summaryrefslogtreecommitdiff
path: root/giscanner/transformer.py
blob: 4f3abcf44e404e272e8b5054e3adcb548e8f67d2 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
# -*- Mode: Python -*-
# GObject-Introspection - a framework for introspecting GObject libraries
# Copyright (C) 2008  Johan Dahlin
#
# This library is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
# License as published by the Free Software Foundation; either
# version 2 of the License, or (at your option) any later version.
#
# This library is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
# Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public
# License along with this library; if not, write to the
# Free Software Foundation, Inc., 59 Temple Place - Suite 330,
# Boston, MA 02111-1307, USA.
#

import os
import sys
import subprocess

from . import ast
from . import message
from . import utils
from .cachestore import CacheStore
from .girparser import GIRParser
from .sourcescanner import (
    SourceSymbol, ctype_name, CTYPE_POINTER,
    CTYPE_BASIC_TYPE, CTYPE_UNION, CTYPE_ARRAY, CTYPE_TYPEDEF,
    CTYPE_VOID, CTYPE_ENUM, CTYPE_FUNCTION, CTYPE_STRUCT,
    CSYMBOL_TYPE_FUNCTION, CSYMBOL_TYPE_TYPEDEF, CSYMBOL_TYPE_STRUCT,
    CSYMBOL_TYPE_ENUM, CSYMBOL_TYPE_UNION, CSYMBOL_TYPE_OBJECT,
    CSYMBOL_TYPE_MEMBER, CSYMBOL_TYPE_ELLIPSIS, CSYMBOL_TYPE_CONST,
    TYPE_QUALIFIER_CONST, TYPE_QUALIFIER_VOLATILE)


class TransformerException(Exception):
    pass


class Transformer(object):
    namespace = property(lambda self: self._namespace)

    def __init__(self, namespace, accept_unprefixed=False,
                 identifier_filter_cmd=None, symbol_filter_cmd=None):
        self._cachestore = CacheStore()
        self._accept_unprefixed = accept_unprefixed
        self._namespace = namespace
        self._pkg_config_packages = set()
        self._typedefs_ns = {}
        self._parsed_includes = {}  # <string namespace -> Namespace>
        self._includepaths = []
        self._passthrough_mode = False
        self._identifier_filter_cmd = identifier_filter_cmd
        self._symbol_filter_cmd = symbol_filter_cmd

        # Cache a list of struct/unions in C's "tag namespace". This helps
        # manage various orderings of typedefs and structs. See:
        # https://bugzilla.gnome.org/show_bug.cgi?id=581525
        self._tag_ns = {}

    def get_pkgconfig_packages(self):
        return self._pkg_config_packages

    def disable_cache(self):
        self._cachestore = None

    def set_passthrough_mode(self):
        self._passthrough_mode = True

    def _append_new_node(self, node):
        original = self._namespace.get(node.name)
        # Special case constants here; we allow duplication to sort-of
        # handle #ifdef.  But this introduces an arch-dependency in the .gir
        # file.  So far this has only come up scanning glib - in theory, other
        # modules will just depend on that.
        if isinstance(original, ast.Constant) and isinstance(node, ast.Constant):
            pass
        elif original is node:
            # Ignore attempts to add the same node to the namespace. This can
            # happen when parsing typedefs and structs in particular orderings:
            #   typedef struct _Foo Foo;
            #   struct _Foo {...};
            pass
        elif original:
            positions = set()
            positions.update(original.file_positions)
            positions.update(node.file_positions)
            message.fatal("Namespace conflict for '%s'" % (node.name, ),
                          positions)
        else:
            self._namespace.append(node)

    def parse(self, symbols):
        for symbol in symbols:
            # WORKAROUND
            # https://bugzilla.gnome.org/show_bug.cgi?id=550616
            if symbol.ident in ['gst_g_error_get_type']:
                continue

            try:
                node = self._traverse_one(symbol)
            except TransformerException as e:
                message.warn_symbol(symbol, e)
                continue

            if node and node.name:
                self._append_new_node(node)
            if isinstance(node, ast.Compound) and node.tag_name and \
                    node.tag_name not in self._tag_ns:
                self._tag_ns[node.tag_name] = node

        # Run through the tag namespace looking for structs that have not been
        # promoted into the main namespace. In this case we simply promote them
        # with their struct tag.
        for tag_name, struct in self._tag_ns.items():
            if not struct.name:
                try:
                    name = self.strip_identifier(tag_name)
                    struct.name = name
                    self._append_new_node(struct)
                except TransformerException as e:
                    message.warn_node(node, e)

    def set_include_paths(self, paths):
        self._includepaths = list(paths)

    def register_include(self, include):
        if include in self._namespace.includes:
            return
        self._namespace.includes.add(include)
        filename = self._find_include(include)
        self._parse_include(filename)

    def register_include_uninstalled(self, include_path):
        basename = os.path.basename(include_path)
        if not basename.endswith('.gir'):
            raise SystemExit("Include path '%s' must be a filename path "
                             "ending in .gir" % (include_path, ))
        girname = basename[:-4]
        include = ast.Include.from_string(girname)
        if include in self._namespace.includes:
            return
        self._namespace.includes.add(include)
        self._parse_include(include_path, uninstalled=True)

    def lookup_giname(self, name):
        """Given a name of the form Foo or Bar.Foo,
return the corresponding ast.Node, or None if none
available.  Will throw KeyError however for unknown
namespaces."""
        if '.' not in name:
            return self._namespace.get(name)
        else:
            (ns, giname) = name.split('.', 1)
            if ns == self._namespace.name:
                return self._namespace.get(giname)
            # Fallback to the main namespace if not a dependency and matches a prefix
            if ns in self._namespace.identifier_prefixes and ns not in self._parsed_includes:
                message.warn(("Deprecated reference to identifier " +
                              "prefix %s in GIName %s") % (ns, name))
                return self._namespace.get(giname)
            include = self._parsed_includes[ns]
            return include.get(giname)

    def lookup_typenode(self, typeobj):
        """Given a Type object, if it points to a giname,
calls lookup_giname() on the name.  Otherwise return
None."""
        if typeobj.target_giname:
            return self.lookup_giname(typeobj.target_giname)
        return None

    # Private

    def _get_gi_data_dirs(self):
        data_dirs = utils.get_system_data_dirs()
        data_dirs.append(DATADIR)
        data_dirs.append(GIRDIR)
        if os.name != 'nt':
            # For backwards compatibility, was always unconditionally added to the list.
            data_dirs.append('/usr/share')
        return data_dirs

    def _find_include(self, include):
        searchdirs = self._includepaths[:]
        for path in self._get_gi_data_dirs():
            searchdirs.append(os.path.join(path, 'gir-1.0'))
        searchdirs.append(os.path.join(DATADIR, 'gir-1.0'))

        girname = '%s-%s.gir' % (include.name, include.version)
        for d in searchdirs:
            path = os.path.join(d, girname)
            if os.path.exists(path):
                return path
        sys.stderr.write("Couldn't find include '%s' (search path: '%s')\n" %
                         (girname, searchdirs))
        sys.exit(1)

    @classmethod
    def parse_from_gir(cls, filename, extra_include_dirs=None):
        self = cls(None)
        if extra_include_dirs is not None:
            self.set_include_paths(extra_include_dirs)
        self.set_passthrough_mode()
        parser = self._parse_include(filename)
        self._namespace = parser.get_namespace()
        del self._parsed_includes[self._namespace.name]
        return self

    def _parse_include(self, filename, uninstalled=False):
        parser = None
        if self._cachestore is not None:
            parser = self._cachestore.load(filename)
        if parser is None:
            parser = GIRParser(types_only=not self._passthrough_mode)
            parser.parse(filename)
            if self._cachestore is not None:
                self._cachestore.store(filename, parser)

        for include in parser.get_namespace().includes:
            if include.name not in self._parsed_includes:
                dep_filename = self._find_include(include)
                self._parse_include(dep_filename)

        if not uninstalled:
            for pkg in parser.get_namespace().exported_packages:
                self._pkg_config_packages.add(pkg)
        namespace = parser.get_namespace()
        self._parsed_includes[namespace.name] = namespace
        return parser

    def _iter_namespaces(self):
        """Return an iterator over all included namespaces; the
currently-scanned namespace is first."""
        yield self._namespace
        for ns in self._parsed_includes.values():
            yield ns

    def _sort_matches(self, val):
        """Key sort which ensures items in self._namespace are last by returning
        a tuple key starting with 1 for self._namespace entries and 0 for
        everythin else.
        """
        if val[0] == self._namespace:
            return 1, val[2]
        else:
            return 0, val[2]

    def _split_c_string_for_namespace_matches(self, name, is_identifier=False):
        if not is_identifier and self._symbol_filter_cmd:
            proc = subprocess.Popen(self._symbol_filter_cmd,
                                    stdin=subprocess.PIPE,
                                    stdout=subprocess.PIPE,
                                    stderr=subprocess.PIPE)
            proc_name, err = proc.communicate(name.encode())
            proc_name = proc_name.strip()
            if proc.returncode:
                raise ValueError('filter: %r exited: %d with error: %s' %
                                 (self._symbol_filter_cmd, proc.returncode, err))
            name = proc_name.decode('ascii')
            name = name.strip()

        matches = []  # Namespaces which might contain this name
        unprefixed_namespaces = []  # Namespaces with no prefix, last resort
        for ns in self._iter_namespaces():
            if is_identifier:
                prefixes = ns.identifier_prefixes
            elif name[0].isupper():
                prefixes = ns._ucase_symbol_prefixes
            else:
                prefixes = ns.symbol_prefixes
            if prefixes:
                for prefix in prefixes:
                    if (not is_identifier) and (not prefix.endswith('_')):
                        prefix = prefix + '_'
                    if name.startswith(prefix):
                        matches.append((ns, name[len(prefix):], len(prefix)))
                        break
            else:
                unprefixed_namespaces.append(ns)
        if matches:
            matches.sort(key=self._sort_matches)
            return list(map(lambda x: (x[0], x[1]), matches))
        elif self._accept_unprefixed:
            return [(self._namespace, name)]
        elif unprefixed_namespaces:
            # A bit of a hack; this function ideally shouldn't look through the
            # contents of namespaces; but since we aren't scanning anything
            # without a prefix, it's not too bad.
            for ns in unprefixed_namespaces:
                if name in ns:
                    return [(ns, name)]
        raise ValueError("Unknown namespace for %s '%s'"
                         % ('identifier' if is_identifier else 'symbol', name, ))

    def split_ctype_namespaces(self, ident):
        """Given a StudlyCaps string identifier like FooBar, return a
list of (namespace, stripped_identifier) sorted by namespace length,
or raise ValueError.  As a special case, if the current namespace matches,
it is always biggest (i.e. last)."""
        return self._split_c_string_for_namespace_matches(ident, is_identifier=True)

    def split_csymbol_namespaces(self, symbol):
        """Given a C symbol like foo_bar_do_baz, return a list of
(namespace, stripped_symbol) sorted by namespace match probablity, or
raise ValueError."""
        return self._split_c_string_for_namespace_matches(symbol, is_identifier=False)

    def split_csymbol(self, symbol):
        """Given a C symbol like foo_bar_do_baz, return the most probable
(namespace, stripped_symbol) match, or raise ValueError."""
        matches = self._split_c_string_for_namespace_matches(symbol, is_identifier=False)
        return matches[-1]

    def strip_identifier(self, ident):
        if self._identifier_filter_cmd:
            proc = subprocess.Popen(self._identifier_filter_cmd,
                                    stdin=subprocess.PIPE,
                                    stdout=subprocess.PIPE,
                                    stderr=subprocess.PIPE)
            proc_ident, err = proc.communicate(ident.encode())
            if proc.returncode:
                raise ValueError('filter: %r exited: %d with error: %s' %
                                 (self._identifier_filter_cmd, proc.returncode, err))
            ident = proc_ident.decode('ascii').strip()

        hidden = ident.startswith('_')
        if hidden:
            ident = ident[1:]
        try:
            matches = self.split_ctype_namespaces(ident)
        except ValueError as e:
            raise TransformerException(str(e))
        for ns, name in matches:
            if ns is self._namespace:
                if hidden:
                    return '_' + name
                return name
        (ns, name) = matches[-1]
        raise TransformerException(
            "Skipping foreign identifier '%s' from namespace %s" % (ident, ns.name, ))
        return None

    def _strip_symbol(self, symbol):
        ident = symbol.ident
        hidden = ident.startswith('_')
        if hidden:
            ident = ident[1:]
        try:
            (ns, name) = self.split_csymbol(ident)
        except ValueError as e:
            raise TransformerException(str(e))
        if ns != self._namespace:
            raise TransformerException(
                "Skipping foreign symbol from namespace %s" % (ns.name, ))
        if hidden:
            return '_' + name
        return name

    def _traverse_one(self, symbol, stype=None, parent_symbol=None):
        assert isinstance(symbol, SourceSymbol), symbol

        if stype is None:
            stype = symbol.type
        if stype == CSYMBOL_TYPE_FUNCTION:
            return self._create_function(symbol)
        elif stype == CSYMBOL_TYPE_TYPEDEF:
            return self._create_typedef(symbol)
        elif stype == CSYMBOL_TYPE_STRUCT:
            return self._create_tag_ns_compound(ast.Record, symbol)
        elif stype == CSYMBOL_TYPE_ENUM:
            return self._create_enum(symbol)
        elif stype == CSYMBOL_TYPE_MEMBER:
            return self._create_member(symbol, parent_symbol)
        elif stype == CSYMBOL_TYPE_UNION:
            return self._create_tag_ns_compound(ast.Union, symbol)
        elif stype == CSYMBOL_TYPE_CONST:
            return self._create_const(symbol)
        # Ignore variable declarations in the header
        elif stype == CSYMBOL_TYPE_OBJECT:
            pass
        else:
            print("transformer: unhandled symbol: '%s'" % (symbol, ))

    def _enum_common_prefix(self, symbol):
        def common_prefix(a, b):
            commonparts = []
            for aword, bword in zip(a.split('_'), b.split('_')):
                if aword != bword:
                    return '_'.join(commonparts) + '_'
                commonparts.append(aword)
            return min(a, b)

        # Nothing less than 2 has a common prefix
        if len(list(symbol.base_type.child_list)) < 2:
            return None
        prefix = None
        for child in symbol.base_type.child_list:
            if prefix is None:
                prefix = child.ident
            else:
                prefix = common_prefix(prefix, child.ident)
                if prefix == '':
                    return None
        return prefix

    def _create_enum(self, symbol):
        prefix = self._enum_common_prefix(symbol)
        if prefix:
            prefixlen = len(prefix)
        else:
            prefixlen = 0
        members = []
        for child in symbol.base_type.child_list:
            if child.private:
                continue
            if prefixlen > 0:
                name = child.ident[prefixlen:]
            else:
                # Ok, the enum members don't have a consistent prefix
                # among them, so let's just remove the global namespace
                # prefix.
                name = self._strip_symbol(child)
            members.append(ast.Member(name.lower(),
                                      child.const_int,
                                      child.ident,
                                      None))

        enum_name = self.strip_identifier(symbol.ident)
        if symbol.base_type.is_bitfield:
            klass = ast.Bitfield
        else:
            klass = ast.Enum
        node = klass(enum_name, symbol.ident, members=members)
        node.add_symbol_reference(symbol)
        return node

    def _create_function(self, symbol):
        # Drop functions that start with _ very early on here
        if symbol.ident.startswith('_'):
            return None
        parameters = list(self._create_parameters(symbol, symbol.base_type))
        return_ = self._create_return(symbol.base_type.base_type)
        name = self._strip_symbol(symbol)
        func = ast.Function(name, return_, parameters, False, symbol.ident)
        func.add_symbol_reference(symbol)
        return func

    def _create_source_type(self, source_type, is_parameter=False):
        assert source_type is not None
        if source_type.type == CTYPE_VOID:
            value = 'void'
        elif source_type.type == CTYPE_BASIC_TYPE:
            value = source_type.name
        elif source_type.type == CTYPE_TYPEDEF:
            value = source_type.name
        elif (source_type.type == CTYPE_POINTER or
                # Array to pointer adjustment as per 6.7.6.3.
                # This is performed only on the outermost array,
                # so we don't forward is_parameter.
                (source_type.type == CTYPE_ARRAY and is_parameter)):
            value = self._create_source_type(source_type.base_type) + '*'
        elif source_type.type == CTYPE_ARRAY:
            return self._create_source_type(source_type.base_type)
        else:
            value = 'gpointer'
        return value

    def _create_complete_source_type(self, source_type, is_parameter=False):
        assert source_type is not None

        const = (source_type.type_qualifier & TYPE_QUALIFIER_CONST)
        volatile = (source_type.type_qualifier & TYPE_QUALIFIER_VOLATILE)

        if source_type.type == CTYPE_VOID:
            return 'void'
        elif source_type.type in [CTYPE_BASIC_TYPE,
                                  CTYPE_TYPEDEF,
                                  CTYPE_STRUCT,
                                  CTYPE_UNION,
                                  CTYPE_ENUM]:
            value = source_type.name
            if const:
                value = 'const ' + value
            if volatile:
                value = 'volatile ' + value
            return value
        elif (source_type.type == CTYPE_POINTER or
                # Array to pointer adjustment as per 6.7.6.3.
                # This is performed only on the outermost array,
                # so we don't forward is_parameter.
                (source_type.type == CTYPE_ARRAY and is_parameter)):
            value = self._create_complete_source_type(source_type.base_type) + '*'
            # TODO: handle pointer to function as a special case?
            if const:
                value += ' const'
            if volatile:
                value += ' volatile'
            return value
        elif source_type.type == CTYPE_ARRAY:
            return self._create_complete_source_type(source_type.base_type)
        else:
            if const:
                value = 'gconstpointer'
            else:
                value = 'gpointer'
            if volatile:
                value = 'volatile ' + value
            return value

    def _create_parameters(self, symbol, base_type):
        for i, child in enumerate(base_type.child_list):
            yield self._create_parameter(symbol, i, child)

    def _synthesize_union_type(self, symbol, parent_symbol):
        # Synthesize a named union so that it can be referenced.
        parent_ident = parent_symbol.ident
        # FIXME: Should split_ctype_namespaces handle the hidden case?
        hidden = parent_ident.startswith('_')
        if hidden:
            parent_ident = parent_ident[1:]
        matches = self.split_ctype_namespaces(parent_ident)
        (namespace, parent_name) = matches[-1]
        assert namespace and parent_name
        if hidden:
            parent_name = '_' + parent_name
        fake_union = ast.Union("%s__%s__union" % (parent_name, symbol.ident))
        # _parse_fields accesses <type>.base_type.child_list, so we have to
        # pass symbol.base_type even though that refers to the array, not the
        # union.
        self._parse_fields(symbol.base_type, fake_union)
        self._append_new_node(fake_union)
        fake_type = ast.Type(
            target_giname="%s.%s" % (namespace.name, fake_union.name))
        return fake_type

    def _create_member(self, symbol, parent_symbol=None):
        source_type = symbol.base_type
        if (source_type.type == CTYPE_POINTER
        and symbol.base_type.base_type.type == CTYPE_FUNCTION):
            node = self._create_callback(symbol, member=True)
        elif source_type.type == CTYPE_STRUCT and source_type.name is None:
            node = self._create_member_compound(ast.Record, symbol)
        elif source_type.type == CTYPE_UNION and source_type.name is None:
            node = self._create_member_compound(ast.Union, symbol)
        else:
            # Special handling for fields; we don't have annotations on them
            # to apply later, yet.
            if source_type.type == CTYPE_ARRAY:
                # Determine flattened array size and its element type.
                flattened_size = 1
                while source_type.type == CTYPE_ARRAY:
                    for child in source_type.child_list:
                        if flattened_size is not None:
                            flattened_size *= child.const_int
                        break
                    else:
                        flattened_size = None
                    source_type = source_type.base_type

                # If the array contains anonymous unions, like in the GValue
                # struct, we need to handle this specially.  This is necessary
                # to be able to properly calculate the size of the compound
                # type (e.g. GValue) that contains this array, see
                # <https://bugzilla.gnome.org/show_bug.cgi?id=657040>.
                if source_type.type == CTYPE_UNION and source_type.name is None:
                    element_type = self._synthesize_union_type(symbol, parent_symbol)
                else:
                    ctype = self._create_source_type(source_type)
                    complete_ctype = self._create_complete_source_type(source_type)
                    element_type = self.create_type_from_ctype_string(ctype,
                                                                      complete_ctype=complete_ctype)
                ftype = ast.Array(None, element_type)
                ftype.zeroterminated = False
                ftype.size = flattened_size
            else:
                ftype = self._create_type_from_base(symbol.base_type)
            # ast.Fields are assumed to be read-write
            # (except for Objects, see also glibtransformer.py)
            node = ast.Field(symbol.ident, ftype,
                             readable=True, writable=True,
                             bits=symbol.const_int)
            if symbol.private:
                node.readable = False
                node.writable = False
                node.private = True
        return node

    def _create_typedef(self, symbol):
        ctype = symbol.base_type.type
        if (ctype == CTYPE_POINTER and symbol.base_type.base_type.type == CTYPE_FUNCTION):
            node = self._create_typedef_callback(symbol)
        elif (ctype == CTYPE_FUNCTION):
            node = self._create_typedef_callback(symbol)
        elif (ctype == CTYPE_POINTER and symbol.base_type.base_type.type == CTYPE_STRUCT):
            node = self._create_typedef_compound(ast.Record, symbol, disguised=True)
        elif ctype == CTYPE_STRUCT:
            node = self._create_typedef_compound(ast.Record, symbol)
        elif ctype == CTYPE_UNION:
            node = self._create_typedef_compound(ast.Union, symbol)
        elif ctype == CTYPE_ENUM:
            return self._create_enum(symbol)
        elif ctype in (CTYPE_TYPEDEF,
                       CTYPE_POINTER,
                       CTYPE_BASIC_TYPE,
                       CTYPE_VOID):
            name = self.strip_identifier(symbol.ident)
            target = self._create_type_from_base(symbol.base_type)
            if name in ast.type_names:
                return None
            # https://bugzilla.gnome.org/show_bug.cgi?id=755882
            if name.endswith('_autoptr'):
                return None
            node = ast.Alias(name, target, ctype=symbol.ident)
            node.add_symbol_reference(symbol)
        else:
            raise NotImplementedError(
                "symbol '%s' of type %s" % (symbol.ident, ctype_name(ctype)))
        return node

    def _canonicalize_ctype(self, ctype):
        # First look up the ctype including any pointers;
        # a few type names like 'char*' have their own aliases
        # and we need pointer information for those.
        firstpass = ast.type_names.get(ctype)

        # If we have a particular alias for this, skip deep
        # canonicalization to prevent changing
        # e.g. char* -> int8*
        if firstpass:
            return firstpass.target_fundamental

        if not ctype.endswith('*'):
            return ctype

        # We have a pointer type.
        # Strip the end pointer, canonicalize our base type
        base = ctype[:-1]
        canonical_base = self._canonicalize_ctype(base)

        # Append the pointer again
        canonical = canonical_base + '*'

        return canonical

    def _create_type_from_base(self, source_type, is_parameter=False, is_return=False):
        ctype = self._create_source_type(source_type, is_parameter=is_parameter)
        complete_ctype = self._create_complete_source_type(source_type, is_parameter=is_parameter)
        const = ((source_type.type == CTYPE_POINTER) and
                 (source_type.base_type.type_qualifier & TYPE_QUALIFIER_CONST))
        return self.create_type_from_ctype_string(ctype, is_const=const,
                                                  is_parameter=is_parameter, is_return=is_return,
                                                  complete_ctype=complete_ctype)

    def _create_bare_container_type(self, base, ctype=None,
                                    is_const=False, complete_ctype=None):
        if base in ('GList', 'GSList', 'GLib.List', 'GLib.SList'):
            if base in ('GList', 'GSList'):
                name = 'GLib.' + base[1:]
            else:
                name = base
            return ast.List(name, ast.TYPE_ANY, ctype=ctype,
                        is_const=is_const, complete_ctype=complete_ctype)
        elif base in ('GByteArray', 'GLib.ByteArray', 'GObject.ByteArray'):
            return ast.Array('GLib.ByteArray', ast.TYPE_UINT8, ctype=ctype,
                         is_const=is_const, complete_ctype=complete_ctype)
        elif base in ('GArray', 'GPtrArray',
                      'GLib.Array', 'GLib.PtrArray',
                      'GObject.Array', 'GObject.PtrArray'):
            if '.' in base:
                name = 'GLib.' + base.split('.', 1)[1]
            else:
                name = 'GLib.' + base[1:]
            return ast.Array(name, ast.TYPE_ANY, ctype=ctype,
                         is_const=is_const, complete_ctype=complete_ctype)
        elif base in ('GHashTable', 'GLib.HashTable', 'GObject.HashTable'):
            return ast.Map(ast.TYPE_ANY, ast.TYPE_ANY, ctype=ctype, is_const=is_const,
                           complete_ctype=complete_ctype)
        return None

    def create_type_from_ctype_string(self, ctype, is_const=False,
                                      is_parameter=False, is_return=False,
                                      complete_ctype=None):
        canonical = self._canonicalize_ctype(ctype)
        base = canonical.replace('*', '')

        # While gboolean and _Bool are distinct types, they used to be treated
        # by scanner as exactly the same one. In general this is incorrect
        # because of different ABI, but this usually works fine,
        # so for backward compatibility lets continue for now:
        # https://gitlab.gnome.org/GNOME/gobject-introspection/merge_requests/24#note_92792
        if canonical in ('_Bool', 'bool'):
            canonical = 'gboolean'
            base = canonical

        # Special default: char ** -> ast.Array, same for GStrv
        if (is_return and canonical == 'utf8*') or base == 'GStrv':
            bare_utf8 = ast.TYPE_STRING.clone()
            bare_utf8.ctype = None
            return ast.Array(None, bare_utf8, ctype=ctype,
                             is_const=is_const, complete_ctype=complete_ctype)

        fundamental = ast.type_names.get(base)
        if fundamental is not None:
            return ast.Type(target_fundamental=fundamental.target_fundamental,
                        ctype=ctype,
                        is_const=is_const, complete_ctype=complete_ctype)
        container = self._create_bare_container_type(base, ctype=ctype, is_const=is_const,
                                                     complete_ctype=complete_ctype)
        if container:
            return container
        return ast.Type(ctype=ctype, is_const=is_const, complete_ctype=complete_ctype)

    def _create_parameter(self, parent_symbol, index, symbol):
        if symbol.type == CSYMBOL_TYPE_ELLIPSIS:
            return ast.Parameter('...', ast.Varargs())
        else:
            ptype = self._create_type_from_base(symbol.base_type, is_parameter=True)

            if symbol.ident is None:
                if symbol.base_type and symbol.base_type.type != CTYPE_VOID:
                    message.warn_symbol(parent_symbol, "missing parameter name; undocumentable")
                ident = 'arg%d' % (index, )
            else:
                ident = symbol.ident

            return ast.Parameter(ident, ptype)

    def _create_return(self, source_type):
        typeval = self._create_type_from_base(source_type, is_return=True)
        return ast.Return(typeval)

    def _create_const(self, symbol):
        if symbol.ident.startswith('_'):
            return None

        # Don't create constants for non-public things
        # http://bugzilla.gnome.org/show_bug.cgi?id=572790
        if (symbol.source_filename is None or not symbol.source_filename.endswith('.h')):
            return None
        name = self._strip_symbol(symbol)
        if symbol.const_string is not None:
            typeval = ast.TYPE_STRING
            value = symbol.const_string
        elif symbol.const_int is not None:
            if symbol.base_type is not None:
                typeval = self._create_type_from_base(symbol.base_type)
            else:
                typeval = ast.TYPE_INT
            unaliased = typeval
            self._resolve_type_from_ctype(unaliased)
            if typeval.target_giname and typeval.ctype:
                target = self.lookup_giname(typeval.target_giname)
                target = self.resolve_aliases(target)
                if isinstance(target, ast.Type):
                    unaliased = target
            if unaliased == ast.TYPE_UINT64:
                value = str(symbol.const_int % 2 ** 64)
            elif unaliased == ast.TYPE_UINT32:
                value = str(symbol.const_int % 2 ** 32)
            elif unaliased == ast.TYPE_UINT16:
                value = str(symbol.const_int % 2 ** 16)
            elif unaliased == ast.TYPE_UINT8:
                value = str(symbol.const_int % 2 ** 16)
            else:
                value = str(symbol.const_int)
        elif symbol.const_boolean is not None:
            typeval = ast.TYPE_BOOLEAN
            value = "true" if symbol.const_boolean else "false"
        elif symbol.const_double is not None:
            typeval = ast.TYPE_DOUBLE
            value = '%f' % (symbol.const_double, )
        else:
            raise AssertionError()

        const = ast.Constant(name, typeval, value,
                             symbol.ident)
        const.add_symbol_reference(symbol)
        return const

    def _create_typedef_compound(self, compound_class, symbol, disguised=False):
        name = self.strip_identifier(symbol.ident)
        assert symbol.base_type
        if symbol.base_type.name:
            tag_name = symbol.base_type.name
        else:
            tag_name = None

        # If the struct already exists in the tag namespace, use it.
        if tag_name in self._tag_ns:
            compound = self._tag_ns[tag_name]
            if compound.name:
                # If the struct name is set it means the struct has already been
                # promoted from the tag namespace to the main namespace by a
                # prior typedef struct. If we get here it means this is another
                # typedef of that struct. Instead of creating an alias to the
                # primary typedef that has been promoted, we create a new Record
                # with shared fields. This handles the case where we want to
                # give structs like GInitiallyUnowned its own Record:
                #    typedef struct _GObject GObject;
                #    typedef struct _GObject GInitiallyUnowned;
                # See: http://bugzilla.gnome.org/show_bug.cgi?id=569408
                new_compound = compound_class(name, symbol.ident, tag_name=tag_name)
                new_compound.fields = compound.fields
                new_compound.add_symbol_reference(symbol)
                return new_compound
            else:
                # If the struct does not have its name set, it exists only in
                # the tag namespace. Set it here and return it which will
                # promote it to the main namespace. Essentially the first
                # typedef for a struct clobbers its name and ctype which is what
                # will be visible to GI.
                compound.name = name
                compound.ctype = symbol.ident
        else:
            # Create a new struct with a typedef name and tag name when available.
            # Structs with a typedef name are promoted into the main namespace
            # by it being returned to the "parse" function and are also added to
            # the tag namespace if it has a tag_name set.
            compound = compound_class(name, symbol.ident, disguised=disguised, tag_name=tag_name)
            if not tag_name:
                # Case where we have an anonymous struct which is typedef'd:
                #   typedef struct {...} Struct;
                # we need to parse the fields because we never get a struct
                # in the tag namespace which is normally where fields are parsed.
                self._parse_fields(symbol, compound)

        compound.add_symbol_reference(symbol)
        return compound

    def _create_tag_ns_compound(self, compound_class, symbol):
        # Get or create a struct from C's tag namespace
        if symbol.ident in self._tag_ns:
            compound = self._tag_ns[symbol.ident]
        else:
            compound = compound_class(None, symbol.ident, tag_name=symbol.ident)
        # Fields may need to be parsed in either of the above cases because the
        # Record can be created with a typedef prior to the struct definition.
        self._parse_fields(symbol, compound)
        compound.add_symbol_reference(symbol)
        return compound

    def _create_member_compound(self, compound_class, symbol):
        compound = compound_class(symbol.ident, symbol.ident)
        self._parse_fields(symbol, compound)
        compound.add_symbol_reference(symbol)
        return compound

    def _create_typedef_callback(self, symbol):
        callback = self._create_callback(symbol)
        if not callback:
            return None
        return callback

    def _parse_fields(self, symbol, compound):
        for child in symbol.base_type.child_list:
            child_node = self._traverse_one(child, parent_symbol=symbol)
            if not child_node:
                continue
            if isinstance(child_node, ast.Field):
                field = child_node
            else:
                field = ast.Field(child.ident, None, True, False,
                              anonymous_node=child_node)
            compound.fields.append(field)

    def _create_callback(self, symbol, member=False):
        if (symbol.base_type.type == CTYPE_FUNCTION):  # function
            paramtype = symbol.base_type
            retvaltype = symbol.base_type.base_type
        elif (symbol.base_type.type == CTYPE_POINTER):  # function pointer
            paramtype = symbol.base_type.base_type
            retvaltype = symbol.base_type.base_type.base_type
        parameters = list(self._create_parameters(symbol, paramtype))
        retval = self._create_return(retvaltype)

        # Mark the 'user_data' arguments
        for i, param in enumerate(parameters):
            if (param.type.target_fundamental == 'gpointer' and param.argname == 'user_data'):
                param.closure_name = param.argname

        if member:
            name = symbol.ident
        elif symbol.ident.find('_') > 0:
            name = self._strip_symbol(symbol)
        else:
            name = self.strip_identifier(symbol.ident)
        callback = ast.Callback(name, retval, parameters, False,
                                ctype=symbol.ident)
        callback.add_symbol_reference(symbol)

        return callback

    def create_type_from_user_string(self, typestr):
        """Parse a C type string (as might be given from an
        annotation) and resolve it.  For compatibility, we can consume
both GI type string (utf8, Foo.Bar) style, as well as C (char *, FooBar) style.

Note that type resolution may not succeed."""
        if '.' in typestr:
            container = self._create_bare_container_type(typestr)
            if container:
                typeval = container
            else:
                typeval = self._namespace.type_from_name(typestr)
        else:
            typeval = self.create_type_from_ctype_string(typestr)

        self.resolve_type(typeval)
        if typeval.resolved:
            # Explicitly clear out the c_type; there isn't one in this case.
            typeval.ctype = None
        return typeval

    def _resolve_type_from_ctype_all_namespaces(self, typeval, pointer_stripped):
        # If we can't determine the namespace from the type name,
        # fall back to trying all of our includes.  An example of this is mutter,
        # which has nominal namespace of "Meta", but a few classes are
        # "Mutter".  We don't export that data in introspection currently.
        # Basically the library should be fixed, but we'll hack around it here.
        for namespace in self._parsed_includes.values():
            target = namespace.get_by_ctype(pointer_stripped)
            if target:
                typeval.target_giname = '%s.%s' % (namespace.name, target.name)
                return True
        return False

    def _resolve_type_from_ctype(self, typeval):
        assert typeval.ctype is not None
        pointer_stripped = typeval.ctype.replace('*', '')
        try:
            matches = self.split_ctype_namespaces(pointer_stripped)
        except ValueError:
            return self._resolve_type_from_ctype_all_namespaces(typeval, pointer_stripped)
        for namespace, name in matches:
            target = namespace.get(name)
            if not target:
                target = namespace.get_by_ctype(pointer_stripped)
            if target:
                typeval.target_giname = '%s.%s' % (namespace.name, target.name)
                return True
        return False

    def _resolve_type_from_gtype_name(self, typeval):
        assert typeval.gtype_name is not None
        for ns in self._iter_namespaces():
            node = ns.type_names.get(typeval.gtype_name, None)
            if node is not None:
                typeval.target_giname = '%s.%s' % (ns.name, node.name)
                return True
        return False

    def _resolve_type_internal(self, typeval):
        if isinstance(typeval, (ast.Array, ast.List)):
            return self.resolve_type(typeval.element_type)
        elif isinstance(typeval, ast.Map):
            key_resolved = self.resolve_type(typeval.key_type)
            value_resolved = self.resolve_type(typeval.value_type)
            return key_resolved and value_resolved
        elif typeval.resolved:
            return True
        elif typeval.ctype:
            return self._resolve_type_from_ctype(typeval)
        elif typeval.gtype_name:
            return self._resolve_type_from_gtype_name(typeval)

    def resolve_type(self, typeval):
        if not self._resolve_type_internal(typeval):
            return False

        if typeval.target_fundamental or typeval.target_foreign:
            return True

        assert typeval.target_giname is not None

        try:
            type_ = self.lookup_giname(typeval.target_giname)
        except KeyError:
            type_ = None

        if type_ is None:
            typeval.target_giname = None

        return typeval.resolved

    def resolve_aliases(self, typenode):
        """Removes all aliases from typenode, returns first non-alias
        in the typenode alias chain.  Returns typenode argument if it
        is not an alias."""
        while isinstance(typenode, ast.Alias):
            if typenode.target.target_giname is not None:
                typenode = self.lookup_giname(typenode.target.target_giname)
            elif typenode.target.target_fundamental is not None:
                typenode = typenode.target
            else:
                break
        return typenode