496 lines
8.6 KiB
C
496 lines
8.6 KiB
C
/* Scheme/Guile language support routines for GDB, the GNU debugger.
|
|
|
|
Copyright (C) 1995, 1996, 2000, 2003, 2005, 2008, 2009
|
|
Free Software Foundation, Inc.
|
|
|
|
This file is part of GDB.
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program 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 General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>. */
|
|
|
|
#include "defs.h"
|
|
#include "symtab.h"
|
|
#include "gdbtypes.h"
|
|
#include "expression.h"
|
|
#include "parser-defs.h"
|
|
#include "language.h"
|
|
#include "value.h"
|
|
#include "c-lang.h"
|
|
#include "scm-lang.h"
|
|
#include "scm-tags.h"
|
|
|
|
#define USE_EXPRSTRING 0
|
|
|
|
static void scm_lreadparen (int);
|
|
static int scm_skip_ws (void);
|
|
static void scm_read_token (int, int);
|
|
static LONGEST scm_istring2number (char *, int, int);
|
|
static LONGEST scm_istr2int (char *, int, int);
|
|
static void scm_lreadr (int);
|
|
|
|
static LONGEST
|
|
scm_istr2int (char *str, int len, int radix)
|
|
{
|
|
int i = 0;
|
|
LONGEST inum = 0;
|
|
int c;
|
|
int sign = 0;
|
|
|
|
if (0 >= len)
|
|
return SCM_BOOL_F; /* zero scm_length */
|
|
switch (str[0])
|
|
{ /* leading sign */
|
|
case '-':
|
|
case '+':
|
|
sign = str[0];
|
|
if (++i == len)
|
|
return SCM_BOOL_F; /* bad if lone `+' or `-' */
|
|
}
|
|
do
|
|
{
|
|
switch (c = str[i++])
|
|
{
|
|
case '0':
|
|
case '1':
|
|
case '2':
|
|
case '3':
|
|
case '4':
|
|
case '5':
|
|
case '6':
|
|
case '7':
|
|
case '8':
|
|
case '9':
|
|
c = c - '0';
|
|
goto accumulate;
|
|
case 'A':
|
|
case 'B':
|
|
case 'C':
|
|
case 'D':
|
|
case 'E':
|
|
case 'F':
|
|
c = c - 'A' + 10;
|
|
goto accumulate;
|
|
case 'a':
|
|
case 'b':
|
|
case 'c':
|
|
case 'd':
|
|
case 'e':
|
|
case 'f':
|
|
c = c - 'a' + 10;
|
|
accumulate:
|
|
if (c >= radix)
|
|
return SCM_BOOL_F; /* bad digit for radix */
|
|
inum *= radix;
|
|
inum += c;
|
|
break;
|
|
default:
|
|
return SCM_BOOL_F; /* not a digit */
|
|
}
|
|
}
|
|
while (i < len);
|
|
if (sign == '-')
|
|
inum = -inum;
|
|
return SCM_MAKINUM (inum);
|
|
}
|
|
|
|
static LONGEST
|
|
scm_istring2number (char *str, int len, int radix)
|
|
{
|
|
int i = 0;
|
|
char ex = 0;
|
|
char ex_p = 0, rx_p = 0; /* Only allow 1 exactness and 1 radix prefix */
|
|
#if 0
|
|
SCM res;
|
|
#endif
|
|
if (len == 1)
|
|
if (*str == '+' || *str == '-') /* Catches lone `+' and `-' for speed */
|
|
return SCM_BOOL_F;
|
|
|
|
while ((len - i) >= 2 && str[i] == '#' && ++i)
|
|
switch (str[i++])
|
|
{
|
|
case 'b':
|
|
case 'B':
|
|
if (rx_p++)
|
|
return SCM_BOOL_F;
|
|
radix = 2;
|
|
break;
|
|
case 'o':
|
|
case 'O':
|
|
if (rx_p++)
|
|
return SCM_BOOL_F;
|
|
radix = 8;
|
|
break;
|
|
case 'd':
|
|
case 'D':
|
|
if (rx_p++)
|
|
return SCM_BOOL_F;
|
|
radix = 10;
|
|
break;
|
|
case 'x':
|
|
case 'X':
|
|
if (rx_p++)
|
|
return SCM_BOOL_F;
|
|
radix = 16;
|
|
break;
|
|
case 'i':
|
|
case 'I':
|
|
if (ex_p++)
|
|
return SCM_BOOL_F;
|
|
ex = 2;
|
|
break;
|
|
case 'e':
|
|
case 'E':
|
|
if (ex_p++)
|
|
return SCM_BOOL_F;
|
|
ex = 1;
|
|
break;
|
|
default:
|
|
return SCM_BOOL_F;
|
|
}
|
|
|
|
switch (ex)
|
|
{
|
|
case 1:
|
|
return scm_istr2int (&str[i], len - i, radix);
|
|
case 0:
|
|
return scm_istr2int (&str[i], len - i, radix);
|
|
#if 0
|
|
if NFALSEP
|
|
(res) return res;
|
|
#ifdef FLOATS
|
|
case 2:
|
|
return scm_istr2flo (&str[i], len - i, radix);
|
|
#endif
|
|
#endif
|
|
}
|
|
return SCM_BOOL_F;
|
|
}
|
|
|
|
static void
|
|
scm_read_token (int c, int weird)
|
|
{
|
|
while (1)
|
|
{
|
|
c = *lexptr++;
|
|
switch (c)
|
|
{
|
|
case '[':
|
|
case ']':
|
|
case '(':
|
|
case ')':
|
|
case '\"':
|
|
case ';':
|
|
case ' ':
|
|
case '\t':
|
|
case '\r':
|
|
case '\f':
|
|
case '\n':
|
|
if (weird)
|
|
goto default_case;
|
|
case '\0': /* End of line */
|
|
eof_case:
|
|
--lexptr;
|
|
return;
|
|
case '\\':
|
|
if (!weird)
|
|
goto default_case;
|
|
else
|
|
{
|
|
c = *lexptr++;
|
|
if (c == '\0')
|
|
goto eof_case;
|
|
else
|
|
goto default_case;
|
|
}
|
|
case '}':
|
|
if (!weird)
|
|
goto default_case;
|
|
|
|
c = *lexptr++;
|
|
if (c == '#')
|
|
return;
|
|
else
|
|
{
|
|
--lexptr;
|
|
c = '}';
|
|
goto default_case;
|
|
}
|
|
|
|
default:
|
|
default_case:
|
|
;
|
|
}
|
|
}
|
|
}
|
|
|
|
static int
|
|
scm_skip_ws (void)
|
|
{
|
|
int c;
|
|
while (1)
|
|
switch ((c = *lexptr++))
|
|
{
|
|
case '\0':
|
|
goteof:
|
|
return c;
|
|
case ';':
|
|
lp:
|
|
switch ((c = *lexptr++))
|
|
{
|
|
case '\0':
|
|
goto goteof;
|
|
default:
|
|
goto lp;
|
|
case '\n':
|
|
break;
|
|
}
|
|
case ' ':
|
|
case '\t':
|
|
case '\r':
|
|
case '\f':
|
|
case '\n':
|
|
break;
|
|
default:
|
|
return c;
|
|
}
|
|
}
|
|
|
|
static void
|
|
scm_lreadparen (int skipping)
|
|
{
|
|
for (;;)
|
|
{
|
|
int c = scm_skip_ws ();
|
|
if (')' == c || ']' == c)
|
|
return;
|
|
--lexptr;
|
|
if (c == '\0')
|
|
error ("missing close paren");
|
|
scm_lreadr (skipping);
|
|
}
|
|
}
|
|
|
|
static void
|
|
scm_lreadr (int skipping)
|
|
{
|
|
int c, j;
|
|
struct stoken str;
|
|
LONGEST svalue = 0;
|
|
tryagain:
|
|
c = *lexptr++;
|
|
switch (c)
|
|
{
|
|
case '\0':
|
|
lexptr--;
|
|
return;
|
|
case '[':
|
|
case '(':
|
|
scm_lreadparen (skipping);
|
|
return;
|
|
case ']':
|
|
case ')':
|
|
error ("unexpected #\\%c", c);
|
|
goto tryagain;
|
|
case '\'':
|
|
case '`':
|
|
str.ptr = lexptr - 1;
|
|
scm_lreadr (skipping);
|
|
if (!skipping)
|
|
{
|
|
struct value *val = scm_evaluate_string (str.ptr, lexptr - str.ptr);
|
|
if (!is_scmvalue_type (value_type (val)))
|
|
error ("quoted scm form yields non-SCM value");
|
|
svalue = extract_signed_integer (value_contents (val),
|
|
TYPE_LENGTH (value_type (val)));
|
|
goto handle_immediate;
|
|
}
|
|
return;
|
|
case ',':
|
|
c = *lexptr++;
|
|
if ('@' != c)
|
|
lexptr--;
|
|
scm_lreadr (skipping);
|
|
return;
|
|
case '#':
|
|
c = *lexptr++;
|
|
switch (c)
|
|
{
|
|
case '[':
|
|
case '(':
|
|
scm_lreadparen (skipping);
|
|
return;
|
|
case 't':
|
|
case 'T':
|
|
svalue = SCM_BOOL_T;
|
|
goto handle_immediate;
|
|
case 'f':
|
|
case 'F':
|
|
svalue = SCM_BOOL_F;
|
|
goto handle_immediate;
|
|
case 'b':
|
|
case 'B':
|
|
case 'o':
|
|
case 'O':
|
|
case 'd':
|
|
case 'D':
|
|
case 'x':
|
|
case 'X':
|
|
case 'i':
|
|
case 'I':
|
|
case 'e':
|
|
case 'E':
|
|
lexptr--;
|
|
c = '#';
|
|
goto num;
|
|
case '*': /* bitvector */
|
|
scm_read_token (c, 0);
|
|
return;
|
|
case '{':
|
|
scm_read_token (c, 1);
|
|
return;
|
|
case '\\': /* character */
|
|
c = *lexptr++;
|
|
scm_read_token (c, 0);
|
|
return;
|
|
case '|':
|
|
j = 1; /* here j is the comment nesting depth */
|
|
lp:
|
|
c = *lexptr++;
|
|
lpc:
|
|
switch (c)
|
|
{
|
|
case '\0':
|
|
error ("unbalanced comment");
|
|
default:
|
|
goto lp;
|
|
case '|':
|
|
if ('#' != (c = *lexptr++))
|
|
goto lpc;
|
|
if (--j)
|
|
goto lp;
|
|
break;
|
|
case '#':
|
|
if ('|' != (c = *lexptr++))
|
|
goto lpc;
|
|
++j;
|
|
goto lp;
|
|
}
|
|
goto tryagain;
|
|
case '.':
|
|
default:
|
|
#if 0
|
|
callshrp:
|
|
#endif
|
|
scm_lreadr (skipping);
|
|
return;
|
|
}
|
|
case '\"':
|
|
while ('\"' != (c = *lexptr++))
|
|
{
|
|
if (c == '\\')
|
|
switch (c = *lexptr++)
|
|
{
|
|
case '\0':
|
|
error ("non-terminated string literal");
|
|
case '\n':
|
|
continue;
|
|
case '0':
|
|
case 'f':
|
|
case 'n':
|
|
case 'r':
|
|
case 't':
|
|
case 'a':
|
|
case 'v':
|
|
break;
|
|
}
|
|
}
|
|
return;
|
|
case '0':
|
|
case '1':
|
|
case '2':
|
|
case '3':
|
|
case '4':
|
|
case '5':
|
|
case '6':
|
|
case '7':
|
|
case '8':
|
|
case '9':
|
|
case '.':
|
|
case '-':
|
|
case '+':
|
|
num:
|
|
{
|
|
str.ptr = lexptr - 1;
|
|
scm_read_token (c, 0);
|
|
if (!skipping)
|
|
{
|
|
svalue = scm_istring2number (str.ptr, lexptr - str.ptr, 10);
|
|
if (svalue != SCM_BOOL_F)
|
|
goto handle_immediate;
|
|
goto tok;
|
|
}
|
|
}
|
|
return;
|
|
case ':':
|
|
scm_read_token ('-', 0);
|
|
return;
|
|
#if 0
|
|
do_symbol:
|
|
#endif
|
|
default:
|
|
str.ptr = lexptr - 1;
|
|
scm_read_token (c, 0);
|
|
tok:
|
|
if (!skipping)
|
|
{
|
|
str.length = lexptr - str.ptr;
|
|
if (str.ptr[0] == '$')
|
|
{
|
|
write_dollar_variable (str);
|
|
return;
|
|
}
|
|
write_exp_elt_opcode (OP_NAME);
|
|
write_exp_string (str);
|
|
write_exp_elt_opcode (OP_NAME);
|
|
}
|
|
return;
|
|
}
|
|
handle_immediate:
|
|
if (!skipping)
|
|
{
|
|
write_exp_elt_opcode (OP_LONG);
|
|
write_exp_elt_type (builtin_scm_type (parse_gdbarch)->builtin_scm);
|
|
write_exp_elt_longcst (svalue);
|
|
write_exp_elt_opcode (OP_LONG);
|
|
}
|
|
}
|
|
|
|
int
|
|
scm_parse (void)
|
|
{
|
|
char *start;
|
|
while (*lexptr == ' ')
|
|
lexptr++;
|
|
start = lexptr;
|
|
scm_lreadr (USE_EXPRSTRING);
|
|
#if USE_EXPRSTRING
|
|
str.length = lexptr - start;
|
|
str.ptr = start;
|
|
write_exp_elt_opcode (OP_EXPRSTRING);
|
|
write_exp_string (str);
|
|
write_exp_elt_opcode (OP_EXPRSTRING);
|
|
#endif
|
|
return 0;
|
|
}
|