/* Mainly the interface between cpplib and the C front ends.
Copyright (C) 1987, 1988, 1989, 1992, 1994, 1995, 1996, 1997
- 1998, 1999, 2000, 2001, 2002, 2003 Free Software Foundation, Inc.
+ 1998, 1999, 2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
This file is part of GCC.
/* The current line map. */
static const struct line_map *map;
-/* The line used to refresh the lineno global variable after each token. */
-static unsigned int src_lineno;
-
/* We may keep statistics about how long which files took to compile. */
static int header_time, body_time;
static splay_tree file_info_tree;
int pending_lang_change; /* If we need to switch languages - C++ only */
int c_header_level; /* depth in C headers - C++ only */
-/* Nonzero tells yylex to ignore \ in string constants. */
-static int ignore_escape_flag;
-
static tree interpret_integer (const cpp_token *, unsigned int);
static tree interpret_float (const cpp_token *, unsigned int);
static enum integer_type_kind
narrowest_unsigned_type (tree, unsigned int);
static enum integer_type_kind
narrowest_signed_type (tree, unsigned int);
-static tree lex_string (const cpp_string *);
+static enum cpp_ttype lex_string (const cpp_token *, tree *, bool);
static tree lex_charconst (const cpp_token *);
static void update_header_times (const char *);
static int dump_one_header (splay_tree_node, void *);
if (n)
return (struct c_fileinfo *) n->value;
- fi = (struct c_fileinfo *) xmalloc (sizeof (struct c_fileinfo));
+ fi = xmalloc (sizeof (struct c_fileinfo));
fi->time = 0;
fi->interface_only = 0;
fi->interface_unknown = 1;
if (! flag_no_ident)
{
/* Convert escapes in the string. */
- tree value ATTRIBUTE_UNUSED = lex_string (str);
- ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (value));
+ cpp_string cstr = { 0, 0 };
+ if (cpp_interpret_string (pfile, str, 1, &cstr, false))
+ {
+ ASM_OUTPUT_IDENT (asm_out_file, (const char *) cstr.text);
+ free ((void *)cstr.text);
+ }
}
#endif
}
lexed token on the line. Used for diagnostic line numbers. */
static void
cb_line_change (cpp_reader *pfile ATTRIBUTE_UNUSED, const cpp_token *token,
- int parsing_args ATTRIBUTE_UNUSED)
+ int parsing_args)
{
- src_lineno = SOURCE_LINE (map, token->line);
+ if (token->type == CPP_EOF || parsing_args)
+ return;
+
+ input_line = SOURCE_LINE (map, token->line);
}
void
fe_file_change (const struct line_map *new_map)
{
- unsigned int to_line = SOURCE_LINE (new_map, new_map->to_line);
+ if (new_map == NULL)
+ {
+ map = NULL;
+ return;
+ }
if (new_map->reason == LC_ENTER)
{
/* Don't stack the main buffer on the input stack;
we already did in compile_file. */
- if (map == NULL)
- main_input_filename = new_map->to_file;
- else
+ if (map != NULL)
{
int included_at = SOURCE_LINE (new_map - 1, new_map->from_line - 1);
#endif
pop_srcloc ();
- (*debug_hooks->end_source_file) (to_line);
+ (*debug_hooks->end_source_file) (new_map->to_line);
}
update_header_times (new_map->to_file);
in_system_header = new_map->sysp != 0;
input_filename = new_map->to_file;
- input_line = to_line;
+ input_line = new_map->to_line;
map = new_map;
/* Hook for C++. */
(const char *) NODE_NAME (node));
}
\f
-int
-c_lex (tree *value)
+static inline const cpp_token *
+get_nonpadding_token (void)
{
const cpp_token *tok;
-
- retry:
timevar_push (TV_CPP);
do
tok = cpp_get_token (parse_in);
while (tok->type == CPP_PADDING);
timevar_pop (TV_CPP);
- /* The C++ front end does horrible things with the current line
- number. To ensure an accurate line number, we must reset it
- every time we return a token. */
- input_line = src_lineno;
+ return tok;
+}
- *value = NULL_TREE;
+int
+c_lex_with_flags (tree *value, unsigned char *cpp_flags)
+{
+ const cpp_token *tok;
+ location_t atloc;
+ static bool no_more_pch;
+
+ retry:
+ tok = get_nonpadding_token ();
+
+ retry_after_at:
switch (tok->type)
{
case CPP_NAME:
}
break;
+ case CPP_ATSIGN:
+ /* An @ may give the next token special significance in Objective-C. */
+ atloc = input_location;
+ tok = get_nonpadding_token ();
+ if (c_dialect_objc ())
+ {
+ tree val;
+ switch (tok->type)
+ {
+ case CPP_NAME:
+ val = HT_IDENT_TO_GCC_IDENT (HT_NODE (tok->val.node));
+ if (C_IS_RESERVED_WORD (val)
+ && OBJC_IS_AT_KEYWORD (C_RID_CODE (val)))
+ {
+ *value = val;
+ return CPP_AT_NAME;
+ }
+ break;
+
+ case CPP_STRING:
+ case CPP_WSTRING:
+ return lex_string (tok, value, true);
+
+ default: break;
+ }
+ }
+
+ /* ... or not. */
+ error ("%Hstray '@' in program", &atloc);
+ goto retry_after_at;
+
case CPP_OTHER:
{
cppchar_t c = tok->val.str.text[0];
case CPP_STRING:
case CPP_WSTRING:
- *value = lex_string (&tok->val.str);
+ return lex_string (tok, value, false);
break;
/* These tokens should not be visible outside cpplib. */
case CPP_MACRO_ARG:
abort ();
- default: break;
+ default:
+ *value = NULL_TREE;
+ break;
}
+ if (! no_more_pch)
+ {
+ no_more_pch = true;
+ c_common_no_more_pch ();
+ }
+
+ if (cpp_flags)
+ *cpp_flags = tok->flags;
return tok->type;
}
+int
+c_lex (tree *value)
+{
+ return c_lex_with_flags (value, NULL);
+}
+
/* Returns the narrowest C-visible unsigned type, starting with the
minimum specified by FLAGS, that can fit VALUE, or itk_none if
there isn't one. */
return value;
}
-static tree
-lex_string (const cpp_string *str)
+/* Convert a series of STRING and/or WSTRING tokens into a tree,
+ performing string constant concatenation. TOK is the first of
+ these. VALP is the location to write the string into. OBJC_STRING
+ indicates whether an '@' token preceded the incoming token.
+ Returns the CPP token type of the result (CPP_STRING, CPP_WSTRING,
+ or CPP_OBJC_STRING).
+
+ This is unfortunately more work than it should be. If any of the
+ strings in the series has an L prefix, the result is a wide string
+ (6.4.5p4). Whether or not the result is a wide string affects the
+ meaning of octal and hexadecimal escapes (6.4.4.4p6,9). But escape
+ sequences do not continue across the boundary between two strings in
+ a series (6.4.5p7), so we must not lose the boundaries. Therefore
+ cpp_interpret_string takes a vector of cpp_string structures, which
+ we must arrange to provide. */
+
+static enum cpp_ttype
+lex_string (const cpp_token *tok, tree *valp, bool objc_string)
{
- bool wide;
tree value;
- char *buf, *q;
- cppchar_t c;
- const unsigned char *p, *limit;
+ bool wide = false;
+ size_t count = 1;
+ struct obstack str_ob;
+ cpp_string istr;
- wide = str->text[0] == 'L';
- p = str->text + 1 + wide;
- limit = str->text + str->len - 1;
- q = buf = alloca ((str->len + 1) * (wide ? WCHAR_BYTES : 1));
+ /* Try to avoid the overhead of creating and destroying an obstack
+ for the common case of just one string. */
+ cpp_string str = tok->val.str;
+ cpp_string *strs = &str;
- while (p < limit)
- {
- c = *p++;
+ if (tok->type == CPP_WSTRING)
+ wide = true;
- if (c == '\\' && !ignore_escape_flag)
- c = cpp_parse_escape (parse_in, &p, limit, wide);
+ tok = get_nonpadding_token ();
+ if (c_dialect_objc () && tok->type == CPP_ATSIGN)
+ {
+ objc_string = true;
+ tok = get_nonpadding_token ();
+ }
+ if (tok->type == CPP_STRING || tok->type == CPP_WSTRING)
+ {
+ gcc_obstack_init (&str_ob);
+ obstack_grow (&str_ob, &str, sizeof (cpp_string));
- /* Add this single character into the buffer either as a wchar_t,
- a multibyte sequence, or as a single byte. */
- if (wide)
+ do
{
- unsigned charwidth = TYPE_PRECISION (char_type_node);
- unsigned bytemask = (1 << charwidth) - 1;
- int byte;
-
- for (byte = 0; byte < WCHAR_BYTES; ++byte)
+ count++;
+ if (tok->type == CPP_WSTRING)
+ wide = true;
+ obstack_grow (&str_ob, &tok->val.str, sizeof (cpp_string));
+
+ tok = get_nonpadding_token ();
+ if (c_dialect_objc () && tok->type == CPP_ATSIGN)
{
- int n;
- if (byte >= (int) sizeof (c))
- n = 0;
- else
- n = (c >> (byte * charwidth)) & bytemask;
- if (BYTES_BIG_ENDIAN)
- q[WCHAR_BYTES - byte - 1] = n;
- else
- q[byte] = n;
+ objc_string = true;
+ tok = get_nonpadding_token ();
}
- q += WCHAR_BYTES;
- }
- else
- {
- *q++ = c;
}
+ while (tok->type == CPP_STRING || tok->type == CPP_WSTRING);
+ strs = obstack_finish (&str_ob);
}
- /* Terminate the string value, either with a single byte zero
- or with a wide zero. */
+ /* We have read one more token than we want. */
+ _cpp_backup_tokens (parse_in, 1);
- if (wide)
+ if (count > 1 && !objc_string && warn_traditional && !in_system_header)
+ warning ("traditional C rejects string constant concatenation");
+
+ if (cpp_interpret_string (parse_in, strs, count, &istr, wide))
{
- memset (q, 0, WCHAR_BYTES);
- q += WCHAR_BYTES;
+ value = build_string (istr.len, (char *)istr.text);
+ free ((void *)istr.text);
}
else
{
- *q++ = '\0';
+ /* Callers cannot generally handle error_mark_node in this context,
+ so return the empty string instead. cpp_interpret_string has
+ issued an error. */
+ if (wide)
+ value = build_string (TYPE_PRECISION (wchar_type_node)
+ / TYPE_PRECISION (char_type_node),
+ "\0\0\0"); /* widest supported wchar_t
+ is 32 bits */
+ else
+ value = build_string (1, "");
}
- value = build_string (q - buf, buf);
+ TREE_TYPE (value) = wide ? wchar_array_type_node : char_array_type_node;
+ *valp = fix_string_type (value);
- if (wide)
- TREE_TYPE (value) = wchar_array_type_node;
- else
- TREE_TYPE (value) = char_array_type_node;
- return value;
+ if (strs != &str)
+ obstack_free (&str_ob, 0);
+
+ return objc_string ? CPP_OBJC_STRING : wide ? CPP_WSTRING : CPP_STRING;
}
/* Converts a (possibly wide) character constant token into a tree. */
type = wchar_type_node;
/* In C, a character constant has type 'int'.
In C++ 'char', but multi-char charconsts have type 'int'. */
- else if ((c_language == clk_c) || chars_seen > 1)
+ else if (!c_dialect_cxx () || chars_seen > 1)
type = integer_type_node;
else
type = char_type_node;
OSDN Git Service
