2 * Demangle VC++ symbols into C function prototypes
4 * Copyright 2000 Jon Griffiths
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
23 #include "wine/port.h"
30 #include "wine/debug.h"
32 WINE_DEFAULT_DEBUG_CHANNEL(msvcrt
);
35 * - document a bit (grammar + functions)
36 * - back-port this new code into tools/winedump/msmangle.c
39 #define UNDNAME_COMPLETE (0x0000)
40 #define UNDNAME_NO_LEADING_UNDERSCORES (0x0001) /* Don't show __ in calling convention */
41 #define UNDNAME_NO_MS_KEYWORDS (0x0002) /* Don't show calling convention at all */
42 #define UNDNAME_NO_FUNCTION_RETURNS (0x0004) /* Don't show function/method return value */
43 #define UNDNAME_NO_ALLOCATION_MODEL (0x0008)
44 #define UNDNAME_NO_ALLOCATION_LANGUAGE (0x0010)
45 #define UNDNAME_NO_MS_THISTYPE (0x0020)
46 #define UNDNAME_NO_CV_THISTYPE (0x0040)
47 #define UNDNAME_NO_THISTYPE (0x0060)
48 #define UNDNAME_NO_ACCESS_SPECIFIERS (0x0080) /* Don't show access specifier (public/protected/private) */
49 #define UNDNAME_NO_THROW_SIGNATURES (0x0100)
50 #define UNDNAME_NO_MEMBER_TYPE (0x0200) /* Don't show static/virtual specifier */
51 #define UNDNAME_NO_RETURN_UDT_MODEL (0x0400)
52 #define UNDNAME_32_BIT_DECODE (0x0800)
53 #define UNDNAME_NAME_ONLY (0x1000) /* Only report the variable/method name */
54 #define UNDNAME_NO_ARGUMENTS (0x2000) /* Don't show method arguments */
55 #define UNDNAME_NO_SPECIAL_SYMS (0x4000)
56 #define UNDNAME_NO_COMPLEX_TYPE (0x8000)
58 /* How data types modifiers are stored:
59 * M (in the following definitions) is defined for
60 * 'A', 'B', 'C' and 'D' as follows
64 * {<D>}: "const volatile "
71 * same as for arguments and also the following
78 unsigned start
; /* first valid reference in array */
79 unsigned num
; /* total number of used elts */
85 /* Structure holding a parsed symbol */
88 unsigned flags
; /* the UNDNAME_ flags used for demangling */
89 malloc_func_t mem_alloc_ptr
; /* internal allocator */
90 free_func_t mem_free_ptr
; /* internal deallocator */
92 const char* current
; /* pointer in input (mangled) string */
93 char* result
; /* demangled string */
95 struct array names
; /* array of names for back reference */
96 struct array stack
; /* stack of parsed strings */
98 void* alloc_list
; /* linked list of allocated blocks */
99 unsigned avail_in_first
; /* number of available bytes in head block */
102 /* Type for parsing mangled types */
109 /******************************************************************
112 * Internal allocator. Uses a simple linked list of large blocks
113 * where we use a poor-man allocator. It's fast, and since all
114 * allocation is pool, memory management is easy (esp. freeing).
116 static void* und_alloc(struct parsed_symbol
* sym
, unsigned int len
)
120 #define BLOCK_SIZE 1024
121 #define AVAIL_SIZE (1024 - sizeof(void*))
123 if (len
> AVAIL_SIZE
)
125 /* allocate a specific block */
126 ptr
= sym
->mem_alloc_ptr(sizeof(void*) + len
);
127 if (!ptr
) return NULL
;
128 *(void**)ptr
= sym
->alloc_list
;
129 sym
->alloc_list
= ptr
;
130 sym
->avail_in_first
= 0;
131 ptr
= (char*)sym
->alloc_list
+ sizeof(void*);
135 if (len
> sym
->avail_in_first
)
137 /* add a new block */
138 ptr
= sym
->mem_alloc_ptr(BLOCK_SIZE
);
139 if (!ptr
) return NULL
;
140 *(void**)ptr
= sym
->alloc_list
;
141 sym
->alloc_list
= ptr
;
142 sym
->avail_in_first
= AVAIL_SIZE
;
144 /* grab memory from head block */
145 ptr
= (char*)sym
->alloc_list
+ BLOCK_SIZE
- sym
->avail_in_first
;
146 sym
->avail_in_first
-= len
;
153 /******************************************************************
155 * Frees all the blocks in the list of large blocks allocated by
158 static void und_free_all(struct parsed_symbol
* sym
)
162 while (sym
->alloc_list
)
164 next
= *(void**)sym
->alloc_list
;
165 if(sym
->mem_free_ptr
) sym
->mem_free_ptr(sym
->alloc_list
);
166 sym
->alloc_list
= next
;
168 sym
->avail_in_first
= 0;
171 /******************************************************************
173 * Initialises an array of strings
175 static void str_array_init(struct array
* a
)
177 a
->start
= a
->num
= a
->max
= a
->alloc
= 0;
181 /******************************************************************
183 * Adding a new string to an array
185 static BOOL
str_array_push(struct parsed_symbol
* sym
, const char* ptr
, int len
,
195 new = und_alloc(sym
, (a
->alloc
= 32) * sizeof(a
->elts
[0]));
196 if (!new) return FALSE
;
199 else if (a
->max
>= a
->alloc
)
201 new = und_alloc(sym
, (a
->alloc
* 2) * sizeof(a
->elts
[0]));
202 if (!new) return FALSE
;
203 memcpy(new, a
->elts
, a
->alloc
* sizeof(a
->elts
[0]));
207 if (len
== -1) len
= strlen(ptr
);
208 a
->elts
[a
->num
] = und_alloc(sym
, len
+ 1);
209 assert(a
->elts
[a
->num
]);
210 memcpy(a
->elts
[a
->num
], ptr
, len
);
211 a
->elts
[a
->num
][len
] = '\0';
212 if (++a
->num
>= a
->max
) a
->max
= a
->num
;
217 for (i
= a
->max
- 1; i
>= 0; i
--)
220 if (i
< a
->start
) c
= '-';
221 else if (i
>= a
->num
) c
= '}';
222 TRACE("%p\t%d%c %s\n", a
, i
, c
, a
->elts
[i
]);
229 /******************************************************************
231 * Extracts a reference from an existing array (doing proper type
234 static char* str_array_get_ref(struct array
* cref
, unsigned idx
)
237 if (cref
->start
+ idx
>= cref
->max
)
239 WARN("Out of bounds: %p %d + %d >= %d\n",
240 cref
, cref
->start
, idx
, cref
->max
);
243 TRACE("Returning %p[%d] => %s\n",
244 cref
, idx
, cref
->elts
[cref
->start
+ idx
]);
245 return cref
->elts
[cref
->start
+ idx
];
248 /******************************************************************
250 * Helper for printf type of command (only %s and %c are implemented)
251 * while dynamically allocating the buffer
253 static char* str_printf(struct parsed_symbol
* sym
, const char* format
, ...)
256 unsigned int len
= 1, i
, sz
;
261 va_start(args
, format
);
262 for (i
= 0; format
[i
]; i
++)
264 if (format
[i
] == '%')
268 case 's': t
= va_arg(args
, char*); if (t
) len
+= strlen(t
); break;
269 case 'c': (void)va_arg(args
, int); len
++; break;
270 default: i
--; /* fall thru */
271 case '%': len
++; break;
277 if (!(tmp
= und_alloc(sym
, len
))) return NULL
;
278 va_start(args
, format
);
279 for (p
= tmp
, i
= 0; format
[i
]; i
++)
281 if (format
[i
] == '%')
286 t
= va_arg(args
, char*);
295 *p
++ = (char)va_arg(args
, int);
297 default: i
--; /* fall thru */
298 case '%': *p
++ = '%'; break;
301 else *p
++ = format
[i
];
308 /* forward declaration */
309 static BOOL
demangle_datatype(struct parsed_symbol
* sym
, struct datatype_t
* ct
,
310 struct array
* pmt
, BOOL in_args
);
312 static const char* get_number(struct parsed_symbol
* sym
)
317 if (*sym
->current
== '?')
322 if (*sym
->current
>= '0' && *sym
->current
<= '8')
324 ptr
= und_alloc(sym
, 3);
325 if (sgn
) ptr
[0] = '-';
326 ptr
[sgn
? 1 : 0] = *sym
->current
+ 1;
327 ptr
[sgn
? 2 : 1] = '\0';
330 else if (*sym
->current
== '9')
332 ptr
= und_alloc(sym
, 4);
333 if (sgn
) ptr
[0] = '-';
334 ptr
[sgn
? 1 : 0] = '1';
335 ptr
[sgn
? 2 : 1] = '0';
336 ptr
[sgn
? 3 : 2] = '\0';
339 else if (*sym
->current
>= 'A' && *sym
->current
<= 'P')
343 while (*sym
->current
>= 'A' && *sym
->current
<= 'P')
346 ret
+= *sym
->current
++ - 'A';
348 if (*sym
->current
!= '@') return NULL
;
350 ptr
= und_alloc(sym
, 17);
351 sprintf(ptr
, "%s%d", sgn
? "-" : "", ret
);
358 /******************************************************************
360 * Parses a list of function/method arguments, creates a string corresponding
361 * to the arguments' list.
363 static char* get_args(struct parsed_symbol
* sym
, struct array
* pmt_ref
, BOOL z_term
,
364 char open_char
, char close_char
)
367 struct datatype_t ct
;
368 struct array arg_collect
;
369 char* args_str
= NULL
;
373 str_array_init(&arg_collect
);
375 /* Now come the function arguments */
376 while (*sym
->current
)
378 /* Decode each data type and append it to the argument list */
379 if (*sym
->current
== '@')
384 if (!demangle_datatype(sym
, &ct
, pmt_ref
, TRUE
))
386 /* 'void' terminates an argument list in a function */
387 if (z_term
&& !strcmp(ct
.left
, "void")) break;
388 if (!str_array_push(sym
, str_printf(sym
, "%s%s", ct
.left
, ct
.right
), -1,
391 if (!strcmp(ct
.left
, "...")) break;
393 /* Functions are always terminated by 'Z'. If we made it this far and
394 * don't find it, we have incorrectly identified a data type.
396 if (z_term
&& *sym
->current
++ != 'Z') return NULL
;
398 if (arg_collect
.num
== 0 ||
399 (arg_collect
.num
== 1 && !strcmp(arg_collect
.elts
[0], "void")))
400 return str_printf(sym
, "%cvoid%c", open_char
, close_char
);
401 for (i
= 1; i
< arg_collect
.num
; i
++)
403 args_str
= str_printf(sym
, "%s,%s", args_str
, arg_collect
.elts
[i
]);
406 last
= args_str
? args_str
: arg_collect
.elts
[0];
407 if (close_char
== '>' && last
[strlen(last
) - 1] == '>')
408 args_str
= str_printf(sym
, "%c%s%s %c",
409 open_char
, arg_collect
.elts
[0], args_str
, close_char
);
411 args_str
= str_printf(sym
, "%c%s%s%c",
412 open_char
, arg_collect
.elts
[0], args_str
, close_char
);
417 /******************************************************************
419 * Parses the type modifier. Always returns a static string
421 static BOOL
get_modifier(char ch
, const char** ret
)
425 case 'A': *ret
= NULL
; break;
426 case 'B': *ret
= "const"; break;
427 case 'C': *ret
= "volatile"; break;
428 case 'D': *ret
= "const volatile"; break;
429 default: return FALSE
;
434 static BOOL
get_modified_type(struct datatype_t
*ct
, struct parsed_symbol
* sym
,
435 struct array
*pmt_ref
, char modif
, BOOL in_args
)
437 const char* modifier
;
438 const char* str_modif
;
442 case 'A': str_modif
= " &"; break;
443 case 'B': str_modif
= " & volatile"; break;
444 case 'P': str_modif
= " *"; break;
445 case 'Q': str_modif
= " * const"; break;
446 case 'R': str_modif
= " * volatile"; break;
447 case 'S': str_modif
= " * const volatile"; break;
448 case '?': str_modif
= ""; break;
449 default: return FALSE
;
452 if (get_modifier(*sym
->current
++, &modifier
))
454 unsigned mark
= sym
->stack
.num
;
455 struct datatype_t sub_ct
;
457 /* multidimensional arrays */
458 if (*sym
->current
== 'Y')
464 if (!(n1
= get_number(sym
))) return FALSE
;
467 if (str_modif
[0] == ' ' && !modifier
)
472 str_modif
= str_printf(sym
, " (%s%s)", modifier
, str_modif
);
476 str_modif
= str_printf(sym
, " (%s)", str_modif
);
479 str_modif
= str_printf(sym
, "%s[%s]", str_modif
, get_number(sym
));
482 /* Recurse to get the referred-to type */
483 if (!demangle_datatype(sym
, &sub_ct
, pmt_ref
, FALSE
))
486 ct
->left
= str_printf(sym
, "%s %s%s", sub_ct
.left
, modifier
, str_modif
);
489 /* don't insert a space between duplicate '*' */
490 if (!in_args
&& str_modif
[0] && str_modif
[1] == '*' && sub_ct
.left
[strlen(sub_ct
.left
)-1] == '*')
492 ct
->left
= str_printf(sym
, "%s%s", sub_ct
.left
, str_modif
);
494 ct
->right
= sub_ct
.right
;
495 sym
->stack
.num
= mark
;
500 /******************************************************************
502 * Gets the literal name from the current position in the mangled
503 * symbol to the first '@' character. It pushes the parsed name to
504 * the symbol names stack and returns a pointer to it or NULL in
507 static char* get_literal_string(struct parsed_symbol
* sym
)
509 const char *ptr
= sym
->current
;
512 if (!((*sym
->current
>= 'A' && *sym
->current
<= 'Z') ||
513 (*sym
->current
>= 'a' && *sym
->current
<= 'z') ||
514 (*sym
->current
>= '0' && *sym
->current
<= '9') ||
515 *sym
->current
== '_' || *sym
->current
== '$')) {
516 TRACE("Failed at '%c' in %s\n", *sym
->current
, ptr
);
519 } while (*++sym
->current
!= '@');
521 if (!str_array_push(sym
, ptr
, sym
->current
- 1 - ptr
, &sym
->names
))
524 return str_array_get_ref(&sym
->names
, sym
->names
.num
- sym
->names
.start
- 1);
527 /******************************************************************
529 * Parses a name with a template argument list and returns it as
531 * In a template argument list the back reference to the names
532 * table is separately created. '0' points to the class component
533 * name with the template arguments. We use the same stack array
534 * to hold the names but save/restore the stack state before/after
535 * parsing the template argument list.
537 static char* get_template_name(struct parsed_symbol
* sym
)
540 unsigned num_mark
= sym
->names
.num
;
541 unsigned start_mark
= sym
->names
.start
;
542 unsigned stack_mark
= sym
->stack
.num
;
543 struct array array_pmt
;
545 sym
->names
.start
= sym
->names
.num
;
546 if (!(name
= get_literal_string(sym
)))
548 str_array_init(&array_pmt
);
549 args
= get_args(sym
, &array_pmt
, FALSE
, '<', '>');
551 name
= str_printf(sym
, "%s%s", name
, args
);
552 sym
->names
.num
= num_mark
;
553 sym
->names
.start
= start_mark
;
554 sym
->stack
.num
= stack_mark
;
558 /******************************************************************
560 * Parses class as a list of parent-classes, terminated by '@' and stores the
561 * result in 'a' array. Each parent-classes, as well as the inner element
562 * (either field/method name or class name), are represented in the mangled
563 * name by a literal name ([a-zA-Z0-9_]+ terminated by '@') or a back reference
564 * ([0-9]) or a name with template arguments ('?$' literal name followed by the
565 * template argument list). The class name components appear in the reverse
566 * order in the mangled name, e.g aaa@bbb@ccc@@ will be demangled to
568 * For each of these class name components a string will be allocated in the
571 static BOOL
get_class(struct parsed_symbol
* sym
)
573 const char* name
= NULL
;
575 while (*sym
->current
!= '@')
577 switch (*sym
->current
)
579 case '\0': return FALSE
;
581 case '0': case '1': case '2': case '3':
582 case '4': case '5': case '6': case '7':
584 name
= str_array_get_ref(&sym
->names
, *sym
->current
++ - '0');
587 if (*++sym
->current
== '$')
590 if ((name
= get_template_name(sym
)) &&
591 !str_array_push(sym
, name
, -1, &sym
->names
))
596 name
= get_literal_string(sym
);
599 if (!name
|| !str_array_push(sym
, name
, -1, &sym
->stack
))
606 /******************************************************************
608 * From an array collected by get_class in sym->stack, constructs the
609 * corresponding (allocated) string
611 static char* get_class_string(struct parsed_symbol
* sym
, int start
)
614 unsigned int len
, sz
;
616 struct array
*a
= &sym
->stack
;
618 for (len
= 0, i
= start
; i
< a
->num
; i
++)
621 len
+= 2 + strlen(a
->elts
[i
]);
623 if (!(ret
= und_alloc(sym
, len
- 1))) return NULL
;
624 for (len
= 0, i
= a
->num
- 1; i
>= start
; i
--)
626 sz
= strlen(a
->elts
[i
]);
627 memcpy(ret
+ len
, a
->elts
[i
], sz
);
639 /******************************************************************
641 * Wrapper around get_class and get_class_string.
643 static char* get_class_name(struct parsed_symbol
* sym
)
645 unsigned mark
= sym
->stack
.num
;
649 s
= get_class_string(sym
, mark
);
650 sym
->stack
.num
= mark
;
654 /******************************************************************
655 * get_calling_convention
656 * Returns a static string corresponding to the calling convention described
657 * by char 'ch'. Sets export to TRUE iff the calling convention is exported.
659 static BOOL
get_calling_convention(char ch
, const char** call_conv
,
660 const char** exported
, unsigned flags
)
662 *call_conv
= *exported
= NULL
;
664 if (!(flags
& (UNDNAME_NO_MS_KEYWORDS
| UNDNAME_NO_ALLOCATION_LANGUAGE
)))
666 if (flags
& UNDNAME_NO_LEADING_UNDERSCORES
)
668 if (((ch
- 'A') % 2) == 1) *exported
= "dll_export ";
671 case 'A': case 'B': *call_conv
= "cdecl"; break;
672 case 'C': case 'D': *call_conv
= "pascal"; break;
673 case 'E': case 'F': *call_conv
= "thiscall"; break;
674 case 'G': case 'H': *call_conv
= "stdcall"; break;
675 case 'I': case 'J': *call_conv
= "fastcall"; break;
676 case 'K': case 'L': break;
677 case 'M': *call_conv
= "clrcall"; break;
678 default: ERR("Unknown calling convention %c\n", ch
); return FALSE
;
683 if (((ch
- 'A') % 2) == 1) *exported
= "__dll_export ";
686 case 'A': case 'B': *call_conv
= "__cdecl"; break;
687 case 'C': case 'D': *call_conv
= "__pascal"; break;
688 case 'E': case 'F': *call_conv
= "__thiscall"; break;
689 case 'G': case 'H': *call_conv
= "__stdcall"; break;
690 case 'I': case 'J': *call_conv
= "__fastcall"; break;
691 case 'K': case 'L': break;
692 case 'M': *call_conv
= "__clrcall"; break;
693 default: ERR("Unknown calling convention %c\n", ch
); return FALSE
;
700 /*******************************************************************
702 * Return a string containing an allocated string for a simple data type
704 static const char* get_simple_type(char c
)
706 const char* type_string
;
710 case 'C': type_string
= "signed char"; break;
711 case 'D': type_string
= "char"; break;
712 case 'E': type_string
= "unsigned char"; break;
713 case 'F': type_string
= "short"; break;
714 case 'G': type_string
= "unsigned short"; break;
715 case 'H': type_string
= "int"; break;
716 case 'I': type_string
= "unsigned int"; break;
717 case 'J': type_string
= "long"; break;
718 case 'K': type_string
= "unsigned long"; break;
719 case 'M': type_string
= "float"; break;
720 case 'N': type_string
= "double"; break;
721 case 'O': type_string
= "long double"; break;
722 case 'X': type_string
= "void"; break;
723 case 'Z': type_string
= "..."; break;
724 default: type_string
= NULL
; break;
729 /*******************************************************************
731 * Return a string containing an allocated string for a simple data type
733 static const char* get_extended_type(char c
)
735 const char* type_string
;
739 case 'D': type_string
= "__int8"; break;
740 case 'E': type_string
= "unsigned __int8"; break;
741 case 'F': type_string
= "__int16"; break;
742 case 'G': type_string
= "unsigned __int16"; break;
743 case 'H': type_string
= "__int32"; break;
744 case 'I': type_string
= "unsigned __int32"; break;
745 case 'J': type_string
= "__int64"; break;
746 case 'K': type_string
= "unsigned __int64"; break;
747 case 'L': type_string
= "__int128"; break;
748 case 'M': type_string
= "unsigned __int128"; break;
749 case 'N': type_string
= "bool"; break;
750 case 'W': type_string
= "wchar_t"; break;
751 default: type_string
= NULL
; break;
756 /*******************************************************************
759 * Attempt to demangle a C++ data type, which may be datatype.
760 * a datatype type is made up of a number of simple types. e.g:
761 * char** = (pointer to (pointer to (char)))
763 static BOOL
demangle_datatype(struct parsed_symbol
* sym
, struct datatype_t
* ct
,
764 struct array
* pmt_ref
, BOOL in_args
)
770 ct
->left
= ct
->right
= NULL
;
772 switch (dt
= *sym
->current
++)
775 /* MS type: __int8,__int16 etc */
776 ct
->left
= get_extended_type(*sym
->current
++);
778 case 'C': case 'D': case 'E': case 'F': case 'G':
779 case 'H': case 'I': case 'J': case 'K': case 'M':
780 case 'N': case 'O': case 'X': case 'Z':
781 /* Simple data types */
782 ct
->left
= get_simple_type(dt
);
785 case 'T': /* union */
786 case 'U': /* struct */
787 case 'V': /* class */
788 case 'Y': /* cointerface */
789 /* Class/struct/union/cointerface */
791 const char* struct_name
= NULL
;
792 const char* type_name
= NULL
;
794 if (!(struct_name
= get_class_name(sym
)))
796 if (!(sym
->flags
& UNDNAME_NO_COMPLEX_TYPE
))
800 case 'T': type_name
= "union "; break;
801 case 'U': type_name
= "struct "; break;
802 case 'V': type_name
= "class "; break;
803 case 'Y': type_name
= "cointerface "; break;
806 ct
->left
= str_printf(sym
, "%s%s", type_name
, struct_name
);
810 /* not all the time is seems */
814 if (!(ptr
= get_number(sym
))) goto done
;
815 ct
->left
= str_printf(sym
, "`template-parameter-%s'", ptr
);
819 if (!get_modified_type(ct
, sym
, pmt_ref
, '?', in_args
)) goto done
;
822 case 'A': /* reference */
823 case 'B': /* volatile reference */
824 if (!get_modified_type(ct
, sym
, pmt_ref
, dt
, in_args
)) goto done
;
826 case 'Q': /* const pointer */
827 case 'R': /* volatile pointer */
828 case 'S': /* const volatile pointer */
829 if (!get_modified_type(ct
, sym
, pmt_ref
, in_args
? dt
: 'P', in_args
)) goto done
;
831 case 'P': /* Pointer */
832 if (isdigit(*sym
->current
))
834 /* FIXME: P6 = Function pointer, others who knows.. */
835 if (*sym
->current
++ == '6')
838 const char* call_conv
;
839 const char* exported
;
840 struct datatype_t sub_ct
;
841 unsigned mark
= sym
->stack
.num
;
843 if (!get_calling_convention(*sym
->current
++,
844 &call_conv
, &exported
,
845 sym
->flags
& ~UNDNAME_NO_ALLOCATION_LANGUAGE
) ||
846 !demangle_datatype(sym
, &sub_ct
, pmt_ref
, FALSE
))
849 args
= get_args(sym
, pmt_ref
, TRUE
, '(', ')');
850 if (!args
) goto done
;
851 sym
->stack
.num
= mark
;
853 ct
->left
= str_printf(sym
, "%s%s (%s*",
854 sub_ct
.left
, sub_ct
.right
, call_conv
);
855 ct
->right
= str_printf(sym
, ")%s", args
);
859 else if (!get_modified_type(ct
, sym
, pmt_ref
, 'P', in_args
)) goto done
;
862 if (*sym
->current
== '4')
866 if (!(enum_name
= get_class_name(sym
)))
868 if (sym
->flags
& UNDNAME_NO_COMPLEX_TYPE
)
869 ct
->left
= enum_name
;
871 ct
->left
= str_printf(sym
, "enum %s", enum_name
);
875 case '0': case '1': case '2': case '3': case '4':
876 case '5': case '6': case '7': case '8': case '9':
877 /* Referring back to previously parsed type */
878 /* left and right are pushed as two separate strings */
879 ct
->left
= str_array_get_ref(pmt_ref
, (dt
- '0') * 2);
880 ct
->right
= str_array_get_ref(pmt_ref
, (dt
- '0') * 2 + 1);
881 if (!ct
->left
) goto done
;
885 switch (*sym
->current
++)
888 if (!(ct
->left
= get_number(sym
))) goto done
;
893 if (!(ptr
= get_number(sym
))) goto done
;
894 ct
->left
= str_printf(sym
, "`template-parameter%s'", ptr
);
901 if (!(p1
= get_number(sym
))) goto done
;
902 if (!(p2
= get_number(sym
))) goto done
;
903 ct
->left
= str_printf(sym
, "{%s,%s}", p1
, p2
);
911 if (!(p1
= get_number(sym
))) goto done
;
912 if (!(p2
= get_number(sym
))) goto done
;
913 if (!(p3
= get_number(sym
))) goto done
;
914 ct
->left
= str_printf(sym
, "{%s,%s,%s}", p1
, p2
, p3
);
920 if (!(ptr
= get_number(sym
))) goto done
;
921 ct
->left
= str_printf(sym
, "`non-type-template-parameter%s'", ptr
);
925 if (*sym
->current
== 'C')
930 if (!get_modifier(*sym
->current
++, &ptr
)) goto done
;
931 if (!demangle_datatype(sym
, ct
, pmt_ref
, in_args
)) goto done
;
932 ct
->left
= str_printf(sym
, "%s %s", ct
->left
, ptr
);
938 ERR("Unknown type %c\n", dt
);
941 if (add_pmt
&& pmt_ref
&& in_args
)
943 /* left and right are pushed as two separate strings */
944 if (!str_array_push(sym
, ct
->left
? ct
->left
: "", -1, pmt_ref
) ||
945 !str_array_push(sym
, ct
->right
? ct
->right
: "", -1, pmt_ref
))
950 return ct
->left
!= NULL
;
953 /******************************************************************
955 * Does the final parsing and handling for a variable or a field in
958 static BOOL
handle_data(struct parsed_symbol
* sym
)
960 const char* access
= NULL
;
961 const char* member_type
= NULL
;
962 const char* modifier
= NULL
;
963 struct datatype_t ct
;
970 * 3 private non-static
971 * 4 protected non-static
972 * 5 public non-static
977 if (!(sym
->flags
& UNDNAME_NO_ACCESS_SPECIFIERS
))
979 /* we only print the access for static members */
980 switch (*sym
->current
)
982 case '0': access
= "private: "; break;
983 case '1': access
= "protected: "; break;
984 case '2': access
= "public: "; break;
988 if (!(sym
->flags
& UNDNAME_NO_MEMBER_TYPE
))
990 if (*sym
->current
>= '0' && *sym
->current
<= '2')
991 member_type
= "static ";
994 name
= get_class_string(sym
, 0);
996 switch (*sym
->current
++)
998 case '0': case '1': case '2':
999 case '3': case '4': case '5':
1001 unsigned mark
= sym
->stack
.num
;
1004 str_array_init(&pmt
);
1006 if (!demangle_datatype(sym
, &ct
, &pmt
, FALSE
)) goto done
;
1007 if (!get_modifier(*sym
->current
++, &modifier
)) goto done
;
1008 sym
->stack
.num
= mark
;
1011 case '6' : /* compiler generated static */
1012 case '7' : /* compiler generated static */
1013 ct
.left
= ct
.right
= NULL
;
1014 if (!get_modifier(*sym
->current
++, &modifier
)) goto done
;
1015 if (*sym
->current
!= '@')
1019 if (!(cls
= get_class_name(sym
)))
1021 ct
.right
= str_printf(sym
, "{for `%s'}", cls
);
1026 modifier
= ct
.left
= ct
.right
= NULL
;
1030 if (sym
->flags
& UNDNAME_NAME_ONLY
) ct
.left
= ct
.right
= modifier
= NULL
;
1032 sym
->result
= str_printf(sym
, "%s%s%s%s%s%s%s%s", access
,
1033 member_type
, ct
.left
,
1034 modifier
&& ct
.left
? " " : NULL
, modifier
,
1035 modifier
|| ct
.left
? " " : NULL
, name
, ct
.right
);
1041 /******************************************************************
1043 * Does the final parsing and handling for a function or a method in
1046 static BOOL
handle_method(struct parsed_symbol
* sym
, BOOL cast_op
)
1049 const char* access
= NULL
;
1050 const char* member_type
= NULL
;
1051 struct datatype_t ct_ret
;
1052 const char* call_conv
;
1053 const char* modifier
= NULL
;
1054 const char* exported
;
1055 const char* args_str
= NULL
;
1056 const char* name
= NULL
;
1059 struct array array_pmt
;
1061 /* FIXME: why 2 possible letters for each option?
1064 * 'C' private: static
1065 * 'D' private: static
1066 * 'E' private: virtual
1067 * 'F' private: virtual
1068 * 'G' private: thunk
1069 * 'H' private: thunk
1072 * 'K' protected: static
1073 * 'L' protected: static
1074 * 'M' protected: virtual
1075 * 'N' protected: virtual
1076 * 'O' protected: thunk
1077 * 'P' protected: thunk
1080 * 'S' public: static
1081 * 'T' public: static
1082 * 'U' public: virtual
1083 * 'V' public: virtual
1089 accmem
= *sym
->current
++;
1090 if (accmem
< 'A' || accmem
> 'Z') goto done
;
1092 if (!(sym
->flags
& UNDNAME_NO_ACCESS_SPECIFIERS
))
1094 switch ((accmem
- 'A') / 8)
1096 case 0: access
= "private: "; break;
1097 case 1: access
= "protected: "; break;
1098 case 2: access
= "public: "; break;
1101 if (!(sym
->flags
& UNDNAME_NO_MEMBER_TYPE
))
1105 switch ((accmem
- 'A') % 8)
1107 case 2: case 3: member_type
= "static "; break;
1108 case 4: case 5: member_type
= "virtual "; break;
1110 access
= str_printf(sym
, "[thunk]:%s", access
);
1111 member_type
= "virtual ";
1117 name
= get_class_string(sym
, 0);
1119 if ((accmem
- 'A') % 8 == 6 || (accmem
- '8') % 8 == 7) /* a thunk */
1120 name
= str_printf(sym
, "%s`adjustor{%s}' ", name
, get_number(sym
));
1124 if (((accmem
- 'A') % 8) != 2 && ((accmem
- 'A') % 8) != 3)
1126 /* Implicit 'this' pointer */
1127 /* If there is an implicit this pointer, const modifier follows */
1128 if (!get_modifier(*sym
->current
, &modifier
)) goto done
;
1133 if (!get_calling_convention(*sym
->current
++, &call_conv
, &exported
,
1137 str_array_init(&array_pmt
);
1139 /* Return type, or @ if 'void' */
1140 if (*sym
->current
== '@')
1142 ct_ret
.left
= "void";
1143 ct_ret
.right
= NULL
;
1148 if (!demangle_datatype(sym
, &ct_ret
, &array_pmt
, FALSE
))
1151 if (sym
->flags
& UNDNAME_NO_FUNCTION_RETURNS
)
1152 ct_ret
.left
= ct_ret
.right
= NULL
;
1155 name
= str_printf(sym
, "%s%s%s", name
, ct_ret
.left
, ct_ret
.right
);
1156 ct_ret
.left
= ct_ret
.right
= NULL
;
1159 mark
= sym
->stack
.num
;
1160 if (!(args_str
= get_args(sym
, &array_pmt
, TRUE
, '(', ')'))) goto done
;
1161 if (sym
->flags
& UNDNAME_NAME_ONLY
) args_str
= modifier
= NULL
;
1162 sym
->stack
.num
= mark
;
1164 /* Note: '()' after 'Z' means 'throws', but we don't care here
1167 sym
->result
= str_printf(sym
, "%s%s%s%s%s%s%s%s%s%s%s%s",
1168 access
, member_type
, ct_ret
.left
,
1169 (ct_ret
.left
&& !ct_ret
.right
) ? " " : NULL
,
1170 call_conv
, call_conv
? " " : NULL
, exported
,
1171 name
, args_str
, modifier
,
1172 modifier
? " " : NULL
, ct_ret
.right
);
1178 /******************************************************************
1180 * Does the final parsing and handling for a name with templates
1182 static BOOL
handle_template(struct parsed_symbol
* sym
)
1187 assert(*sym
->current
++ == '$');
1188 if (!(name
= get_literal_string(sym
))) return FALSE
;
1189 if (!(args
= get_args(sym
, NULL
, FALSE
, '<', '>'))) return FALSE
;
1190 sym
->result
= str_printf(sym
, "%s%s", name
, args
);
1194 /*******************************************************************
1196 * Demangle a C++ linker symbol
1198 static BOOL
symbol_demangle(struct parsed_symbol
* sym
)
1201 unsigned do_after
= 0;
1202 static CHAR dashed_null
[] = "--null--";
1204 /* FIXME seems wrong as name, as it demangles a simple data type */
1205 if (sym
->flags
& UNDNAME_NO_ARGUMENTS
)
1207 struct datatype_t ct
;
1209 if (demangle_datatype(sym
, &ct
, NULL
, FALSE
))
1211 sym
->result
= str_printf(sym
, "%s%s", ct
.left
, ct
.right
);
1217 /* MS mangled names always begin with '?' */
1218 if (*sym
->current
!= '?') return FALSE
;
1219 str_array_init(&sym
->names
);
1220 str_array_init(&sym
->stack
);
1223 /* Then function name or operator code */
1224 if (*sym
->current
== '?' && sym
->current
[1] != '$')
1226 const char* function_name
= NULL
;
1228 /* C++ operator code (one character, or two if the first is '_') */
1229 switch (*++sym
->current
)
1231 case '0': do_after
= 1; break;
1232 case '1': do_after
= 2; break;
1233 case '2': function_name
= "operator new"; break;
1234 case '3': function_name
= "operator delete"; break;
1235 case '4': function_name
= "operator="; break;
1236 case '5': function_name
= "operator>>"; break;
1237 case '6': function_name
= "operator<<"; break;
1238 case '7': function_name
= "operator!"; break;
1239 case '8': function_name
= "operator=="; break;
1240 case '9': function_name
= "operator!="; break;
1241 case 'A': function_name
= "operator[]"; break;
1242 case 'B': function_name
= "operator "; do_after
= 3; break;
1243 case 'C': function_name
= "operator->"; break;
1244 case 'D': function_name
= "operator*"; break;
1245 case 'E': function_name
= "operator++"; break;
1246 case 'F': function_name
= "operator--"; break;
1247 case 'G': function_name
= "operator-"; break;
1248 case 'H': function_name
= "operator+"; break;
1249 case 'I': function_name
= "operator&"; break;
1250 case 'J': function_name
= "operator->*"; break;
1251 case 'K': function_name
= "operator/"; break;
1252 case 'L': function_name
= "operator%"; break;
1253 case 'M': function_name
= "operator<"; break;
1254 case 'N': function_name
= "operator<="; break;
1255 case 'O': function_name
= "operator>"; break;
1256 case 'P': function_name
= "operator>="; break;
1257 case 'Q': function_name
= "operator,"; break;
1258 case 'R': function_name
= "operator()"; break;
1259 case 'S': function_name
= "operator~"; break;
1260 case 'T': function_name
= "operator^"; break;
1261 case 'U': function_name
= "operator|"; break;
1262 case 'V': function_name
= "operator&&"; break;
1263 case 'W': function_name
= "operator||"; break;
1264 case 'X': function_name
= "operator*="; break;
1265 case 'Y': function_name
= "operator+="; break;
1266 case 'Z': function_name
= "operator-="; break;
1268 switch (*++sym
->current
)
1270 case '0': function_name
= "operator/="; break;
1271 case '1': function_name
= "operator%="; break;
1272 case '2': function_name
= "operator>>="; break;
1273 case '3': function_name
= "operator<<="; break;
1274 case '4': function_name
= "operator&="; break;
1275 case '5': function_name
= "operator|="; break;
1276 case '6': function_name
= "operator^="; break;
1277 case '7': function_name
= "`vftable'"; break;
1278 case '8': function_name
= "`vbtable'"; break;
1279 case '9': function_name
= "`vcall'"; break;
1280 case 'A': function_name
= "`typeof'"; break;
1281 case 'B': function_name
= "`local static guard'"; break;
1282 case 'C': function_name
= "`string'"; do_after
= 4; break;
1283 case 'D': function_name
= "`vbase destructor'"; break;
1284 case 'E': function_name
= "`vector deleting destructor'"; break;
1285 case 'F': function_name
= "`default constructor closure'"; break;
1286 case 'G': function_name
= "`scalar deleting destructor'"; break;
1287 case 'H': function_name
= "`vector constructor iterator'"; break;
1288 case 'I': function_name
= "`vector destructor iterator'"; break;
1289 case 'J': function_name
= "`vector vbase constructor iterator'"; break;
1290 case 'K': function_name
= "`virtual displacement map'"; break;
1291 case 'L': function_name
= "`eh vector constructor iterator'"; break;
1292 case 'M': function_name
= "`eh vector destructor iterator'"; break;
1293 case 'N': function_name
= "`eh vector vbase constructor iterator'"; break;
1294 case 'O': function_name
= "`copy constructor closure'"; break;
1296 sym
->flags
|= UNDNAME_NO_FUNCTION_RETURNS
;
1297 switch (*++sym
->current
)
1301 struct datatype_t ct
;
1305 str_array_init(&pmt
);
1306 demangle_datatype(sym
, &ct
, &pmt
, FALSE
);
1307 function_name
= str_printf(sym
, "%s%s `RTTI Type Descriptor'",
1314 const char* n1
, *n2
, *n3
, *n4
;
1316 n1
= get_number(sym
);
1317 n2
= get_number(sym
);
1318 n3
= get_number(sym
);
1319 n4
= get_number(sym
);
1321 function_name
= str_printf(sym
, "`RTTI Base Class Descriptor at (%s,%s,%s,%s)'",
1325 case '2': function_name
= "`RTTI Base Class Array'"; break;
1326 case '3': function_name
= "`RTTI Class Hierarchy Descriptor'"; break;
1327 case '4': function_name
= "`RTTI Complete Object Locator'"; break;
1329 ERR("Unknown RTTI operator: _R%c\n", *sym
->current
);
1333 case 'S': function_name
= "`local vftable'"; break;
1334 case 'T': function_name
= "`local vftable constructor closure'"; break;
1335 case 'U': function_name
= "operator new[]"; break;
1336 case 'V': function_name
= "operator delete[]"; break;
1337 case 'X': function_name
= "`placement delete closure'"; break;
1338 case 'Y': function_name
= "`placement delete[] closure'"; break;
1340 ERR("Unknown operator: _%c\n", *sym
->current
);
1345 /* FIXME: Other operators */
1346 ERR("Unknown operator: %c\n", *sym
->current
);
1353 if (!str_array_push(sym
, dashed_null
, -1, &sym
->stack
))
1357 sym
->result
= (char*)function_name
;
1361 if (!str_array_push(sym
, function_name
, -1, &sym
->stack
))
1366 else if (*sym
->current
== '$')
1368 /* Strange construct, it's a name with a template argument list
1371 ret
= (sym
->result
= get_template_name(sym
)) != NULL
;
1374 else if (*sym
->current
== '?' && sym
->current
[1] == '$')
1377 /* Either a class name, or '@' if the symbol is not a class member */
1378 switch (*sym
->current
)
1380 case '@': sym
->current
++; break;
1383 /* Class the function is associated with, terminated by '@@' */
1384 if (!get_class(sym
)) goto done
;
1390 case 0: default: break;
1392 /* it's time to set the member name for ctor & dtor */
1393 if (sym
->stack
.num
<= 1) goto done
;
1395 sym
->stack
.elts
[0] = sym
->stack
.elts
[1];
1397 sym
->stack
.elts
[0] = str_printf(sym
, "~%s", sym
->stack
.elts
[1]);
1398 /* ctors and dtors don't have return type */
1399 sym
->flags
|= UNDNAME_NO_FUNCTION_RETURNS
;
1402 sym
->flags
&= ~UNDNAME_NO_FUNCTION_RETURNS
;
1405 sym
->names
.start
= 1;
1409 /* Function/Data type and access level */
1410 if (*sym
->current
>= '0' && *sym
->current
<= '9')
1411 ret
= handle_data(sym
);
1412 else if (*sym
->current
>= 'A' && *sym
->current
<= 'Z')
1413 ret
= handle_method(sym
, do_after
== 3);
1414 else if (*sym
->current
== '$')
1415 ret
= handle_template(sym
);
1418 if (ret
) assert(sym
->result
);
1419 else WARN("Failed at %s\n", sym
->current
);
1424 /*********************************************************************
1425 * __unDNameEx (MSVCRT.@)
1427 * Demangle a C++ identifier.
1430 * buffer [O] If not NULL, the place to put the demangled string
1431 * mangled [I] Mangled name of the function
1432 * buflen [I] Length of buffer
1433 * memget [I] Function to allocate memory with
1434 * memfree [I] Function to free memory with
1435 * unknown [?] Unknown, possibly a call back
1436 * flags [I] Flags determining demangled format
1439 * Success: A string pointing to the unmangled name, allocated with memget.
1442 char* CDECL
__unDNameEx(char* buffer
, const char* mangled
, int buflen
,
1443 malloc_func_t memget
, free_func_t memfree
,
1444 void* unknown
, unsigned short int flags
)
1446 struct parsed_symbol sym
;
1449 TRACE("(%p,%s,%d,%p,%p,%p,%x)\n",
1450 buffer
, mangled
, buflen
, memget
, memfree
, unknown
, flags
);
1452 /* The flags details is not documented by MS. However, it looks exactly
1453 * like the UNDNAME_ manifest constants from imagehlp.h and dbghelp.h
1454 * So, we copied those (on top of the file)
1456 memset(&sym
, 0, sizeof(struct parsed_symbol
));
1457 if (flags
& UNDNAME_NAME_ONLY
)
1458 flags
|= UNDNAME_NO_FUNCTION_RETURNS
| UNDNAME_NO_ACCESS_SPECIFIERS
|
1459 UNDNAME_NO_MEMBER_TYPE
| UNDNAME_NO_ALLOCATION_LANGUAGE
|
1460 UNDNAME_NO_COMPLEX_TYPE
;
1463 sym
.mem_alloc_ptr
= memget
;
1464 sym
.mem_free_ptr
= memfree
;
1465 sym
.current
= mangled
;
1467 result
= symbol_demangle(&sym
) ? sym
.result
: mangled
;
1468 if (buffer
&& buflen
)
1470 lstrcpynA( buffer
, result
, buflen
);
1474 buffer
= memget(strlen(result
) + 1);
1475 if (buffer
) strcpy(buffer
, result
);
1484 /*********************************************************************
1485 * __unDName (MSVCRT.@)
1487 char* CDECL
__unDName(char* buffer
, const char* mangled
, int buflen
,
1488 malloc_func_t memget
, free_func_t memfree
,
1489 unsigned short int flags
)
1491 return __unDNameEx(buffer
, mangled
, buflen
, memget
, memfree
, NULL
, flags
);