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 deserialized source location. 22class SourceLocation(object): 23 def __init__(self, json_loc): 24 super(SourceLocation, self).__init__() 25 self.line = json_loc['line'] 26 self.col = json_loc['column'] 27 self.filename = json_loc['filename'] \ 28 if 'filename' in json_loc else '(main file)' 29 30 31# A deserialized program point. 32class ProgramPoint(object): 33 def __init__(self, json_pp): 34 super(ProgramPoint, self).__init__() 35 self.kind = json_pp['kind'] 36 self.tag = json_pp['tag'] 37 if self.kind == 'Edge': 38 self.src_id = json_pp['src_id'] 39 self.dst_id = json_pp['dst_id'] 40 elif self.kind == 'Statement': 41 self.stmt_kind = json_pp['stmt_kind'] 42 self.pointer = json_pp['pointer'] 43 self.pretty = json_pp['pretty'] 44 self.loc = SourceLocation(json_pp['location']) \ 45 if json_pp['location'] is not None else None 46 elif self.kind == 'BlockEntrance': 47 self.block_id = json_pp['block_id'] 48 49 50# A value of a single expression in a deserialized Environment. 51class EnvironmentBinding(object): 52 def __init__(self, json_eb): 53 super(EnvironmentBinding, self).__init__() 54 self.stmt_id = json_eb['stmt_id'] 55 self.pretty = json_eb['pretty'] 56 self.value = json_eb['value'] 57 58 59# Deserialized description of a location context. 60class LocationContext(object): 61 def __init__(self, json_frame): 62 super(LocationContext, self).__init__() 63 self.lctx_id = json_frame['lctx_id'] 64 self.caption = json_frame['location_context'] 65 self.decl = json_frame['calling'] 66 self.line = json_frame['call_line'] 67 68 69# A group of deserialized Environment bindings that correspond to a specific 70# location context. 71class EnvironmentFrame(object): 72 def __init__(self, json_frame): 73 super(EnvironmentFrame, self).__init__() 74 self.location_context = LocationContext(json_frame) 75 self.bindings = [EnvironmentBinding(b) for b in json_frame['items']] \ 76 if json_frame['items'] is not None else [] 77 78 79# A deserialized Environment. 80class Environment(object): 81 def __init__(self, json_e): 82 super(Environment, self).__init__() 83 self.frames = [EnvironmentFrame(f) for f in json_e] 84 85 86# A single binding in a deserialized RegionStore cluster. 87class StoreBinding(object): 88 def __init__(self, json_sb): 89 super(StoreBinding, self).__init__() 90 self.kind = json_sb['kind'] 91 self.offset = json_sb['offset'] 92 self.value = json_sb['value'] 93 94 95# A single cluster of the deserialized RegionStore. 96class StoreCluster(object): 97 def __init__(self, json_sc): 98 super(StoreCluster, self).__init__() 99 self.base_region = json_sc['cluster'] 100 self.bindings = [StoreBinding(b) for b in json_sc['items']] 101 102 103# A deserialized RegionStore. 104class Store(object): 105 def __init__(self, json_s): 106 super(Store, self).__init__() 107 self.clusters = [StoreCluster(c) for c in json_s] 108 109 110# A deserialized program state. 111class ProgramState(object): 112 def __init__(self, state_id, json_ps): 113 super(ProgramState, self).__init__() 114 logging.debug('Adding ProgramState ' + str(state_id)) 115 116 self.state_id = state_id 117 self.store = Store(json_ps['store']) \ 118 if json_ps['store'] is not None else None 119 self.environment = Environment(json_ps['environment']) \ 120 if json_ps['environment'] is not None else None 121 # TODO: Objects under construction. 122 # TODO: Constraint ranges. 123 # TODO: Dynamic types of objects. 124 # TODO: Checker messages. 125 126 127# A deserialized exploded graph node. Has a default constructor because it 128# may be referenced as part of an edge before its contents are deserialized, 129# and in this moment we already need a room for predecessors and successors. 130class ExplodedNode(object): 131 def __init__(self): 132 super(ExplodedNode, self).__init__() 133 self.predecessors = [] 134 self.successors = [] 135 136 def construct(self, node_id, json_node): 137 logging.debug('Adding ' + node_id) 138 self.node_id = json_node['node_id'] 139 self.ptr = json_node['pointer'] 140 self.points = [ProgramPoint(p) for p in json_node['program_points']] 141 self.state = ProgramState(json_node['state_id'], 142 json_node['program_state']) \ 143 if json_node['program_state'] is not None else None 144 145 assert self.node_name() == node_id 146 147 def node_name(self): 148 return 'Node' + self.ptr 149 150 151# A deserialized ExplodedGraph. Constructed by consuming a .dot file 152# line-by-line. 153class ExplodedGraph(object): 154 # Parse .dot files with regular expressions. 155 node_re = re.compile( 156 '^(Node0x[0-9a-f]*) \\[shape=record,.*label="{(.*)\\\\l}"\\];$') 157 edge_re = re.compile( 158 '^(Node0x[0-9a-f]*) -> (Node0x[0-9a-f]*);$') 159 160 def __init__(self): 161 super(ExplodedGraph, self).__init__() 162 self.nodes = collections.defaultdict(ExplodedNode) 163 self.root_id = None 164 self.incomplete_line = '' 165 166 def add_raw_line(self, raw_line): 167 if raw_line.startswith('//'): 168 return 169 170 # Allow line breaks by waiting for ';'. This is not valid in 171 # a .dot file, but it is useful for writing tests. 172 if len(raw_line) > 0 and raw_line[-1] != ';': 173 self.incomplete_line += raw_line 174 return 175 raw_line = self.incomplete_line + raw_line 176 self.incomplete_line = '' 177 178 # Apply regexps one by one to see if it's a node or an edge 179 # and extract contents if necessary. 180 logging.debug('Line: ' + raw_line) 181 result = self.edge_re.match(raw_line) 182 if result is not None: 183 logging.debug('Classified as edge line.') 184 pred = result.group(1) 185 succ = result.group(2) 186 self.nodes[pred].successors.append(succ) 187 self.nodes[succ].predecessors.append(pred) 188 return 189 result = self.node_re.match(raw_line) 190 if result is not None: 191 logging.debug('Classified as node line.') 192 node_id = result.group(1) 193 if len(self.nodes) == 0: 194 self.root_id = node_id 195 # Note: when writing tests you don't need to escape everything, 196 # even though in a valid dot file everything is escaped. 197 node_label = result.group(2).replace('\\l', '') \ 198 .replace(' ', '') \ 199 .replace('\\"', '"') \ 200 .replace('\\{', '{') \ 201 .replace('\\}', '}') \ 202 .replace('\\\\', '\\') \ 203 .replace('\\<', '\\\\<') \ 204 .replace('\\>', '\\\\>') \ 205 .rstrip(',') 206 logging.debug(node_label) 207 json_node = json.loads(node_label) 208 self.nodes[node_id].construct(node_id, json_node) 209 return 210 logging.debug('Skipping.') 211 212 213# A visitor that dumps the ExplodedGraph into a DOT file with fancy HTML-based 214# syntax highlighing. 215class DotDumpVisitor(object): 216 def __init__(self): 217 super(DotDumpVisitor, self).__init__() 218 219 @staticmethod 220 def _dump_raw(s): 221 print(s, end='') 222 223 @staticmethod 224 def _dump(s): 225 print(s.replace('&', '&') 226 .replace('{', '\\{') 227 .replace('}', '\\}') 228 .replace('\\<', '<') 229 .replace('\\>', '>') 230 .replace('\\l', '<br />') 231 .replace('|', ''), end='') 232 233 def visit_begin_graph(self, graph): 234 self._graph = graph 235 self._dump_raw('digraph "ExplodedGraph" {\n') 236 self._dump_raw('label="";\n') 237 238 def visit_program_point(self, p): 239 if p.kind in ['Edge', 'BlockEntrance', 'BlockExit']: 240 color = 'gold3' 241 elif p.kind in ['PreStmtPurgeDeadSymbols', 242 'PostStmtPurgeDeadSymbols']: 243 color = 'red' 244 elif p.kind in ['CallEnter', 'CallExitBegin', 'CallExitEnd']: 245 color = 'blue' 246 elif p.kind in ['Statement']: 247 color = 'cyan3' 248 else: 249 color = 'forestgreen' 250 251 if p.kind == 'Statement': 252 if p.loc is not None: 253 self._dump('<tr><td align="left" width="0">' 254 '%s:<b>%s</b>:<b>%s</b>:</td>' 255 '<td align="left" width="0"><font color="%s">' 256 '%s</font></td><td>%s</td></tr>' 257 % (p.loc.filename, p.loc.line, 258 p.loc.col, color, p.stmt_kind, p.pretty)) 259 else: 260 self._dump('<tr><td align="left" width="0">' 261 '<i>Invalid Source Location</i>:</td>' 262 '<td align="left" width="0">' 263 '<font color="%s">%s</font></td><td>%s</td></tr>' 264 % (color, p.stmt_kind, p.pretty)) 265 elif p.kind == 'Edge': 266 self._dump('<tr><td width="0">-</td>' 267 '<td align="left" width="0">' 268 '<font color="%s">%s</font></td><td align="left">' 269 '[B%d] -\\> [B%d]</td></tr>' 270 % (color, p.kind, p.src_id, p.dst_id)) 271 else: 272 # TODO: Print more stuff for other kinds of points. 273 self._dump('<tr><td width="0">-</td>' 274 '<td align="left" width="0" colspan="2">' 275 '<font color="%s">%s</font></td></tr>' 276 % (color, p.kind)) 277 278 def visit_environment(self, e): 279 self._dump('<table border="0">') 280 281 for f in e.frames: 282 self._dump('<tr><td align="left"><b>%s</b></td>' 283 '<td align="left"><font color="grey60">%s </font>' 284 '%s</td></tr>' 285 % (f.location_context.caption, 286 f.location_context.decl, 287 ('(line %s)' % f.location_context.line) 288 if f.location_context.line is not None else '')) 289 for b in f.bindings: 290 self._dump('<tr><td align="left"><i>S%s</i></td>' 291 '<td align="left">%s</td>' 292 '<td align="left">%s</td></tr>' 293 % (b.stmt_id, b.pretty, b.value)) 294 295 self._dump('</table>') 296 297 def visit_store(self, s): 298 self._dump('<table border="0">') 299 300 for c in s.clusters: 301 for b in c.bindings: 302 self._dump('<tr><td align="left">%s</td>' 303 '<td align="left">%s</td>' 304 '<td align="left">%s</td>' 305 '<td align="left">%s</td></tr>' 306 % (c.base_region, b.offset, 307 '(<i>Default</i>)' if b.kind == 'Default' 308 else '', 309 b.value)) 310 311 self._dump('</table>') 312 313 def visit_state(self, s): 314 self._dump('<tr><td align="left">' 315 '<b>Store: </b>') 316 if s.store is None: 317 self._dump('<i> Nothing!</i>') 318 else: 319 self._dump('</td></tr>' 320 '<tr><td align="left">') 321 self.visit_store(s.store) 322 323 self._dump('</td></tr><hr />' 324 '<tr><td align="left">' 325 '<b>Environment: </b>') 326 if s.environment is None: 327 self._dump('<i> Nothing!</i>') 328 else: 329 self._dump('</td></tr>' 330 '<tr><td align="left">') 331 self.visit_environment(s.environment) 332 333 self._dump('</td></tr>') 334 335 def visit_node(self, node): 336 self._dump('%s [shape=record,label=<<table border="0">' 337 % (node.node_name())) 338 339 self._dump('<tr><td bgcolor="grey"><b>Node %d (%s) - ' 340 'State %s</b></td></tr>' 341 % (node.node_id, node.ptr, node.state.state_id 342 if node.state is not None else 'Unspecified')) 343 self._dump('<tr><td align="left" width="0">') 344 if len(node.points) > 1: 345 self._dump('<b>Program points:</b></td></tr>') 346 else: 347 self._dump('<b>Program point:</b></td></tr>') 348 self._dump('<tr><td align="left" width="0">' 349 '<table border="0" align="left" width="0">') 350 for p in node.points: 351 self.visit_program_point(p) 352 self._dump('</table></td></tr>') 353 354 if node.state is not None: 355 self._dump('<hr />') 356 self.visit_state(node.state) 357 self._dump_raw('</table>>];\n') 358 359 def visit_edge(self, pred, succ): 360 self._dump_raw('%s -> %s;\n' % (pred.node_name(), succ.node_name())) 361 362 def visit_end_of_graph(self): 363 self._dump_raw('}\n') 364 365 366# A class that encapsulates traversal of the ExplodedGraph. Different explorer 367# kinds could potentially traverse specific sub-graphs. 368class Explorer(object): 369 def __init__(self): 370 super(Explorer, self).__init__() 371 372 def explore(self, graph, visitor): 373 visitor.visit_begin_graph(graph) 374 for node in sorted(graph.nodes): 375 logging.debug('Visiting ' + node) 376 visitor.visit_node(graph.nodes[node]) 377 for succ in sorted(graph.nodes[node].successors): 378 logging.debug('Visiting edge: %s -> %s ' % (node, succ)) 379 visitor.visit_edge(graph.nodes[node], graph.nodes[succ]) 380 visitor.visit_end_of_graph() 381 382 383def main(): 384 parser = argparse.ArgumentParser() 385 parser.add_argument('filename', type=str) 386 parser.add_argument('-d', '--debug', action='store_const', dest='loglevel', 387 const=logging.DEBUG, default=logging.WARNING, 388 help='enable debug prints') 389 parser.add_argument('-v', '--verbose', action='store_const', 390 dest='loglevel', const=logging.INFO, 391 default=logging.WARNING, 392 help='enable info prints') 393 args = parser.parse_args() 394 logging.basicConfig(level=args.loglevel) 395 396 graph = ExplodedGraph() 397 with open(args.filename) as fd: 398 for raw_line in fd: 399 raw_line = raw_line.strip() 400 graph.add_raw_line(raw_line) 401 402 explorer = Explorer() 403 visitor = DotDumpVisitor() 404 explorer.explore(graph, visitor) 405 406 407if __name__ == '__main__': 408 main() 409