xref: /llvm-project/clang/utils/analyzer/exploded-graph-rewriter.py (revision ed035ff8264fcfd8ba380a6966c8f29097deb251)
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(object):
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(object):
50    def __init__(self, json_loc):
51        super(SourceLocation, self).__init__()
52        logging.debug('json: %s' % json_loc)
53        self.line = json_loc['line']
54        self.col = json_loc['column']
55        self.filename = os.path.basename(json_loc['file']) \
56            if 'file' in json_loc else '(main file)'
57        self.spelling = SourceLocation(json_loc['spelling']) \
58            if 'spelling' in json_loc else None
59
60    def is_macro(self):
61        return self.spelling is not None
62
63
64# A deserialized program point.
65class ProgramPoint(object):
66    def __init__(self, json_pp):
67        super(ProgramPoint, self).__init__()
68        self.kind = json_pp['kind']
69        self.tag = json_pp['tag']
70        if self.kind == 'Edge':
71            self.src_id = json_pp['src_id']
72            self.dst_id = json_pp['dst_id']
73        elif self.kind == 'Statement':
74            logging.debug(json_pp)
75            self.stmt_kind = json_pp['stmt_kind']
76            self.stmt_point_kind = json_pp['stmt_point_kind']
77            self.stmt_id = json_pp['stmt_id']
78            self.pointer = json_pp['pointer']
79            self.pretty = json_pp['pretty']
80            self.loc = SourceLocation(json_pp['location']) \
81                if json_pp['location'] is not None else None
82        elif self.kind == 'BlockEntrance':
83            self.block_id = json_pp['block_id']
84
85
86# A single expression acting as a key in a deserialized Environment.
87class EnvironmentBindingKey(object):
88    def __init__(self, json_ek):
89        super(EnvironmentBindingKey, self).__init__()
90        # CXXCtorInitializer is not a Stmt!
91        self.stmt_id = json_ek['stmt_id'] if 'stmt_id' in json_ek \
92            else json_ek['init_id']
93        self.pretty = json_ek['pretty']
94        self.kind = json_ek['kind'] if 'kind' in json_ek else None
95
96    def _key(self):
97        return self.stmt_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# Deserialized description of a location context.
107class LocationContext(object):
108    def __init__(self, json_frame):
109        super(LocationContext, self).__init__()
110        self.lctx_id = json_frame['lctx_id']
111        self.caption = json_frame['location_context']
112        self.decl = json_frame['calling']
113        self.loc = SourceLocation(json_frame['location']) \
114            if json_frame['location'] is not None else None
115
116    def _key(self):
117        return self.lctx_id
118
119    def __eq__(self, other):
120        return self._key() == other._key()
121
122    def __hash__(self):
123        return hash(self._key())
124
125
126# A group of deserialized Environment bindings that correspond to a specific
127# location context.
128class EnvironmentFrame(object):
129    def __init__(self, json_frame):
130        super(EnvironmentFrame, self).__init__()
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.
147class GenericEnvironment(object):
148    def __init__(self, json_e):
149        super(GenericEnvironment, self).__init__()
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(object):
179    def __init__(self, json_sk):
180        super(StoreBindingKey, self).__init__()
181        self.kind = json_sk['kind']
182        self.offset = json_sk['offset']
183
184    def _key(self):
185        return (self.kind, self.offset)
186
187    def __eq__(self, other):
188        return self._key() == other._key()
189
190    def __hash__(self):
191        return hash(self._key())
192
193
194# A single cluster of the deserialized RegionStore.
195class StoreCluster(object):
196    def __init__(self, json_sc):
197        super(StoreCluster, self).__init__()
198        self.base_region = json_sc['cluster']
199        self.bindings = collections.OrderedDict(
200            [(StoreBindingKey(b), b['value']) for b in json_sc['items']])
201
202    def diff_bindings(self, prev):
203        return diff_dicts(self.bindings, prev.bindings)
204
205    def is_different(self, prev):
206        removed, added = self.diff_bindings(prev)
207        return len(removed) != 0 or len(added) != 0
208
209
210# A deserialized RegionStore.
211class Store(object):
212    def __init__(self, json_s):
213        super(Store, self).__init__()
214        self.ptr = json_s['pointer']
215        self.clusters = collections.OrderedDict(
216            [(c['pointer'], StoreCluster(c)) for c in json_s['items']])
217
218    def diff_clusters(self, prev):
219        removed = [k for k in prev.clusters if k not in self.clusters]
220        added = [k for k in self.clusters if k not in prev.clusters]
221        updated = [k for k in prev.clusters if k in self.clusters
222                   and prev.clusters[k].is_different(self.clusters[k])]
223        return (removed, added, updated)
224
225    def is_different(self, prev):
226        removed, added, updated = self.diff_clusters(prev)
227        return len(removed) != 0 or len(added) != 0 or len(updated) != 0
228
229
230# Deserialized messages from a single checker in a single program state.
231# Basically a list of raw strings.
232class CheckerLines(object):
233    def __init__(self, json_lines):
234        super(CheckerLines, self).__init__()
235        self.lines = json_lines
236
237    def diff_lines(self, prev):
238        lines = difflib.ndiff(prev.lines, self.lines)
239        return [l.strip() for l in lines
240                if l.startswith('+') or l.startswith('-')]
241
242    def is_different(self, prev):
243        return len(self.diff_lines(prev)) > 0
244
245
246# Deserialized messages of all checkers, separated by checker.
247class CheckerMessages(object):
248    def __init__(self, json_m):
249        super(CheckerMessages, self).__init__()
250        self.items = collections.OrderedDict(
251            [(m['checker'], CheckerLines(m['messages'])) for m in json_m])
252
253    def diff_messages(self, prev):
254        removed = [k for k in prev.items if k not in self.items]
255        added = [k for k in self.items if k not in prev.items]
256        updated = [k for k in prev.items if k in self.items
257                   and prev.items[k].is_different(self.items[k])]
258        return (removed, added, updated)
259
260    def is_different(self, prev):
261        removed, added, updated = self.diff_messages(prev)
262        return len(removed) != 0 or len(added) != 0 or len(updated) != 0
263
264
265# A deserialized program state.
266class ProgramState(object):
267    def __init__(self, state_id, json_ps):
268        super(ProgramState, self).__init__()
269        logging.debug('Adding ProgramState ' + str(state_id))
270
271        self.state_id = state_id
272
273        self.store = Store(json_ps['store']) \
274            if json_ps['store'] is not None else None
275
276        self.environment = \
277            GenericEnvironment(json_ps['environment']['items']) \
278            if json_ps['environment'] is not None else None
279
280        self.constraints = GenericMap([
281            (c['symbol'], c['range']) for c in json_ps['constraints']
282        ]) if json_ps['constraints'] is not None else None
283
284        self.dynamic_types = GenericMap([
285                (t['region'], '%s%s' % (t['dyn_type'],
286                                        ' (or a sub-class)'
287                                        if t['sub_classable'] else ''))
288                for t in json_ps['dynamic_types']]) \
289            if json_ps['dynamic_types'] is not None else None
290
291        self.constructing_objects = \
292            GenericEnvironment(json_ps['constructing_objects']) \
293            if json_ps['constructing_objects'] is not None else None
294
295        self.checker_messages = CheckerMessages(json_ps['checker_messages']) \
296            if json_ps['checker_messages'] is not None else None
297
298
299# A deserialized exploded graph node. Has a default constructor because it
300# may be referenced as part of an edge before its contents are deserialized,
301# and in this moment we already need a room for predecessors and successors.
302class ExplodedNode(object):
303    def __init__(self):
304        super(ExplodedNode, self).__init__()
305        self.predecessors = []
306        self.successors = []
307
308    def construct(self, node_id, json_node):
309        logging.debug('Adding ' + node_id)
310        self.node_id = json_node['node_id']
311        self.ptr = json_node['pointer']
312        self.has_report = json_node['has_report']
313        self.is_sink = json_node['is_sink']
314        self.points = [ProgramPoint(p) for p in json_node['program_points']]
315        self.state = ProgramState(json_node['state_id'],
316                                  json_node['program_state']) \
317            if json_node['program_state'] is not None else None
318
319        assert self.node_name() == node_id
320
321    def node_name(self):
322        return 'Node' + self.ptr
323
324
325# A deserialized ExplodedGraph. Constructed by consuming a .dot file
326# line-by-line.
327class ExplodedGraph(object):
328    # Parse .dot files with regular expressions.
329    node_re = re.compile(
330        '^(Node0x[0-9a-f]*) \\[shape=record,.*label="{(.*)\\\\l}"\\];$')
331    edge_re = re.compile(
332        '^(Node0x[0-9a-f]*) -> (Node0x[0-9a-f]*);$')
333
334    def __init__(self):
335        super(ExplodedGraph, self).__init__()
336        self.nodes = collections.defaultdict(ExplodedNode)
337        self.root_id = None
338        self.incomplete_line = ''
339
340    def add_raw_line(self, raw_line):
341        if raw_line.startswith('//'):
342            return
343
344        # Allow line breaks by waiting for ';'. This is not valid in
345        # a .dot file, but it is useful for writing tests.
346        if len(raw_line) > 0 and raw_line[-1] != ';':
347            self.incomplete_line += raw_line
348            return
349        raw_line = self.incomplete_line + raw_line
350        self.incomplete_line = ''
351
352        # Apply regexps one by one to see if it's a node or an edge
353        # and extract contents if necessary.
354        logging.debug('Line: ' + raw_line)
355        result = self.edge_re.match(raw_line)
356        if result is not None:
357            logging.debug('Classified as edge line.')
358            pred = result.group(1)
359            succ = result.group(2)
360            self.nodes[pred].successors.append(succ)
361            self.nodes[succ].predecessors.append(pred)
362            return
363        result = self.node_re.match(raw_line)
364        if result is not None:
365            logging.debug('Classified as node line.')
366            node_id = result.group(1)
367            if len(self.nodes) == 0:
368                self.root_id = node_id
369            # Note: when writing tests you don't need to escape everything,
370            # even though in a valid dot file everything is escaped.
371            node_label = result.group(2).replace('\\l', '') \
372                                        .replace(' ', '') \
373                                        .replace('\\"', '"') \
374                                        .replace('\\{', '{') \
375                                        .replace('\\}', '}') \
376                                        .replace('\\\\', '\\') \
377                                        .replace('\\|', '|') \
378                                        .replace('\\<', '\\\\<') \
379                                        .replace('\\>', '\\\\>') \
380                                        .rstrip(',')
381            logging.debug(node_label)
382            json_node = json.loads(node_label)
383            self.nodes[node_id].construct(node_id, json_node)
384            return
385        logging.debug('Skipping.')
386
387
388#===-----------------------------------------------------------------------===#
389# Visitors traverse a deserialized ExplodedGraph and do different things
390# with every node and edge.
391#===-----------------------------------------------------------------------===#
392
393
394# A visitor that dumps the ExplodedGraph into a DOT file with fancy HTML-based
395# syntax highlighing.
396class DotDumpVisitor(object):
397    def __init__(self, do_diffs, dark_mode, gray_mode, topo_mode):
398        super(DotDumpVisitor, self).__init__()
399        self._do_diffs = do_diffs
400        self._dark_mode = dark_mode
401        self._gray_mode = gray_mode
402        self._topo_mode = topo_mode
403
404    @staticmethod
405    def _dump_raw(s):
406        print(s, end='')
407
408    def _dump(self, s):
409        s = s.replace('&', '&amp;') \
410             .replace('{', '\\{') \
411             .replace('}', '\\}') \
412             .replace('\\<', '&lt;') \
413             .replace('\\>', '&gt;') \
414             .replace('\\l', '<br />') \
415             .replace('|', '\\|')
416        if self._gray_mode:
417            s = re.sub(r'<font color="[a-z0-9]*">', '', s)
418            s = re.sub(r'</font>', '', s)
419        self._dump_raw(s)
420
421    @staticmethod
422    def _diff_plus_minus(is_added):
423        if is_added is None:
424            return ''
425        if is_added:
426            return '<font color="forestgreen">+</font>'
427        return '<font color="red">-</font>'
428
429    @staticmethod
430    def _short_pretty(s):
431        if s is None:
432            return None
433        if len(s) < 20:
434            return s
435        left = s.find('{')
436        right = s.rfind('}')
437        if left == -1 or right == -1 or left >= right:
438            return s
439        candidate = s[0:left + 1] + ' ... ' + s[right:]
440        if len(candidate) >= len(s):
441            return s
442        return candidate
443
444    @staticmethod
445    def _make_sloc(loc):
446        if loc is None:
447            return '<i>Invalid Source Location</i>'
448
449        def make_plain_loc(loc):
450            return '%s:<b>%s</b>:<b>%s</b>' \
451                % (loc.filename, loc.line, loc.col)
452
453        if loc.is_macro():
454            return '%s <font color="royalblue1">' \
455                   '(<i>spelling at </i> %s)</font>' \
456                % (make_plain_loc(loc), make_plain_loc(loc.spelling))
457
458        return make_plain_loc(loc)
459
460    def visit_begin_graph(self, graph):
461        self._graph = graph
462        self._dump_raw('digraph "ExplodedGraph" {\n')
463        if self._dark_mode:
464            self._dump_raw('bgcolor="gray10";\n')
465        self._dump_raw('label="";\n')
466
467    def visit_program_point(self, p):
468        if p.kind in ['Edge', 'BlockEntrance', 'BlockExit']:
469            color = 'gold3'
470        elif p.kind in ['PreStmtPurgeDeadSymbols',
471                        'PostStmtPurgeDeadSymbols']:
472            color = 'red'
473        elif p.kind in ['CallEnter', 'CallExitBegin', 'CallExitEnd']:
474            color = 'dodgerblue' if self._dark_mode else 'blue'
475        elif p.kind in ['Statement']:
476            color = 'cyan4'
477        else:
478            color = 'forestgreen'
479
480        if p.kind == 'Statement':
481            # This avoids pretty-printing huge statements such as CompoundStmt.
482            # Such statements show up only at [Pre|Post]StmtPurgeDeadSymbols
483            skip_pretty = 'PurgeDeadSymbols' in p.stmt_point_kind
484            stmt_color = 'cyan3'
485            self._dump('<tr><td align="left" width="0">%s:</td>'
486                       '<td align="left" width="0"><font color="%s">'
487                       '%s</font> </td>'
488                       '<td align="left"><i>S%s</i></td>'
489                       '<td align="left"><font color="%s">%s</font></td>'
490                       '<td align="left">%s</td></tr>'
491                       % (self._make_sloc(p.loc), color, p.stmt_kind,
492                          p.stmt_id, stmt_color, p.stmt_point_kind,
493                          self._short_pretty(p.pretty)
494                          if not skip_pretty else ''))
495        elif p.kind == 'Edge':
496            self._dump('<tr><td width="0"></td>'
497                       '<td align="left" width="0">'
498                       '<font color="%s">%s</font></td><td align="left">'
499                       '[B%d] -\\> [B%d]</td></tr>'
500                       % (color, 'BlockEdge', p.src_id, p.dst_id))
501        elif p.kind == 'BlockEntrance':
502            self._dump('<tr><td width="0"></td>'
503                       '<td align="left" width="0">'
504                       '<font color="%s">%s</font></td>'
505                       '<td align="left">[B%d]</td></tr>'
506                       % (color, p.kind, p.block_id))
507        else:
508            # TODO: Print more stuff for other kinds of points.
509            self._dump('<tr><td width="0"></td>'
510                       '<td align="left" width="0" colspan="2">'
511                       '<font color="%s">%s</font></td></tr>'
512                       % (color, p.kind))
513
514        if p.tag is not None:
515            self._dump('<tr><td width="0"></td>'
516                       '<td colspan="3" align="left">'
517                       '<b>Tag: </b> <font color="crimson">'
518                       '%s</font></td></tr>' % p.tag)
519
520    def visit_environment(self, e, prev_e=None):
521        self._dump('<table border="0">')
522
523        def dump_location_context(lc, is_added=None):
524            self._dump('<tr><td>%s</td>'
525                       '<td align="left"><b>%s</b></td>'
526                       '<td align="left" colspan="2">'
527                       '<font color="gray60">%s </font>'
528                       '%s</td></tr>'
529                       % (self._diff_plus_minus(is_added),
530                          lc.caption, lc.decl,
531                          ('(%s)' % self._make_sloc(lc.loc))
532                          if lc.loc is not None else ''))
533
534        def dump_binding(f, b, is_added=None):
535            self._dump('<tr><td>%s</td>'
536                       '<td align="left"><i>S%s</i></td>'
537                       '%s'
538                       '<td align="left">%s</td>'
539                       '<td align="left">%s</td></tr>'
540                       % (self._diff_plus_minus(is_added),
541                          b.stmt_id,
542                          '<td align="left"><font color="%s"><i>'
543                          '%s</i></font></td>' % (
544                              'lavender' if self._dark_mode else 'darkgreen',
545                              ('(%s)' % b.kind) if b.kind is not None else ' '
546                          ),
547                          self._short_pretty(b.pretty), f.bindings[b]))
548
549        frames_updated = e.diff_frames(prev_e) if prev_e is not None else None
550        if frames_updated:
551            for i in frames_updated:
552                f = e.frames[i]
553                prev_f = prev_e.frames[i]
554                dump_location_context(f.location_context)
555                bindings_removed, bindings_added = f.diff_bindings(prev_f)
556                for b in bindings_removed:
557                    dump_binding(prev_f, b, False)
558                for b in bindings_added:
559                    dump_binding(f, b, True)
560        else:
561            for f in e.frames:
562                dump_location_context(f.location_context)
563                for b in f.bindings:
564                    dump_binding(f, b)
565
566        self._dump('</table>')
567
568    def visit_environment_in_state(self, selector, title, s, prev_s=None):
569        e = getattr(s, selector)
570        prev_e = getattr(prev_s, selector) if prev_s is not None else None
571        if e is None and prev_e is None:
572            return
573
574        self._dump('<hr /><tr><td align="left"><b>%s: </b>' % title)
575        if e is None:
576            self._dump('<i> Nothing!</i>')
577        else:
578            if prev_e is not None:
579                if e.is_different(prev_e):
580                    self._dump('</td></tr><tr><td align="left">')
581                    self.visit_environment(e, prev_e)
582                else:
583                    self._dump('<i> No changes!</i>')
584            else:
585                self._dump('</td></tr><tr><td align="left">')
586                self.visit_environment(e)
587
588        self._dump('</td></tr>')
589
590    def visit_store(self, s, prev_s=None):
591        self._dump('<table border="0">')
592
593        def dump_binding(s, c, b, is_added=None):
594            self._dump('<tr><td>%s</td>'
595                       '<td align="left">%s</td>'
596                       '<td align="left">%s</td>'
597                       '<td align="left">%s</td>'
598                       '<td align="left">%s</td></tr>'
599                       % (self._diff_plus_minus(is_added),
600                          s.clusters[c].base_region, b.offset,
601                          '(<i>Default</i>)' if b.kind == 'Default'
602                          else '',
603                          s.clusters[c].bindings[b]))
604
605        if prev_s is not None:
606            clusters_removed, clusters_added, clusters_updated = \
607                s.diff_clusters(prev_s)
608            for c in clusters_removed:
609                for b in prev_s.clusters[c].bindings:
610                    dump_binding(prev_s, c, b, False)
611            for c in clusters_updated:
612                bindings_removed, bindings_added = \
613                    s.clusters[c].diff_bindings(prev_s.clusters[c])
614                for b in bindings_removed:
615                    dump_binding(prev_s, c, b, False)
616                for b in bindings_added:
617                    dump_binding(s, c, b, True)
618            for c in clusters_added:
619                for b in s.clusters[c].bindings:
620                    dump_binding(s, c, b, True)
621        else:
622            for c in s.clusters:
623                for b in s.clusters[c].bindings:
624                    dump_binding(s, c, b)
625
626        self._dump('</table>')
627
628    def visit_store_in_state(self, s, prev_s=None):
629        st = s.store
630        prev_st = prev_s.store if prev_s is not None else None
631        if st is None and prev_st is None:
632            return
633
634        self._dump('<hr /><tr><td align="left"><b>Store: </b>')
635        if st is None:
636            self._dump('<i> Nothing!</i>')
637        else:
638            if prev_st is not None:
639                if s.store.is_different(prev_st):
640                    self._dump('</td></tr><tr><td align="left">')
641                    self.visit_store(st, prev_st)
642                else:
643                    self._dump('<i> No changes!</i>')
644            else:
645                self._dump('</td></tr><tr><td align="left">')
646                self.visit_store(st)
647        self._dump('</td></tr>')
648
649    def visit_generic_map(self, m, prev_m=None):
650        self._dump('<table border="0">')
651
652        def dump_pair(m, k, is_added=None):
653            self._dump('<tr><td>%s</td>'
654                       '<td align="left">%s</td>'
655                       '<td align="left">%s</td></tr>'
656                       % (self._diff_plus_minus(is_added),
657                          k, m.generic_map[k]))
658
659        if prev_m is not None:
660            removed, added = m.diff(prev_m)
661            for k in removed:
662                dump_pair(prev_m, k, False)
663            for k in added:
664                dump_pair(m, k, True)
665        else:
666            for k in m.generic_map:
667                dump_pair(m, k, None)
668
669        self._dump('</table>')
670
671    def visit_generic_map_in_state(self, selector, title, s, prev_s=None):
672        m = getattr(s, selector)
673        prev_m = getattr(prev_s, selector) if prev_s is not None else None
674        if m is None and prev_m is None:
675            return
676
677        self._dump('<hr />')
678        self._dump('<tr><td align="left">'
679                   '<b>%s: </b>' % title)
680        if m is None:
681            self._dump('<i> Nothing!</i>')
682        else:
683            if prev_m is not None:
684                if m.is_different(prev_m):
685                    self._dump('</td></tr><tr><td align="left">')
686                    self.visit_generic_map(m, prev_m)
687                else:
688                    self._dump('<i> No changes!</i>')
689            else:
690                self._dump('</td></tr><tr><td align="left">')
691                self.visit_generic_map(m)
692
693        self._dump('</td></tr>')
694
695    def visit_checker_messages(self, m, prev_m=None):
696        self._dump('<table border="0">')
697
698        def dump_line(l, is_added=None):
699            self._dump('<tr><td>%s</td>'
700                       '<td align="left">%s</td></tr>'
701                       % (self._diff_plus_minus(is_added), l))
702
703        def dump_chk(chk, is_added=None):
704            dump_line('<i>%s</i>:' % chk, is_added)
705
706        if prev_m is not None:
707            removed, added, updated = m.diff_messages(prev_m)
708            for chk in removed:
709                dump_chk(chk, False)
710                for l in prev_m.items[chk].lines:
711                    dump_line(l, False)
712            for chk in updated:
713                dump_chk(chk)
714                for l in m.items[chk].diff_lines(prev_m.items[chk]):
715                    dump_line(l[1:], l.startswith('+'))
716            for chk in added:
717                dump_chk(chk, True)
718                for l in m.items[chk].lines:
719                    dump_line(l, True)
720        else:
721            for chk in m.items:
722                dump_chk(chk)
723                for l in m.items[chk].lines:
724                    dump_line(l)
725
726        self._dump('</table>')
727
728    def visit_checker_messages_in_state(self, s, prev_s=None):
729        m = s.checker_messages
730        prev_m = prev_s.checker_messages if prev_s is not None else None
731        if m is None and prev_m is None:
732            return
733
734        self._dump('<hr />')
735        self._dump('<tr><td align="left">'
736                   '<b>Checker State: </b>')
737        if m is None:
738            self._dump('<i> Nothing!</i>')
739        else:
740            if prev_m is not None:
741                if m.is_different(prev_m):
742                    self._dump('</td></tr><tr><td align="left">')
743                    self.visit_checker_messages(m, prev_m)
744                else:
745                    self._dump('<i> No changes!</i>')
746            else:
747                self._dump('</td></tr><tr><td align="left">')
748                self.visit_checker_messages(m)
749
750        self._dump('</td></tr>')
751
752    def visit_state(self, s, prev_s):
753        self.visit_store_in_state(s, prev_s)
754        self.visit_environment_in_state('environment', 'Environment',
755                                        s, prev_s)
756        self.visit_generic_map_in_state('constraints', 'Ranges',
757                                        s, prev_s)
758        self.visit_generic_map_in_state('dynamic_types', 'Dynamic Types',
759                                        s, prev_s)
760        self.visit_environment_in_state('constructing_objects',
761                                        'Objects Under Construction',
762                                        s, prev_s)
763        self.visit_checker_messages_in_state(s, prev_s)
764
765    def visit_node(self, node):
766        self._dump('%s [shape=record,'
767                   % (node.node_name()))
768        if self._dark_mode:
769            self._dump('color="white",fontcolor="gray80",')
770        self._dump('label=<<table border="0">')
771
772        self._dump('<tr><td bgcolor="%s"><b>Node %d (%s) - '
773                   'State %s</b></td></tr>'
774                   % ("gray20" if self._dark_mode else "gray",
775                      node.node_id, node.ptr, node.state.state_id
776                      if node.state is not None else 'Unspecified'))
777        if node.has_report:
778            self._dump('<tr><td><font color="red"><b>Bug Report Attached'
779                       '</b></font></td></tr>')
780        if node.is_sink:
781            self._dump('<tr><td><font color="cornflowerblue"><b>Sink Node'
782                       '</b></font></td></tr>')
783        if not self._topo_mode:
784            self._dump('<tr><td align="left" width="0">')
785            if len(node.points) > 1:
786                self._dump('<b>Program points:</b></td></tr>')
787            else:
788                self._dump('<b>Program point:</b></td></tr>')
789        self._dump('<tr><td align="left" width="0">'
790                   '<table border="0" align="left" width="0">')
791        for p in node.points:
792            self.visit_program_point(p)
793        self._dump('</table></td></tr>')
794
795        if node.state is not None and not self._topo_mode:
796            prev_s = None
797            # Do diffs only when we have a unique predecessor.
798            # Don't do diffs on the leaf nodes because they're
799            # the important ones.
800            if self._do_diffs and len(node.predecessors) == 1 \
801               and len(node.successors) > 0:
802                prev_s = self._graph.nodes[node.predecessors[0]].state
803            self.visit_state(node.state, prev_s)
804        self._dump_raw('</table>>];\n')
805
806    def visit_edge(self, pred, succ):
807        self._dump_raw('%s -> %s%s;\n' % (
808            pred.node_name(), succ.node_name(),
809            ' [color="white"]' if self._dark_mode else ''
810        ))
811
812    def visit_end_of_graph(self):
813        self._dump_raw('}\n')
814
815
816#===-----------------------------------------------------------------------===#
817# Explorers know how to traverse the ExplodedGraph in a certain order.
818# They would invoke a Visitor on every node or edge they encounter.
819#===-----------------------------------------------------------------------===#
820
821
822# BasicExplorer explores the whole graph in no particular order.
823class BasicExplorer(object):
824    def __init__(self):
825        super(BasicExplorer, self).__init__()
826
827    def explore(self, graph, visitor):
828        visitor.visit_begin_graph(graph)
829        for node in sorted(graph.nodes):
830            logging.debug('Visiting ' + node)
831            visitor.visit_node(graph.nodes[node])
832            for succ in sorted(graph.nodes[node].successors):
833                logging.debug('Visiting edge: %s -> %s ' % (node, succ))
834                visitor.visit_edge(graph.nodes[node], graph.nodes[succ])
835        visitor.visit_end_of_graph()
836
837
838# SinglePathExplorer traverses only a single path - the leftmost path
839# from the root. Useful when the trimmed graph is still too large
840# due to a large amount of equivalent reports.
841class SinglePathExplorer(object):
842    def __init__(self):
843        super(SinglePathExplorer, self).__init__()
844
845    def explore(self, graph, visitor):
846        visitor.visit_begin_graph(graph)
847
848        # Keep track of visited nodes in order to avoid loops.
849        visited = set()
850        node_id = graph.root_id
851        while True:
852            visited.add(node_id)
853            node = graph.nodes[node_id]
854            logging.debug('Visiting ' + node_id)
855            visitor.visit_node(node)
856            if len(node.successors) == 0:
857                break
858
859            succ_id = node.successors[0]
860            succ = graph.nodes[succ_id]
861            logging.debug('Visiting edge: %s -> %s ' % (node_id, succ_id))
862            visitor.visit_edge(node, succ)
863            if succ_id in visited:
864                break
865
866            node_id = succ_id
867
868        visitor.visit_end_of_graph()
869
870
871#===-----------------------------------------------------------------------===#
872# The entry point to the script.
873#===-----------------------------------------------------------------------===#
874
875
876def main():
877    parser = argparse.ArgumentParser()
878    parser.add_argument('filename', type=str)
879    parser.add_argument('-v', '--verbose', action='store_const',
880                        dest='loglevel', const=logging.DEBUG,
881                        default=logging.WARNING,
882                        help='enable info prints')
883    parser.add_argument('-d', '--diff', action='store_const', dest='diff',
884                        const=True, default=False,
885                        help='display differences between states')
886    parser.add_argument('-t', '--topology', action='store_const',
887                        dest='topology', const=True, default=False,
888                        help='only display program points, omit states')
889    parser.add_argument('-s', '--single-path', action='store_const',
890                        dest='single_path', const=True, default=False,
891                        help='only display the leftmost path in the graph '
892                             '(useful for trimmed graphs that still '
893                             'branch too much)')
894    parser.add_argument('--dark', action='store_const', dest='dark',
895                        const=True, default=False,
896                        help='dark mode')
897    parser.add_argument('--gray', action='store_const', dest='gray',
898                        const=True, default=False,
899                        help='black-and-white mode')
900    args = parser.parse_args()
901    logging.basicConfig(level=args.loglevel)
902
903    graph = ExplodedGraph()
904    with open(args.filename) as fd:
905        for raw_line in fd:
906            raw_line = raw_line.strip()
907            graph.add_raw_line(raw_line)
908
909    explorer = SinglePathExplorer() if args.single_path else BasicExplorer()
910    visitor = DotDumpVisitor(args.diff, args.dark, args.gray, args.topology)
911
912    explorer.explore(graph, visitor)
913
914
915if __name__ == '__main__':
916    main()
917