12cf91fef0
gcc: * cpphash.h (U): New define, to correct type of string constants. (ustrcmp, ustrncmp, ustrlen, uxstrdup, ustrchr): New wrapper routines, to do casts when passing unsigned strings to libc. * cppexp.c, cppfiles.c, cpphash.c, cppinit.c, cpplib.c: Use them. * cppfiles.c (_cpp_execute_include): Make filename an U_CHAR *. * cpphash.c (_cpp_quote_string): Make string an U_CHAR *. * cppinit.c (dump_special_to_buffer): Make macro name an U_CHAR *. * cpplex.c (parse_ifdef, parse_include, validate_else): Make second argument an U_CHAR *. * cppinit.c (builtin_array): Make name and value U_CHAR *, add length field, clean up initializer. (ISTABLE): Add __extension__ to designated- initializers version. * cpplex.c (CHARTAB): Likewise. * mbchar.c: Add dummy external declaration to the !MULTIBYTE_CHARS case so the file won't be empty. include: * symcat.h: Remove #endif label. From-SVN: r33657
1716 lines
45 KiB
C
1716 lines
45 KiB
C
/* CPP Library.
|
|
Copyright (C) 1986, 1987, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
|
|
1999, 2000 Free Software Foundation, Inc.
|
|
Contributed by Per Bothner, 1994-95.
|
|
Based on CCCP program by Paul Rubin, June 1986
|
|
Adapted to ANSI C, Richard Stallman, Jan 1987
|
|
|
|
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, 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, 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
|
|
|
|
#include "config.h"
|
|
#include "system.h"
|
|
|
|
#include "cpplib.h"
|
|
#include "cpphash.h"
|
|
#include "intl.h"
|
|
#include "symcat.h"
|
|
|
|
/* `struct directive' defines one #-directive, including how to handle it. */
|
|
|
|
struct directive
|
|
{
|
|
directive_handler func; /* Function to handle directive. */
|
|
const U_CHAR *name; /* Name of directive. */
|
|
unsigned short length; /* Length of name. */
|
|
unsigned short flags; /* Flags describing this directive. */
|
|
};
|
|
|
|
/* Stack of conditionals currently in progress
|
|
(including both successful and failing conditionals). */
|
|
|
|
struct if_stack
|
|
{
|
|
struct if_stack *next;
|
|
int lineno; /* line number where condition started */
|
|
int if_succeeded; /* truth of last condition in this group */
|
|
const U_CHAR *control_macro; /* macro name for #ifndef around entire file */
|
|
int type; /* type of last directive seen in this group */
|
|
};
|
|
typedef struct if_stack IF_STACK;
|
|
|
|
/* Forward declarations. */
|
|
|
|
static void validate_else PARAMS ((cpp_reader *, const U_CHAR *));
|
|
static int parse_ifdef PARAMS ((cpp_reader *, const U_CHAR *));
|
|
static unsigned int parse_include PARAMS ((cpp_reader *, const U_CHAR *));
|
|
static int conditional_skip PARAMS ((cpp_reader *, int, int,
|
|
U_CHAR *));
|
|
static int skip_if_group PARAMS ((cpp_reader *));
|
|
static void pass_thru_directive PARAMS ((const U_CHAR *, size_t,
|
|
cpp_reader *, int));
|
|
static int read_line_number PARAMS ((cpp_reader *, int *));
|
|
static U_CHAR *detect_if_not_defined PARAMS ((cpp_reader *));
|
|
static int consider_directive_while_skipping
|
|
PARAMS ((cpp_reader *, IF_STACK *));
|
|
|
|
/* Values for the flags field of the table below. KANDR and COND
|
|
directives come from traditional (K&R) C. The difference is, if we
|
|
care about it while skipping a failed conditional block, its origin
|
|
is COND. STDC89 directives come from the 1989 C standard.
|
|
EXTENSION directives are extensions, with origins noted below. */
|
|
|
|
#define KANDR 0
|
|
#define COND 1
|
|
#define STDC89 2
|
|
#define EXTENSION 3
|
|
|
|
#define ORIGIN_MASK 3
|
|
#define ORIGIN(f) ((f) & ORIGIN_MASK)
|
|
#define TRAD_DIRECT_P(f) (ORIGIN (f) == KANDR || ORIGIN (f) == COND)
|
|
|
|
/* This is the table of directive handlers. It is ordered by
|
|
frequency of occurrence; the numbers at the end are directive
|
|
counts from all the source code I have lying around (egcs and libc
|
|
CVS as of 1999-05-18, plus grub-0.5.91, linux-2.2.9, and
|
|
pcmcia-cs-3.0.9).
|
|
|
|
The entries with a dash and a name after the count are extensions,
|
|
of which all but #warning and #include_next are deprecated. The name
|
|
is where the extension appears to have come from. */
|
|
|
|
/* #sccs is not always recognized. */
|
|
#ifdef SCCS_DIRECTIVE
|
|
# define SCCS_ENTRY D(sccs, T_SCCS, EXTENSION) /* 0 - SVR2? */
|
|
#else
|
|
# define SCCS_ENTRY /* nothing */
|
|
#endif
|
|
|
|
#define DIRECTIVE_TABLE \
|
|
D(define, T_DEFINE = 0, KANDR) /* 270554 */ \
|
|
D(include, T_INCLUDE, KANDR | SYNTAX_INCLUDE) /* 52262 */ \
|
|
D(endif, T_ENDIF, COND) /* 45855 */ \
|
|
D(ifdef, T_IFDEF, COND) /* 22000 */ \
|
|
D(if, T_IF, COND) /* 18162 */ \
|
|
D(else, T_ELSE, COND) /* 9863 */ \
|
|
D(ifndef, T_IFNDEF, COND) /* 9675 */ \
|
|
D(undef, T_UNDEF, KANDR) /* 4837 */ \
|
|
D(line, T_LINE, KANDR) /* 2465 */ \
|
|
D(elif, T_ELIF, COND) /* 610 */ \
|
|
D(error, T_ERROR, STDC89) /* 475 */ \
|
|
D(pragma, T_PRAGMA, STDC89) /* 195 */ \
|
|
D(warning, T_WARNING, EXTENSION) /* 22 GNU */ \
|
|
D(include_next, T_INCLUDE_NEXT, EXTENSION | SYNTAX_INCLUDE) /* 19 GNU */ \
|
|
D(ident, T_IDENT, EXTENSION) /* 11 SVR4 */ \
|
|
D(import, T_IMPORT, EXTENSION | SYNTAX_INCLUDE) /* 0 ObjC */ \
|
|
D(assert, T_ASSERT, EXTENSION | SYNTAX_ASSERT) /* 0 SVR4 */ \
|
|
D(unassert, T_UNASSERT, EXTENSION | SYNTAX_ASSERT) /* 0 SVR4 */ \
|
|
SCCS_ENTRY
|
|
|
|
/* Use the table to generate a series of prototypes, an enum for the
|
|
directive names, and an array of directive handlers. */
|
|
|
|
/* The directive-processing functions are declared to return int
|
|
instead of void, because some old compilers have trouble with
|
|
pointers to functions returning void. */
|
|
|
|
/* Don't invoke CONCAT2 with any whitespace or K&R cc will fail. */
|
|
#define D(name, t, f) static int CONCAT2(do_,name) PARAMS ((cpp_reader *));
|
|
DIRECTIVE_TABLE
|
|
#undef D
|
|
|
|
#define D(n, tag, f) tag,
|
|
enum
|
|
{
|
|
DIRECTIVE_TABLE
|
|
N_DIRECTIVES
|
|
};
|
|
#undef D
|
|
|
|
/* Don't invoke CONCAT2 with any whitespace or K&R cc will fail. */
|
|
#define D(name, t, flags) \
|
|
{ CONCAT2(do_,name), (const U_CHAR *) STRINGX(name), \
|
|
sizeof STRINGX(name) - 1, flags },
|
|
static const struct directive dtable[] =
|
|
{
|
|
DIRECTIVE_TABLE
|
|
};
|
|
#undef D
|
|
#undef DIRECTIVE_TABLE
|
|
|
|
/* Check if a token's name matches that of a known directive. Put in
|
|
this file to save exporting dtable and other unneeded information. */
|
|
void
|
|
_cpp_check_directive (list, token)
|
|
cpp_toklist *list;
|
|
cpp_token *token;
|
|
{
|
|
const U_CHAR *name = list->namebuf + token->val.name.offset;
|
|
size_t len = token->val.name.len;
|
|
unsigned int i;
|
|
|
|
list->dir_handler = 0;
|
|
list->dir_flags = 0;
|
|
|
|
for (i = 0; i < N_DIRECTIVES; i++)
|
|
if (dtable[i].length == len && !ustrncmp (dtable[i].name, name, len))
|
|
{
|
|
list->dir_handler = dtable[i].func;
|
|
list->dir_flags = dtable[i].flags;
|
|
break;
|
|
}
|
|
}
|
|
|
|
/* Handle a possible # directive.
|
|
'#' has already been read. */
|
|
|
|
int
|
|
_cpp_handle_directive (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
int i;
|
|
int hash_at_bol;
|
|
unsigned int len;
|
|
U_CHAR *ident;
|
|
long old_written = CPP_WRITTEN (pfile);
|
|
enum cpp_ttype tok;
|
|
|
|
if (CPP_IS_MACRO_BUFFER (CPP_BUFFER (pfile)))
|
|
{
|
|
cpp_ice (pfile, "handle_directive called on macro buffer");
|
|
return 0;
|
|
}
|
|
|
|
/* -traditional directives are recognized only with the # in column 1. */
|
|
hash_at_bol = CPP_IN_COLUMN_1 (pfile);
|
|
|
|
/* Scan the next token, then pretend we didn't. */
|
|
CPP_SET_MARK (pfile);
|
|
pfile->no_macro_expand++;
|
|
tok = _cpp_get_directive_token (pfile);
|
|
pfile->no_macro_expand--;
|
|
|
|
ident = pfile->token_buffer + old_written;
|
|
len = CPP_PWRITTEN (pfile) - ident;
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
CPP_GOTO_MARK (pfile);
|
|
|
|
/* # followed by a number is equivalent to #line. Do not recognize
|
|
this form in assembly language source files. Complain about this
|
|
form if we're being pedantic, but not if this is regurgitated
|
|
input (preprocessed or fed back in by the C++ frontend). */
|
|
if (tok == CPP_NUMBER)
|
|
{
|
|
if (CPP_OPTION (pfile, lang_asm))
|
|
return 0;
|
|
|
|
if (CPP_PEDANTIC (pfile)
|
|
&& CPP_BUFFER (pfile)->ihash
|
|
&& ! CPP_OPTION (pfile, preprocessed))
|
|
cpp_pedwarn (pfile, "# followed by integer");
|
|
do_line (pfile);
|
|
return 1;
|
|
}
|
|
|
|
/* If we are rescanning preprocessed input, don't obey any directives
|
|
other than # nnn. */
|
|
else if (CPP_OPTION (pfile, preprocessed))
|
|
return 0;
|
|
|
|
/* A line of just # becomes blank. */
|
|
else if (tok == CPP_VSPACE)
|
|
return 1;
|
|
|
|
/* A NAME token might in fact be a directive! */
|
|
else if (tok == CPP_NAME)
|
|
{
|
|
for (i = 0; i < N_DIRECTIVES; i++)
|
|
{
|
|
if (dtable[i].length == len
|
|
&& !ustrncmp (dtable[i].name, ident, len))
|
|
goto real_directive;
|
|
}
|
|
/* Don't complain about invalid directives in assembly source,
|
|
we don't know where the comments are, and # may introduce
|
|
assembler pseudo-ops. */
|
|
if (!CPP_OPTION (pfile, lang_asm))
|
|
cpp_error (pfile, "invalid preprocessing directive #%s", ident);
|
|
return 0;
|
|
}
|
|
/* And anything else means the # wasn't a directive marker. */
|
|
else
|
|
return 0;
|
|
|
|
real_directive:
|
|
|
|
/* In -traditional mode, a directive is ignored unless its # is in
|
|
column 1. */
|
|
if (CPP_TRADITIONAL (pfile) && !hash_at_bol)
|
|
{
|
|
if (CPP_WTRADITIONAL (pfile))
|
|
cpp_warning (pfile, "ignoring #%s because of its indented #",
|
|
dtable[i].name);
|
|
return 0;
|
|
}
|
|
|
|
/* no_directives is set when we are parsing macro arguments. Directives
|
|
in macro arguments are undefined behavior (C99 6.10.3.11); this
|
|
implementation chooses to make them hard errors. */
|
|
if (pfile->no_directives)
|
|
{
|
|
cpp_error (pfile, "#%s may not be used inside a macro argument",
|
|
dtable[i].name);
|
|
_cpp_skip_rest_of_line (pfile);
|
|
return 1;
|
|
}
|
|
|
|
/* Issue -pedantic warnings for extended directives. */
|
|
if (CPP_PEDANTIC (pfile) && ORIGIN (dtable[i].flags) == EXTENSION)
|
|
cpp_pedwarn (pfile, "ISO C does not allow #%s", dtable[i].name);
|
|
|
|
/* -Wtraditional gives warnings about directives with inappropriate
|
|
indentation of #. */
|
|
if (CPP_WTRADITIONAL (pfile))
|
|
{
|
|
if (!hash_at_bol && TRAD_DIRECT_P (dtable[i].flags))
|
|
cpp_warning (pfile, "traditional C ignores #%s with the # indented",
|
|
dtable[i].name);
|
|
else if (hash_at_bol && ! TRAD_DIRECT_P (dtable[i].flags))
|
|
cpp_warning (pfile,
|
|
"suggest hiding #%s from traditional C with an indented #",
|
|
dtable[i].name);
|
|
}
|
|
|
|
/* Unfortunately, it's necessary to scan the directive name again,
|
|
now we know we're going to consume it. FIXME. */
|
|
|
|
pfile->no_macro_expand++;
|
|
_cpp_get_directive_token (pfile);
|
|
pfile->no_macro_expand--;
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
|
|
/* Some directives (e.g. #if) may return a request to execute
|
|
another directive handler immediately. No directive ever
|
|
requests that #define be executed immediately, so it is safe for
|
|
the loop to terminate when some function returns 0 (== T_DEFINE). */
|
|
while ((i = dtable[i].func (pfile)));
|
|
return 1;
|
|
}
|
|
|
|
/* Pass a directive through to the output file.
|
|
BUF points to the contents of the directive, as a contiguous string.
|
|
LEN is the length of the string pointed to by BUF.
|
|
KEYWORD is the keyword-table entry for the directive. */
|
|
|
|
static void
|
|
pass_thru_directive (buf, len, pfile, keyword)
|
|
const U_CHAR *buf;
|
|
size_t len;
|
|
cpp_reader *pfile;
|
|
int keyword;
|
|
{
|
|
const struct directive *kt = &dtable[keyword];
|
|
register unsigned klen = kt->length;
|
|
|
|
CPP_RESERVE (pfile, 1 + klen + len);
|
|
CPP_PUTC_Q (pfile, '#');
|
|
CPP_PUTS_Q (pfile, kt->name, klen);
|
|
if (len != 0 && buf[0] != ' ')
|
|
CPP_PUTC_Q (pfile, ' ');
|
|
CPP_PUTS_Q (pfile, buf, len);
|
|
}
|
|
|
|
/* Process a #define command. */
|
|
|
|
static int
|
|
do_define (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
HASHNODE *node;
|
|
int len;
|
|
U_CHAR *sym;
|
|
cpp_toklist *list = &pfile->directbuf;
|
|
|
|
pfile->no_macro_expand++;
|
|
pfile->parsing_define_directive++;
|
|
CPP_OPTION (pfile, discard_comments)++;
|
|
|
|
_cpp_scan_line (pfile, list);
|
|
|
|
/* First token on the line must be a NAME. There must be at least
|
|
one token (the VSPACE at the end). */
|
|
if (TOK_TYPE (list, 0) != CPP_NAME)
|
|
{
|
|
cpp_error_with_line (pfile, list->line, TOK_COL (list, 0),
|
|
"#define must be followed by an identifier");
|
|
goto out;
|
|
}
|
|
|
|
sym = TOK_NAME (list, 0);
|
|
len = TOK_LEN (list, 0);
|
|
|
|
/* That NAME is not allowed to be "defined". (Not clear if the
|
|
standard requires this.) */
|
|
if (len == 7 && !ustrncmp (sym, U"defined", 7))
|
|
{
|
|
cpp_error_with_line (pfile, list->line, TOK_COL (list, 0),
|
|
"\"defined\" is not a legal macro name");
|
|
goto out;
|
|
}
|
|
|
|
node = _cpp_lookup (pfile, sym, len);
|
|
/* Check for poisoned identifiers now. All other checks
|
|
are done in cpphash.c. */
|
|
if (node->type == T_POISON)
|
|
{
|
|
cpp_error (pfile, "redefining poisoned `%.*s'", len, sym);
|
|
goto out;
|
|
}
|
|
|
|
if (_cpp_create_definition (pfile, list, node) == 0)
|
|
goto out;
|
|
|
|
if (CPP_OPTION (pfile, debug_output)
|
|
|| CPP_OPTION (pfile, dump_macros) == dump_definitions)
|
|
_cpp_dump_definition (pfile, node);
|
|
else if (CPP_OPTION (pfile, dump_macros) == dump_names)
|
|
pass_thru_directive (sym, len, pfile, T_DEFINE);
|
|
|
|
out:
|
|
pfile->no_macro_expand--;
|
|
pfile->parsing_define_directive--;
|
|
CPP_OPTION (pfile, discard_comments)--;
|
|
return 0;
|
|
}
|
|
|
|
/* Handle #include and #import. */
|
|
|
|
static unsigned int
|
|
parse_include (pfile, name)
|
|
cpp_reader *pfile;
|
|
const U_CHAR *name;
|
|
{
|
|
long old_written = CPP_WRITTEN (pfile);
|
|
enum cpp_ttype token;
|
|
int len;
|
|
|
|
pfile->parsing_include_directive++;
|
|
token = _cpp_get_directive_token (pfile);
|
|
pfile->parsing_include_directive--;
|
|
|
|
len = CPP_WRITTEN (pfile) - old_written;
|
|
|
|
if (token == CPP_STRING)
|
|
; /* No special treatment required. */
|
|
#ifdef VMS
|
|
else if (token == CPP_NAME)
|
|
{
|
|
/* Support '#include xyz' like VAX-C. It is taken as
|
|
'#include <xyz.h>' and generates a warning. */
|
|
cpp_warning (pfile, "#%s filename is obsolete, use #%s <filename.h>",
|
|
name, name);
|
|
|
|
/* Rewrite the token to <xyz.h>. */
|
|
CPP_RESERVE (pfile, 4);
|
|
len += 4;
|
|
memmove (pfile->token_buffer + old_written + 1,
|
|
pfile->token_buffer + old_written,
|
|
CPP_WRITTEN (pfile) - old_written);
|
|
pfile->token_buffer[old_written] = '<';
|
|
CPP_PUTS_Q (pfile, ".h>", 2);
|
|
}
|
|
#endif
|
|
else
|
|
{
|
|
cpp_error (pfile, "`#%s' expects \"FILENAME\" or <FILENAME>", name);
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
_cpp_skip_rest_of_line (pfile);
|
|
return 0;
|
|
}
|
|
|
|
if (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
{
|
|
cpp_error (pfile, "junk at end of `#%s'", name);
|
|
_cpp_skip_rest_of_line (pfile);
|
|
}
|
|
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
|
|
if (len == 0)
|
|
cpp_error (pfile, "empty file name in `#%s'", name);
|
|
|
|
return len;
|
|
}
|
|
|
|
static int
|
|
do_include (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
unsigned int len;
|
|
U_CHAR *token;
|
|
|
|
len = parse_include (pfile, dtable[T_INCLUDE].name);
|
|
if (len == 0)
|
|
return 0;
|
|
token = (U_CHAR *) alloca (len + 1);
|
|
memcpy (token, CPP_PWRITTEN (pfile), len);
|
|
token[len] = '\0';
|
|
|
|
if (CPP_OPTION (pfile, dump_includes))
|
|
pass_thru_directive (token, len, pfile, T_INCLUDE);
|
|
|
|
_cpp_execute_include (pfile, token, len, 0, 0);
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
do_import (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
unsigned int len;
|
|
U_CHAR *token;
|
|
|
|
if (CPP_OPTION (pfile, warn_import)
|
|
&& !CPP_BUFFER (pfile)->system_header_p && !pfile->import_warning)
|
|
{
|
|
pfile->import_warning = 1;
|
|
cpp_warning (pfile,
|
|
"#import is obsolete, use an #ifndef wrapper in the header file");
|
|
}
|
|
|
|
len = parse_include (pfile, dtable[T_IMPORT].name);
|
|
if (len == 0)
|
|
return 0;
|
|
token = (U_CHAR *) alloca (len + 1);
|
|
memcpy (token, CPP_PWRITTEN (pfile), len);
|
|
token[len] = '\0';
|
|
|
|
if (CPP_OPTION (pfile, dump_includes))
|
|
pass_thru_directive (token, len, pfile, T_IMPORT);
|
|
|
|
_cpp_execute_include (pfile, token, len, 1, 0);
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
do_include_next (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
unsigned int len;
|
|
U_CHAR *token;
|
|
struct file_name_list *search_start = 0;
|
|
|
|
len = parse_include (pfile, dtable[T_INCLUDE_NEXT].name);
|
|
if (len == 0)
|
|
return 0;
|
|
token = (U_CHAR *) alloca (len + 1);
|
|
memcpy (token, CPP_PWRITTEN (pfile), len);
|
|
token[len] = '\0';
|
|
|
|
if (CPP_OPTION (pfile, dump_includes))
|
|
pass_thru_directive (token, len, pfile, T_INCLUDE_NEXT);
|
|
|
|
/* For #include_next, skip in the search path past the dir in which the
|
|
containing file was found. Treat files specified using an absolute path
|
|
as if there are no more directories to search. Treat the primary source
|
|
file like any other included source, but generate a warning. */
|
|
if (CPP_PREV_BUFFER (CPP_BUFFER (pfile)))
|
|
{
|
|
if (CPP_BUFFER (pfile)->ihash->foundhere != ABSOLUTE_PATH)
|
|
search_start = CPP_BUFFER (pfile)->ihash->foundhere->next;
|
|
}
|
|
else
|
|
cpp_warning (pfile, "#include_next in primary source file");
|
|
|
|
_cpp_execute_include (pfile, token, len, 0, search_start);
|
|
return 0;
|
|
}
|
|
|
|
/* Subroutine of do_line. Read next token from PFILE without adding it to
|
|
the output buffer. If it is a number between 1 and 4, store it in *NUM
|
|
and return 1; otherwise, return 0 and complain if we aren't at the end
|
|
of the directive. */
|
|
|
|
static int
|
|
read_line_number (pfile, num)
|
|
cpp_reader *pfile;
|
|
int *num;
|
|
{
|
|
long save_written = CPP_WRITTEN (pfile);
|
|
U_CHAR *p;
|
|
enum cpp_ttype token = _cpp_get_directive_token (pfile);
|
|
p = pfile->token_buffer + save_written;
|
|
|
|
if (token == CPP_NUMBER && p + 1 == CPP_PWRITTEN (pfile)
|
|
&& p[0] >= '1' && p[0] <= '4')
|
|
{
|
|
*num = p[0] - '0';
|
|
CPP_SET_WRITTEN (pfile, save_written);
|
|
return 1;
|
|
}
|
|
else
|
|
{
|
|
if (token != CPP_VSPACE && token != CPP_EOF)
|
|
cpp_error (pfile, "invalid format `#line' command");
|
|
CPP_SET_WRITTEN (pfile, save_written);
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
/* Interpret #line command.
|
|
Note that the filename string (if any) is treated as if it were an
|
|
include filename. That means no escape handling. */
|
|
|
|
static int
|
|
do_line (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
cpp_buffer *ip = CPP_BUFFER (pfile);
|
|
unsigned int new_lineno;
|
|
long old_written = CPP_WRITTEN (pfile);
|
|
enum cpp_ttype token;
|
|
char *x;
|
|
|
|
token = _cpp_get_directive_token (pfile);
|
|
|
|
if (token != CPP_NUMBER)
|
|
{
|
|
cpp_error (pfile, "token after `#line' is not an integer");
|
|
goto bad_line_directive;
|
|
}
|
|
|
|
CPP_PUTC (pfile, '\0'); /* not terminated for us */
|
|
new_lineno = strtoul ((const char *) (pfile->token_buffer + old_written),
|
|
&x, 10);
|
|
if (x[0] != '\0')
|
|
{
|
|
cpp_error (pfile, "token after `#line' is not an integer");
|
|
goto bad_line_directive;
|
|
}
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
|
|
if (CPP_PEDANTIC (pfile) && (new_lineno <= 0 || new_lineno > 32767))
|
|
cpp_pedwarn (pfile, "line number out of range in `#line' command");
|
|
|
|
token = _cpp_get_directive_token (pfile);
|
|
|
|
if (token == CPP_STRING)
|
|
{
|
|
U_CHAR *fname = pfile->token_buffer + old_written + 1;
|
|
U_CHAR *end_name = CPP_PWRITTEN (pfile) - 1;
|
|
int action_number = 0;
|
|
|
|
if (read_line_number (pfile, &action_number))
|
|
{
|
|
if (CPP_PEDANTIC (pfile))
|
|
cpp_pedwarn (pfile, "garbage at end of `#line' command");
|
|
|
|
/* This is somewhat questionable: change the buffer stack
|
|
depth so that output_line_command thinks we've stacked
|
|
another buffer. */
|
|
if (action_number == 1)
|
|
{
|
|
pfile->buffer_stack_depth++;
|
|
ip->system_header_p = 0;
|
|
read_line_number (pfile, &action_number);
|
|
}
|
|
else if (action_number == 2)
|
|
{
|
|
pfile->buffer_stack_depth--;
|
|
ip->system_header_p = 0;
|
|
read_line_number (pfile, &action_number);
|
|
}
|
|
if (action_number == 3)
|
|
{
|
|
ip->system_header_p = 1;
|
|
read_line_number (pfile, &action_number);
|
|
}
|
|
if (action_number == 4)
|
|
{
|
|
ip->system_header_p = 2;
|
|
read_line_number (pfile, &action_number);
|
|
}
|
|
}
|
|
|
|
*end_name = '\0';
|
|
|
|
if (strcmp ((const char *)fname, ip->nominal_fname))
|
|
{
|
|
if (!strcmp ((const char *)fname, ip->ihash->name))
|
|
ip->nominal_fname = ip->ihash->name;
|
|
else
|
|
ip->nominal_fname = _cpp_fake_ihash (pfile, (const char *)fname);
|
|
}
|
|
}
|
|
else if (token != CPP_VSPACE && token != CPP_EOF)
|
|
{
|
|
cpp_error (pfile, "token after `#line %d' is not a string", new_lineno);
|
|
goto bad_line_directive;
|
|
}
|
|
|
|
/* The Newline at the end of this line remains to be processed.
|
|
To put the next line at the specified line number,
|
|
we must store a line number now that is one less. */
|
|
ip->lineno = new_lineno - 1;
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
return 0;
|
|
|
|
bad_line_directive:
|
|
_cpp_skip_rest_of_line (pfile);
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
return 0;
|
|
}
|
|
|
|
/* Remove the definition of a symbol from the symbol table.
|
|
According to the C standard, it is not an error to undef
|
|
something that has no definitions. */
|
|
static int
|
|
do_undef (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
int len;
|
|
HASHNODE *hp;
|
|
U_CHAR *name;
|
|
long here = CPP_WRITTEN (pfile);
|
|
enum cpp_ttype token;
|
|
|
|
pfile->no_macro_expand++;
|
|
token = _cpp_get_directive_token (pfile);
|
|
pfile->no_macro_expand--;
|
|
|
|
if (token != CPP_NAME)
|
|
{
|
|
cpp_error (pfile, "token after #undef is not an identifier");
|
|
_cpp_skip_rest_of_line (pfile);
|
|
return 0;
|
|
}
|
|
len = CPP_WRITTEN (pfile) - here;
|
|
|
|
token = _cpp_get_directive_token (pfile);
|
|
if (token != CPP_VSPACE)
|
|
{
|
|
cpp_pedwarn (pfile, "junk on line after #undef");
|
|
_cpp_skip_rest_of_line (pfile);
|
|
}
|
|
|
|
name = pfile->token_buffer + here;
|
|
CPP_SET_WRITTEN (pfile, here);
|
|
|
|
hp = _cpp_lookup (pfile, name, len);
|
|
if (hp->type == T_VOID)
|
|
; /* Not defined in the first place - do nothing. */
|
|
else if (hp->type == T_POISON)
|
|
cpp_error (pfile, "cannot undefine poisoned \"%s\"", hp->name);
|
|
else
|
|
{
|
|
/* If we are generating additional info for debugging (with -g) we
|
|
need to pass through all effective #undef commands. */
|
|
if (CPP_OPTION (pfile, debug_output))
|
|
pass_thru_directive (hp->name, len, pfile, T_UNDEF);
|
|
|
|
if (hp->type != T_MACRO && hp->type != T_FMACRO
|
|
&& hp->type != T_EMPTY && hp->type != T_IDENTITY)
|
|
cpp_warning (pfile, "undefining `%s'", hp->name);
|
|
|
|
_cpp_free_definition (hp);
|
|
hp->type = T_VOID;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Report an error detected by the program we are processing.
|
|
* Use the text of the line in the error message.
|
|
* (We use error because it prints the filename & line#.)
|
|
*/
|
|
|
|
static int
|
|
do_error (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
const U_CHAR *text, *limit;
|
|
|
|
_cpp_skip_hspace (pfile);
|
|
text = CPP_BUFFER (pfile)->cur;
|
|
_cpp_skip_rest_of_line (pfile);
|
|
limit = CPP_BUFFER (pfile)->cur;
|
|
|
|
cpp_error (pfile, "#error %.*s", (int)(limit - text), text);
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Report a warning detected by the program we are processing.
|
|
* Use the text of the line in the warning message, then continue.
|
|
*/
|
|
|
|
static int
|
|
do_warning (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
const U_CHAR *text, *limit;
|
|
|
|
_cpp_skip_hspace (pfile);
|
|
text = CPP_BUFFER (pfile)->cur;
|
|
_cpp_skip_rest_of_line (pfile);
|
|
limit = CPP_BUFFER (pfile)->cur;
|
|
|
|
cpp_warning (pfile, "#warning %.*s", (int)(limit - text), text);
|
|
return 0;
|
|
}
|
|
|
|
/* Report program identification. */
|
|
|
|
static int
|
|
do_ident (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
long old_written = CPP_WRITTEN (pfile);
|
|
|
|
CPP_PUTS (pfile, "#ident ", 7);
|
|
|
|
/* Next token should be a string constant. */
|
|
if (_cpp_get_directive_token (pfile) == CPP_STRING)
|
|
/* And then a newline. */
|
|
if (_cpp_get_directive_token (pfile) == CPP_VSPACE)
|
|
/* Good - ship it. */
|
|
return 0;
|
|
|
|
cpp_error (pfile, "invalid #ident");
|
|
_cpp_skip_rest_of_line (pfile);
|
|
CPP_SET_WRITTEN (pfile, old_written); /* discard directive */
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* Pragmata handling. We handle some of these, and pass the rest on
|
|
to the front end. C99 defines three pragmas and says that no macro
|
|
expansion is to be performed on them; whether or not macro
|
|
expansion happens for other pragmas is implementation defined.
|
|
This implementation never macro-expands the text after #pragma.
|
|
|
|
We currently do not support the _Pragma operator. Support for that
|
|
has to be coordinated with the front end. Proposed implementation:
|
|
both #pragma blah blah and _Pragma("blah blah") become
|
|
__builtin_pragma(blah blah) and we teach the parser about that. */
|
|
|
|
/* Sub-handlers for the pragmas needing treatment here.
|
|
They return 1 if the token buffer is to be popped, 0 if not. */
|
|
static int do_pragma_once PARAMS ((cpp_reader *));
|
|
static int do_pragma_implementation PARAMS ((cpp_reader *));
|
|
static int do_pragma_poison PARAMS ((cpp_reader *));
|
|
static int do_pragma_default PARAMS ((cpp_reader *));
|
|
|
|
static int
|
|
do_pragma (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
long here, key;
|
|
U_CHAR *buf;
|
|
int pop;
|
|
enum cpp_ttype token;
|
|
|
|
here = CPP_WRITTEN (pfile);
|
|
CPP_PUTS (pfile, "#pragma ", 8);
|
|
|
|
key = CPP_WRITTEN (pfile);
|
|
pfile->no_macro_expand++;
|
|
token = _cpp_get_directive_token (pfile);
|
|
if (token != CPP_NAME)
|
|
{
|
|
if (token == CPP_VSPACE)
|
|
goto empty;
|
|
else
|
|
goto skip;
|
|
}
|
|
|
|
buf = pfile->token_buffer + key;
|
|
CPP_PUTC (pfile, ' ');
|
|
|
|
#define tokis(x) !strncmp((char *) buf, x, sizeof(x) - 1)
|
|
if (tokis ("once"))
|
|
pop = do_pragma_once (pfile);
|
|
else if (tokis ("implementation"))
|
|
pop = do_pragma_implementation (pfile);
|
|
else if (tokis ("poison"))
|
|
pop = do_pragma_poison (pfile);
|
|
else
|
|
pop = do_pragma_default (pfile);
|
|
#undef tokis
|
|
|
|
if (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
goto skip;
|
|
|
|
if (pop)
|
|
CPP_SET_WRITTEN (pfile, here);
|
|
pfile->no_macro_expand--;
|
|
return 0;
|
|
|
|
skip:
|
|
cpp_error (pfile, "malformed #pragma directive");
|
|
_cpp_skip_rest_of_line (pfile);
|
|
empty:
|
|
CPP_SET_WRITTEN (pfile, here);
|
|
pfile->no_macro_expand--;
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
do_pragma_default (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
while (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
CPP_PUTC (pfile, ' ');
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
do_pragma_once (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
cpp_buffer *ip = CPP_BUFFER (pfile);
|
|
|
|
/* Allow #pragma once in system headers, since that's not the user's
|
|
fault. */
|
|
if (!ip->system_header_p)
|
|
cpp_warning (pfile, "`#pragma once' is obsolete");
|
|
|
|
if (CPP_PREV_BUFFER (ip) == NULL)
|
|
cpp_warning (pfile, "`#pragma once' outside include file");
|
|
else
|
|
ip->ihash->control_macro = U""; /* never repeat */
|
|
|
|
return 1;
|
|
}
|
|
|
|
static int
|
|
do_pragma_implementation (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
/* Be quiet about `#pragma implementation' for a file only if it hasn't
|
|
been included yet. */
|
|
enum cpp_ttype token;
|
|
long written = CPP_WRITTEN (pfile);
|
|
U_CHAR *name;
|
|
char *copy;
|
|
size_t len;
|
|
|
|
token = _cpp_get_directive_token (pfile);
|
|
if (token == CPP_VSPACE)
|
|
return 0;
|
|
else if (token != CPP_STRING)
|
|
{
|
|
cpp_error (pfile, "malformed #pragma implementation");
|
|
return 1;
|
|
}
|
|
|
|
/* Trim the leading and trailing quote marks from the string. */
|
|
name = pfile->token_buffer + written + 1;
|
|
len = CPP_PWRITTEN (pfile) - name;
|
|
copy = alloca (len);
|
|
memcpy (copy, name, len - 1);
|
|
copy[len - 1] = '\0';
|
|
|
|
if (cpp_included (pfile, copy))
|
|
cpp_warning (pfile,
|
|
"`#pragma implementation' for `%s' appears after file is included",
|
|
copy);
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
do_pragma_poison (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
/* Poison these symbols so that all subsequent usage produces an
|
|
error message. */
|
|
U_CHAR *p;
|
|
HASHNODE *hp;
|
|
long written;
|
|
size_t len;
|
|
enum cpp_ttype token;
|
|
int writeit;
|
|
|
|
/* As a rule, don't include #pragma poison commands in output,
|
|
unless the user asks for them. */
|
|
writeit = (CPP_OPTION (pfile, debug_output)
|
|
|| CPP_OPTION (pfile, dump_macros) == dump_definitions
|
|
|| CPP_OPTION (pfile, dump_macros) == dump_names);
|
|
|
|
for (;;)
|
|
{
|
|
written = CPP_WRITTEN (pfile);
|
|
token = _cpp_get_directive_token (pfile);
|
|
if (token == CPP_VSPACE)
|
|
break;
|
|
if (token != CPP_NAME)
|
|
{
|
|
cpp_error (pfile, "invalid #pragma poison directive");
|
|
_cpp_skip_rest_of_line (pfile);
|
|
return 1;
|
|
}
|
|
|
|
p = pfile->token_buffer + written;
|
|
len = CPP_PWRITTEN (pfile) - p;
|
|
hp = _cpp_lookup (pfile, p, len);
|
|
if (hp->type == T_POISON)
|
|
; /* It is allowed to poison the same identifier twice. */
|
|
else
|
|
{
|
|
if (hp->type != T_VOID)
|
|
cpp_warning (pfile, "poisoning existing macro `%s'", hp->name);
|
|
_cpp_free_definition (hp);
|
|
hp->type = T_POISON;
|
|
}
|
|
if (writeit)
|
|
CPP_PUTC (pfile, ' ');
|
|
}
|
|
return !writeit;
|
|
}
|
|
|
|
/* Just ignore #sccs, on systems where we define it at all. */
|
|
#ifdef SCCS_DIRECTIVE
|
|
static int
|
|
do_sccs (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
_cpp_skip_rest_of_line (pfile);
|
|
return 0;
|
|
}
|
|
#endif
|
|
|
|
/* We've found an `#if' directive. If the only thing before it in
|
|
this file is white space, and if it is of the form
|
|
`#if ! defined SYMBOL', then SYMBOL is a possible controlling macro
|
|
for inclusion of this file. (See redundant_include_p in cppfiles.c
|
|
for an explanation of controlling macros.) If so, return a
|
|
malloced copy of SYMBOL. Otherwise, return NULL. */
|
|
|
|
static U_CHAR *
|
|
detect_if_not_defined (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
U_CHAR *control_macro = 0;
|
|
enum cpp_ttype token;
|
|
unsigned int base_offset;
|
|
unsigned int token_offset;
|
|
unsigned int need_rparen = 0;
|
|
unsigned int token_len;
|
|
|
|
if (pfile->only_seen_white != 2)
|
|
return NULL;
|
|
|
|
/* Save state required for restore. */
|
|
pfile->no_macro_expand++;
|
|
CPP_SET_MARK (pfile);
|
|
base_offset = CPP_WRITTEN (pfile);
|
|
|
|
/* Look for `!', */
|
|
if (_cpp_get_directive_token (pfile) != CPP_OTHER
|
|
|| CPP_WRITTEN (pfile) != (size_t) base_offset + 1
|
|
|| CPP_PWRITTEN (pfile)[-1] != '!')
|
|
goto restore;
|
|
|
|
/* ...then `defined', */
|
|
token_offset = CPP_WRITTEN (pfile);
|
|
token = _cpp_get_directive_token (pfile);
|
|
if (token != CPP_NAME)
|
|
goto restore;
|
|
if (ustrncmp (pfile->token_buffer + token_offset, U"defined", 7))
|
|
goto restore;
|
|
|
|
/* ...then an optional '(' and the name, */
|
|
token_offset = CPP_WRITTEN (pfile);
|
|
token = _cpp_get_directive_token (pfile);
|
|
if (token == CPP_OPEN_PAREN)
|
|
{
|
|
token_offset = CPP_WRITTEN (pfile);
|
|
need_rparen = 1;
|
|
token = _cpp_get_directive_token (pfile);
|
|
}
|
|
if (token != CPP_NAME)
|
|
goto restore;
|
|
|
|
token_len = CPP_WRITTEN (pfile) - token_offset;
|
|
|
|
/* ...then the ')', if necessary, */
|
|
if (need_rparen && _cpp_get_directive_token (pfile) != CPP_CLOSE_PAREN)
|
|
goto restore;
|
|
|
|
/* ...and make sure there's nothing else on the line. */
|
|
if (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
goto restore;
|
|
|
|
/* We have a legitimate controlling macro for this header. */
|
|
control_macro = (U_CHAR *) xmalloc (token_len + 1);
|
|
memcpy (control_macro, pfile->token_buffer + token_offset, token_len);
|
|
control_macro[token_len] = '\0';
|
|
|
|
restore:
|
|
CPP_SET_WRITTEN (pfile, base_offset);
|
|
pfile->no_macro_expand--;
|
|
CPP_GOTO_MARK (pfile);
|
|
|
|
return control_macro;
|
|
}
|
|
|
|
/*
|
|
* #if is straightforward; just call _cpp_parse_expr, then conditional_skip.
|
|
* Also, check for a reinclude preventer of the form #if !defined (MACRO).
|
|
*/
|
|
|
|
static int
|
|
do_if (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
U_CHAR *control_macro = detect_if_not_defined (pfile);
|
|
int value = _cpp_parse_expr (pfile);
|
|
return conditional_skip (pfile, value == 0, T_IF, control_macro);
|
|
}
|
|
|
|
/*
|
|
* handle a #elif directive by not changing if_stack either.
|
|
* see the comment above do_else.
|
|
*/
|
|
|
|
static int
|
|
do_elif (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
if (pfile->if_stack == CPP_BUFFER (pfile)->if_stack)
|
|
{
|
|
cpp_error (pfile, "`#elif' not within a conditional");
|
|
return 0;
|
|
}
|
|
else
|
|
{
|
|
if (pfile->if_stack->type == T_ELSE)
|
|
{
|
|
cpp_error (pfile, "`#elif' after `#else'");
|
|
cpp_error_with_line (pfile, pfile->if_stack->lineno, 0,
|
|
"the conditional began here");
|
|
}
|
|
pfile->if_stack->type = T_ELIF;
|
|
}
|
|
|
|
if (pfile->if_stack->if_succeeded)
|
|
{
|
|
_cpp_skip_rest_of_line (pfile);
|
|
return skip_if_group (pfile);
|
|
}
|
|
if (_cpp_parse_expr (pfile) == 0)
|
|
return skip_if_group (pfile);
|
|
|
|
++pfile->if_stack->if_succeeded; /* continue processing input */
|
|
return 0;
|
|
}
|
|
|
|
/* Parse an #ifdef or #ifndef directive. Returns 1 for defined, 0 for
|
|
not defined; the macro tested is left in the token buffer (but
|
|
popped). */
|
|
|
|
static int
|
|
parse_ifdef (pfile, name)
|
|
cpp_reader *pfile;
|
|
const U_CHAR *name;
|
|
{
|
|
U_CHAR *ident;
|
|
unsigned int len;
|
|
enum cpp_ttype token;
|
|
long old_written = CPP_WRITTEN (pfile);
|
|
int defined;
|
|
|
|
pfile->no_macro_expand++;
|
|
token = _cpp_get_directive_token (pfile);
|
|
pfile->no_macro_expand--;
|
|
|
|
ident = pfile->token_buffer + old_written;
|
|
len = CPP_WRITTEN (pfile) - old_written;
|
|
|
|
if (token == CPP_VSPACE)
|
|
{
|
|
if (! CPP_TRADITIONAL (pfile))
|
|
cpp_pedwarn (pfile, "`#%s' with no argument", name);
|
|
defined = 0;
|
|
goto done;
|
|
}
|
|
else if (token == CPP_NAME)
|
|
{
|
|
defined = cpp_defined (pfile, ident, len);
|
|
CPP_PUTC (pfile, '\0'); /* so it can be copied with xstrdup */
|
|
}
|
|
else
|
|
{
|
|
defined = 0;
|
|
if (! CPP_TRADITIONAL (pfile))
|
|
cpp_error (pfile, "`#%s' with invalid argument", name);
|
|
}
|
|
|
|
if (!CPP_TRADITIONAL (pfile))
|
|
{
|
|
if (_cpp_get_directive_token (pfile) == CPP_VSPACE)
|
|
goto done;
|
|
|
|
cpp_pedwarn (pfile, "garbage at end of `#%s' argument", name);
|
|
}
|
|
_cpp_skip_rest_of_line (pfile);
|
|
|
|
done:
|
|
CPP_SET_WRITTEN (pfile, old_written); /* Pop */
|
|
return defined;
|
|
}
|
|
|
|
/* #ifdef is dead simple. */
|
|
|
|
static int
|
|
do_ifdef (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
int skip = ! parse_ifdef (pfile, dtable[T_IFDEF].name);
|
|
return conditional_skip (pfile, skip, T_IFDEF, 0);
|
|
}
|
|
|
|
/* #ifndef is a tad more complex, because we need to check for a
|
|
no-reinclusion wrapper. */
|
|
|
|
static int
|
|
do_ifndef (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
int start_of_file, skip;
|
|
U_CHAR *control_macro = 0;
|
|
|
|
start_of_file = pfile->only_seen_white == 2;
|
|
skip = parse_ifdef (pfile, dtable[T_IFNDEF].name);
|
|
|
|
if (start_of_file && !skip)
|
|
control_macro = uxstrdup (CPP_PWRITTEN (pfile));
|
|
|
|
return conditional_skip (pfile, skip, T_IFNDEF, control_macro);
|
|
}
|
|
|
|
/* Push TYPE on stack; then, if SKIP is nonzero, skip ahead.
|
|
If this is a #ifndef starting at the beginning of a file,
|
|
CONTROL_MACRO is the macro name tested by the #ifndef.
|
|
Otherwise, CONTROL_MACRO is 0. */
|
|
|
|
static int
|
|
conditional_skip (pfile, skip, type, control_macro)
|
|
cpp_reader *pfile;
|
|
int skip;
|
|
int type;
|
|
U_CHAR *control_macro;
|
|
{
|
|
IF_STACK *temp;
|
|
|
|
temp = (IF_STACK *) xcalloc (1, sizeof (IF_STACK));
|
|
temp->lineno = CPP_BUFFER (pfile)->lineno;
|
|
temp->next = pfile->if_stack;
|
|
temp->control_macro = control_macro;
|
|
pfile->if_stack = temp;
|
|
|
|
pfile->if_stack->type = type;
|
|
|
|
if (skip != 0)
|
|
return skip_if_group (pfile);
|
|
|
|
++pfile->if_stack->if_succeeded;
|
|
return 0;
|
|
}
|
|
|
|
/* Subroutine of skip_if_group. Examine one preprocessing directive
|
|
and return 0 if skipping should continue, or the directive number
|
|
of the directive that ends the block if it should halt.
|
|
|
|
Also adjusts the if_stack as appropriate. The `#' has been read,
|
|
but not the identifier. */
|
|
|
|
static int
|
|
consider_directive_while_skipping (pfile, stack)
|
|
cpp_reader *pfile;
|
|
IF_STACK *stack;
|
|
{
|
|
long ident;
|
|
int i, hash_at_bol;
|
|
unsigned int len;
|
|
IF_STACK *temp;
|
|
|
|
/* -traditional directives are recognized only with the # in column 1. */
|
|
hash_at_bol = CPP_IN_COLUMN_1 (pfile);
|
|
|
|
ident = CPP_WRITTEN (pfile);
|
|
if (_cpp_get_directive_token (pfile) != CPP_NAME)
|
|
return 0;
|
|
len = CPP_WRITTEN (pfile) - ident;
|
|
|
|
for (i = 0; i < N_DIRECTIVES; i++)
|
|
{
|
|
if (dtable[i].length == len
|
|
&& !ustrncmp (dtable[i].name, pfile->token_buffer + ident, len))
|
|
goto real_directive;
|
|
}
|
|
return 0;
|
|
|
|
real_directive:
|
|
|
|
/* If it's not a directive of interest to us, return now. */
|
|
if (ORIGIN (dtable[i].flags) != COND)
|
|
return 0;
|
|
|
|
/* First, deal with -traditional and -Wtraditional.
|
|
All COND directives are from K+R. */
|
|
|
|
if (! hash_at_bol)
|
|
{
|
|
if (CPP_TRADITIONAL (pfile))
|
|
{
|
|
if (CPP_WTRADITIONAL (pfile))
|
|
cpp_warning (pfile, "ignoring #%s because of its indented #",
|
|
dtable[i].name);
|
|
return 0;
|
|
}
|
|
if (CPP_WTRADITIONAL (pfile))
|
|
cpp_warning (pfile, "traditional C ignores %s with the # indented",
|
|
dtable[i].name);
|
|
}
|
|
|
|
switch (i)
|
|
{
|
|
default:
|
|
cpp_ice (pfile, "non COND directive in switch in c_d_w_s");
|
|
return 0;
|
|
|
|
case T_IF:
|
|
case T_IFDEF:
|
|
case T_IFNDEF:
|
|
temp = (IF_STACK *) xcalloc (1, sizeof (IF_STACK));
|
|
temp->lineno = CPP_BUFFER (pfile)->lineno;
|
|
temp->next = pfile->if_stack;
|
|
temp->type = i;
|
|
pfile->if_stack = temp;
|
|
return 0;
|
|
|
|
case T_ELSE:
|
|
if (pfile->if_stack != stack)
|
|
validate_else (pfile, dtable[i].name);
|
|
/* fall through */
|
|
case T_ELIF:
|
|
if (pfile->if_stack == stack)
|
|
return i;
|
|
|
|
pfile->if_stack->type = i;
|
|
return 0;
|
|
|
|
case T_ENDIF:
|
|
if (pfile->if_stack != stack)
|
|
validate_else (pfile, dtable[i].name);
|
|
|
|
if (pfile->if_stack == stack)
|
|
return i;
|
|
|
|
temp = pfile->if_stack;
|
|
pfile->if_stack = temp->next;
|
|
free (temp);
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
/* Skip to #endif, #else, or #elif. Consumes the directive that
|
|
causes it to stop, but not its argument. Returns the number of
|
|
that directive, which must be passed back up to
|
|
_cpp_handle_directive, which will execute it. */
|
|
static int
|
|
skip_if_group (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
enum cpp_ttype token;
|
|
IF_STACK *save_if_stack = pfile->if_stack; /* don't pop past here */
|
|
long old_written;
|
|
int ret = 0;
|
|
|
|
/* We are no longer at the start of the file. */
|
|
pfile->only_seen_white = 0;
|
|
|
|
old_written = CPP_WRITTEN (pfile);
|
|
pfile->no_macro_expand++;
|
|
for (;;)
|
|
{
|
|
/* We are at the end of a line.
|
|
XXX Serious layering violation here. */
|
|
int c = CPP_BUF_PEEK (CPP_BUFFER (pfile));
|
|
if (c == EOF)
|
|
break; /* Caller will issue error. */
|
|
else if (c != '\n')
|
|
cpp_ice (pfile, "character %c at end of line in skip_if_group", c);
|
|
CPP_BUFFER (pfile)->cur++;
|
|
CPP_BUMP_LINE (pfile);
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
pfile->only_seen_white = 1;
|
|
|
|
token = _cpp_get_directive_token (pfile);
|
|
|
|
if (token == CPP_DIRECTIVE)
|
|
{
|
|
ret = consider_directive_while_skipping (pfile, save_if_stack);
|
|
if (ret)
|
|
break;
|
|
}
|
|
|
|
if (token != CPP_VSPACE)
|
|
_cpp_skip_rest_of_line (pfile);
|
|
}
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
pfile->no_macro_expand--;
|
|
return ret;
|
|
}
|
|
|
|
/*
|
|
* handle a #else directive. Do this by just continuing processing
|
|
* without changing if_stack ; this is so that the error message
|
|
* for missing #endif's etc. will point to the original #if. It
|
|
* is possible that something different would be better.
|
|
*/
|
|
|
|
static int
|
|
do_else (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
validate_else (pfile, dtable[T_ELSE].name);
|
|
_cpp_skip_rest_of_line (pfile);
|
|
|
|
if (pfile->if_stack == CPP_BUFFER (pfile)->if_stack)
|
|
{
|
|
cpp_error (pfile, "`#else' not within a conditional");
|
|
return 0;
|
|
}
|
|
else
|
|
{
|
|
/* #ifndef can't have its special treatment for containing the whole file
|
|
if it has a #else clause. */
|
|
pfile->if_stack->control_macro = 0;
|
|
|
|
if (pfile->if_stack->type == T_ELSE)
|
|
{
|
|
cpp_error (pfile, "`#else' after `#else'");
|
|
cpp_error_with_line (pfile, pfile->if_stack->lineno, 0,
|
|
"the conditional began here");
|
|
}
|
|
pfile->if_stack->type = T_ELSE;
|
|
}
|
|
|
|
if (pfile->if_stack->if_succeeded)
|
|
return skip_if_group (pfile);
|
|
|
|
++pfile->if_stack->if_succeeded; /* continue processing input */
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* unstack after #endif command
|
|
*/
|
|
|
|
static int
|
|
do_endif (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
validate_else (pfile, dtable[T_ENDIF].name);
|
|
_cpp_skip_rest_of_line (pfile);
|
|
|
|
if (pfile->if_stack == CPP_BUFFER (pfile)->if_stack)
|
|
cpp_error (pfile, "`#endif' not within a conditional");
|
|
else
|
|
{
|
|
IF_STACK *temp = pfile->if_stack;
|
|
pfile->if_stack = temp->next;
|
|
if (temp->control_macro != 0)
|
|
pfile->potential_control_macro = temp->control_macro;
|
|
free (temp);
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/* Issue -pedantic warning for text which is not a comment following
|
|
an #else or #endif. Do not warn in system headers, as this is harmless
|
|
and very common on old systems. */
|
|
|
|
static void
|
|
validate_else (pfile, directive)
|
|
cpp_reader *pfile;
|
|
const U_CHAR *directive;
|
|
{
|
|
long old_written;
|
|
if (! CPP_PEDANTIC (pfile))
|
|
return;
|
|
|
|
old_written = CPP_WRITTEN (pfile);
|
|
pfile->no_macro_expand++;
|
|
if (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
cpp_pedwarn (pfile,
|
|
"text following `#%s' violates ANSI standard", directive);
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
pfile->no_macro_expand--;
|
|
}
|
|
|
|
/* Called when we reach the end of a macro buffer. Walk back up the
|
|
conditional stack till we reach its level at entry to this file,
|
|
issuing error messages. */
|
|
void
|
|
_cpp_unwind_if_stack (pfile, pbuf)
|
|
cpp_reader *pfile;
|
|
cpp_buffer *pbuf;
|
|
{
|
|
struct if_stack *ifs, *nifs;
|
|
|
|
for (ifs = pfile->if_stack;
|
|
ifs != pbuf->if_stack;
|
|
ifs = nifs)
|
|
{
|
|
cpp_error_with_line (pfile, ifs->lineno, 0,
|
|
"unterminated `#%s' conditional",
|
|
dtable[ifs->type].name);
|
|
|
|
nifs = ifs->next;
|
|
free (ifs);
|
|
}
|
|
pfile->if_stack = ifs;
|
|
}
|
|
|
|
static int
|
|
do_assert (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
long old_written;
|
|
U_CHAR *sym;
|
|
int ret;
|
|
HASHNODE *base, *this;
|
|
size_t blen, tlen;
|
|
|
|
old_written = CPP_WRITTEN (pfile); /* remember where it starts */
|
|
ret = _cpp_parse_assertion (pfile);
|
|
if (ret == 0)
|
|
goto error;
|
|
else if (ret == 1)
|
|
{
|
|
cpp_error (pfile, "missing token-sequence in #assert");
|
|
goto error;
|
|
}
|
|
tlen = CPP_WRITTEN (pfile) - old_written;
|
|
|
|
if (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
{
|
|
cpp_error (pfile, "junk at end of #assert");
|
|
goto error;
|
|
}
|
|
sym = pfile->token_buffer + old_written;
|
|
|
|
this = _cpp_lookup (pfile, sym, tlen);
|
|
if (this->type == T_ASSERT)
|
|
{
|
|
cpp_warning (pfile, "%s re-asserted", sym);
|
|
goto error;
|
|
}
|
|
|
|
blen = ustrchr (sym, '(') - sym;
|
|
base = _cpp_lookup (pfile, sym, blen);
|
|
if (base->type == T_VOID)
|
|
{
|
|
base->type = T_ASSERT;
|
|
base->value.aschain = 0;
|
|
}
|
|
|
|
this->type = T_ASSERT;
|
|
this->value.aschain = base->value.aschain;
|
|
base->value.aschain = this;
|
|
|
|
error:
|
|
_cpp_skip_rest_of_line (pfile);
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
return 0;
|
|
}
|
|
|
|
static int
|
|
do_unassert (pfile)
|
|
cpp_reader *pfile;
|
|
{
|
|
int ret;
|
|
long old_written;
|
|
U_CHAR *sym;
|
|
long baselen, thislen;
|
|
HASHNODE *base, *this, *next;
|
|
|
|
old_written = CPP_WRITTEN (pfile);
|
|
ret = _cpp_parse_assertion (pfile);
|
|
if (ret == 0)
|
|
goto out;
|
|
thislen = CPP_WRITTEN (pfile) - old_written;
|
|
|
|
if (_cpp_get_directive_token (pfile) != CPP_VSPACE)
|
|
{
|
|
cpp_error (pfile, "junk at end of #unassert");
|
|
goto out;
|
|
}
|
|
sym = pfile->token_buffer + old_written;
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
|
|
if (ret == 1)
|
|
{
|
|
base = _cpp_lookup (pfile, sym, thislen);
|
|
if (base->type == T_VOID)
|
|
goto out; /* It isn't an error to #undef what isn't #defined,
|
|
so it isn't an error to #unassert what isn't
|
|
#asserted either. */
|
|
|
|
for (this = base->value.aschain; this; this = next)
|
|
{
|
|
next = this->value.aschain;
|
|
this->value.aschain = NULL;
|
|
this->type = T_VOID;
|
|
}
|
|
base->value.aschain = NULL;
|
|
base->type = T_VOID;
|
|
}
|
|
else
|
|
{
|
|
baselen = ustrchr (sym, '(') - sym;
|
|
base = _cpp_lookup (pfile, sym, baselen);
|
|
if (base->type == T_VOID) goto out;
|
|
this = _cpp_lookup (pfile, sym, thislen);
|
|
if (this->type == T_VOID) goto out;
|
|
|
|
next = base;
|
|
while (next->value.aschain != this)
|
|
next = next->value.aschain;
|
|
|
|
next->value.aschain = this->value.aschain;
|
|
this->value.aschain = NULL;
|
|
this->type = T_VOID;
|
|
|
|
if (base->value.aschain == NULL)
|
|
/* Last answer for this predicate deleted. */
|
|
base->type = T_VOID;
|
|
}
|
|
return 0;
|
|
|
|
out:
|
|
_cpp_skip_rest_of_line (pfile);
|
|
CPP_SET_WRITTEN (pfile, old_written);
|
|
return 0;
|
|
}
|
|
|
|
/* These are for -D, -U, -A. */
|
|
|
|
/* Process the string STR as if it appeared as the body of a #define.
|
|
If STR is just an identifier, define it with value 1.
|
|
If STR has anything after the identifier, then it should
|
|
be identifier=definition. */
|
|
|
|
void
|
|
cpp_define (pfile, str)
|
|
cpp_reader *pfile;
|
|
const char *str;
|
|
{
|
|
char *buf, *p;
|
|
size_t count;
|
|
|
|
p = strchr (str, '=');
|
|
/* Copy the entire option so we can modify it.
|
|
Change the first "=" in the string to a space. If there is none,
|
|
tack " 1" on the end. Then add a newline and a NUL. */
|
|
|
|
if (p)
|
|
{
|
|
count = strlen (str) + 2;
|
|
buf = (char *) alloca (count);
|
|
memcpy (buf, str, count - 2);
|
|
buf[p - str] = ' ';
|
|
buf[count - 2] = '\n';
|
|
buf[count - 1] = '\0';
|
|
}
|
|
else
|
|
{
|
|
count = strlen (str) + 4;
|
|
buf = (char *) alloca (count);
|
|
memcpy (buf, str, count - 4);
|
|
strcpy (&buf[count-4], " 1\n");
|
|
}
|
|
|
|
if (cpp_push_buffer (pfile, (U_CHAR *)buf, count - 1) != NULL)
|
|
{
|
|
do_define (pfile);
|
|
cpp_pop_buffer (pfile);
|
|
}
|
|
}
|
|
|
|
/* Process MACRO as if it appeared as the body of an #undef. */
|
|
void
|
|
cpp_undef (pfile, macro)
|
|
cpp_reader *pfile;
|
|
const char *macro;
|
|
{
|
|
/* Copy the string so we can append a newline. */
|
|
size_t len = strlen (macro);
|
|
char *buf = (char *) alloca (len + 2);
|
|
memcpy (buf, macro, len);
|
|
buf[len] = '\n';
|
|
buf[len + 1] = '\0';
|
|
if (cpp_push_buffer (pfile, (U_CHAR *)buf, len + 1) != NULL)
|
|
{
|
|
do_undef (pfile);
|
|
cpp_pop_buffer (pfile);
|
|
}
|
|
}
|
|
|
|
/* Process the string STR as if it appeared as the body of a #assert. */
|
|
void
|
|
cpp_assert (pfile, str)
|
|
cpp_reader *pfile;
|
|
const char *str;
|
|
{
|
|
if (cpp_push_buffer (pfile, (const U_CHAR *)str, strlen (str)) != NULL)
|
|
{
|
|
do_assert (pfile);
|
|
cpp_pop_buffer (pfile);
|
|
}
|
|
}
|
|
|
|
/* Process STR as if it appeared as the body of an #unassert. */
|
|
void
|
|
cpp_unassert (pfile, str)
|
|
cpp_reader *pfile;
|
|
const char *str;
|
|
{
|
|
if (cpp_push_buffer (pfile, (const U_CHAR *)str, strlen (str)) != NULL)
|
|
{
|
|
do_unassert (pfile);
|
|
cpp_pop_buffer (pfile);
|
|
}
|
|
}
|
|
|
|
/* Determine whether the identifier ID, of length LEN, is a defined macro. */
|
|
int
|
|
cpp_defined (pfile, id, len)
|
|
cpp_reader *pfile;
|
|
const U_CHAR *id;
|
|
int len;
|
|
{
|
|
HASHNODE *hp = _cpp_lookup (pfile, id, len);
|
|
if (hp->type == T_POISON)
|
|
{
|
|
cpp_error (pfile, "attempt to use poisoned `%s'", hp->name);
|
|
return 0;
|
|
}
|
|
return (hp->type != T_VOID);
|
|
}
|