', '\\{') 329 .replace('}', '\\}') 330 .replace('\\<', '<') 331 .replace('\\>', '>') 332 .replace('\\l', '<br />') 333 .replace('|', '\\|'), end='') 334 335 @staticmethod 336 def _diff_plus_minus(is_added): 337 if is_added is None: 338 return '' 339 if is_added: 340 return '<font color="forestgreen">+</font>' 341 return '<font color="red">-</font>' 342 343 def visit_begin_graph(self, graph): 344 self._graph = graph 345 self._dump_raw('digraph "ExplodedGraph" {\n') 346 self._dump_raw('label="";\n') 347 348 def visit_program_point(self, p): 349 if p.kind in ['Edge', 'BlockEntrance', 'BlockExit']: 350 color = 'gold3' 351 elif p.kind in ['PreStmtPurgeDeadSymbols', 352 'PostStmtPurgeDeadSymbols']: 353 color = 'red' 354 elif p.kind in ['CallEnter', 'CallExitBegin', 'CallExitEnd']: 355 color = 'blue' 356 elif p.kind in ['Statement']: 357 color = 'cyan3' 358 else: 359 color = 'forestgreen' 360 361 if p.kind == 'Statement': 362 if p.loc is not None: 363 self._dump('<tr><td align="left" width="0">' 364 '%s:<b>%s</b>:<b>%s</b>:</td>' 365 '<td align="left" width="0"><font color="%s">' 366 '%s</font></td><td>%s</td></tr>' 367 % (p.loc.filename, p.loc.line, 368 p.loc.col, color, p.stmt_kind, p.pretty)) 369 else: 370 self._dump('<tr><td align="left" width="0">' 371 '<i>Invalid Source Location</i>:</td>' 372 '<td align="left" width="0">' 373 '<font color="%s">%s</font></td><td>%s</td></tr>' 374 % (color, p.stmt_kind, p.pretty)) 375 elif p.kind == 'Edge': 376 self._dump('<tr><td width="0"></td>' 377 '<td align="left" width="0">' 378 '<font color="%s">%s</font></td><td align="left">' 379 '[B%d] -\\> [B%d]</td></tr>' 380 % (color, p.kind, p.src_id, p.dst_id)) 381 else: 382 # TODO: Print more stuff for other kinds of points. 383 self._dump('<tr><td width="0"></td>' 384 '<td align="left" width="0" colspan="2">' 385 '<font color="%s">%s</font></td></tr>' 386 % (color, p.kind)) 387 388 def visit_environment(self, e, prev_e=None): 389 self._dump('<table border="0">') 390 391 def dump_location_context(lc, is_added=None): 392 self._dump('<tr><td>%s</td>' 393 '<td align="left"><b>%s</b></td>' 394 '<td align="left"><font color="grey60">%s </font>' 395 '%s</td></tr>' 396 % (self._diff_plus_minus(is_added), 397 lc.caption, lc.decl, 398 ('(line %s)' % lc.line) if lc.line is not None 399 else '')) 400 401 def dump_binding(f, b, is_added=None): 402 self._dump('<tr><td>%s</td>' 403 '<td align="left"><i>S%s</i></td>' 404 '<td align="left">%s</td>' 405 '<td align="left">%s</td></tr>' 406 % (self._diff_plus_minus(is_added), 407 b.stmt_id, b.pretty, f.bindings[b])) 408 409 frames_updated = e.diff_frames(prev_e) if prev_e is not None else None 410 if frames_updated: 411 for i in frames_updated: 412 f = e.frames[i] 413 prev_f = prev_e.frames[i] 414 dump_location_context(f.location_context) 415 bindings_removed, bindings_added = f.diff_bindings(prev_f) 416 for b in bindings_removed: 417 dump_binding(prev_f, b, False) 418 for b in bindings_added: 419 dump_binding(f, b, True) 420 else: 421 for f in e.frames: 422 dump_location_context(f.location_context) 423 for b in f.bindings: 424 dump_binding(f, b) 425 426 self._dump('</table>') 427 428 def visit_environment_in_state(self, s, prev_s=None): 429 self._dump('<tr><td align="left">' 430 '<b>Environment: </b>') 431 if s.environment is None: 432 self._dump('<i> Nothing!</i>') 433 else: 434 if prev_s is not None and prev_s.environment is not None: 435 if s.environment.is_different(prev_s.environment): 436 self._dump('</td></tr><tr><td align="left">') 437 self.visit_environment(s.environment, prev_s.environment) 438 else: 439 self._dump('<i> No changes!</i>') 440 else: 441 self._dump('</td></tr><tr><td align="left">') 442 self.visit_environment(s.environment) 443 444 self._dump('</td></tr>') 445 446 def visit_store(self, s, prev_s=None): 447 self._dump('<table border="0">') 448 449 def dump_binding(s, c, b, is_added=None): 450 self._dump('<tr><td>%s</td>' 451 '<td align="left">%s</td>' 452 '<td align="left">%s</td>' 453 '<td align="left">%s</td>' 454 '<td align="left">%s</td></tr>' 455 % (self._diff_plus_minus(is_added), 456 s.clusters[c].base_region, b.offset, 457 '(<i>Default</i>)' if b.kind == 'Default' 458 else '', 459 s.clusters[c].bindings[b])) 460 461 if prev_s is not None: 462 clusters_removed, clusters_added, clusters_updated = \ 463 s.diff_clusters(prev_s) 464 for c in clusters_removed: 465 for b in prev_s.clusters[c].bindings: 466 dump_binding(prev_s, c, b, False) 467 for c in clusters_updated: 468 bindings_removed, bindings_added = \ 469 s.clusters[c].diff_bindings(prev_s.clusters[c]) 470 for b in bindings_removed: 471 dump_binding(prev_s, c, b, False) 472 for b in bindings_added: 473 dump_binding(s, c, b, True) 474 for c in clusters_added: 475 for b in s.clusters[c].bindings: 476 dump_binding(s, c, b, True) 477 else: 478 for c in s.clusters: 479 for b in s.clusters[c].bindings: 480 dump_binding(s, c, b) 481 482 self._dump('</table>') 483 484 def visit_store_in_state(self, s, prev_s=None): 485 self._dump('<tr><td align="left"><b>Store: </b>') 486 if s.store is None: 487 self._dump('<i> Nothing!</i>') 488 else: 489 if prev_s is not None and prev_s.store is not None: 490 if s.store.is_different(prev_s.store): 491 self._dump('</td></tr><tr><td align="left">') 492 self.visit_store(s.store, prev_s.store) 493 else: 494 self._dump('<i> No changes!</i>') 495 else: 496 self._dump('</td></tr><tr><td align="left">') 497 self.visit_store(s.store) 498 self._dump('</td></tr>') 499 500 def visit_generic_map(self, m, prev_m=None): 501 self._dump('<table border="0">') 502 503 def dump_pair(m, k, is_added=None): 504 self._dump('<tr><td>%s</td>' 505 '<td align="left">%s</td>' 506 '<td align="left">%s</td></tr>' 507 % (self._diff_plus_minus(is_added), 508 k, m.generic_map[k])) 509 510 if prev_m is not None: 511 removed, added = m.diff(prev_m) 512 for k in removed: 513 dump_pair(prev_m, k, False) 514 for k in added: 515 dump_pair(m, k, True) 516 else: 517 for k in m.generic_map: 518 dump_pair(m, k, None) 519 520 self._dump('</table>') 521 522 def visit_generic_map_in_state(self, selector, s, prev_s=None): 523 self._dump('<tr><td align="left">' 524 '<b>Ranges: </b>') 525 m = getattr(s, selector) 526 if m is None: 527 self._dump('<i> Nothing!</i>') 528 else: 529 prev_m = None 530 if prev_s is not None: 531 prev_m = getattr(prev_s, selector) 532 if prev_m is not None: 533 if m.is_different(prev_m): 534 self._dump('</td></tr><tr><td align="left">') 535 self.visit_generic_map(m, prev_m) 536 else: 537 self._dump('<i> No changes!</i>') 538 if prev_m is None: 539 self._dump('</td></tr><tr><td align="left">') 540 self.visit_generic_map(m) 541 self._dump('</td></tr>') 542 543 def visit_state(self, s, prev_s): 544 self.visit_store_in_state(s, prev_s) 545 self._dump('<hr />') 546 self.visit_environment_in_state(s, prev_s) 547 self._dump('<hr />') 548 self.visit_generic_map_in_state('constraints', s, prev_s) 549 550 def visit_node(self, node): 551 self._dump('%s [shape=record,label=<<table border="0">' 552 % (node.node_name())) 553 554 self._dump('<tr><td bgcolor="grey"><b>Node %d (%s) - ' 555 'State %s</b></td></tr>' 556 % (node.node_id, node.ptr, node.state.state_id 557 if node.state is not None else 'Unspecified')) 558 self._dump('<tr><td align="left" width="0">') 559 if len(node.points) > 1: 560 self._dump('<b>Program points:</b></td></tr>') 561 else: 562 self._dump('<b>Program point:</b></td></tr>') 563 self._dump('<tr><td align="left" width="0">' 564 '<table border="0" align="left" width="0">') 565 for p in node.points: 566 self.visit_program_point(p) 567 self._dump('</table></td></tr>') 568 569 if node.state is not None: 570 self._dump('<hr />') 571 prev_s = None 572 # Do diffs only when we have a unique predecessor. 573 # Don't do diffs on the leaf nodes because they're 574 # the important ones. 575 if self._do_diffs and len(node.predecessors) == 1 \ 576 and len(node.successors) > 0: 577 prev_s = self._graph.nodes[node.predecessors[0]].state 578 self.visit_state(node.state, prev_s) 579 self._dump_raw('</table>>];\n') 580 581 def visit_edge(self, pred, succ): 582 self._dump_raw('%s -> %s;\n' % (pred.node_name(), succ.node_name())) 583 584 def visit_end_of_graph(self): 585 self._dump_raw('}\n') 586 587 588# A class that encapsulates traversal of the ExplodedGraph. Different explorer 589# kinds could potentially traverse specific sub-graphs. 590class Explorer(object): 591 def __init__(self): 592 super(Explorer, self).__init__() 593 594 def explore(self, graph, visitor): 595 visitor.visit_begin_graph(graph) 596 for node in sorted(graph.nodes): 597 logging.debug('Visiting ' + node) 598 visitor.visit_node(graph.nodes[node]) 599 for succ in sorted(graph.nodes[node].successors): 600 logging.debug('Visiting edge: %s -> %s ' % (node, succ)) 601 visitor.visit_edge(graph.nodes[node], graph.nodes[succ]) 602 visitor.visit_end_of_graph() 603 604 605def main(): 606 parser = argparse.ArgumentParser() 607 parser.add_argument('filename', type=str) 608 parser.add_argument('-v', '--verbose', action='store_const', 609 dest='loglevel', const=logging.DEBUG, 610 default=logging.WARNING, 611 help='enable info prints') 612 parser.add_argument('-d', '--diff', action='store_const', dest='diff', 613 const=True, default=False, 614 help='display differences between states') 615 args = parser.parse_args() 616 logging.basicConfig(level=args.loglevel) 617 618 graph = ExplodedGraph() 619 with open(args.filename) as fd: 620 for raw_line in fd: 621 raw_line = raw_line.strip() 622 graph.add_raw_line(raw_line) 623 624 explorer = Explorer() 625 visitor = DotDumpVisitor(args.diff) 626 explorer.explore(graph, visitor) 627 628 629if __name__ == '__main__': 630 main() 631