xref: /llvm-project/clang/utils/analyzer/exploded-graph-rewriter.py (revision 78566e454f68248ab18ed6b43cde64950096c98e)
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 re
20
21
22#===-----------------------------------------------------------------------===#
23# These data structures represent a deserialized ExplodedGraph.
24#===-----------------------------------------------------------------------===#
25
26
27# A helper function for finding the difference between two dictionaries.
28def diff_dicts(curr, prev):
29    removed = [k for k in prev if k not in curr or curr[k] != prev[k]]
30    added = [k for k in curr if k not in prev or curr[k] != prev[k]]
31    return (removed, added)
32
33
34# Represents any program state trait that is a dictionary of key-value pairs.
35class GenericMap(object):
36    def __init__(self, items):
37        self.generic_map = collections.OrderedDict(items)
38
39    def diff(self, prev):
40        return diff_dicts(self.generic_map, prev.generic_map)
41
42    def is_different(self, prev):
43        removed, added = self.diff(prev)
44        return len(removed) != 0 or len(added) != 0
45
46
47# A deserialized source location.
48class SourceLocation(object):
49    def __init__(self, json_loc):
50        super(SourceLocation, self).__init__()
51        self.line = json_loc['line']
52        self.col = json_loc['column']
53        self.filename = json_loc['filename'] \
54            if 'filename' in json_loc else '(main file)'
55
56
57# A deserialized program point.
58class ProgramPoint(object):
59    def __init__(self, json_pp):
60        super(ProgramPoint, self).__init__()
61        self.kind = json_pp['kind']
62        self.tag = json_pp['tag']
63        if self.kind == 'Edge':
64            self.src_id = json_pp['src_id']
65            self.dst_id = json_pp['dst_id']
66        elif self.kind == 'Statement':
67            self.stmt_kind = json_pp['stmt_kind']
68            self.stmt_point_kind = json_pp['stmt_point_kind']
69            self.pointer = json_pp['pointer']
70            self.pretty = json_pp['pretty']
71            self.loc = SourceLocation(json_pp['location']) \
72                if json_pp['location'] is not None else None
73        elif self.kind == 'BlockEntrance':
74            self.block_id = json_pp['block_id']
75
76
77# A single expression acting as a key in a deserialized Environment.
78class EnvironmentBindingKey(object):
79    def __init__(self, json_ek):
80        super(EnvironmentBindingKey, self).__init__()
81        # CXXCtorInitializer is not a Stmt!
82        self.stmt_id = json_ek['stmt_id'] if 'stmt_id' in json_ek \
83            else json_ek['init_id']
84        self.pretty = json_ek['pretty']
85        self.kind = json_ek['kind'] if 'kind' in json_ek else None
86
87    def _key(self):
88        return self.stmt_id
89
90    def __eq__(self, other):
91        return self._key() == other._key()
92
93    def __hash__(self):
94        return hash(self._key())
95
96
97# Deserialized description of a location context.
98class LocationContext(object):
99    def __init__(self, json_frame):
100        super(LocationContext, self).__init__()
101        self.lctx_id = json_frame['lctx_id']
102        self.caption = json_frame['location_context']
103        self.decl = json_frame['calling']
104        self.line = json_frame['call_line']
105
106    def _key(self):
107        return self.lctx_id
108
109    def __eq__(self, other):
110        return self._key() == other._key()
111
112    def __hash__(self):
113        return hash(self._key())
114
115
116# A group of deserialized Environment bindings that correspond to a specific
117# location context.
118class EnvironmentFrame(object):
119    def __init__(self, json_frame):
120        super(EnvironmentFrame, self).__init__()
121        self.location_context = LocationContext(json_frame)
122        self.bindings = collections.OrderedDict(
123            [(EnvironmentBindingKey(b),
124              b['value']) for b in json_frame['items']]
125            if json_frame['items'] is not None else [])
126
127    def diff_bindings(self, prev):
128        return diff_dicts(self.bindings, prev.bindings)
129
130    def is_different(self, prev):
131        removed, added = self.diff_bindings(prev)
132        return len(removed) != 0 or len(added) != 0
133
134
135# A deserialized Environment. This class can also hold other entities that
136# are similar to Environment, such as Objects Under Construction.
137class GenericEnvironment(object):
138    def __init__(self, json_e):
139        super(GenericEnvironment, self).__init__()
140        self.frames = [EnvironmentFrame(f) for f in json_e]
141
142    def diff_frames(self, prev):
143        # TODO: It's difficult to display a good diff when frame numbers shift.
144        if len(self.frames) != len(prev.frames):
145            return None
146
147        updated = []
148        for i in range(len(self.frames)):
149            f = self.frames[i]
150            prev_f = prev.frames[i]
151            if f.location_context == prev_f.location_context:
152                if f.is_different(prev_f):
153                    updated.append(i)
154            else:
155                # We have the whole frame replaced with another frame.
156                # TODO: Produce a nice diff.
157                return None
158
159        # TODO: Add support for added/removed.
160        return updated
161
162    def is_different(self, prev):
163        updated = self.diff_frames(prev)
164        return updated is None or len(updated) > 0
165
166
167# A single binding key in a deserialized RegionStore cluster.
168class StoreBindingKey(object):
169    def __init__(self, json_sk):
170        super(StoreBindingKey, self).__init__()
171        self.kind = json_sk['kind']
172        self.offset = json_sk['offset']
173
174    def _key(self):
175        return (self.kind, self.offset)
176
177    def __eq__(self, other):
178        return self._key() == other._key()
179
180    def __hash__(self):
181        return hash(self._key())
182
183
184# A single cluster of the deserialized RegionStore.
185class StoreCluster(object):
186    def __init__(self, json_sc):
187        super(StoreCluster, self).__init__()
188        self.base_region = json_sc['cluster']
189        self.bindings = collections.OrderedDict(
190            [(StoreBindingKey(b), b['value']) for b in json_sc['items']])
191
192    def diff_bindings(self, prev):
193        return diff_dicts(self.bindings, prev.bindings)
194
195    def is_different(self, prev):
196        removed, added = self.diff_bindings(prev)
197        return len(removed) != 0 or len(added) != 0
198
199
200# A deserialized RegionStore.
201class Store(object):
202    def __init__(self, json_s):
203        super(Store, self).__init__()
204        self.ptr = json_s['pointer']
205        self.clusters = collections.OrderedDict(
206            [(c['pointer'], StoreCluster(c)) for c in json_s['items']])
207
208    def diff_clusters(self, prev):
209        removed = [k for k in prev.clusters if k not in self.clusters]
210        added = [k for k in self.clusters if k not in prev.clusters]
211        updated = [k for k in prev.clusters if k in self.clusters
212                   and prev.clusters[k].is_different(self.clusters[k])]
213        return (removed, added, updated)
214
215    def is_different(self, prev):
216        removed, added, updated = self.diff_clusters(prev)
217        return len(removed) != 0 or len(added) != 0 or len(updated) != 0
218
219
220# Deserialized messages from a single checker in a single program state.
221# Basically a list of raw strings.
222class CheckerLines(object):
223    def __init__(self, json_lines):
224        super(CheckerLines, self).__init__()
225        self.lines = json_lines
226
227    def diff_lines(self, prev):
228        lines = difflib.ndiff(prev.lines, self.lines)
229        return [l.strip() for l in lines
230                if l.startswith('+') or l.startswith('-')]
231
232    def is_different(self, prev):
233        return len(self.diff_lines(prev)) > 0
234
235
236# Deserialized messages of all checkers, separated by checker.
237class CheckerMessages(object):
238    def __init__(self, json_m):
239        super(CheckerMessages, self).__init__()
240        self.items = collections.OrderedDict(
241            [(m['checker'], CheckerLines(m['messages'])) for m in json_m])
242
243    def diff_messages(self, prev):
244        removed = [k for k in prev.items if k not in self.items]
245        added = [k for k in self.items if k not in prev.items]
246        updated = [k for k in prev.items if k in self.items
247                   and prev.items[k].is_different(self.items[k])]
248        return (removed, added, updated)
249
250    def is_different(self, prev):
251        removed, added, updated = self.diff_messages(prev)
252        return len(removed) != 0 or len(added) != 0 or len(updated) != 0
253
254
255# A deserialized program state.
256class ProgramState(object):
257    def __init__(self, state_id, json_ps):
258        super(ProgramState, self).__init__()
259        logging.debug('Adding ProgramState ' + str(state_id))
260
261        self.state_id = state_id
262
263        self.store = Store(json_ps['store']) \
264            if json_ps['store'] is not None else None
265
266        self.environment = \
267            GenericEnvironment(json_ps['environment']['items']) \
268            if json_ps['environment'] is not None else None
269
270        self.constraints = GenericMap([
271            (c['symbol'], c['range']) for c in json_ps['constraints']
272        ]) if json_ps['constraints'] is not None else None
273
274        self.dynamic_types = GenericMap([
275                (t['region'], '%s%s' % (t['dyn_type'],
276                                        ' (or a sub-class)'
277                                        if t['sub_classable'] else ''))
278                for t in json_ps['dynamic_types']]) \
279            if json_ps['dynamic_types'] is not None else None
280
281        self.constructing_objects = \
282            GenericEnvironment(json_ps['constructing_objects']) \
283            if json_ps['constructing_objects'] is not None else None
284
285        self.checker_messages = CheckerMessages(json_ps['checker_messages']) \
286            if json_ps['checker_messages'] is not None else None
287
288
289# A deserialized exploded graph node. Has a default constructor because it
290# may be referenced as part of an edge before its contents are deserialized,
291# and in this moment we already need a room for predecessors and successors.
292class ExplodedNode(object):
293    def __init__(self):
294        super(ExplodedNode, self).__init__()
295        self.predecessors = []
296        self.successors = []
297
298    def construct(self, node_id, json_node):
299        logging.debug('Adding ' + node_id)
300        self.node_id = json_node['node_id']
301        self.ptr = json_node['pointer']
302        self.has_report = json_node['has_report']
303        self.is_sink = json_node['is_sink']
304        self.points = [ProgramPoint(p) for p in json_node['program_points']]
305        self.state = ProgramState(json_node['state_id'],
306                                  json_node['program_state']) \
307            if json_node['program_state'] is not None else None
308
309        assert self.node_name() == node_id
310
311    def node_name(self):
312        return 'Node' + self.ptr
313
314
315# A deserialized ExplodedGraph. Constructed by consuming a .dot file
316# line-by-line.
317class ExplodedGraph(object):
318    # Parse .dot files with regular expressions.
319    node_re = re.compile(
320        '^(Node0x[0-9a-f]*) \\[shape=record,.*label="{(.*)\\\\l}"\\];$')
321    edge_re = re.compile(
322        '^(Node0x[0-9a-f]*) -> (Node0x[0-9a-f]*);$')
323
324    def __init__(self):
325        super(ExplodedGraph, self).__init__()
326        self.nodes = collections.defaultdict(ExplodedNode)
327        self.root_id = None
328        self.incomplete_line = ''
329
330    def add_raw_line(self, raw_line):
331        if raw_line.startswith('//'):
332            return
333
334        # Allow line breaks by waiting for ';'. This is not valid in
335        # a .dot file, but it is useful for writing tests.
336        if len(raw_line) > 0 and raw_line[-1] != ';':
337            self.incomplete_line += raw_line
338            return
339        raw_line = self.incomplete_line + raw_line
340        self.incomplete_line = ''
341
342        # Apply regexps one by one to see if it's a node or an edge
343        # and extract contents if necessary.
344        logging.debug('Line: ' + raw_line)
345        result = self.edge_re.match(raw_line)
346        if result is not None:
347            logging.debug('Classified as edge line.')
348            pred = result.group(1)
349            succ = result.group(2)
350            self.nodes[pred].successors.append(succ)
351            self.nodes[succ].predecessors.append(pred)
352            return
353        result = self.node_re.match(raw_line)
354        if result is not None:
355            logging.debug('Classified as node line.')
356            node_id = result.group(1)
357            if len(self.nodes) == 0:
358                self.root_id = node_id
359            # Note: when writing tests you don't need to escape everything,
360            # even though in a valid dot file everything is escaped.
361            node_label = result.group(2).replace('\\l', '') \
362                                        .replace(' ', '') \
363                                        .replace('\\"', '"') \
364                                        .replace('\\{', '{') \
365                                        .replace('\\}', '}') \
366                                        .replace('\\\\', '\\') \
367                                        .replace('\\|', '|') \
368                                        .replace('\\<', '\\\\<') \
369                                        .replace('\\>', '\\\\>') \
370                                        .rstrip(',')
371            logging.debug(node_label)
372            json_node = json.loads(node_label)
373            self.nodes[node_id].construct(node_id, json_node)
374            return
375        logging.debug('Skipping.')
376
377
378#===-----------------------------------------------------------------------===#
379# Visitors traverse a deserialized ExplodedGraph and do different things
380# with every node and edge.
381#===-----------------------------------------------------------------------===#
382
383
384# A visitor that dumps the ExplodedGraph into a DOT file with fancy HTML-based
385# syntax highlighing.
386class DotDumpVisitor(object):
387    def __init__(self, do_diffs, dark_mode, gray_mode):
388        super(DotDumpVisitor, self).__init__()
389        self._do_diffs = do_diffs
390        self._dark_mode = dark_mode
391        self._gray_mode = gray_mode
392
393    @staticmethod
394    def _dump_raw(s):
395        print(s, end='')
396
397    def _dump(self, s):
398        s = s.replace('&', '&amp;') \
399             .replace('{', '\\{') \
400             .replace('}', '\\}') \
401             .replace('\\<', '&lt;') \
402             .replace('\\>', '&gt;') \
403             .replace('\\l', '<br />') \
404             .replace('|', '\\|')
405        if self._gray_mode:
406            s = re.sub(r'<font color="[a-z0-9]*">', '', s)
407            s = re.sub(r'</font>', '', s)
408        self._dump_raw(s)
409
410    @staticmethod
411    def _diff_plus_minus(is_added):
412        if is_added is None:
413            return ''
414        if is_added:
415            return '<font color="forestgreen">+</font>'
416        return '<font color="red">-</font>'
417
418    @staticmethod
419    def _short_pretty(s):
420        if s is None:
421            return None
422        if len(s) < 20:
423            return s
424        left = s.find('{')
425        right = s.rfind('}')
426        if left == -1 or right == -1 or left >= right:
427            return s
428        candidate = s[0:left + 1] + ' ... ' + s[right:]
429        if len(candidate) >= len(s):
430            return s
431        return candidate
432
433    def visit_begin_graph(self, graph):
434        self._graph = graph
435        self._dump_raw('digraph "ExplodedGraph" {\n')
436        if self._dark_mode:
437            self._dump_raw('bgcolor="gray10";\n')
438        self._dump_raw('label="";\n')
439
440    def visit_program_point(self, p):
441        if p.kind in ['Edge', 'BlockEntrance', 'BlockExit']:
442            color = 'gold3'
443        elif p.kind in ['PreStmtPurgeDeadSymbols',
444                        'PostStmtPurgeDeadSymbols']:
445            color = 'red'
446        elif p.kind in ['CallEnter', 'CallExitBegin', 'CallExitEnd']:
447            color = 'dodgerblue' if self._dark_mode else 'blue'
448        elif p.kind in ['Statement']:
449            color = 'cyan4'
450        else:
451            color = 'forestgreen'
452
453        if p.kind == 'Statement':
454            # This avoids pretty-printing huge statements such as CompoundStmt.
455            # Such statements show up only at [Pre|Post]StmtPurgeDeadSymbols
456            skip_pretty = 'PurgeDeadSymbols' in p.stmt_point_kind
457            stmt_color = 'cyan3'
458            if p.loc is not None:
459                self._dump('<tr><td align="left" width="0">'
460                           '%s:<b>%s</b>:<b>%s</b>:</td>'
461                           '<td align="left" width="0"><font color="%s">'
462                           '%s</font></td>'
463                           '<td align="left"><font color="%s">%s</font></td>'
464                           '<td>%s</td></tr>'
465                           % (p.loc.filename, p.loc.line,
466                              p.loc.col, color, p.stmt_kind,
467                              stmt_color, p.stmt_point_kind,
468                              self._short_pretty(p.pretty)
469                              if not skip_pretty else ''))
470            else:
471                self._dump('<tr><td align="left" width="0">'
472                           '<i>Invalid Source Location</i>:</td>'
473                           '<td align="left" width="0">'
474                           '<font color="%s">%s</font></td>'
475                           '<td align="left"><font color="%s">%s</font></td>'
476                           '<td>%s</td></tr>'
477                           % (color, p.stmt_kind,
478                              stmt_color, p.stmt_point_kind,
479                              self._short_pretty(p.pretty)
480                              if not skip_pretty else ''))
481        elif p.kind == 'Edge':
482            self._dump('<tr><td width="0"></td>'
483                       '<td align="left" width="0">'
484                       '<font color="%s">%s</font></td><td align="left">'
485                       '[B%d] -\\> [B%d]</td></tr>'
486                       % (color, 'BlockEdge', p.src_id, p.dst_id))
487        elif p.kind == 'BlockEntrance':
488            self._dump('<tr><td width="0"></td>'
489                       '<td align="left" width="0">'
490                       '<font color="%s">%s</font></td>'
491                       '<td align="left">[B%d]</td></tr>'
492                       % (color, p.kind, p.block_id))
493        else:
494            # TODO: Print more stuff for other kinds of points.
495            self._dump('<tr><td width="0"></td>'
496                       '<td align="left" width="0" colspan="2">'
497                       '<font color="%s">%s</font></td></tr>'
498                       % (color, p.kind))
499
500        if p.tag is not None:
501            self._dump('<tr><td width="0"></td>'
502                       '<td colspan="3" align="left">'
503                       '<b>Tag: </b> <font color="crimson">'
504                       '%s</font></td></tr>' % p.tag)
505
506    def visit_environment(self, e, prev_e=None):
507        self._dump('<table border="0">')
508
509        def dump_location_context(lc, is_added=None):
510            self._dump('<tr><td>%s</td>'
511                       '<td align="left"><b>%s</b></td>'
512                       '<td align="left" colspan="2">'
513                       '<font color="gray60">%s </font>'
514                       '%s</td></tr>'
515                       % (self._diff_plus_minus(is_added),
516                          lc.caption, lc.decl,
517                          ('(line %s)' % lc.line) if lc.line is not None
518                          else ''))
519
520        def dump_binding(f, b, is_added=None):
521            self._dump('<tr><td>%s</td>'
522                       '<td align="left"><i>S%s</i></td>'
523                       '%s'
524                       '<td align="left">%s</td>'
525                       '<td align="left">%s</td></tr>'
526                       % (self._diff_plus_minus(is_added),
527                          b.stmt_id,
528                          '<td align="left"><font color="%s"><i>'
529                          '%s</i></font></td>' % (
530                              'lavender' if self._dark_mode else 'darkgreen',
531                              ('(%s)' % b.kind) if b.kind is not None else ' '
532                          ),
533                          self._short_pretty(b.pretty), f.bindings[b]))
534
535        frames_updated = e.diff_frames(prev_e) if prev_e is not None else None
536        if frames_updated:
537            for i in frames_updated:
538                f = e.frames[i]
539                prev_f = prev_e.frames[i]
540                dump_location_context(f.location_context)
541                bindings_removed, bindings_added = f.diff_bindings(prev_f)
542                for b in bindings_removed:
543                    dump_binding(prev_f, b, False)
544                for b in bindings_added:
545                    dump_binding(f, b, True)
546        else:
547            for f in e.frames:
548                dump_location_context(f.location_context)
549                for b in f.bindings:
550                    dump_binding(f, b)
551
552        self._dump('</table>')
553
554    def visit_environment_in_state(self, selector, title, s, prev_s=None):
555        e = getattr(s, selector)
556        prev_e = getattr(prev_s, selector) if prev_s is not None else None
557        if e is None and prev_e is None:
558            return
559
560        self._dump('<hr /><tr><td align="left"><b>%s: </b>' % title)
561        if e is None:
562            self._dump('<i> Nothing!</i>')
563        else:
564            if prev_e is not None:
565                if e.is_different(prev_e):
566                    self._dump('</td></tr><tr><td align="left">')
567                    self.visit_environment(e, prev_e)
568                else:
569                    self._dump('<i> No changes!</i>')
570            else:
571                self._dump('</td></tr><tr><td align="left">')
572                self.visit_environment(e)
573
574        self._dump('</td></tr>')
575
576    def visit_store(self, s, prev_s=None):
577        self._dump('<table border="0">')
578
579        def dump_binding(s, c, b, is_added=None):
580            self._dump('<tr><td>%s</td>'
581                       '<td align="left">%s</td>'
582                       '<td align="left">%s</td>'
583                       '<td align="left">%s</td>'
584                       '<td align="left">%s</td></tr>'
585                       % (self._diff_plus_minus(is_added),
586                          s.clusters[c].base_region, b.offset,
587                          '(<i>Default</i>)' if b.kind == 'Default'
588                          else '',
589                          s.clusters[c].bindings[b]))
590
591        if prev_s is not None:
592            clusters_removed, clusters_added, clusters_updated = \
593                s.diff_clusters(prev_s)
594            for c in clusters_removed:
595                for b in prev_s.clusters[c].bindings:
596                    dump_binding(prev_s, c, b, False)
597            for c in clusters_updated:
598                bindings_removed, bindings_added = \
599                    s.clusters[c].diff_bindings(prev_s.clusters[c])
600                for b in bindings_removed:
601                    dump_binding(prev_s, c, b, False)
602                for b in bindings_added:
603                    dump_binding(s, c, b, True)
604            for c in clusters_added:
605                for b in s.clusters[c].bindings:
606                    dump_binding(s, c, b, True)
607        else:
608            for c in s.clusters:
609                for b in s.clusters[c].bindings:
610                    dump_binding(s, c, b)
611
612        self._dump('</table>')
613
614    def visit_store_in_state(self, s, prev_s=None):
615        st = s.store
616        prev_st = prev_s.store if prev_s is not None else None
617        if st is None and prev_st is None:
618            return
619
620        self._dump('<hr /><tr><td align="left"><b>Store: </b>')
621        if st is None:
622            self._dump('<i> Nothing!</i>')
623        else:
624            if prev_st is not None:
625                if s.store.is_different(prev_st):
626                    self._dump('</td></tr><tr><td align="left">')
627                    self.visit_store(st, prev_st)
628                else:
629                    self._dump('<i> No changes!</i>')
630            else:
631                self._dump('</td></tr><tr><td align="left">')
632                self.visit_store(st)
633        self._dump('</td></tr>')
634
635    def visit_generic_map(self, m, prev_m=None):
636        self._dump('<table border="0">')
637
638        def dump_pair(m, k, is_added=None):
639            self._dump('<tr><td>%s</td>'
640                       '<td align="left">%s</td>'
641                       '<td align="left">%s</td></tr>'
642                       % (self._diff_plus_minus(is_added),
643                          k, m.generic_map[k]))
644
645        if prev_m is not None:
646            removed, added = m.diff(prev_m)
647            for k in removed:
648                dump_pair(prev_m, k, False)
649            for k in added:
650                dump_pair(m, k, True)
651        else:
652            for k in m.generic_map:
653                dump_pair(m, k, None)
654
655        self._dump('</table>')
656
657    def visit_generic_map_in_state(self, selector, title, s, prev_s=None):
658        m = getattr(s, selector)
659        prev_m = getattr(prev_s, selector) if prev_s is not None else None
660        if m is None and prev_m is None:
661            return
662
663        self._dump('<hr />')
664        self._dump('<tr><td align="left">'
665                   '<b>%s: </b>' % title)
666        if m is None:
667            self._dump('<i> Nothing!</i>')
668        else:
669            if prev_m is not None:
670                if m.is_different(prev_m):
671                    self._dump('</td></tr><tr><td align="left">')
672                    self.visit_generic_map(m, prev_m)
673                else:
674                    self._dump('<i> No changes!</i>')
675            else:
676                self._dump('</td></tr><tr><td align="left">')
677                self.visit_generic_map(m)
678
679        self._dump('</td></tr>')
680
681    def visit_checker_messages(self, m, prev_m=None):
682        self._dump('<table border="0">')
683
684        def dump_line(l, is_added=None):
685            self._dump('<tr><td>%s</td>'
686                       '<td align="left">%s</td></tr>'
687                       % (self._diff_plus_minus(is_added), l))
688
689        def dump_chk(chk, is_added=None):
690            dump_line('<i>%s</i>:' % chk, is_added)
691
692        if prev_m is not None:
693            removed, added, updated = m.diff_messages(prev_m)
694            for chk in removed:
695                dump_chk(chk, False)
696                for l in prev_m.items[chk].lines:
697                    dump_line(l, False)
698            for chk in updated:
699                dump_chk(chk)
700                for l in m.items[chk].diff_lines(prev_m.items[chk]):
701                    dump_line(l[1:], l.startswith('+'))
702            for chk in added:
703                dump_chk(chk, True)
704                for l in m.items[chk].lines:
705                    dump_line(l, True)
706        else:
707            for chk in m.items:
708                dump_chk(chk)
709                for l in m.items[chk].lines:
710                    dump_line(l)
711
712        self._dump('</table>')
713
714    def visit_checker_messages_in_state(self, s, prev_s=None):
715        m = s.checker_messages
716        prev_m = prev_s.checker_messages if prev_s is not None else None
717        if m is None and prev_m is None:
718            return
719
720        self._dump('<hr />')
721        self._dump('<tr><td align="left">'
722                   '<b>Checker State: </b>')
723        if m is None:
724            self._dump('<i> Nothing!</i>')
725        else:
726            if prev_m is not None:
727                if m.is_different(prev_m):
728                    self._dump('</td></tr><tr><td align="left">')
729                    self.visit_checker_messages(m, prev_m)
730                else:
731                    self._dump('<i> No changes!</i>')
732            else:
733                self._dump('</td></tr><tr><td align="left">')
734                self.visit_checker_messages(m)
735
736        self._dump('</td></tr>')
737
738    def visit_state(self, s, prev_s):
739        self.visit_store_in_state(s, prev_s)
740        self.visit_environment_in_state('environment', 'Environment',
741                                        s, prev_s)
742        self.visit_generic_map_in_state('constraints', 'Ranges',
743                                        s, prev_s)
744        self.visit_generic_map_in_state('dynamic_types', 'Dynamic Types',
745                                        s, prev_s)
746        self.visit_environment_in_state('constructing_objects',
747                                        'Objects Under Construction',
748                                        s, prev_s)
749        self.visit_checker_messages_in_state(s, prev_s)
750
751    def visit_node(self, node):
752        self._dump('%s [shape=record,'
753                   % (node.node_name()))
754        if self._dark_mode:
755            self._dump('color="white",fontcolor="gray80",')
756        self._dump('label=<<table border="0">')
757
758        self._dump('<tr><td bgcolor="%s"><b>Node %d (%s) - '
759                   'State %s</b></td></tr>'
760                   % ("gray20" if self._dark_mode else "gray",
761                      node.node_id, node.ptr, node.state.state_id
762                      if node.state is not None else 'Unspecified'))
763        if node.has_report:
764            self._dump('<tr><td><font color="red"><b>Bug Report Attached'
765                       '</b></font></td></tr>')
766        if node.is_sink:
767            self._dump('<tr><td><font color="cornflowerblue"><b>Sink Node'
768                       '</b></font></td></tr>')
769        self._dump('<tr><td align="left" width="0">')
770        if len(node.points) > 1:
771            self._dump('<b>Program points:</b></td></tr>')
772        else:
773            self._dump('<b>Program point:</b></td></tr>')
774        self._dump('<tr><td align="left" width="0">'
775                   '<table border="0" align="left" width="0">')
776        for p in node.points:
777            self.visit_program_point(p)
778        self._dump('</table></td></tr>')
779
780        if node.state is not None:
781            prev_s = None
782            # Do diffs only when we have a unique predecessor.
783            # Don't do diffs on the leaf nodes because they're
784            # the important ones.
785            if self._do_diffs and len(node.predecessors) == 1 \
786               and len(node.successors) > 0:
787                prev_s = self._graph.nodes[node.predecessors[0]].state
788            self.visit_state(node.state, prev_s)
789        self._dump_raw('</table>>];\n')
790
791    def visit_edge(self, pred, succ):
792        self._dump_raw('%s -> %s%s;\n' % (
793            pred.node_name(), succ.node_name(),
794            ' [color="white"]' if self._dark_mode else ''
795        ))
796
797    def visit_end_of_graph(self):
798        self._dump_raw('}\n')
799
800
801#===-----------------------------------------------------------------------===#
802# Explorers know how to traverse the ExplodedGraph in a certain order.
803# They would invoke a Visitor on every node or edge they encounter.
804#===-----------------------------------------------------------------------===#
805
806
807# BasicExplorer explores the whole graph in no particular order.
808class BasicExplorer(object):
809    def __init__(self):
810        super(BasicExplorer, self).__init__()
811
812    def explore(self, graph, visitor):
813        visitor.visit_begin_graph(graph)
814        for node in sorted(graph.nodes):
815            logging.debug('Visiting ' + node)
816            visitor.visit_node(graph.nodes[node])
817            for succ in sorted(graph.nodes[node].successors):
818                logging.debug('Visiting edge: %s -> %s ' % (node, succ))
819                visitor.visit_edge(graph.nodes[node], graph.nodes[succ])
820        visitor.visit_end_of_graph()
821
822
823# SinglePathExplorer traverses only a single path - the leftmost path
824# from the root. Useful when the trimmed graph is still too large
825# due to a large amount of equivalent reports.
826class SinglePathExplorer(object):
827    def __init__(self):
828        super(SinglePathExplorer, self).__init__()
829
830    def explore(self, graph, visitor):
831        visitor.visit_begin_graph(graph)
832
833        # Keep track of visited nodes in order to avoid loops.
834        visited = set()
835        node_id = graph.root_id
836        while True:
837            visited.add(node_id)
838            node = graph.nodes[node_id]
839            logging.debug('Visiting ' + node_id)
840            visitor.visit_node(node)
841            if len(node.successors) == 0:
842                break
843
844            succ_id = node.successors[0]
845            succ = graph.nodes[succ_id]
846            logging.debug('Visiting edge: %s -> %s ' % (node_id, succ_id))
847            visitor.visit_edge(node, succ)
848            if succ_id in visited:
849                break
850
851            node_id = succ_id
852
853        visitor.visit_end_of_graph()
854
855
856#===-----------------------------------------------------------------------===#
857# The entry point to the script.
858#===-----------------------------------------------------------------------===#
859
860
861def main():
862    parser = argparse.ArgumentParser()
863    parser.add_argument('filename', type=str)
864    parser.add_argument('-v', '--verbose', action='store_const',
865                        dest='loglevel', const=logging.DEBUG,
866                        default=logging.WARNING,
867                        help='enable info prints')
868    parser.add_argument('-d', '--diff', action='store_const', dest='diff',
869                        const=True, default=False,
870                        help='display differences between states')
871    parser.add_argument('-s', '--single-path', action='store_const',
872                        dest='single_path', const=True, default=False,
873                        help='only display the leftmost path in the graph '
874                             '(useful for trimmed graphs that still '
875                             'branch too much)')
876    parser.add_argument('--dark', action='store_const', dest='dark',
877                        const=True, default=False,
878                        help='dark mode')
879    parser.add_argument('--gray', action='store_const', dest='gray',
880                        const=True, default=False,
881                        help='black-and-white mode')
882    args = parser.parse_args()
883    logging.basicConfig(level=args.loglevel)
884
885    graph = ExplodedGraph()
886    with open(args.filename) as fd:
887        for raw_line in fd:
888            raw_line = raw_line.strip()
889            graph.add_raw_line(raw_line)
890
891    explorer = SinglePathExplorer() if args.single_path else BasicExplorer()
892    visitor = DotDumpVisitor(args.diff, args.dark, args.gray)
893
894    explorer.explore(graph, visitor)
895
896
897if __name__ == '__main__':
898    main()
899