Merge Chromium + Blink git repositories
[chromium-blink-merge.git] / base / third_party / symbolize / demangle.cc
blobe858181a68f4ec83b4a3e30e71ba13c5bb082cd8
1 // Copyright (c) 2006, Google Inc.
2 // All rights reserved.
3 //
4 // Redistribution and use in source and binary forms, with or without
5 // modification, are permitted provided that the following conditions are
6 // met:
7 //
8 // * Redistributions of source code must retain the above copyright
9 // notice, this list of conditions and the following disclaimer.
10 // * Redistributions in binary form must reproduce the above
11 // copyright notice, this list of conditions and the following disclaimer
12 // in the documentation and/or other materials provided with the
13 // distribution.
14 // * Neither the name of Google Inc. nor the names of its
15 // contributors may be used to endorse or promote products derived from
16 // this software without specific prior written permission.
18 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
21 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
22 // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
23 // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
24 // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
25 // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
26 // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27 // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
28 // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 // Author: Satoru Takabayashi
32 // For reference check out:
33 // http://www.codesourcery.com/public/cxx-abi/abi.html#mangling
35 // Note that we only have partial C++0x support yet.
37 #include <stdio.h> // for NULL
38 #include "demangle.h"
40 _START_GOOGLE_NAMESPACE_
42 typedef struct {
43 const char *abbrev;
44 const char *real_name;
45 } AbbrevPair;
47 // List of operators from Itanium C++ ABI.
48 static const AbbrevPair kOperatorList[] = {
49 { "nw", "new" },
50 { "na", "new[]" },
51 { "dl", "delete" },
52 { "da", "delete[]" },
53 { "ps", "+" },
54 { "ng", "-" },
55 { "ad", "&" },
56 { "de", "*" },
57 { "co", "~" },
58 { "pl", "+" },
59 { "mi", "-" },
60 { "ml", "*" },
61 { "dv", "/" },
62 { "rm", "%" },
63 { "an", "&" },
64 { "or", "|" },
65 { "eo", "^" },
66 { "aS", "=" },
67 { "pL", "+=" },
68 { "mI", "-=" },
69 { "mL", "*=" },
70 { "dV", "/=" },
71 { "rM", "%=" },
72 { "aN", "&=" },
73 { "oR", "|=" },
74 { "eO", "^=" },
75 { "ls", "<<" },
76 { "rs", ">>" },
77 { "lS", "<<=" },
78 { "rS", ">>=" },
79 { "eq", "==" },
80 { "ne", "!=" },
81 { "lt", "<" },
82 { "gt", ">" },
83 { "le", "<=" },
84 { "ge", ">=" },
85 { "nt", "!" },
86 { "aa", "&&" },
87 { "oo", "||" },
88 { "pp", "++" },
89 { "mm", "--" },
90 { "cm", "," },
91 { "pm", "->*" },
92 { "pt", "->" },
93 { "cl", "()" },
94 { "ix", "[]" },
95 { "qu", "?" },
96 { "st", "sizeof" },
97 { "sz", "sizeof" },
98 { NULL, NULL },
101 // List of builtin types from Itanium C++ ABI.
102 static const AbbrevPair kBuiltinTypeList[] = {
103 { "v", "void" },
104 { "w", "wchar_t" },
105 { "b", "bool" },
106 { "c", "char" },
107 { "a", "signed char" },
108 { "h", "unsigned char" },
109 { "s", "short" },
110 { "t", "unsigned short" },
111 { "i", "int" },
112 { "j", "unsigned int" },
113 { "l", "long" },
114 { "m", "unsigned long" },
115 { "x", "long long" },
116 { "y", "unsigned long long" },
117 { "n", "__int128" },
118 { "o", "unsigned __int128" },
119 { "f", "float" },
120 { "d", "double" },
121 { "e", "long double" },
122 { "g", "__float128" },
123 { "z", "ellipsis" },
124 { NULL, NULL }
127 // List of substitutions Itanium C++ ABI.
128 static const AbbrevPair kSubstitutionList[] = {
129 { "St", "" },
130 { "Sa", "allocator" },
131 { "Sb", "basic_string" },
132 // std::basic_string<char, std::char_traits<char>,std::allocator<char> >
133 { "Ss", "string"},
134 // std::basic_istream<char, std::char_traits<char> >
135 { "Si", "istream" },
136 // std::basic_ostream<char, std::char_traits<char> >
137 { "So", "ostream" },
138 // std::basic_iostream<char, std::char_traits<char> >
139 { "Sd", "iostream" },
140 { NULL, NULL }
143 // State needed for demangling.
144 typedef struct {
145 const char *mangled_cur; // Cursor of mangled name.
146 char *out_cur; // Cursor of output string.
147 const char *out_begin; // Beginning of output string.
148 const char *out_end; // End of output string.
149 const char *prev_name; // For constructors/destructors.
150 int prev_name_length; // For constructors/destructors.
151 short nest_level; // For nested names.
152 bool append; // Append flag.
153 bool overflowed; // True if output gets overflowed.
154 } State;
156 // We don't use strlen() in libc since it's not guaranteed to be async
157 // signal safe.
158 static size_t StrLen(const char *str) {
159 size_t len = 0;
160 while (*str != '\0') {
161 ++str;
162 ++len;
164 return len;
167 // Returns true if "str" has at least "n" characters remaining.
168 static bool AtLeastNumCharsRemaining(const char *str, int n) {
169 for (int i = 0; i < n; ++i) {
170 if (str[i] == '\0') {
171 return false;
174 return true;
177 // Returns true if "str" has "prefix" as a prefix.
178 static bool StrPrefix(const char *str, const char *prefix) {
179 size_t i = 0;
180 while (str[i] != '\0' && prefix[i] != '\0' &&
181 str[i] == prefix[i]) {
182 ++i;
184 return prefix[i] == '\0'; // Consumed everything in "prefix".
187 static void InitState(State *state, const char *mangled,
188 char *out, int out_size) {
189 state->mangled_cur = mangled;
190 state->out_cur = out;
191 state->out_begin = out;
192 state->out_end = out + out_size;
193 state->prev_name = NULL;
194 state->prev_name_length = -1;
195 state->nest_level = -1;
196 state->append = true;
197 state->overflowed = false;
200 // Returns true and advances "mangled_cur" if we find "one_char_token"
201 // at "mangled_cur" position. It is assumed that "one_char_token" does
202 // not contain '\0'.
203 static bool ParseOneCharToken(State *state, const char one_char_token) {
204 if (state->mangled_cur[0] == one_char_token) {
205 ++state->mangled_cur;
206 return true;
208 return false;
211 // Returns true and advances "mangled_cur" if we find "two_char_token"
212 // at "mangled_cur" position. It is assumed that "two_char_token" does
213 // not contain '\0'.
214 static bool ParseTwoCharToken(State *state, const char *two_char_token) {
215 if (state->mangled_cur[0] == two_char_token[0] &&
216 state->mangled_cur[1] == two_char_token[1]) {
217 state->mangled_cur += 2;
218 return true;
220 return false;
223 // Returns true and advances "mangled_cur" if we find any character in
224 // "char_class" at "mangled_cur" position.
225 static bool ParseCharClass(State *state, const char *char_class) {
226 const char *p = char_class;
227 for (; *p != '\0'; ++p) {
228 if (state->mangled_cur[0] == *p) {
229 ++state->mangled_cur;
230 return true;
233 return false;
236 // This function is used for handling an optional non-terminal.
237 static bool Optional(bool) {
238 return true;
241 // This function is used for handling <non-terminal>+ syntax.
242 typedef bool (*ParseFunc)(State *);
243 static bool OneOrMore(ParseFunc parse_func, State *state) {
244 if (parse_func(state)) {
245 while (parse_func(state)) {
247 return true;
249 return false;
252 // This function is used for handling <non-terminal>* syntax. The function
253 // always returns true and must be followed by a termination token or a
254 // terminating sequence not handled by parse_func (e.g.
255 // ParseOneCharToken(state, 'E')).
256 static bool ZeroOrMore(ParseFunc parse_func, State *state) {
257 while (parse_func(state)) {
259 return true;
262 // Append "str" at "out_cur". If there is an overflow, "overflowed"
263 // is set to true for later use. The output string is ensured to
264 // always terminate with '\0' as long as there is no overflow.
265 static void Append(State *state, const char * const str, const int length) {
266 int i;
267 for (i = 0; i < length; ++i) {
268 if (state->out_cur + 1 < state->out_end) { // +1 for '\0'
269 *state->out_cur = str[i];
270 ++state->out_cur;
271 } else {
272 state->overflowed = true;
273 break;
276 if (!state->overflowed) {
277 *state->out_cur = '\0'; // Terminate it with '\0'
281 // We don't use equivalents in libc to avoid locale issues.
282 static bool IsLower(char c) {
283 return c >= 'a' && c <= 'z';
286 static bool IsAlpha(char c) {
287 return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z');
290 static bool IsDigit(char c) {
291 return c >= '0' && c <= '9';
294 // Returns true if "str" is a function clone suffix. These suffixes are used
295 // by GCC 4.5.x and later versions to indicate functions which have been
296 // cloned during optimization. We treat any sequence (.<alpha>+.<digit>+)+ as
297 // a function clone suffix.
298 static bool IsFunctionCloneSuffix(const char *str) {
299 size_t i = 0;
300 while (str[i] != '\0') {
301 // Consume a single .<alpha>+.<digit>+ sequence.
302 if (str[i] != '.' || !IsAlpha(str[i + 1])) {
303 return false;
305 i += 2;
306 while (IsAlpha(str[i])) {
307 ++i;
309 if (str[i] != '.' || !IsDigit(str[i + 1])) {
310 return false;
312 i += 2;
313 while (IsDigit(str[i])) {
314 ++i;
317 return true; // Consumed everything in "str".
320 // Append "str" with some tweaks, iff "append" state is true.
321 // Returns true so that it can be placed in "if" conditions.
322 static void MaybeAppendWithLength(State *state, const char * const str,
323 const int length) {
324 if (state->append && length > 0) {
325 // Append a space if the output buffer ends with '<' and "str"
326 // starts with '<' to avoid <<<.
327 if (str[0] == '<' && state->out_begin < state->out_cur &&
328 state->out_cur[-1] == '<') {
329 Append(state, " ", 1);
331 // Remember the last identifier name for ctors/dtors.
332 if (IsAlpha(str[0]) || str[0] == '_') {
333 state->prev_name = state->out_cur;
334 state->prev_name_length = length;
336 Append(state, str, length);
340 // A convenient wrapper arount MaybeAppendWithLength().
341 static bool MaybeAppend(State *state, const char * const str) {
342 if (state->append) {
343 int length = StrLen(str);
344 MaybeAppendWithLength(state, str, length);
346 return true;
349 // This function is used for handling nested names.
350 static bool EnterNestedName(State *state) {
351 state->nest_level = 0;
352 return true;
355 // This function is used for handling nested names.
356 static bool LeaveNestedName(State *state, short prev_value) {
357 state->nest_level = prev_value;
358 return true;
361 // Disable the append mode not to print function parameters, etc.
362 static bool DisableAppend(State *state) {
363 state->append = false;
364 return true;
367 // Restore the append mode to the previous state.
368 static bool RestoreAppend(State *state, bool prev_value) {
369 state->append = prev_value;
370 return true;
373 // Increase the nest level for nested names.
374 static void MaybeIncreaseNestLevel(State *state) {
375 if (state->nest_level > -1) {
376 ++state->nest_level;
380 // Appends :: for nested names if necessary.
381 static void MaybeAppendSeparator(State *state) {
382 if (state->nest_level >= 1) {
383 MaybeAppend(state, "::");
387 // Cancel the last separator if necessary.
388 static void MaybeCancelLastSeparator(State *state) {
389 if (state->nest_level >= 1 && state->append &&
390 state->out_begin <= state->out_cur - 2) {
391 state->out_cur -= 2;
392 *state->out_cur = '\0';
396 // Returns true if the identifier of the given length pointed to by
397 // "mangled_cur" is anonymous namespace.
398 static bool IdentifierIsAnonymousNamespace(State *state, int length) {
399 static const char anon_prefix[] = "_GLOBAL__N_";
400 return (length > (int)sizeof(anon_prefix) - 1 && // Should be longer.
401 StrPrefix(state->mangled_cur, anon_prefix));
404 // Forward declarations of our parsing functions.
405 static bool ParseMangledName(State *state);
406 static bool ParseEncoding(State *state);
407 static bool ParseName(State *state);
408 static bool ParseUnscopedName(State *state);
409 static bool ParseUnscopedTemplateName(State *state);
410 static bool ParseNestedName(State *state);
411 static bool ParsePrefix(State *state);
412 static bool ParseUnqualifiedName(State *state);
413 static bool ParseSourceName(State *state);
414 static bool ParseLocalSourceName(State *state);
415 static bool ParseNumber(State *state, int *number_out);
416 static bool ParseFloatNumber(State *state);
417 static bool ParseSeqId(State *state);
418 static bool ParseIdentifier(State *state, int length);
419 static bool ParseOperatorName(State *state);
420 static bool ParseSpecialName(State *state);
421 static bool ParseCallOffset(State *state);
422 static bool ParseNVOffset(State *state);
423 static bool ParseVOffset(State *state);
424 static bool ParseCtorDtorName(State *state);
425 static bool ParseType(State *state);
426 static bool ParseCVQualifiers(State *state);
427 static bool ParseBuiltinType(State *state);
428 static bool ParseFunctionType(State *state);
429 static bool ParseBareFunctionType(State *state);
430 static bool ParseClassEnumType(State *state);
431 static bool ParseArrayType(State *state);
432 static bool ParsePointerToMemberType(State *state);
433 static bool ParseTemplateParam(State *state);
434 static bool ParseTemplateTemplateParam(State *state);
435 static bool ParseTemplateArgs(State *state);
436 static bool ParseTemplateArg(State *state);
437 static bool ParseExpression(State *state);
438 static bool ParseExprPrimary(State *state);
439 static bool ParseLocalName(State *state);
440 static bool ParseDiscriminator(State *state);
441 static bool ParseSubstitution(State *state);
443 // Implementation note: the following code is a straightforward
444 // translation of the Itanium C++ ABI defined in BNF with a couple of
445 // exceptions.
447 // - Support GNU extensions not defined in the Itanium C++ ABI
448 // - <prefix> and <template-prefix> are combined to avoid infinite loop
449 // - Reorder patterns to shorten the code
450 // - Reorder patterns to give greedier functions precedence
451 // We'll mark "Less greedy than" for these cases in the code
453 // Each parsing function changes the state and returns true on
454 // success. Otherwise, don't change the state and returns false. To
455 // ensure that the state isn't changed in the latter case, we save the
456 // original state before we call more than one parsing functions
457 // consecutively with &&, and restore the state if unsuccessful. See
458 // ParseEncoding() as an example of this convention. We follow the
459 // convention throughout the code.
461 // Originally we tried to do demangling without following the full ABI
462 // syntax but it turned out we needed to follow the full syntax to
463 // parse complicated cases like nested template arguments. Note that
464 // implementing a full-fledged demangler isn't trivial (libiberty's
465 // cp-demangle.c has +4300 lines).
467 // Note that (foo) in <(foo) ...> is a modifier to be ignored.
469 // Reference:
470 // - Itanium C++ ABI
471 // <http://www.codesourcery.com/cxx-abi/abi.html#mangling>
473 // <mangled-name> ::= _Z <encoding>
474 static bool ParseMangledName(State *state) {
475 return ParseTwoCharToken(state, "_Z") && ParseEncoding(state);
478 // <encoding> ::= <(function) name> <bare-function-type>
479 // ::= <(data) name>
480 // ::= <special-name>
481 static bool ParseEncoding(State *state) {
482 State copy = *state;
483 if (ParseName(state) && ParseBareFunctionType(state)) {
484 return true;
486 *state = copy;
488 if (ParseName(state) || ParseSpecialName(state)) {
489 return true;
491 return false;
494 // <name> ::= <nested-name>
495 // ::= <unscoped-template-name> <template-args>
496 // ::= <unscoped-name>
497 // ::= <local-name>
498 static bool ParseName(State *state) {
499 if (ParseNestedName(state) || ParseLocalName(state)) {
500 return true;
503 State copy = *state;
504 if (ParseUnscopedTemplateName(state) &&
505 ParseTemplateArgs(state)) {
506 return true;
508 *state = copy;
510 // Less greedy than <unscoped-template-name> <template-args>.
511 if (ParseUnscopedName(state)) {
512 return true;
514 return false;
517 // <unscoped-name> ::= <unqualified-name>
518 // ::= St <unqualified-name>
519 static bool ParseUnscopedName(State *state) {
520 if (ParseUnqualifiedName(state)) {
521 return true;
524 State copy = *state;
525 if (ParseTwoCharToken(state, "St") &&
526 MaybeAppend(state, "std::") &&
527 ParseUnqualifiedName(state)) {
528 return true;
530 *state = copy;
531 return false;
534 // <unscoped-template-name> ::= <unscoped-name>
535 // ::= <substitution>
536 static bool ParseUnscopedTemplateName(State *state) {
537 return ParseUnscopedName(state) || ParseSubstitution(state);
540 // <nested-name> ::= N [<CV-qualifiers>] <prefix> <unqualified-name> E
541 // ::= N [<CV-qualifiers>] <template-prefix> <template-args> E
542 static bool ParseNestedName(State *state) {
543 State copy = *state;
544 if (ParseOneCharToken(state, 'N') &&
545 EnterNestedName(state) &&
546 Optional(ParseCVQualifiers(state)) &&
547 ParsePrefix(state) &&
548 LeaveNestedName(state, copy.nest_level) &&
549 ParseOneCharToken(state, 'E')) {
550 return true;
552 *state = copy;
553 return false;
556 // This part is tricky. If we literally translate them to code, we'll
557 // end up infinite loop. Hence we merge them to avoid the case.
559 // <prefix> ::= <prefix> <unqualified-name>
560 // ::= <template-prefix> <template-args>
561 // ::= <template-param>
562 // ::= <substitution>
563 // ::= # empty
564 // <template-prefix> ::= <prefix> <(template) unqualified-name>
565 // ::= <template-param>
566 // ::= <substitution>
567 static bool ParsePrefix(State *state) {
568 bool has_something = false;
569 while (true) {
570 MaybeAppendSeparator(state);
571 if (ParseTemplateParam(state) ||
572 ParseSubstitution(state) ||
573 ParseUnscopedName(state)) {
574 has_something = true;
575 MaybeIncreaseNestLevel(state);
576 continue;
578 MaybeCancelLastSeparator(state);
579 if (has_something && ParseTemplateArgs(state)) {
580 return ParsePrefix(state);
581 } else {
582 break;
585 return true;
588 // <unqualified-name> ::= <operator-name>
589 // ::= <ctor-dtor-name>
590 // ::= <source-name>
591 // ::= <local-source-name>
592 static bool ParseUnqualifiedName(State *state) {
593 return (ParseOperatorName(state) ||
594 ParseCtorDtorName(state) ||
595 ParseSourceName(state) ||
596 ParseLocalSourceName(state));
599 // <source-name> ::= <positive length number> <identifier>
600 static bool ParseSourceName(State *state) {
601 State copy = *state;
602 int length = -1;
603 if (ParseNumber(state, &length) && ParseIdentifier(state, length)) {
604 return true;
606 *state = copy;
607 return false;
610 // <local-source-name> ::= L <source-name> [<discriminator>]
612 // References:
613 // http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31775
614 // http://gcc.gnu.org/viewcvs?view=rev&revision=124467
615 static bool ParseLocalSourceName(State *state) {
616 State copy = *state;
617 if (ParseOneCharToken(state, 'L') && ParseSourceName(state) &&
618 Optional(ParseDiscriminator(state))) {
619 return true;
621 *state = copy;
622 return false;
625 // <number> ::= [n] <non-negative decimal integer>
626 // If "number_out" is non-null, then *number_out is set to the value of the
627 // parsed number on success.
628 static bool ParseNumber(State *state, int *number_out) {
629 int sign = 1;
630 if (ParseOneCharToken(state, 'n')) {
631 sign = -1;
633 const char *p = state->mangled_cur;
634 int number = 0;
635 for (;*p != '\0'; ++p) {
636 if (IsDigit(*p)) {
637 number = number * 10 + (*p - '0');
638 } else {
639 break;
642 if (p != state->mangled_cur) { // Conversion succeeded.
643 state->mangled_cur = p;
644 if (number_out != NULL) {
645 *number_out = number * sign;
647 return true;
649 return false;
652 // Floating-point literals are encoded using a fixed-length lowercase
653 // hexadecimal string.
654 static bool ParseFloatNumber(State *state) {
655 const char *p = state->mangled_cur;
656 for (;*p != '\0'; ++p) {
657 if (!IsDigit(*p) && !(*p >= 'a' && *p <= 'f')) {
658 break;
661 if (p != state->mangled_cur) { // Conversion succeeded.
662 state->mangled_cur = p;
663 return true;
665 return false;
668 // The <seq-id> is a sequence number in base 36,
669 // using digits and upper case letters
670 static bool ParseSeqId(State *state) {
671 const char *p = state->mangled_cur;
672 for (;*p != '\0'; ++p) {
673 if (!IsDigit(*p) && !(*p >= 'A' && *p <= 'Z')) {
674 break;
677 if (p != state->mangled_cur) { // Conversion succeeded.
678 state->mangled_cur = p;
679 return true;
681 return false;
684 // <identifier> ::= <unqualified source code identifier> (of given length)
685 static bool ParseIdentifier(State *state, int length) {
686 if (length == -1 ||
687 !AtLeastNumCharsRemaining(state->mangled_cur, length)) {
688 return false;
690 if (IdentifierIsAnonymousNamespace(state, length)) {
691 MaybeAppend(state, "(anonymous namespace)");
692 } else {
693 MaybeAppendWithLength(state, state->mangled_cur, length);
695 state->mangled_cur += length;
696 return true;
699 // <operator-name> ::= nw, and other two letters cases
700 // ::= cv <type> # (cast)
701 // ::= v <digit> <source-name> # vendor extended operator
702 static bool ParseOperatorName(State *state) {
703 if (!AtLeastNumCharsRemaining(state->mangled_cur, 2)) {
704 return false;
706 // First check with "cv" (cast) case.
707 State copy = *state;
708 if (ParseTwoCharToken(state, "cv") &&
709 MaybeAppend(state, "operator ") &&
710 EnterNestedName(state) &&
711 ParseType(state) &&
712 LeaveNestedName(state, copy.nest_level)) {
713 return true;
715 *state = copy;
717 // Then vendor extended operators.
718 if (ParseOneCharToken(state, 'v') && ParseCharClass(state, "0123456789") &&
719 ParseSourceName(state)) {
720 return true;
722 *state = copy;
724 // Other operator names should start with a lower alphabet followed
725 // by a lower/upper alphabet.
726 if (!(IsLower(state->mangled_cur[0]) &&
727 IsAlpha(state->mangled_cur[1]))) {
728 return false;
730 // We may want to perform a binary search if we really need speed.
731 const AbbrevPair *p;
732 for (p = kOperatorList; p->abbrev != NULL; ++p) {
733 if (state->mangled_cur[0] == p->abbrev[0] &&
734 state->mangled_cur[1] == p->abbrev[1]) {
735 MaybeAppend(state, "operator");
736 if (IsLower(*p->real_name)) { // new, delete, etc.
737 MaybeAppend(state, " ");
739 MaybeAppend(state, p->real_name);
740 state->mangled_cur += 2;
741 return true;
744 return false;
747 // <special-name> ::= TV <type>
748 // ::= TT <type>
749 // ::= TI <type>
750 // ::= TS <type>
751 // ::= Tc <call-offset> <call-offset> <(base) encoding>
752 // ::= GV <(object) name>
753 // ::= T <call-offset> <(base) encoding>
754 // G++ extensions:
755 // ::= TC <type> <(offset) number> _ <(base) type>
756 // ::= TF <type>
757 // ::= TJ <type>
758 // ::= GR <name>
759 // ::= GA <encoding>
760 // ::= Th <call-offset> <(base) encoding>
761 // ::= Tv <call-offset> <(base) encoding>
763 // Note: we don't care much about them since they don't appear in
764 // stack traces. The are special data.
765 static bool ParseSpecialName(State *state) {
766 State copy = *state;
767 if (ParseOneCharToken(state, 'T') &&
768 ParseCharClass(state, "VTIS") &&
769 ParseType(state)) {
770 return true;
772 *state = copy;
774 if (ParseTwoCharToken(state, "Tc") && ParseCallOffset(state) &&
775 ParseCallOffset(state) && ParseEncoding(state)) {
776 return true;
778 *state = copy;
780 if (ParseTwoCharToken(state, "GV") &&
781 ParseName(state)) {
782 return true;
784 *state = copy;
786 if (ParseOneCharToken(state, 'T') && ParseCallOffset(state) &&
787 ParseEncoding(state)) {
788 return true;
790 *state = copy;
792 // G++ extensions
793 if (ParseTwoCharToken(state, "TC") && ParseType(state) &&
794 ParseNumber(state, NULL) && ParseOneCharToken(state, '_') &&
795 DisableAppend(state) &&
796 ParseType(state)) {
797 RestoreAppend(state, copy.append);
798 return true;
800 *state = copy;
802 if (ParseOneCharToken(state, 'T') && ParseCharClass(state, "FJ") &&
803 ParseType(state)) {
804 return true;
806 *state = copy;
808 if (ParseTwoCharToken(state, "GR") && ParseName(state)) {
809 return true;
811 *state = copy;
813 if (ParseTwoCharToken(state, "GA") && ParseEncoding(state)) {
814 return true;
816 *state = copy;
818 if (ParseOneCharToken(state, 'T') && ParseCharClass(state, "hv") &&
819 ParseCallOffset(state) && ParseEncoding(state)) {
820 return true;
822 *state = copy;
823 return false;
826 // <call-offset> ::= h <nv-offset> _
827 // ::= v <v-offset> _
828 static bool ParseCallOffset(State *state) {
829 State copy = *state;
830 if (ParseOneCharToken(state, 'h') &&
831 ParseNVOffset(state) && ParseOneCharToken(state, '_')) {
832 return true;
834 *state = copy;
836 if (ParseOneCharToken(state, 'v') &&
837 ParseVOffset(state) && ParseOneCharToken(state, '_')) {
838 return true;
840 *state = copy;
842 return false;
845 // <nv-offset> ::= <(offset) number>
846 static bool ParseNVOffset(State *state) {
847 return ParseNumber(state, NULL);
850 // <v-offset> ::= <(offset) number> _ <(virtual offset) number>
851 static bool ParseVOffset(State *state) {
852 State copy = *state;
853 if (ParseNumber(state, NULL) && ParseOneCharToken(state, '_') &&
854 ParseNumber(state, NULL)) {
855 return true;
857 *state = copy;
858 return false;
861 // <ctor-dtor-name> ::= C1 | C2 | C3
862 // ::= D0 | D1 | D2
863 static bool ParseCtorDtorName(State *state) {
864 State copy = *state;
865 if (ParseOneCharToken(state, 'C') &&
866 ParseCharClass(state, "123")) {
867 const char * const prev_name = state->prev_name;
868 const int prev_name_length = state->prev_name_length;
869 MaybeAppendWithLength(state, prev_name, prev_name_length);
870 return true;
872 *state = copy;
874 if (ParseOneCharToken(state, 'D') &&
875 ParseCharClass(state, "012")) {
876 const char * const prev_name = state->prev_name;
877 const int prev_name_length = state->prev_name_length;
878 MaybeAppend(state, "~");
879 MaybeAppendWithLength(state, prev_name, prev_name_length);
880 return true;
882 *state = copy;
883 return false;
886 // <type> ::= <CV-qualifiers> <type>
887 // ::= P <type> # pointer-to
888 // ::= R <type> # reference-to
889 // ::= O <type> # rvalue reference-to (C++0x)
890 // ::= C <type> # complex pair (C 2000)
891 // ::= G <type> # imaginary (C 2000)
892 // ::= U <source-name> <type> # vendor extended type qualifier
893 // ::= <builtin-type>
894 // ::= <function-type>
895 // ::= <class-enum-type>
896 // ::= <array-type>
897 // ::= <pointer-to-member-type>
898 // ::= <template-template-param> <template-args>
899 // ::= <template-param>
900 // ::= <substitution>
901 // ::= Dp <type> # pack expansion of (C++0x)
902 // ::= Dt <expression> E # decltype of an id-expression or class
903 // # member access (C++0x)
904 // ::= DT <expression> E # decltype of an expression (C++0x)
906 static bool ParseType(State *state) {
907 // We should check CV-qualifers, and PRGC things first.
908 State copy = *state;
909 if (ParseCVQualifiers(state) && ParseType(state)) {
910 return true;
912 *state = copy;
914 if (ParseCharClass(state, "OPRCG") && ParseType(state)) {
915 return true;
917 *state = copy;
919 if (ParseTwoCharToken(state, "Dp") && ParseType(state)) {
920 return true;
922 *state = copy;
924 if (ParseOneCharToken(state, 'D') && ParseCharClass(state, "tT") &&
925 ParseExpression(state) && ParseOneCharToken(state, 'E')) {
926 return true;
928 *state = copy;
930 if (ParseOneCharToken(state, 'U') && ParseSourceName(state) &&
931 ParseType(state)) {
932 return true;
934 *state = copy;
936 if (ParseBuiltinType(state) ||
937 ParseFunctionType(state) ||
938 ParseClassEnumType(state) ||
939 ParseArrayType(state) ||
940 ParsePointerToMemberType(state) ||
941 ParseSubstitution(state)) {
942 return true;
945 if (ParseTemplateTemplateParam(state) &&
946 ParseTemplateArgs(state)) {
947 return true;
949 *state = copy;
951 // Less greedy than <template-template-param> <template-args>.
952 if (ParseTemplateParam(state)) {
953 return true;
956 return false;
959 // <CV-qualifiers> ::= [r] [V] [K]
960 // We don't allow empty <CV-qualifiers> to avoid infinite loop in
961 // ParseType().
962 static bool ParseCVQualifiers(State *state) {
963 int num_cv_qualifiers = 0;
964 num_cv_qualifiers += ParseOneCharToken(state, 'r');
965 num_cv_qualifiers += ParseOneCharToken(state, 'V');
966 num_cv_qualifiers += ParseOneCharToken(state, 'K');
967 return num_cv_qualifiers > 0;
970 // <builtin-type> ::= v, etc.
971 // ::= u <source-name>
972 static bool ParseBuiltinType(State *state) {
973 const AbbrevPair *p;
974 for (p = kBuiltinTypeList; p->abbrev != NULL; ++p) {
975 if (state->mangled_cur[0] == p->abbrev[0]) {
976 MaybeAppend(state, p->real_name);
977 ++state->mangled_cur;
978 return true;
982 State copy = *state;
983 if (ParseOneCharToken(state, 'u') && ParseSourceName(state)) {
984 return true;
986 *state = copy;
987 return false;
990 // <function-type> ::= F [Y] <bare-function-type> E
991 static bool ParseFunctionType(State *state) {
992 State copy = *state;
993 if (ParseOneCharToken(state, 'F') &&
994 Optional(ParseOneCharToken(state, 'Y')) &&
995 ParseBareFunctionType(state) && ParseOneCharToken(state, 'E')) {
996 return true;
998 *state = copy;
999 return false;
1002 // <bare-function-type> ::= <(signature) type>+
1003 static bool ParseBareFunctionType(State *state) {
1004 State copy = *state;
1005 DisableAppend(state);
1006 if (OneOrMore(ParseType, state)) {
1007 RestoreAppend(state, copy.append);
1008 MaybeAppend(state, "()");
1009 return true;
1011 *state = copy;
1012 return false;
1015 // <class-enum-type> ::= <name>
1016 static bool ParseClassEnumType(State *state) {
1017 return ParseName(state);
1020 // <array-type> ::= A <(positive dimension) number> _ <(element) type>
1021 // ::= A [<(dimension) expression>] _ <(element) type>
1022 static bool ParseArrayType(State *state) {
1023 State copy = *state;
1024 if (ParseOneCharToken(state, 'A') && ParseNumber(state, NULL) &&
1025 ParseOneCharToken(state, '_') && ParseType(state)) {
1026 return true;
1028 *state = copy;
1030 if (ParseOneCharToken(state, 'A') && Optional(ParseExpression(state)) &&
1031 ParseOneCharToken(state, '_') && ParseType(state)) {
1032 return true;
1034 *state = copy;
1035 return false;
1038 // <pointer-to-member-type> ::= M <(class) type> <(member) type>
1039 static bool ParsePointerToMemberType(State *state) {
1040 State copy = *state;
1041 if (ParseOneCharToken(state, 'M') && ParseType(state) &&
1042 ParseType(state)) {
1043 return true;
1045 *state = copy;
1046 return false;
1049 // <template-param> ::= T_
1050 // ::= T <parameter-2 non-negative number> _
1051 static bool ParseTemplateParam(State *state) {
1052 if (ParseTwoCharToken(state, "T_")) {
1053 MaybeAppend(state, "?"); // We don't support template substitutions.
1054 return true;
1057 State copy = *state;
1058 if (ParseOneCharToken(state, 'T') && ParseNumber(state, NULL) &&
1059 ParseOneCharToken(state, '_')) {
1060 MaybeAppend(state, "?"); // We don't support template substitutions.
1061 return true;
1063 *state = copy;
1064 return false;
1068 // <template-template-param> ::= <template-param>
1069 // ::= <substitution>
1070 static bool ParseTemplateTemplateParam(State *state) {
1071 return (ParseTemplateParam(state) ||
1072 ParseSubstitution(state));
1075 // <template-args> ::= I <template-arg>+ E
1076 static bool ParseTemplateArgs(State *state) {
1077 State copy = *state;
1078 DisableAppend(state);
1079 if (ParseOneCharToken(state, 'I') &&
1080 OneOrMore(ParseTemplateArg, state) &&
1081 ParseOneCharToken(state, 'E')) {
1082 RestoreAppend(state, copy.append);
1083 MaybeAppend(state, "<>");
1084 return true;
1086 *state = copy;
1087 return false;
1090 // <template-arg> ::= <type>
1091 // ::= <expr-primary>
1092 // ::= I <template-arg>* E # argument pack
1093 // ::= X <expression> E
1094 static bool ParseTemplateArg(State *state) {
1095 State copy = *state;
1096 if (ParseOneCharToken(state, 'I') &&
1097 ZeroOrMore(ParseTemplateArg, state) &&
1098 ParseOneCharToken(state, 'E')) {
1099 return true;
1101 *state = copy;
1103 if (ParseType(state) ||
1104 ParseExprPrimary(state)) {
1105 return true;
1107 *state = copy;
1109 if (ParseOneCharToken(state, 'X') && ParseExpression(state) &&
1110 ParseOneCharToken(state, 'E')) {
1111 return true;
1113 *state = copy;
1114 return false;
1117 // <expression> ::= <template-param>
1118 // ::= <expr-primary>
1119 // ::= <unary operator-name> <expression>
1120 // ::= <binary operator-name> <expression> <expression>
1121 // ::= <trinary operator-name> <expression> <expression>
1122 // <expression>
1123 // ::= st <type>
1124 // ::= sr <type> <unqualified-name> <template-args>
1125 // ::= sr <type> <unqualified-name>
1126 static bool ParseExpression(State *state) {
1127 if (ParseTemplateParam(state) || ParseExprPrimary(state)) {
1128 return true;
1131 State copy = *state;
1132 if (ParseOperatorName(state) &&
1133 ParseExpression(state) &&
1134 ParseExpression(state) &&
1135 ParseExpression(state)) {
1136 return true;
1138 *state = copy;
1140 if (ParseOperatorName(state) &&
1141 ParseExpression(state) &&
1142 ParseExpression(state)) {
1143 return true;
1145 *state = copy;
1147 if (ParseOperatorName(state) &&
1148 ParseExpression(state)) {
1149 return true;
1151 *state = copy;
1153 if (ParseTwoCharToken(state, "st") && ParseType(state)) {
1154 return true;
1156 *state = copy;
1158 if (ParseTwoCharToken(state, "sr") && ParseType(state) &&
1159 ParseUnqualifiedName(state) &&
1160 ParseTemplateArgs(state)) {
1161 return true;
1163 *state = copy;
1165 if (ParseTwoCharToken(state, "sr") && ParseType(state) &&
1166 ParseUnqualifiedName(state)) {
1167 return true;
1169 *state = copy;
1170 return false;
1173 // <expr-primary> ::= L <type> <(value) number> E
1174 // ::= L <type> <(value) float> E
1175 // ::= L <mangled-name> E
1176 // // A bug in g++'s C++ ABI version 2 (-fabi-version=2).
1177 // ::= LZ <encoding> E
1178 static bool ParseExprPrimary(State *state) {
1179 State copy = *state;
1180 if (ParseOneCharToken(state, 'L') && ParseType(state) &&
1181 ParseNumber(state, NULL) &&
1182 ParseOneCharToken(state, 'E')) {
1183 return true;
1185 *state = copy;
1187 if (ParseOneCharToken(state, 'L') && ParseType(state) &&
1188 ParseFloatNumber(state) &&
1189 ParseOneCharToken(state, 'E')) {
1190 return true;
1192 *state = copy;
1194 if (ParseOneCharToken(state, 'L') && ParseMangledName(state) &&
1195 ParseOneCharToken(state, 'E')) {
1196 return true;
1198 *state = copy;
1200 if (ParseTwoCharToken(state, "LZ") && ParseEncoding(state) &&
1201 ParseOneCharToken(state, 'E')) {
1202 return true;
1204 *state = copy;
1206 return false;
1209 // <local-name> := Z <(function) encoding> E <(entity) name>
1210 // [<discriminator>]
1211 // := Z <(function) encoding> E s [<discriminator>]
1212 static bool ParseLocalName(State *state) {
1213 State copy = *state;
1214 if (ParseOneCharToken(state, 'Z') && ParseEncoding(state) &&
1215 ParseOneCharToken(state, 'E') && MaybeAppend(state, "::") &&
1216 ParseName(state) && Optional(ParseDiscriminator(state))) {
1217 return true;
1219 *state = copy;
1221 if (ParseOneCharToken(state, 'Z') && ParseEncoding(state) &&
1222 ParseTwoCharToken(state, "Es") && Optional(ParseDiscriminator(state))) {
1223 return true;
1225 *state = copy;
1226 return false;
1229 // <discriminator> := _ <(non-negative) number>
1230 static bool ParseDiscriminator(State *state) {
1231 State copy = *state;
1232 if (ParseOneCharToken(state, '_') && ParseNumber(state, NULL)) {
1233 return true;
1235 *state = copy;
1236 return false;
1239 // <substitution> ::= S_
1240 // ::= S <seq-id> _
1241 // ::= St, etc.
1242 static bool ParseSubstitution(State *state) {
1243 if (ParseTwoCharToken(state, "S_")) {
1244 MaybeAppend(state, "?"); // We don't support substitutions.
1245 return true;
1248 State copy = *state;
1249 if (ParseOneCharToken(state, 'S') && ParseSeqId(state) &&
1250 ParseOneCharToken(state, '_')) {
1251 MaybeAppend(state, "?"); // We don't support substitutions.
1252 return true;
1254 *state = copy;
1256 // Expand abbreviations like "St" => "std".
1257 if (ParseOneCharToken(state, 'S')) {
1258 const AbbrevPair *p;
1259 for (p = kSubstitutionList; p->abbrev != NULL; ++p) {
1260 if (state->mangled_cur[0] == p->abbrev[1]) {
1261 MaybeAppend(state, "std");
1262 if (p->real_name[0] != '\0') {
1263 MaybeAppend(state, "::");
1264 MaybeAppend(state, p->real_name);
1266 ++state->mangled_cur;
1267 return true;
1271 *state = copy;
1272 return false;
1275 // Parse <mangled-name>, optionally followed by either a function-clone suffix
1276 // or version suffix. Returns true only if all of "mangled_cur" was consumed.
1277 static bool ParseTopLevelMangledName(State *state) {
1278 if (ParseMangledName(state)) {
1279 if (state->mangled_cur[0] != '\0') {
1280 // Drop trailing function clone suffix, if any.
1281 if (IsFunctionCloneSuffix(state->mangled_cur)) {
1282 return true;
1284 // Append trailing version suffix if any.
1285 // ex. _Z3foo@@GLIBCXX_3.4
1286 if (state->mangled_cur[0] == '@') {
1287 MaybeAppend(state, state->mangled_cur);
1288 return true;
1290 return false; // Unconsumed suffix.
1292 return true;
1294 return false;
1297 // The demangler entry point.
1298 bool Demangle(const char *mangled, char *out, int out_size) {
1299 State state;
1300 InitState(&state, mangled, out, out_size);
1301 return ParseTopLevelMangledName(&state) && !state.overflowed;
1304 _END_GOOGLE_NAMESPACE_