Sweden-Number/dlls/dbghelp/stabs.c

1534 lines
51 KiB
C

/*
* File stabs.c - read stabs information from the modules
*
* Copyright (C) 1996, Eric Youngdale.
* 1999-2005, Eric Pouech
*
* 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 2.1 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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
*
*
* Maintenance Information
* -----------------------
*
* For documentation on the stabs format see for example
* The "stabs" debug format
* by Julia Menapace, Jim Kingdon, David Mackenzie
* of Cygnus Support
* available (hopefully) from http:\\sources.redhat.com\gdb\onlinedocs
*/
#include "config.h"
#include "wine/port.h"
#include <sys/types.h>
#include <fcntl.h>
#ifdef HAVE_SYS_STAT_H
# include <sys/stat.h>
#endif
#ifdef HAVE_SYS_MMAN_H
#include <sys/mman.h>
#endif
#include <limits.h>
#include <stdlib.h>
#include <string.h>
#ifdef HAVE_UNISTD_H
# include <unistd.h>
#endif
#include <stdio.h>
#ifndef PATH_MAX
#define PATH_MAX MAX_PATH
#endif
#include <assert.h>
#include <stdarg.h>
#include "windef.h"
#include "winbase.h"
#include "winnls.h"
#include "dbghelp_private.h"
#include "wine/debug.h"
WINE_DEFAULT_DEBUG_CHANNEL(dbghelp_stabs);
#ifndef N_UNDF
#define N_UNDF 0x00
#endif
#define N_GSYM 0x20
#define N_FUN 0x24
#define N_STSYM 0x26
#define N_LCSYM 0x28
#define N_MAIN 0x2a
#define N_ROSYM 0x2c
#define N_BNSYM 0x2e
#define N_OPT 0x3c
#define N_RSYM 0x40
#define N_SLINE 0x44
#define N_ENSYM 0x4e
#define N_SO 0x64
#define N_LSYM 0x80
#define N_BINCL 0x82
#define N_SOL 0x84
#define N_PSYM 0xa0
#define N_EINCL 0xa2
#define N_LBRAC 0xc0
#define N_EXCL 0xc2
#define N_RBRAC 0xe0
struct stab_nlist
{
union
{
char* n_name;
struct stab_nlist* n_next;
long n_strx;
} n_un;
unsigned char n_type;
char n_other;
short n_desc;
unsigned long n_value;
};
static void stab_strcpy(char* dest, int sz, const char* source)
{
char* ptr = dest;
/*
* A strcpy routine that stops when we hit the ':' character.
* Faster than copying the whole thing, and then nuking the
* ':'.
* Takes also care of (valid) a::b constructs
*/
while (*source != '\0')
{
if (source[0] != ':' && sz-- > 0) *ptr++ = *source++;
else if (source[1] == ':' && (sz -= 2) > 0)
{
*ptr++ = *source++;
*ptr++ = *source++;
}
else break;
}
*ptr-- = '\0';
/* GCC emits, in some cases, a .<digit>+ suffix.
* This is used for static variable inside functions, so
* that we can have several such variables with same name in
* the same compilation unit
* We simply ignore that suffix when present (we also get rid
* of it in ELF symtab parsing)
*/
if (ptr >= dest && isdigit(*ptr))
{
while (ptr > dest && isdigit(*ptr)) ptr--;
if (*ptr == '.') *ptr = '\0';
}
assert(sz > 0);
}
typedef struct
{
char* name;
unsigned long value;
struct symt** vector;
int nrofentries;
} include_def;
#define MAX_INCLUDES 5120
static include_def* include_defs = NULL;
static int num_include_def = 0;
static int num_alloc_include_def = 0;
static int cu_include_stack[MAX_INCLUDES];
static int cu_include_stk_idx = 0;
static struct symt** cu_vector = NULL;
static int cu_nrofentries = 0;
static struct symt_basic* stabs_basic[36];
static int stabs_new_include(const char* file, unsigned long val)
{
if (num_include_def == num_alloc_include_def)
{
num_alloc_include_def += 256;
if (!include_defs)
include_defs = HeapAlloc(GetProcessHeap(), 0,
sizeof(include_defs[0]) * num_alloc_include_def);
else
include_defs = HeapReAlloc(GetProcessHeap(), 0, include_defs,
sizeof(include_defs[0]) * num_alloc_include_def);
memset(include_defs + num_include_def, 0, sizeof(include_defs[0]) * 256);
}
include_defs[num_include_def].name = strcpy(HeapAlloc(GetProcessHeap(), 0, strlen(file) + 1), file);
include_defs[num_include_def].value = val;
include_defs[num_include_def].vector = NULL;
include_defs[num_include_def].nrofentries = 0;
return num_include_def++;
}
static int stabs_find_include(const char* file, unsigned long val)
{
int i;
for (i = 0; i < num_include_def; i++)
{
if (val == include_defs[i].value &&
strcmp(file, include_defs[i].name) == 0)
return i;
}
return -1;
}
static int stabs_add_include(int idx)
{
if (idx < 0) return -1;
cu_include_stk_idx++;
/* if this happens, just bump MAX_INCLUDES */
/* we could also handle this as another dynarray */
assert(cu_include_stk_idx < MAX_INCLUDES);
cu_include_stack[cu_include_stk_idx] = idx;
return cu_include_stk_idx;
}
static void stabs_reset_includes(void)
{
/*
* The struct symt:s that we would need to use are reset when
* we start a new file. (at least the ones in filenr == 0)
*/
cu_include_stk_idx = 0;/* keep 0 as index for the .c file itself */
memset(cu_vector, 0, sizeof(cu_vector[0]) * cu_nrofentries);
}
static void stabs_free_includes(void)
{
int i;
stabs_reset_includes();
for (i = 0; i < num_include_def; i++)
{
HeapFree(GetProcessHeap(), 0, include_defs[i].name);
HeapFree(GetProcessHeap(), 0, include_defs[i].vector);
}
HeapFree(GetProcessHeap(), 0, include_defs);
include_defs = NULL;
num_include_def = 0;
num_alloc_include_def = 0;
HeapFree(GetProcessHeap(), 0, cu_vector);
cu_vector = NULL;
cu_nrofentries = 0;
}
static struct symt** stabs_find_ref(long filenr, long subnr)
{
struct symt** ret;
/* FIXME: I could perhaps create a dummy include_def for each compilation
* unit which would allow not to handle those two cases separately
*/
if (filenr == 0)
{
if (cu_nrofentries <= subnr)
{
if (!cu_vector)
cu_vector = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
sizeof(cu_vector[0]) * (subnr+1));
else
cu_vector = HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
cu_vector, sizeof(cu_vector[0]) * (subnr+1));
cu_nrofentries = subnr + 1;
}
ret = &cu_vector[subnr];
}
else
{
include_def* idef;
assert(filenr <= cu_include_stk_idx);
idef = &include_defs[cu_include_stack[filenr]];
if (idef->nrofentries <= subnr)
{
if (!idef->vector)
idef->vector = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
sizeof(idef->vector[0]) * (subnr+1));
else
idef->vector = HeapReAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY,
idef->vector, sizeof(idef->vector[0]) * (subnr+1));
idef->nrofentries = subnr + 1;
}
ret = &idef->vector[subnr];
}
TRACE("(%ld,%ld) => %p (%p)\n", filenr, subnr, ret, *ret);
return ret;
}
static struct symt** stabs_read_type_enum(const char** x)
{
long filenr, subnr;
const char* iter;
char* end;
iter = *x;
if (*iter == '(')
{
++iter; /* '(' */
filenr = strtol(iter, &end, 10); /* <int> */
iter = ++end; /* ',' */
subnr = strtol(iter, &end, 10); /* <int> */
iter = ++end; /* ')' */
}
else
{
filenr = 0;
subnr = strtol(iter, &end, 10); /* <int> */
iter = end;
}
*x = iter;
return stabs_find_ref(filenr, subnr);
}
#define PTS_DEBUG
struct ParseTypedefData
{
const char* ptr;
char buf[1024];
int idx;
struct module* module;
#ifdef PTS_DEBUG
struct PTS_Error
{
const char* ptr;
unsigned line;
} errors[16];
int err_idx;
#endif
};
#ifdef PTS_DEBUG
static void stabs_pts_push(struct ParseTypedefData* ptd, unsigned line)
{
assert(ptd->err_idx < sizeof(ptd->errors) / sizeof(ptd->errors[0]));
ptd->errors[ptd->err_idx].line = line;
ptd->errors[ptd->err_idx].ptr = ptd->ptr;
ptd->err_idx++;
}
#define PTS_ABORTIF(ptd, t) do { if (t) { stabs_pts_push((ptd), __LINE__); return -1;} } while (0)
#else
#define PTS_ABORTIF(ptd, t) do { if (t) return -1; } while (0)
#endif
static int stabs_get_basic(struct ParseTypedefData* ptd, unsigned basic, struct symt** symt)
{
PTS_ABORTIF(ptd, basic >= sizeof(stabs_basic) / sizeof(stabs_basic[0]));
if (!stabs_basic[basic])
{
switch (basic)
{
case 1: stabs_basic[basic] = symt_new_basic(ptd->module, btInt, "int", 4); break;
case 2: stabs_basic[basic] = symt_new_basic(ptd->module, btChar, "char", 1); break;
case 3: stabs_basic[basic] = symt_new_basic(ptd->module, btInt, "short int", 2); break;
case 4: stabs_basic[basic] = symt_new_basic(ptd->module, btInt, "long int", 4); break;
case 5: stabs_basic[basic] = symt_new_basic(ptd->module, btUInt, "unsigned char", 1); break;
case 6: stabs_basic[basic] = symt_new_basic(ptd->module, btInt, "signed char", 1); break;
case 7: stabs_basic[basic] = symt_new_basic(ptd->module, btUInt, "unsigned short int", 2); break;
case 8: stabs_basic[basic] = symt_new_basic(ptd->module, btUInt, "unsigned int", 4); break;
case 9: stabs_basic[basic] = symt_new_basic(ptd->module, btUInt, "unsigned", 2); break;
case 10: stabs_basic[basic] = symt_new_basic(ptd->module, btUInt, "unsigned long int", 2); break;
case 11: stabs_basic[basic] = symt_new_basic(ptd->module, btVoid, "void", 0); break;
case 12: stabs_basic[basic] = symt_new_basic(ptd->module, btFloat, "float", 4); break;
case 13: stabs_basic[basic] = symt_new_basic(ptd->module, btFloat, "double", 8); break;
case 14: stabs_basic[basic] = symt_new_basic(ptd->module, btFloat, "long double", 12); break;
case 15: stabs_basic[basic] = symt_new_basic(ptd->module, btInt, "integer", 4); break;
case 16: stabs_basic[basic] = symt_new_basic(ptd->module, btBool, "bool", 1); break;
/* case 17: short real */
/* case 18: real */
case 25: stabs_basic[basic] = symt_new_basic(ptd->module, btComplex, "float complex", 8); break;
case 26: stabs_basic[basic] = symt_new_basic(ptd->module, btComplex, "double complex", 16); break;
case 30: stabs_basic[basic] = symt_new_basic(ptd->module, btWChar, "wchar_t", 2); break;
case 31: stabs_basic[basic] = symt_new_basic(ptd->module, btInt, "long long int", 8); break;
case 32: stabs_basic[basic] = symt_new_basic(ptd->module, btUInt, "long long unsigned", 8); break;
/* starting at 35 are wine extensions (especially for R implementation) */
case 35: stabs_basic[basic] = symt_new_basic(ptd->module, btComplex, "long double complex", 24); break;
default: PTS_ABORTIF(ptd, 1);
}
}
*symt = &stabs_basic[basic]->symt;
return 0;
}
static int stabs_pts_read_type_def(struct ParseTypedefData* ptd,
const char* typename, struct symt** dt);
static int stabs_pts_read_id(struct ParseTypedefData* ptd)
{
const char* first = ptd->ptr;
unsigned int template = 0;
char ch;
while ((ch = *ptd->ptr++) != '\0')
{
switch (ch)
{
case ':':
if (template == 0)
{
unsigned int len = ptd->ptr - first - 1;
PTS_ABORTIF(ptd, len >= sizeof(ptd->buf) - ptd->idx);
memcpy(ptd->buf + ptd->idx, first, len);
ptd->buf[ptd->idx + len] = '\0';
ptd->idx += len + 1;
return 0;
}
break;
case '<': template++; break;
case '>': PTS_ABORTIF(ptd, template == 0); template--; break;
}
}
return -1;
}
static int stabs_pts_read_number(struct ParseTypedefData* ptd, long* v)
{
char* last;
*v = strtol(ptd->ptr, &last, 10);
PTS_ABORTIF(ptd, last == ptd->ptr);
ptd->ptr = last;
return 0;
}
static int stabs_pts_read_type_reference(struct ParseTypedefData* ptd,
long* filenr, long* subnr)
{
if (*ptd->ptr == '(')
{
/* '(' <int> ',' <int> ')' */
ptd->ptr++;
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, filenr) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, subnr) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ')');
}
else
{
*filenr = 0;
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, subnr) == -1);
}
return 0;
}
struct pts_range_value
{
ULONGLONG val;
int sign;
};
static int stabs_pts_read_range_value(struct ParseTypedefData* ptd, struct pts_range_value* prv)
{
char* last;
switch (*ptd->ptr)
{
case '0':
while (*ptd->ptr == '0') ptd->ptr++;
if (*ptd->ptr >= '1' && *ptd->ptr <= '7')
{
switch (ptd->ptr[1])
{
case '0':
PTS_ABORTIF(ptd, ptd->ptr[0] != '1');
prv->sign = -1;
prv->val = 0;
while (isdigit(*ptd->ptr)) prv->val = (prv->val << 3) + *ptd->ptr++ - '0';
break;
case '7':
prv->sign = 1;
prv->val = 0;
while (isdigit(*ptd->ptr)) prv->val = (prv->val << 3) + *ptd->ptr++ - '0';
break;
default: PTS_ABORTIF(ptd, 1); break;
}
} else prv->sign = 0;
break;
case '-':
prv->sign = -1;
prv->val = strtoull(++ptd->ptr, &last, 10);
ptd->ptr = last;
break;
case '+':
default:
prv->sign = 1;
prv->val = strtoull(ptd->ptr, &last, 10);
ptd->ptr = last;
break;
}
return 0;
}
static int stabs_pts_read_range(struct ParseTypedefData* ptd, const char* typename,
struct symt** dt)
{
struct symt* ref;
struct pts_range_value lo;
struct pts_range_value hi;
unsigned size;
enum BasicType bt;
int i;
ULONGLONG v;
/* type ';' <int> ';' <int> ';' */
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &ref) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_range_value(ptd, &lo) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_range_value(ptd, &hi) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
/* basically, we don't use ref... in some cases, for example, float is declared
* as a derived type of int... which won't help us... so we guess the types
* from the various formats
*/
if (lo.sign == 0 && hi.sign < 0)
{
bt = btUInt;
size = hi.val;
}
else if (lo.sign < 0 && hi.sign == 0)
{
bt = btUInt;
size = lo.val;
}
else if (lo.sign > 0 && hi.sign == 0)
{
bt = btFloat;
size = lo.val;
}
else if (lo.sign < 0 && hi.sign > 0)
{
v = 1 << 7;
for (i = 7; i < 64; i += 8)
{
if (lo.val == v && hi.val == v - 1)
{
bt = btInt;
size = (i + 1) / 8;
break;
}
v <<= 8;
}
PTS_ABORTIF(ptd, i >= 64);
}
else if (lo.sign == 0 && hi.sign > 0)
{
if (hi.val == 127) /* specific case for char... */
{
bt = btChar;
size = 1;
}
else
{
v = 1;
for (i = 8; i <= 64; i += 8)
{
v <<= 8;
if (hi.val + 1 == v)
{
bt = btUInt;
size = (i + 1) / 8;
break;
}
}
PTS_ABORTIF(ptd, i > 64);
}
}
else PTS_ABORTIF(ptd, 1);
*dt = &symt_new_basic(ptd->module, bt, typename, size)->symt;
return 0;
}
static inline int stabs_pts_read_method_info(struct ParseTypedefData* ptd)
{
struct symt* dt;
char* tmp;
char mthd;
do
{
/* get type of return value */
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &dt) == -1);
if (*ptd->ptr == ';') ptd->ptr++;
/* get types of parameters */
if (*ptd->ptr == ':')
{
PTS_ABORTIF(ptd, !(tmp = strchr(ptd->ptr + 1, ';')));
ptd->ptr = tmp + 1;
}
PTS_ABORTIF(ptd, !(*ptd->ptr >= '0' && *ptd->ptr <= '9'));
ptd->ptr++;
PTS_ABORTIF(ptd, !(ptd->ptr[0] >= 'A' && *ptd->ptr <= 'D'));
mthd = *++ptd->ptr;
PTS_ABORTIF(ptd, mthd != '.' && mthd != '?' && mthd != '*');
ptd->ptr++;
if (mthd == '*')
{
long int ofs;
struct symt* dt;
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &ofs) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
}
} while (*ptd->ptr != ';');
ptd->ptr++;
return 0;
}
static inline int stabs_pts_read_aggregate(struct ParseTypedefData* ptd,
struct symt_udt* sdt)
{
long sz, ofs;
struct symt* adt;
struct symt* dt = NULL;
int idx;
int doadd;
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &sz) == -1);
doadd = symt_set_udt_size(ptd->module, sdt, sz);
if (*ptd->ptr == '!') /* C++ inheritence */
{
long num_classes;
ptd->ptr++;
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &num_classes) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
while (--num_classes >= 0)
{
ptd->ptr += 2; /* skip visibility and inheritence */
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &ofs) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &adt) == -1);
if (doadd)
{
char tmp[256];
WCHAR* name;
DWORD64 size;
symt_get_info(adt, TI_GET_SYMNAME, &name);
strcpy(tmp, "__inherited_class_");
WideCharToMultiByte(CP_ACP, 0, name, -1,
tmp + strlen(tmp), sizeof(tmp) - strlen(tmp),
NULL, NULL);
HeapFree(GetProcessHeap(), 0, name);
/* FIXME: TI_GET_LENGTH will not always work, especially when adt
* has just been seen as a forward definition and not the real stuff
* yet.
* As we don't use much the size of members in structs, this may not
* be much of a problem
*/
symt_get_info(adt, TI_GET_LENGTH, &size);
symt_add_udt_element(ptd->module, sdt, tmp, adt, ofs, (DWORD)size * 8);
}
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
}
}
/* if the structure has already been filled, just redo the parsing
* but don't store results into the struct
* FIXME: there's a quite ugly memory leak in there...
*/
/* Now parse the individual elements of the structure/union. */
while (*ptd->ptr != ';')
{
/* agg_name : type ',' <int:offset> ',' <int:size> */
idx = ptd->idx;
if (ptd->ptr[0] == '$' && ptd->ptr[1] == 'v')
{
long x;
if (ptd->ptr[2] == 'f')
{
/* C++ virtual method table */
ptd->ptr += 3;
stabs_read_type_enum(&ptd->ptr);
PTS_ABORTIF(ptd, *ptd->ptr++ != ':');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &x) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
ptd->idx = idx;
continue;
}
else if (ptd->ptr[2] == 'b')
{
ptd->ptr += 3;
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ':');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &x) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
ptd->idx = idx;
continue;
}
}
PTS_ABORTIF(ptd, stabs_pts_read_id(ptd) == -1);
/* Ref. TSDF R2.130 Section 7.4. When the field name is a method name
* it is followed by two colons rather than one.
*/
if (*ptd->ptr == ':')
{
ptd->ptr++;
stabs_pts_read_method_info(ptd);
ptd->idx = idx;
continue;
}
else
{
/* skip C++ member protection /0 /1 or /2 */
if (*ptd->ptr == '/') ptd->ptr += 2;
}
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &adt) == -1);
switch (*ptd->ptr++)
{
case ',':
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &ofs) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &sz) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
if (doadd) symt_add_udt_element(ptd->module, sdt, ptd->buf + idx, adt, ofs, sz);
break;
case ':':
{
char* tmp;
/* method parameters... terminated by ';' */
PTS_ABORTIF(ptd, !(tmp = strchr(ptd->ptr, ';')));
ptd->ptr = tmp + 1;
}
break;
default:
PTS_ABORTIF(ptd, TRUE);
}
ptd->idx = idx;
}
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
if (*ptd->ptr == '~')
{
ptd->ptr++;
PTS_ABORTIF(ptd, *ptd->ptr++ != '%');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
}
return 0;
}
static inline int stabs_pts_read_enum(struct ParseTypedefData* ptd,
struct symt_enum* edt)
{
long value;
int idx;
while (*ptd->ptr != ';')
{
idx = ptd->idx;
PTS_ABORTIF(ptd, stabs_pts_read_id(ptd) == -1);
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &value) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
symt_add_enum_element(ptd->module, edt, ptd->buf + idx, value);
ptd->idx = idx;
}
ptd->ptr++;
return 0;
}
static inline int stabs_pts_read_array(struct ParseTypedefData* ptd,
struct symt** adt)
{
long lo, hi;
struct symt* range_dt;
struct symt* base_dt;
/* ar<typeinfo_nodef>;<int>;<int>;<typeinfo> */
PTS_ABORTIF(ptd, *ptd->ptr++ != 'r');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &range_dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &lo) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &hi) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &base_dt) == -1);
*adt = &symt_new_array(ptd->module, lo, hi, base_dt, range_dt)->symt;
return 0;
}
static int stabs_pts_read_type_def(struct ParseTypedefData* ptd, const char* typename,
struct symt** ret_dt)
{
int idx;
long sz = -1;
struct symt* new_dt = NULL; /* newly created data type */
struct symt* ref_dt; /* referenced data type (pointer...) */
long filenr1, subnr1, tmp;
/* things are a bit complicated because of the way the typedefs are stored inside
* the file, because addresses can change when realloc is done, so we must call
* over and over stabs_find_ref() to keep the correct values around
*/
PTS_ABORTIF(ptd, stabs_pts_read_type_reference(ptd, &filenr1, &subnr1) == -1);
while (*ptd->ptr == '=')
{
ptd->ptr++;
PTS_ABORTIF(ptd, new_dt != NULL);
/* first handle attribute if any */
switch (*ptd->ptr)
{
case '@':
if (*++ptd->ptr == 's')
{
ptd->ptr++;
if (stabs_pts_read_number(ptd, &sz) == -1)
{
ERR("Not an attribute... NIY\n");
ptd->ptr -= 2;
return -1;
}
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
}
break;
}
/* then the real definitions */
switch (*ptd->ptr++)
{
case '*':
case '&':
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &ref_dt) == -1);
new_dt = &symt_new_pointer(ptd->module, ref_dt)->symt;
break;
case 'k': /* 'const' modifier */
case 'B': /* 'volatile' modifier */
/* just kinda ignore the modifier, I guess -gmt */
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, typename, &new_dt) == -1);
break;
case '(':
ptd->ptr--;
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, typename, &new_dt) == -1);
break;
case 'a':
PTS_ABORTIF(ptd, stabs_pts_read_array(ptd, &new_dt) == -1);
break;
case 'r':
PTS_ABORTIF(ptd, stabs_pts_read_range(ptd, typename, &new_dt) == -1);
assert(!*stabs_find_ref(filenr1, subnr1));
*stabs_find_ref(filenr1, subnr1) = new_dt;
break;
case 'f':
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &ref_dt) == -1);
new_dt = &symt_new_function_signature(ptd->module, ref_dt, -1)->symt;
break;
case 'e':
stabs_get_basic(ptd, 1 /* int */, &ref_dt);
new_dt = &symt_new_enum(ptd->module, typename, ref_dt)->symt;
PTS_ABORTIF(ptd, stabs_pts_read_enum(ptd, (struct symt_enum*)new_dt) == -1);
break;
case 's':
case 'u':
{
struct symt_udt* udt;
enum UdtKind kind = (ptd->ptr[-1] == 's') ? UdtStruct : UdtUnion;
/* udt can have been already defined in a forward definition */
udt = (struct symt_udt*)*stabs_find_ref(filenr1, subnr1);
if (!udt)
{
udt = symt_new_udt(ptd->module, typename, 0, kind);
/* we need to set it here, because a struct can hold a pointer
* to itself
*/
new_dt = *stabs_find_ref(filenr1, subnr1) = &udt->symt;
}
else
{
unsigned l1, l2;
if (udt->symt.tag != SymTagUDT)
{
ERR("Forward declaration (%p/%s) is not an aggregate (%u)\n",
udt, symt_get_name(&udt->symt), udt->symt.tag);
return -1;
}
/* FIXME: we currently don't correctly construct nested C++
* classes names. Therefore, we could be here with either:
* - typename and udt->hash_elt.name being the same string
* (non embedded case)
* - typename being foo::bar while udt->hash_elt.name being
* just bar
* So, we twist the comparison to test both occurrences. When
* we have proper C++ types in this file, this twist has to be
* removed
*/
l1 = strlen(udt->hash_elt.name);
l2 = strlen(typename);
if (l1 > l2 || strcmp(udt->hash_elt.name, typename + l2 - l1))
ERR("Forward declaration name mismatch %s <> %s\n",
udt->hash_elt.name, typename);
new_dt = &udt->symt;
}
PTS_ABORTIF(ptd, stabs_pts_read_aggregate(ptd, udt) == -1);
}
break;
case 'x':
idx = ptd->idx;
tmp = *ptd->ptr++;
PTS_ABORTIF(ptd, stabs_pts_read_id(ptd) == -1);
switch (tmp)
{
case 'e':
stabs_get_basic(ptd, 1 /* int */, &ref_dt);
new_dt = &symt_new_enum(ptd->module, ptd->buf + idx, ref_dt)->symt;
break;
case 's':
new_dt = &symt_new_udt(ptd->module, ptd->buf + idx, 0, UdtStruct)->symt;
break;
case 'u':
new_dt = &symt_new_udt(ptd->module, ptd->buf + idx, 0, UdtUnion)->symt;
break;
default:
return -1;
}
ptd->idx = idx;
break;
case '-':
{
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &tmp) == -1);
PTS_ABORTIF(ptd, stabs_get_basic(ptd, tmp, &new_dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';');
}
break;
case '#':
if (*ptd->ptr == '#')
{
ptd->ptr++;
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &ref_dt) == -1);
new_dt = &symt_new_function_signature(ptd->module, ref_dt, -1)->symt;
}
else
{
struct symt* cls_dt;
struct symt* pmt_dt;
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &cls_dt) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ',');
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &ref_dt) == -1);
new_dt = &symt_new_function_signature(ptd->module, ref_dt, -1)->symt;
while (*ptd->ptr == ',')
{
ptd->ptr++;
PTS_ABORTIF(ptd, stabs_pts_read_type_def(ptd, NULL, &pmt_dt) == -1);
}
}
break;
case 'R':
{
long type, len, unk;
int basic;
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &type) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &len) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
PTS_ABORTIF(ptd, stabs_pts_read_number(ptd, &unk) == -1);
PTS_ABORTIF(ptd, *ptd->ptr++ != ';'); /* ';' */
switch (type) /* see stabs_get_basic for the details */
{
case 1: basic = 12; break;
case 2: basic = 13; break;
case 3: basic = 25; break;
case 4: basic = 26; break;
case 5: basic = 35; break;
case 6: basic = 14; break;
default: PTS_ABORTIF(ptd, 1);
}
PTS_ABORTIF(ptd, stabs_get_basic(ptd, basic, &new_dt) == -1);
}
break;
default:
ERR("Unknown type '%c'\n", ptd->ptr[-1]);
return -1;
}
}
if (!new_dt)
{
/* is it a forward declaration that has been filled ? */
new_dt = *stabs_find_ref(filenr1, subnr1);
/* if not, this should be void (which is defined as a ref to itself, but we
* don't correctly catch it)
*/
if (!new_dt && typename)
{
new_dt = &symt_new_basic(ptd->module, btVoid, typename, 0)->symt;
PTS_ABORTIF(ptd, strcmp(typename, "void"));
}
}
*stabs_find_ref(filenr1, subnr1) = *ret_dt = new_dt;
TRACE("Adding (%ld,%ld) %s\n", filenr1, subnr1, debugstr_a(typename));
return 0;
}
static int stabs_parse_typedef(struct module* module, const char* ptr,
const char* typename)
{
struct ParseTypedefData ptd;
struct symt* dt;
int ret = -1;
/* check for already existing definition */
TRACE("%s => %s\n", typename, debugstr_a(ptr));
ptd.module = module;
ptd.idx = 0;
#ifdef PTS_DEBUG
ptd.err_idx = 0;
#endif
for (ptd.ptr = ptr - 1; ;)
{
ptd.ptr = strchr(ptd.ptr + 1, ':');
if (ptd.ptr == NULL || *++ptd.ptr != ':') break;
}
if (ptd.ptr)
{
if (*ptd.ptr != '(') ptd.ptr++;
/* most of type definitions take one char, except Tt */
if (*ptd.ptr != '(') ptd.ptr++;
ret = stabs_pts_read_type_def(&ptd, typename, &dt);
}
if (ret == -1 || *ptd.ptr)
{
#ifdef PTS_DEBUG
int i;
TRACE("Failure on %s\n", debugstr_a(ptr));
if (ret == -1)
{
for (i = 0; i < ptd.err_idx; i++)
{
TRACE("[%d]: line %d => %s\n",
i, ptd.errors[i].line, debugstr_a(ptd.errors[i].ptr));
}
}
else
TRACE("[0]: => %s\n", debugstr_a(ptd.ptr));
#else
ERR("Failure on %s at %s\n", debugstr_a(ptr), debugstr_a(ptd.ptr));
#endif
return FALSE;
}
return TRUE;
}
static struct symt* stabs_parse_type(const char* stab)
{
const char* c = stab - 1;
/*
* Look through the stab definition, and figure out what struct symt
* this represents. If we have something we know about, assign the
* type.
* According to "The \"stabs\" debug format" (Rev 2.130) the name may be
* a C++ name and contain double colons e.g. foo::bar::baz:t5=*6.
*/
do
{
if ((c = strchr(c + 1, ':')) == NULL) return NULL;
} while (*++c == ':');
/*
* The next characters say more about the type (i.e. data, function, etc)
* of symbol. Skip them. (C++ for example may have Tt).
* Actually this is a very weak description; I think Tt is the only
* multiple combination we should see.
*/
while (*c && *c != '(' && !isdigit(*c))
c++;
/*
* The next is either an integer or a (integer,integer).
* The stabs_read_type_enum() takes care that stab_types is large enough.
*/
return *stabs_read_type_enum(&c);
}
struct pending_loc_var
{
char name[256];
struct symt* type;
enum DataKind kind;
struct location loc;
};
struct pending_block
{
struct pending_loc_var* vars;
unsigned num;
unsigned allocated;
};
static inline void pending_add(struct pending_block* pending, const char* name,
enum DataKind dt, const struct location* loc)
{
if (pending->num == pending->allocated)
{
pending->allocated += 8;
if (!pending->vars)
pending->vars = HeapAlloc(GetProcessHeap(), 0,
pending->allocated * sizeof(pending->vars[0]));
else
pending->vars = HeapReAlloc(GetProcessHeap(), 0, pending->vars,
pending->allocated * sizeof(pending->vars[0]));
}
stab_strcpy(pending->vars[pending->num].name,
sizeof(pending->vars[pending->num].name), name);
pending->vars[pending->num].type = stabs_parse_type(name);
pending->vars[pending->num].kind = dt;
pending->vars[pending->num].loc = *loc;
pending->num++;
}
static void pending_flush(struct pending_block* pending, struct module* module,
struct symt_function* func, struct symt_block* block)
{
int i;
for (i = 0; i < pending->num; i++)
{
symt_add_func_local(module, func,
pending->vars[i].kind, &pending->vars[i].loc,
block, pending->vars[i].type, pending->vars[i].name);
}
pending->num = 0;
}
/******************************************************************
* stabs_finalize_function
*
* Ends function creation: mainly:
* - cleans up line number information
* - tries to set up a debug-start tag (FIXME: heuristic to be enhanced)
* - for stabs which have absolute address in them, initializes the size of the
* function (assuming that current function ends where next function starts)
*/
static void stabs_finalize_function(struct module* module, struct symt_function* func,
unsigned long size)
{
IMAGEHLP_LINE il;
struct location loc;
if (!func) return;
symt_normalize_function(module, func);
/* To define the debug-start of the function, we use the second line number.
* Not 100% bullet proof, but better than nothing
*/
if (symt_fill_func_line_info(module, func, func->address, &il) &&
symt_get_func_line_next(module, &il))
{
loc.kind = loc_absolute;
loc.offset = il.Address - func->address;
symt_add_function_point(module, func, SymTagFuncDebugStart,
&loc, NULL);
}
if (size) func->size = size;
}
BOOL stabs_parse(struct module* module, unsigned long load_offset,
const void* pv_stab_ptr, int stablen,
const char* strs, int strtablen)
{
struct symt_function* curr_func = NULL;
struct symt_block* block = NULL;
struct symt_compiland* compiland = NULL;
char srcpath[PATH_MAX]; /* path to directory source file is in */
int i;
int nstab;
const char* ptr;
char* stabbuff;
unsigned int stabbufflen;
const struct stab_nlist* stab_ptr = pv_stab_ptr;
const char* strs_end;
int strtabinc;
char symname[4096];
unsigned incl[32];
int incl_stk = -1;
int source_idx = -1;
struct pending_block pending;
BOOL ret = TRUE;
struct location loc;
nstab = stablen / sizeof(struct stab_nlist);
strs_end = strs + strtablen;
memset(srcpath, 0, sizeof(srcpath));
memset(stabs_basic, 0, sizeof(stabs_basic));
memset(&pending, 0, sizeof(pending));
/*
* Allocate a buffer into which we can build stab strings for cases
* where the stab is continued over multiple lines.
*/
stabbufflen = 65536;
stabbuff = HeapAlloc(GetProcessHeap(), 0, stabbufflen);
strtabinc = 0;
stabbuff[0] = '\0';
for (i = 0; i < nstab; i++, stab_ptr++)
{
ptr = strs + stab_ptr->n_un.n_strx;
if ((ptr > strs_end) || (ptr + strlen(ptr) > strs_end))
{
WARN("Bad stabs string %p\n", ptr);
continue;
}
if (*ptr != '\0' && (ptr[strlen(ptr) - 1] == '\\'))
{
/*
* Indicates continuation. Append this to the buffer, and go onto the
* next record. Repeat the process until we find a stab without the
* '/' character, as this indicates we have the whole thing.
*/
unsigned len = strlen(ptr);
if (strlen(stabbuff) + len > stabbufflen)
{
stabbufflen += 65536;
stabbuff = HeapReAlloc(GetProcessHeap(), 0, stabbuff, stabbufflen);
}
strncat(stabbuff, ptr, len - 1);
continue;
}
else if (stabbuff[0] != '\0')
{
strcat(stabbuff, ptr);
ptr = stabbuff;
}
/* only symbol entries contain a typedef */
switch (stab_ptr->n_type)
{
case N_GSYM:
case N_LCSYM:
case N_STSYM:
case N_RSYM:
case N_LSYM:
case N_ROSYM:
case N_PSYM:
if (strchr(ptr, '=') != NULL)
{
/*
* The stabs aren't in writable memory, so copy it over so we are
* sure we can scribble on it.
*/
if (ptr != stabbuff)
{
strcpy(stabbuff, ptr);
ptr = stabbuff;
}
stab_strcpy(symname, sizeof(symname), ptr);
if (!stabs_parse_typedef(module, ptr, symname))
{
/* skip this definition */
stabbuff[0] = '\0';
continue;
}
}
}
switch (stab_ptr->n_type)
{
case N_GSYM:
/*
* These are useless with ELF. They have no value, and you have to
* read the normal symbol table to get the address. Thus we
* ignore them, and when we process the normal symbol table
* we should do the right thing.
*
* With a.out or mingw, they actually do make some amount of sense.
*/
stab_strcpy(symname, sizeof(symname), ptr);
symt_new_global_variable(module, compiland, symname, TRUE /* FIXME */,
load_offset + stab_ptr->n_value, 0,
stabs_parse_type(ptr));
break;
case N_LCSYM:
case N_STSYM:
/* These are static symbols and BSS symbols. */
stab_strcpy(symname, sizeof(symname), ptr);
symt_new_global_variable(module, compiland, symname, TRUE /* FIXME */,
load_offset + stab_ptr->n_value, 0,
stabs_parse_type(ptr));
break;
case N_LBRAC:
if (curr_func)
{
block = symt_open_func_block(module, curr_func, block,
stab_ptr->n_value, 0);
pending_flush(&pending, module, curr_func, block);
}
break;
case N_RBRAC:
if (curr_func)
block = symt_close_func_block(module, curr_func, block,
stab_ptr->n_value);
break;
case N_PSYM:
/* These are function parameters. */
if (curr_func != NULL)
{
struct symt* param_type = stabs_parse_type(ptr);
stab_strcpy(symname, sizeof(symname), ptr);
loc.kind = loc_regrel;
loc.reg = 0; /* FIXME */
loc.offset = stab_ptr->n_value;
symt_add_func_local(module, curr_func,
(long)stab_ptr->n_value >= 0 ? DataIsParam : DataIsLocal,
&loc, NULL, param_type, symname);
symt_add_function_signature_parameter(module,
(struct symt_function_signature*)curr_func->type,
param_type);
}
break;
case N_RSYM:
/* These are registers (as local variables) */
if (curr_func != NULL)
{
loc.kind = loc_register;
loc.offset = 0;
switch (stab_ptr->n_value)
{
case 0: loc.reg = CV_REG_EAX; break;
case 1: loc.reg = CV_REG_ECX; break;
case 2: loc.reg = CV_REG_EDX; break;
case 3: loc.reg = CV_REG_EBX; break;
case 4: loc.reg = CV_REG_ESP; break;
case 5: loc.reg = CV_REG_EBP; break;
case 6: loc.reg = CV_REG_ESI; break;
case 7: loc.reg = CV_REG_EDI; break;
case 11:
case 12:
case 13:
case 14:
case 15:
case 16:
case 17:
case 18:
case 19: loc.reg = CV_REG_ST0 + stab_ptr->n_value - 12; break;
default:
FIXME("Unknown register value (%lu)\n", stab_ptr->n_value);
loc.reg = CV_REG_NONE;
break;
}
stab_strcpy(symname, sizeof(symname), ptr);
if (ptr[strlen(symname) + 1] == 'P')
{
struct symt* param_type = stabs_parse_type(ptr);
stab_strcpy(symname, sizeof(symname), ptr);
symt_add_func_local(module, curr_func, DataIsParam, &loc,
NULL, param_type, symname);
symt_add_function_signature_parameter(module,
(struct symt_function_signature*)curr_func->type,
param_type);
}
else
pending_add(&pending, ptr, DataIsLocal, &loc);
}
break;
case N_LSYM:
/* These are local variables */
loc.kind = loc_regrel;
loc.reg = 0; /* FIXME */
loc.offset = stab_ptr->n_value;
if (curr_func != NULL) pending_add(&pending, ptr, DataIsLocal, &loc);
break;
case N_SLINE:
/*
* This is a line number. These are always relative to the start
* of the function (N_FUN), and this makes the lookup easier.
*/
if (curr_func != NULL)
{
assert(source_idx >= 0);
symt_add_func_line(module, curr_func, source_idx,
stab_ptr->n_desc, stab_ptr->n_value);
}
break;
case N_FUN:
/*
* For now, just declare the various functions. Later
* on, we will add the line number information and the
* local symbols.
*/
/*
* Copy the string to a temp buffer so we
* can kill everything after the ':'. We do
* it this way because otherwise we end up dirtying
* all of the pages related to the stabs, and that
* sucks up swap space like crazy.
*/
stab_strcpy(symname, sizeof(symname), ptr);
if (*symname)
{
struct symt_function_signature* func_type;
if (curr_func)
{
/* First, clean up the previous function we were working on.
* Assume size of the func is the delta between current offset
* and offset of last function
*/
stabs_finalize_function(module, curr_func,
stab_ptr->n_value ?
(load_offset + stab_ptr->n_value - curr_func->address) : 0);
}
func_type = symt_new_function_signature(module,
stabs_parse_type(ptr), -1);
curr_func = symt_new_function(module, compiland, symname,
load_offset + stab_ptr->n_value, 0,
&func_type->symt);
}
else
{
/* some versions of GCC to use a N_FUN "" to mark the end of a function
* and n_value contains the size of the func
*/
stabs_finalize_function(module, curr_func, stab_ptr->n_value);
curr_func = NULL;
}
break;
case N_SO:
/*
* This indicates a new source file. Append the records
* together, to build the correct path name.
*/
if (*ptr == '\0') /* end of N_SO file */
{
/* Nuke old path. */
srcpath[0] = '\0';
stabs_finalize_function(module, curr_func, 0);
curr_func = NULL;
source_idx = -1;
incl_stk = -1;
assert(block == NULL);
compiland = NULL;
}
else
{
int len = strlen(ptr);
if (ptr[len-1] != '/')
{
stabs_reset_includes();
source_idx = source_new(module, srcpath, ptr);
compiland = symt_new_compiland(module, 0 /* FIXME */, source_idx);
}
else
strcpy(srcpath, ptr);
}
break;
case N_SOL:
source_idx = source_new(module, srcpath, ptr);
break;
case N_UNDF:
strs += strtabinc;
strtabinc = stab_ptr->n_value;
/* I'm not sure this is needed, so trace it before we obsolete it */
if (curr_func)
{
FIXME("UNDF: curr_func %s\n", curr_func->hash_elt.name);
stabs_finalize_function(module, curr_func, 0); /* FIXME */
curr_func = NULL;
}
break;
case N_OPT:
/* Ignore this. We don't care what it points to. */
break;
case N_BINCL:
stabs_add_include(stabs_new_include(ptr, stab_ptr->n_value));
assert(incl_stk < (int)(sizeof(incl) / sizeof(incl[0])) - 1);
incl[++incl_stk] = source_idx;
source_idx = source_new(module, NULL, ptr);
break;
case N_EINCL:
assert(incl_stk >= 0);
source_idx = incl[incl_stk--];
break;
case N_EXCL:
if (stabs_add_include(stabs_find_include(ptr, stab_ptr->n_value)) < 0)
{
ERR("Excluded header not found (%s,%ld)\n", ptr, stab_ptr->n_value);
module_reset_debug_info(module);
ret = FALSE;
goto done;
}
break;
case N_MAIN:
/* Always ignore these. GCC doesn't even generate them. */
break;
case N_BNSYM:
case N_ENSYM:
/* Always ignore these, they seem to be used only on Darwin. */
break;
default:
ERR("Unknown stab type 0x%02x\n", stab_ptr->n_type);
break;
}
stabbuff[0] = '\0';
TRACE("0x%02x %lx %s\n",
stab_ptr->n_type, stab_ptr->n_value, debugstr_a(strs + stab_ptr->n_un.n_strx));
}
module->module.SymType = SymDia;
module->module.CVSig = 'S' | ('T' << 8) | ('A' << 16) | ('B' << 24);
/* FIXME: we could have a finer grain here */
module->module.LineNumbers = TRUE;
module->module.GlobalSymbols = TRUE;
module->module.TypeInfo = TRUE;
module->module.SourceIndexed = TRUE;
module->module.Publics = TRUE;
done:
HeapFree(GetProcessHeap(), 0, stabbuff);
stabs_free_includes();
HeapFree(GetProcessHeap(), 0, pending.vars);
return ret;
}