OpenCores
URL https://opencores.org/ocsvn/or1k/or1k/trunk

Subversion Repositories or1k

[/] [or1k/] [trunk/] [insight/] [gdb/] [parse.c] - Rev 1782

Go to most recent revision | Compare with Previous | Blame | View Log

/* Parse expressions for GDB.
   Copyright 1986, 1989, 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997,
   1998, 1999, 2000, 2001 Free Software Foundation, Inc.
   Modified from expread.y by the Department of Computer Science at the
   State University of New York at Buffalo, 1991.
 
   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 2 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, write to the Free Software
   Foundation, Inc., 59 Temple Place - Suite 330,
   Boston, MA 02111-1307, USA.  */
 
/* Parse an expression from text in a string,
   and return the result as a  struct expression  pointer.
   That structure contains arithmetic operations in reverse polish,
   with constants represented by operations that are followed by special data.
   See expression.h for the details of the format.
   What is important here is that it can be built up sequentially
   during the process of parsing; the lower levels of the tree always
   come first in the result.  */
 
#include <ctype.h>
 
#include "defs.h"
#include "gdb_string.h"
#include "symtab.h"
#include "gdbtypes.h"
#include "frame.h"
#include "expression.h"
#include "value.h"
#include "command.h"
#include "language.h"
#include "parser-defs.h"
#include "gdbcmd.h"
#include "symfile.h"		/* for overlay functions */
#include "inferior.h"		/* for NUM_PSEUDO_REGS.  NOTE: replace 
				   with "gdbarch.h" when appropriate.  */
 

/* Symbols which architectures can redefine.  */
 
/* Some systems have routines whose names start with `$'.  Giving this
   macro a non-zero value tells GDB's expression parser to check for
   such routines when parsing tokens that begin with `$'.
 
   On HP-UX, certain system routines (millicode) have names beginning
   with `$' or `$$'.  For example, `$$dyncall' is a millicode routine
   that handles inter-space procedure calls on PA-RISC.  */
#ifndef SYMBOLS_CAN_START_WITH_DOLLAR
#define SYMBOLS_CAN_START_WITH_DOLLAR (0)
#endif
 
 

/* Global variables declared in parser-defs.h (and commented there).  */
struct expression *expout;
int expout_size;
int expout_ptr;
struct block *expression_context_block;
struct block *innermost_block;
int arglist_len;
union type_stack_elt *type_stack;
int type_stack_depth, type_stack_size;
char *lexptr;
char *namecopy;
int paren_depth;
int comma_terminates;

static int expressiondebug = 0;
 
extern int hp_som_som_object_present;
 
static void free_funcalls (void *ignore);
 
static void prefixify_expression (struct expression *);
 
static void
prefixify_subexp (struct expression *, struct expression *, int, int);
 
void _initialize_parse (void);
 
/* Data structure for saving values of arglist_len for function calls whose
   arguments contain other function calls.  */
 
struct funcall
  {
    struct funcall *next;
    int arglist_len;
  };
 
static struct funcall *funcall_chain;
 
/* Assign machine-independent names to certain registers 
   (unless overridden by the REGISTER_NAMES table) */
 
unsigned num_std_regs = 0;
struct std_regs *std_regs;
 
/* The generic method for targets to specify how their registers are
   named.  The mapping can be derived from three sources:
   REGISTER_NAME; std_regs; or a target specific alias hook. */
 
int
target_map_name_to_register (char *str, int len)
{
  int i;
 
  /* First try target specific aliases. We try these first because on some 
     systems standard names can be context dependent (eg. $pc on a 
     multiprocessor can be could be any of several PCs).  */
#ifdef REGISTER_NAME_ALIAS_HOOK
  i = REGISTER_NAME_ALIAS_HOOK (str, len);
  if (i >= 0)
    return i;
#endif
 
  /* Search architectural register name space. */
  for (i = 0; i < NUM_REGS; i++)
    if (REGISTER_NAME (i) && len == strlen (REGISTER_NAME (i))
	&& STREQN (str, REGISTER_NAME (i), len))
      {
	return i;
      }
 
  /* Try pseudo-registers, if any. */
  for (i = NUM_REGS; i < NUM_REGS + NUM_PSEUDO_REGS; i++)
    if (REGISTER_NAME (i) && len == strlen (REGISTER_NAME (i))
	&& STREQN (str, REGISTER_NAME (i), len))
      {
	return i;
      }
 
  /* Try standard aliases. */
  for (i = 0; i < num_std_regs; i++)
    if (std_regs[i].name && len == strlen (std_regs[i].name)
	&& STREQN (str, std_regs[i].name, len))
      {
	return std_regs[i].regnum;
      }
 
  return -1;
}
 
/* Begin counting arguments for a function call,
   saving the data about any containing call.  */
 
void
start_arglist (void)
{
  register struct funcall *new;
 
  new = (struct funcall *) xmalloc (sizeof (struct funcall));
  new->next = funcall_chain;
  new->arglist_len = arglist_len;
  arglist_len = 0;
  funcall_chain = new;
}
 
/* Return the number of arguments in a function call just terminated,
   and restore the data for the containing function call.  */
 
int
end_arglist (void)
{
  register int val = arglist_len;
  register struct funcall *call = funcall_chain;
  funcall_chain = call->next;
  arglist_len = call->arglist_len;
  xfree (call);
  return val;
}
 
/* Free everything in the funcall chain.
   Used when there is an error inside parsing.  */
 
static void
free_funcalls (void *ignore)
{
  register struct funcall *call, *next;
 
  for (call = funcall_chain; call; call = next)
    {
      next = call->next;
      xfree (call);
    }
}

/* This page contains the functions for adding data to the  struct expression
   being constructed.  */
 
/* Add one element to the end of the expression.  */
 
/* To avoid a bug in the Sun 4 compiler, we pass things that can fit into
   a register through here */
 
void
write_exp_elt (union exp_element expelt)
{
  if (expout_ptr >= expout_size)
    {
      expout_size *= 2;
      expout = (struct expression *)
	xrealloc ((char *) expout, sizeof (struct expression)
		  + EXP_ELEM_TO_BYTES (expout_size));
    }
  expout->elts[expout_ptr++] = expelt;
}
 
void
write_exp_elt_opcode (enum exp_opcode expelt)
{
  union exp_element tmp;
 
  tmp.opcode = expelt;
 
  write_exp_elt (tmp);
}
 
void
write_exp_elt_sym (struct symbol *expelt)
{
  union exp_element tmp;
 
  tmp.symbol = expelt;
 
  write_exp_elt (tmp);
}
 
void
write_exp_elt_block (struct block *b)
{
  union exp_element tmp;
  tmp.block = b;
  write_exp_elt (tmp);
}
 
void
write_exp_elt_longcst (LONGEST expelt)
{
  union exp_element tmp;
 
  tmp.longconst = expelt;
 
  write_exp_elt (tmp);
}
 
void
write_exp_elt_dblcst (DOUBLEST expelt)
{
  union exp_element tmp;
 
  tmp.doubleconst = expelt;
 
  write_exp_elt (tmp);
}
 
void
write_exp_elt_type (struct type *expelt)
{
  union exp_element tmp;
 
  tmp.type = expelt;
 
  write_exp_elt (tmp);
}
 
void
write_exp_elt_intern (struct internalvar *expelt)
{
  union exp_element tmp;
 
  tmp.internalvar = expelt;
 
  write_exp_elt (tmp);
}
 
/* Add a string constant to the end of the expression.
 
   String constants are stored by first writing an expression element
   that contains the length of the string, then stuffing the string
   constant itself into however many expression elements are needed
   to hold it, and then writing another expression element that contains
   the length of the string.  I.E. an expression element at each end of
   the string records the string length, so you can skip over the 
   expression elements containing the actual string bytes from either
   end of the string.  Note that this also allows gdb to handle
   strings with embedded null bytes, as is required for some languages.
 
   Don't be fooled by the fact that the string is null byte terminated,
   this is strictly for the convenience of debugging gdb itself.  Gdb
   Gdb does not depend up the string being null terminated, since the
   actual length is recorded in expression elements at each end of the
   string.  The null byte is taken into consideration when computing how
   many expression elements are required to hold the string constant, of
   course. */
 
 
void
write_exp_string (struct stoken str)
{
  register int len = str.length;
  register int lenelt;
  register char *strdata;
 
  /* Compute the number of expression elements required to hold the string
     (including a null byte terminator), along with one expression element
     at each end to record the actual string length (not including the
     null byte terminator). */
 
  lenelt = 2 + BYTES_TO_EXP_ELEM (len + 1);
 
  /* Ensure that we have enough available expression elements to store
     everything. */
 
  if ((expout_ptr + lenelt) >= expout_size)
    {
      expout_size = max (expout_size * 2, expout_ptr + lenelt + 10);
      expout = (struct expression *)
	xrealloc ((char *) expout, (sizeof (struct expression)
				    + EXP_ELEM_TO_BYTES (expout_size)));
    }
 
  /* Write the leading length expression element (which advances the current
     expression element index), then write the string constant followed by a
     terminating null byte, and then write the trailing length expression
     element. */
 
  write_exp_elt_longcst ((LONGEST) len);
  strdata = (char *) &expout->elts[expout_ptr];
  memcpy (strdata, str.ptr, len);
  *(strdata + len) = '\0';
  expout_ptr += lenelt - 2;
  write_exp_elt_longcst ((LONGEST) len);
}
 
/* Add a bitstring constant to the end of the expression.
 
   Bitstring constants are stored by first writing an expression element
   that contains the length of the bitstring (in bits), then stuffing the
   bitstring constant itself into however many expression elements are
   needed to hold it, and then writing another expression element that
   contains the length of the bitstring.  I.E. an expression element at
   each end of the bitstring records the bitstring length, so you can skip
   over the expression elements containing the actual bitstring bytes from
   either end of the bitstring. */
 
void
write_exp_bitstring (struct stoken str)
{
  register int bits = str.length;	/* length in bits */
  register int len = (bits + HOST_CHAR_BIT - 1) / HOST_CHAR_BIT;
  register int lenelt;
  register char *strdata;
 
  /* Compute the number of expression elements required to hold the bitstring,
     along with one expression element at each end to record the actual
     bitstring length in bits. */
 
  lenelt = 2 + BYTES_TO_EXP_ELEM (len);
 
  /* Ensure that we have enough available expression elements to store
     everything. */
 
  if ((expout_ptr + lenelt) >= expout_size)
    {
      expout_size = max (expout_size * 2, expout_ptr + lenelt + 10);
      expout = (struct expression *)
	xrealloc ((char *) expout, (sizeof (struct expression)
				    + EXP_ELEM_TO_BYTES (expout_size)));
    }
 
  /* Write the leading length expression element (which advances the current
     expression element index), then write the bitstring constant, and then
     write the trailing length expression element. */
 
  write_exp_elt_longcst ((LONGEST) bits);
  strdata = (char *) &expout->elts[expout_ptr];
  memcpy (strdata, str.ptr, len);
  expout_ptr += lenelt - 2;
  write_exp_elt_longcst ((LONGEST) bits);
}
 
/* Add the appropriate elements for a minimal symbol to the end of
   the expression.  The rationale behind passing in text_symbol_type and
   data_symbol_type was so that Modula-2 could pass in WORD for
   data_symbol_type.  Perhaps it still is useful to have those types vary
   based on the language, but they no longer have names like "int", so
   the initial rationale is gone.  */
 
static struct type *msym_text_symbol_type;
static struct type *msym_data_symbol_type;
static struct type *msym_unknown_symbol_type;
 
void
write_exp_msymbol (struct minimal_symbol *msymbol, 
		   struct type *text_symbol_type, 
		   struct type *data_symbol_type)
{
  CORE_ADDR addr;
 
  write_exp_elt_opcode (OP_LONG);
  /* Let's make the type big enough to hold a 64-bit address.  */
  write_exp_elt_type (builtin_type_CORE_ADDR);
 
  addr = SYMBOL_VALUE_ADDRESS (msymbol);
  if (overlay_debugging)
    addr = symbol_overlayed_address (addr, SYMBOL_BFD_SECTION (msymbol));
  write_exp_elt_longcst ((LONGEST) addr);
 
  write_exp_elt_opcode (OP_LONG);
 
  write_exp_elt_opcode (UNOP_MEMVAL);
  switch (msymbol->type)
    {
    case mst_text:
    case mst_file_text:
    case mst_solib_trampoline:
      write_exp_elt_type (msym_text_symbol_type);
      break;
 
    case mst_data:
    case mst_file_data:
    case mst_bss:
    case mst_file_bss:
      write_exp_elt_type (msym_data_symbol_type);
      break;
 
    default:
      write_exp_elt_type (msym_unknown_symbol_type);
      break;
    }
  write_exp_elt_opcode (UNOP_MEMVAL);
}

/* Recognize tokens that start with '$'.  These include:
 
   $regname     A native register name or a "standard
   register name".
 
   $variable    A convenience variable with a name chosen
   by the user.
 
   $digits              Value history with index <digits>, starting
   from the first value which has index 1.
 
   $$digits     Value history with index <digits> relative
   to the last value.  I.E. $$0 is the last
   value, $$1 is the one previous to that, $$2
   is the one previous to $$1, etc.
 
   $ | $0 | $$0 The last value in the value history.
 
   $$           An abbreviation for the second to the last
   value in the value history, I.E. $$1
 
 */
 
void
write_dollar_variable (struct stoken str)
{
  /* Handle the tokens $digits; also $ (short for $0) and $$ (short for $$1)
     and $$digits (equivalent to $<-digits> if you could type that). */
 
  int negate = 0;
  int i = 1;
  /* Double dollar means negate the number and add -1 as well.
     Thus $$ alone means -1.  */
  if (str.length >= 2 && str.ptr[1] == '$')
    {
      negate = 1;
      i = 2;
    }
  if (i == str.length)
    {
      /* Just dollars (one or two) */
      i = -negate;
      goto handle_last;
    }
  /* Is the rest of the token digits?  */
  for (; i < str.length; i++)
    if (!(str.ptr[i] >= '0' && str.ptr[i] <= '9'))
      break;
  if (i == str.length)
    {
      i = atoi (str.ptr + 1 + negate);
      if (negate)
	i = -i;
      goto handle_last;
    }
 
  /* Handle tokens that refer to machine registers:
     $ followed by a register name.  */
  i = target_map_name_to_register (str.ptr + 1, str.length - 1);
  if (i >= 0)
    goto handle_register;
 
  if (SYMBOLS_CAN_START_WITH_DOLLAR)
    {
      struct symbol *sym = NULL;
      struct minimal_symbol *msym = NULL;
 
      /* On HP-UX, certain system routines (millicode) have names beginning
	 with $ or $$, e.g. $$dyncall, which handles inter-space procedure
	 calls on PA-RISC. Check for those, first. */
 
      /* This code is not enabled on non HP-UX systems, since worst case 
	 symbol table lookup performance is awful, to put it mildly. */
 
      sym = lookup_symbol (copy_name (str), (struct block *) NULL,
			   VAR_NAMESPACE, (int *) NULL, (struct symtab **) NULL);
      if (sym)
	{
	  write_exp_elt_opcode (OP_VAR_VALUE);
	  write_exp_elt_block (block_found);	/* set by lookup_symbol */
	  write_exp_elt_sym (sym);
	  write_exp_elt_opcode (OP_VAR_VALUE);
	  return;
	}
      msym = lookup_minimal_symbol (copy_name (str), NULL, NULL);
      if (msym)
	{
	  write_exp_msymbol (msym,
			     lookup_function_type (builtin_type_int),
			     builtin_type_int);
	  return;
	}
    }
 
  /* Any other names starting in $ are debugger internal variables.  */
 
  write_exp_elt_opcode (OP_INTERNALVAR);
  write_exp_elt_intern (lookup_internalvar (copy_name (str) + 1));
  write_exp_elt_opcode (OP_INTERNALVAR);
  return;
handle_last:
  write_exp_elt_opcode (OP_LAST);
  write_exp_elt_longcst ((LONGEST) i);
  write_exp_elt_opcode (OP_LAST);
  return;
handle_register:
  write_exp_elt_opcode (OP_REGISTER);
  write_exp_elt_longcst (i);
  write_exp_elt_opcode (OP_REGISTER);
  return;
}
 
 
/* Parse a string that is possibly a namespace / nested class
   specification, i.e., something of the form A::B::C::x.  Input
   (NAME) is the entire string; LEN is the current valid length; the
   output is a string, TOKEN, which points to the largest recognized
   prefix which is a series of namespaces or classes.  CLASS_PREFIX is
   another output, which records whether a nested class spec was
   recognized (= 1) or a fully qualified variable name was found (=
   0).  ARGPTR is side-effected (if non-NULL) to point to beyond the
   string recognized and consumed by this routine.
 
   The return value is a pointer to the symbol for the base class or
   variable if found, or NULL if not found.  Callers must check this
   first -- if NULL, the outputs may not be correct. 
 
   This function is used c-exp.y.  This is used specifically to get
   around HP aCC (and possibly other compilers), which insists on
   generating names with embedded colons for namespace or nested class
   members.
 
   (Argument LEN is currently unused. 1997-08-27)
 
   Callers must free memory allocated for the output string TOKEN.  */
 
static const char coloncolon[2] =
{':', ':'};
 
struct symbol *
parse_nested_classes_for_hpacc (char *name, int len, char **token,
				int *class_prefix, char **argptr)
{
  /* Comment below comes from decode_line_1 which has very similar
     code, which is called for "break" command parsing. */
 
  /* We have what looks like a class or namespace
     scope specification (A::B), possibly with many
     levels of namespaces or classes (A::B::C::D).
 
     Some versions of the HP ANSI C++ compiler (as also possibly
     other compilers) generate class/function/member names with
     embedded double-colons if they are inside namespaces. To
     handle this, we loop a few times, considering larger and
     larger prefixes of the string as though they were single
     symbols.  So, if the initially supplied string is
     A::B::C::D::foo, we have to look up "A", then "A::B",
     then "A::B::C", then "A::B::C::D", and finally
     "A::B::C::D::foo" as single, monolithic symbols, because
     A, B, C or D may be namespaces.
 
     Note that namespaces can nest only inside other
     namespaces, and not inside classes.  So we need only
     consider *prefixes* of the string; there is no need to look up
     "B::C" separately as a symbol in the previous example. */
 
  register char *p;
  char *start, *end;
  char *prefix = NULL;
  char *tmp;
  struct symbol *sym_class = NULL;
  struct symbol *sym_var = NULL;
  struct type *t;
  int prefix_len = 0;
  int done = 0;
  char *q;
 
  /* Check for HP-compiled executable -- in other cases
     return NULL, and caller must default to standard GDB
     behaviour. */
 
  if (!hp_som_som_object_present)
    return (struct symbol *) NULL;
 
  p = name;
 
  /* Skip over whitespace and possible global "::" */
  while (*p && (*p == ' ' || *p == '\t'))
    p++;
  if (p[0] == ':' && p[1] == ':')
    p += 2;
  while (*p && (*p == ' ' || *p == '\t'))
    p++;
 
  while (1)
    {
      /* Get to the end of the next namespace or class spec. */
      /* If we're looking at some non-token, fail immediately */
      start = p;
      if (!(isalpha (*p) || *p == '$' || *p == '_'))
	return (struct symbol *) NULL;
      p++;
      while (*p && (isalnum (*p) || *p == '$' || *p == '_'))
	p++;
 
      if (*p == '<')
	{
	  /* If we have the start of a template specification,
	     scan right ahead to its end */
	  q = find_template_name_end (p);
	  if (q)
	    p = q;
	}
 
      end = p;
 
      /* Skip over "::" and whitespace for next time around */
      while (*p && (*p == ' ' || *p == '\t'))
	p++;
      if (p[0] == ':' && p[1] == ':')
	p += 2;
      while (*p && (*p == ' ' || *p == '\t'))
	p++;
 
      /* Done with tokens? */
      if (!*p || !(isalpha (*p) || *p == '$' || *p == '_'))
	done = 1;
 
      tmp = (char *) alloca (prefix_len + end - start + 3);
      if (prefix)
	{
	  memcpy (tmp, prefix, prefix_len);
	  memcpy (tmp + prefix_len, coloncolon, 2);
	  memcpy (tmp + prefix_len + 2, start, end - start);
	  tmp[prefix_len + 2 + end - start] = '\000';
	}
      else
	{
	  memcpy (tmp, start, end - start);
	  tmp[end - start] = '\000';
	}
 
      prefix = tmp;
      prefix_len = strlen (prefix);
 
      /* See if the prefix we have now is something we know about */
 
      if (!done)
	{
	  /* More tokens to process, so this must be a class/namespace */
	  sym_class = lookup_symbol (prefix, 0, STRUCT_NAMESPACE,
				     0, (struct symtab **) NULL);
	}
      else
	{
	  /* No more tokens, so try as a variable first */
	  sym_var = lookup_symbol (prefix, 0, VAR_NAMESPACE,
				   0, (struct symtab **) NULL);
	  /* If failed, try as class/namespace */
	  if (!sym_var)
	    sym_class = lookup_symbol (prefix, 0, STRUCT_NAMESPACE,
				       0, (struct symtab **) NULL);
	}
 
      if (sym_var ||
	  (sym_class &&
	   (t = check_typedef (SYMBOL_TYPE (sym_class)),
	    (TYPE_CODE (t) == TYPE_CODE_STRUCT
	     || TYPE_CODE (t) == TYPE_CODE_UNION))))
	{
	  /* We found a valid token */
	  *token = (char *) xmalloc (prefix_len + 1);
	  memcpy (*token, prefix, prefix_len);
	  (*token)[prefix_len] = '\000';
	  break;
	}
 
      /* No variable or class/namespace found, no more tokens */
      if (done)
	return (struct symbol *) NULL;
    }
 
  /* Out of loop, so we must have found a valid token */
  if (sym_var)
    *class_prefix = 0;
  else
    *class_prefix = 1;
 
  if (argptr)
    *argptr = done ? p : end;
 
  return sym_var ? sym_var : sym_class;		/* found */
}
 
char *
find_template_name_end (char *p)
{
  int depth = 1;
  int just_seen_right = 0;
  int just_seen_colon = 0;
  int just_seen_space = 0;
 
  if (!p || (*p != '<'))
    return 0;
 
  while (*++p)
    {
      switch (*p)
	{
	case '\'':
	case '\"':
	case '{':
	case '}':
	  /* In future, may want to allow these?? */
	  return 0;
	case '<':
	  depth++;		/* start nested template */
	  if (just_seen_colon || just_seen_right || just_seen_space)
	    return 0;		/* but not after : or :: or > or space */
	  break;
	case '>':
	  if (just_seen_colon || just_seen_right)
	    return 0;		/* end a (nested?) template */
	  just_seen_right = 1;	/* but not after : or :: */
	  if (--depth == 0)	/* also disallow >>, insist on > > */
	    return ++p;		/* if outermost ended, return */
	  break;
	case ':':
	  if (just_seen_space || (just_seen_colon > 1))
	    return 0;		/* nested class spec coming up */
	  just_seen_colon++;	/* we allow :: but not :::: */
	  break;
	case ' ':
	  break;
	default:
	  if (!((*p >= 'a' && *p <= 'z') ||	/* allow token chars */
		(*p >= 'A' && *p <= 'Z') ||
		(*p >= '0' && *p <= '9') ||
		(*p == '_') || (*p == ',') ||	/* commas for template args */
		(*p == '&') || (*p == '*') ||	/* pointer and ref types */
		(*p == '(') || (*p == ')') ||	/* function types */
		(*p == '[') || (*p == ']')))	/* array types */
	    return 0;
	}
      if (*p != ' ')
	just_seen_space = 0;
      if (*p != ':')
	just_seen_colon = 0;
      if (*p != '>')
	just_seen_right = 0;
    }
  return 0;
}

 
 
/* Return a null-terminated temporary copy of the name
   of a string token.  */
 
char *
copy_name (struct stoken token)
{
  memcpy (namecopy, token.ptr, token.length);
  namecopy[token.length] = 0;
  return namecopy;
}

/* Reverse an expression from suffix form (in which it is constructed)
   to prefix form (in which we can conveniently print or execute it).  */
 
static void
prefixify_expression (register struct expression *expr)
{
  register int len =
  sizeof (struct expression) + EXP_ELEM_TO_BYTES (expr->nelts);
  register struct expression *temp;
  register int inpos = expr->nelts, outpos = 0;
 
  temp = (struct expression *) alloca (len);
 
  /* Copy the original expression into temp.  */
  memcpy (temp, expr, len);
 
  prefixify_subexp (temp, expr, inpos, outpos);
}
 
/* Return the number of exp_elements in the subexpression of EXPR
   whose last exp_element is at index ENDPOS - 1 in EXPR.  */
 
int
length_of_subexp (register struct expression *expr, register int endpos)
{
  register int oplen = 1;
  register int args = 0;
  register int i;
 
  if (endpos < 1)
    error ("?error in length_of_subexp");
 
  i = (int) expr->elts[endpos - 1].opcode;
 
  switch (i)
    {
      /* C++  */
    case OP_SCOPE:
      oplen = longest_to_int (expr->elts[endpos - 2].longconst);
      oplen = 5 + BYTES_TO_EXP_ELEM (oplen + 1);
      break;
 
    case OP_LONG:
    case OP_DOUBLE:
    case OP_VAR_VALUE:
      oplen = 4;
      break;
 
    case OP_TYPE:
    case OP_BOOL:
    case OP_LAST:
    case OP_REGISTER:
    case OP_INTERNALVAR:
      oplen = 3;
      break;
 
    case OP_COMPLEX:
      oplen = 1;
      args = 2;
      break;
 
    case OP_FUNCALL:
    case OP_F77_UNDETERMINED_ARGLIST:
      oplen = 3;
      args = 1 + longest_to_int (expr->elts[endpos - 2].longconst);
      break;
 
    case UNOP_MAX:
    case UNOP_MIN:
      oplen = 3;
      break;
 
    case BINOP_VAL:
    case UNOP_CAST:
    case UNOP_MEMVAL:
      oplen = 3;
      args = 1;
      break;
 
    case UNOP_ABS:
    case UNOP_CAP:
    case UNOP_CHR:
    case UNOP_FLOAT:
    case UNOP_HIGH:
    case UNOP_ODD:
    case UNOP_ORD:
    case UNOP_TRUNC:
      oplen = 1;
      args = 1;
      break;
 
    case OP_LABELED:
    case STRUCTOP_STRUCT:
    case STRUCTOP_PTR:
      args = 1;
      /* fall through */
    case OP_M2_STRING:
    case OP_STRING:
    case OP_NAME:
    case OP_EXPRSTRING:
      oplen = longest_to_int (expr->elts[endpos - 2].longconst);
      oplen = 4 + BYTES_TO_EXP_ELEM (oplen + 1);
      break;
 
    case OP_BITSTRING:
      oplen = longest_to_int (expr->elts[endpos - 2].longconst);
      oplen = (oplen + HOST_CHAR_BIT - 1) / HOST_CHAR_BIT;
      oplen = 4 + BYTES_TO_EXP_ELEM (oplen);
      break;
 
    case OP_ARRAY:
      oplen = 4;
      args = longest_to_int (expr->elts[endpos - 2].longconst);
      args -= longest_to_int (expr->elts[endpos - 3].longconst);
      args += 1;
      break;
 
    case TERNOP_COND:
    case TERNOP_SLICE:
    case TERNOP_SLICE_COUNT:
      args = 3;
      break;
 
      /* Modula-2 */
    case MULTI_SUBSCRIPT:
      oplen = 3;
      args = 1 + longest_to_int (expr->elts[endpos - 2].longconst);
      break;
 
    case BINOP_ASSIGN_MODIFY:
      oplen = 3;
      args = 2;
      break;
 
      /* C++ */
    case OP_THIS:
      oplen = 2;
      break;
 
    default:
      args = 1 + (i < (int) BINOP_END);
    }
 
  while (args > 0)
    {
      oplen += length_of_subexp (expr, endpos - oplen);
      args--;
    }
 
  return oplen;
}
 
/* Copy the subexpression ending just before index INEND in INEXPR
   into OUTEXPR, starting at index OUTBEG.
   In the process, convert it from suffix to prefix form.  */
 
static void
prefixify_subexp (register struct expression *inexpr,
		  struct expression *outexpr, register int inend, int outbeg)
{
  register int oplen = 1;
  register int args = 0;
  register int i;
  int *arglens;
  enum exp_opcode opcode;
 
  /* Compute how long the last operation is (in OPLEN),
     and also how many preceding subexpressions serve as
     arguments for it (in ARGS).  */
 
  opcode = inexpr->elts[inend - 1].opcode;
  switch (opcode)
    {
      /* C++  */
    case OP_SCOPE:
      oplen = longest_to_int (inexpr->elts[inend - 2].longconst);
      oplen = 5 + BYTES_TO_EXP_ELEM (oplen + 1);
      break;
 
    case OP_LONG:
    case OP_DOUBLE:
    case OP_VAR_VALUE:
      oplen = 4;
      break;
 
    case OP_TYPE:
    case OP_BOOL:
    case OP_LAST:
    case OP_REGISTER:
    case OP_INTERNALVAR:
      oplen = 3;
      break;
 
    case OP_COMPLEX:
      oplen = 1;
      args = 2;
      break;
 
    case OP_FUNCALL:
    case OP_F77_UNDETERMINED_ARGLIST:
      oplen = 3;
      args = 1 + longest_to_int (inexpr->elts[inend - 2].longconst);
      break;
 
    case UNOP_MIN:
    case UNOP_MAX:
      oplen = 3;
      break;
 
    case UNOP_CAST:
    case UNOP_MEMVAL:
      oplen = 3;
      args = 1;
      break;
 
    case UNOP_ABS:
    case UNOP_CAP:
    case UNOP_CHR:
    case UNOP_FLOAT:
    case UNOP_HIGH:
    case UNOP_ODD:
    case UNOP_ORD:
    case UNOP_TRUNC:
      oplen = 1;
      args = 1;
      break;
 
    case STRUCTOP_STRUCT:
    case STRUCTOP_PTR:
    case OP_LABELED:
      args = 1;
      /* fall through */
    case OP_M2_STRING:
    case OP_STRING:
    case OP_NAME:
    case OP_EXPRSTRING:
      oplen = longest_to_int (inexpr->elts[inend - 2].longconst);
      oplen = 4 + BYTES_TO_EXP_ELEM (oplen + 1);
      break;
 
    case OP_BITSTRING:
      oplen = longest_to_int (inexpr->elts[inend - 2].longconst);
      oplen = (oplen + HOST_CHAR_BIT - 1) / HOST_CHAR_BIT;
      oplen = 4 + BYTES_TO_EXP_ELEM (oplen);
      break;
 
    case OP_ARRAY:
      oplen = 4;
      args = longest_to_int (inexpr->elts[inend - 2].longconst);
      args -= longest_to_int (inexpr->elts[inend - 3].longconst);
      args += 1;
      break;
 
    case TERNOP_COND:
    case TERNOP_SLICE:
    case TERNOP_SLICE_COUNT:
      args = 3;
      break;
 
    case BINOP_ASSIGN_MODIFY:
      oplen = 3;
      args = 2;
      break;
 
      /* Modula-2 */
    case MULTI_SUBSCRIPT:
      oplen = 3;
      args = 1 + longest_to_int (inexpr->elts[inend - 2].longconst);
      break;
 
      /* C++ */
    case OP_THIS:
      oplen = 2;
      break;
 
    default:
      args = 1 + ((int) opcode < (int) BINOP_END);
    }
 
  /* Copy the final operator itself, from the end of the input
     to the beginning of the output.  */
  inend -= oplen;
  memcpy (&outexpr->elts[outbeg], &inexpr->elts[inend],
	  EXP_ELEM_TO_BYTES (oplen));
  outbeg += oplen;
 
  /* Find the lengths of the arg subexpressions.  */
  arglens = (int *) alloca (args * sizeof (int));
  for (i = args - 1; i >= 0; i--)
    {
      oplen = length_of_subexp (inexpr, inend);
      arglens[i] = oplen;
      inend -= oplen;
    }
 
  /* Now copy each subexpression, preserving the order of
     the subexpressions, but prefixifying each one.
     In this loop, inend starts at the beginning of
     the expression this level is working on
     and marches forward over the arguments.
     outbeg does similarly in the output.  */
  for (i = 0; i < args; i++)
    {
      oplen = arglens[i];
      inend += oplen;
      prefixify_subexp (inexpr, outexpr, inend, outbeg);
      outbeg += oplen;
    }
}

/* This page contains the two entry points to this file.  */
 
/* Read an expression from the string *STRINGPTR points to,
   parse it, and return a pointer to a  struct expression  that we malloc.
   Use block BLOCK as the lexical context for variable names;
   if BLOCK is zero, use the block of the selected stack frame.
   Meanwhile, advance *STRINGPTR to point after the expression,
   at the first nonwhite character that is not part of the expression
   (possibly a null character).
 
   If COMMA is nonzero, stop if a comma is reached.  */
 
struct expression *
parse_exp_1 (char **stringptr, struct block *block, int comma)
{
  struct cleanup *old_chain;
 
  lexptr = *stringptr;
 
  paren_depth = 0;
  type_stack_depth = 0;
 
  comma_terminates = comma;
 
  if (lexptr == 0 || *lexptr == 0)
    error_no_arg ("expression to compute");
 
  old_chain = make_cleanup (free_funcalls, 0 /*ignore*/);
  funcall_chain = 0;
 
  expression_context_block = block ? block : get_selected_block ();
 
  namecopy = (char *) alloca (strlen (lexptr) + 1);
  expout_size = 10;
  expout_ptr = 0;
  expout = (struct expression *)
    xmalloc (sizeof (struct expression) + EXP_ELEM_TO_BYTES (expout_size));
  expout->language_defn = current_language;
  make_cleanup (free_current_contents, &expout);
 
  if (current_language->la_parser ())
    current_language->la_error (NULL);
 
  discard_cleanups (old_chain);
 
  /* Record the actual number of expression elements, and then
     reallocate the expression memory so that we free up any
     excess elements. */
 
  expout->nelts = expout_ptr;
  expout = (struct expression *)
    xrealloc ((char *) expout,
	      sizeof (struct expression) + EXP_ELEM_TO_BYTES (expout_ptr));;
 
  /* Convert expression from postfix form as generated by yacc
     parser, to a prefix form. */
 
  if (expressiondebug)
    dump_prefix_expression (expout, gdb_stdlog,
			    "before conversion to prefix form");
 
  prefixify_expression (expout);
 
  if (expressiondebug)
    dump_postfix_expression (expout, gdb_stdlog,
			     "after conversion to prefix form");
 
  *stringptr = lexptr;
  return expout;
}
 
/* Parse STRING as an expression, and complain if this fails
   to use up all of the contents of STRING.  */
 
struct expression *
parse_expression (char *string)
{
  register struct expression *exp;
  exp = parse_exp_1 (&string, 0, 0);
  if (*string)
    error ("Junk after end of expression.");
  return exp;
}

/* Stuff for maintaining a stack of types.  Currently just used by C, but
   probably useful for any language which declares its types "backwards".  */
 
void
push_type (enum type_pieces tp)
{
  if (type_stack_depth == type_stack_size)
    {
      type_stack_size *= 2;
      type_stack = (union type_stack_elt *)
	xrealloc ((char *) type_stack, type_stack_size * sizeof (*type_stack));
    }
  type_stack[type_stack_depth++].piece = tp;
}
 
void
push_type_int (int n)
{
  if (type_stack_depth == type_stack_size)
    {
      type_stack_size *= 2;
      type_stack = (union type_stack_elt *)
	xrealloc ((char *) type_stack, type_stack_size * sizeof (*type_stack));
    }
  type_stack[type_stack_depth++].int_val = n;
}
 
enum type_pieces
pop_type (void)
{
  if (type_stack_depth)
    return type_stack[--type_stack_depth].piece;
  return tp_end;
}
 
int
pop_type_int (void)
{
  if (type_stack_depth)
    return type_stack[--type_stack_depth].int_val;
  /* "Can't happen".  */
  return 0;
}
 
/* Pop the type stack and return the type which corresponds to FOLLOW_TYPE
   as modified by all the stuff on the stack.  */
struct type *
follow_types (struct type *follow_type)
{
  int done = 0;
  int array_size;
  struct type *range_type;
 
  while (!done)
    switch (pop_type ())
      {
      case tp_end:
	done = 1;
	break;
      case tp_pointer:
	follow_type = lookup_pointer_type (follow_type);
	break;
      case tp_reference:
	follow_type = lookup_reference_type (follow_type);
	break;
      case tp_array:
	array_size = pop_type_int ();
	/* FIXME-type-allocation: need a way to free this type when we are
	   done with it.  */
	range_type =
	  create_range_type ((struct type *) NULL,
			     builtin_type_int, 0,
			     array_size >= 0 ? array_size - 1 : 0);
	follow_type =
	  create_array_type ((struct type *) NULL,
			     follow_type, range_type);
	if (array_size < 0)
	  TYPE_ARRAY_UPPER_BOUND_TYPE (follow_type)
	    = BOUND_CANNOT_BE_DETERMINED;
	break;
      case tp_function:
	/* FIXME-type-allocation: need a way to free this type when we are
	   done with it.  */
	follow_type = lookup_function_type (follow_type);
	break;
      }
  return follow_type;
}

static void build_parse (void);
static void
build_parse (void)
{
  int i;
 
  msym_text_symbol_type =
    init_type (TYPE_CODE_FUNC, 1, 0, "<text variable, no debug info>", NULL);
  TYPE_TARGET_TYPE (msym_text_symbol_type) = builtin_type_int;
  msym_data_symbol_type =
    init_type (TYPE_CODE_INT, TARGET_INT_BIT / HOST_CHAR_BIT, 0,
	       "<data variable, no debug info>", NULL);
  msym_unknown_symbol_type =
    init_type (TYPE_CODE_INT, 1, 0,
	       "<variable (not text or data), no debug info>",
	       NULL);
 
  /* create the std_regs table */
 
  num_std_regs = 0;
#ifdef PC_REGNUM
  if (PC_REGNUM >= 0)
    num_std_regs++;
#endif
#ifdef FP_REGNUM
  if (FP_REGNUM >= 0)
    num_std_regs++;
#endif
#ifdef SP_REGNUM
  if (SP_REGNUM >= 0)
    num_std_regs++;
#endif
#ifdef PS_REGNUM
  if (PS_REGNUM >= 0)
    num_std_regs++;
#endif
  /* create an empty table */
  std_regs = xmalloc ((num_std_regs + 1) * sizeof *std_regs);
  i = 0;
  /* fill it in */
#ifdef PC_REGNUM
  if (PC_REGNUM >= 0)
    {
      std_regs[i].name = "pc";
      std_regs[i].regnum = PC_REGNUM;
      i++;
    }
#endif
#ifdef FP_REGNUM
  if (FP_REGNUM >= 0)
    {
      std_regs[i].name = "fp";
      std_regs[i].regnum = FP_REGNUM;
      i++;
    }
#endif
#ifdef SP_REGNUM
  if (SP_REGNUM >= 0)
    {
      std_regs[i].name = "sp";
      std_regs[i].regnum = SP_REGNUM;
      i++;
    }
#endif
#ifdef PS_REGNUM
  if (PS_REGNUM >= 0)
    {
      std_regs[i].name = "ps";
      std_regs[i].regnum = PS_REGNUM;
      i++;
    }
#endif
  memset (&std_regs[i], 0, sizeof (std_regs[i]));
}
 
void
_initialize_parse (void)
{
  type_stack_size = 80;
  type_stack_depth = 0;
  type_stack = (union type_stack_elt *)
    xmalloc (type_stack_size * sizeof (*type_stack));
 
  build_parse ();
 
  /* FIXME - For the moment, handle types by swapping them in and out.
     Should be using the per-architecture data-pointer and a large
     struct. */
  register_gdbarch_swap (&msym_text_symbol_type, sizeof (msym_text_symbol_type), NULL);
  register_gdbarch_swap (&msym_data_symbol_type, sizeof (msym_data_symbol_type), NULL);
  register_gdbarch_swap (&msym_unknown_symbol_type, sizeof (msym_unknown_symbol_type), NULL);
 
  register_gdbarch_swap (&num_std_regs, sizeof (std_regs), NULL);
  register_gdbarch_swap (&std_regs, sizeof (std_regs), NULL);
  register_gdbarch_swap (NULL, 0, build_parse);
 
  add_show_from_set (
	    add_set_cmd ("expression", class_maintenance, var_zinteger,
			 (char *) &expressiondebug,
			 "Set expression debugging.\n\
When non-zero, the internal representation of expressions will be printed.",
			 &setdebuglist),
		      &showdebuglist);
}
 

Go to most recent revision | Compare with Previous | Blame | View Log

powered by: WebSVN 2.1.0

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.