2 * Do C preprocessing, based on a token list gathered by
5 * This may not be the smartest preprocessor on the planet.
7 * Copyright (C) 2003 Transmeta Corp.
8 * 2003-2004 Linus Torvalds
10 * Licensed under the Open Software License version 1.1
22 #include "pre-process.h"
28 #include "expression.h"
30 static int true_nesting
= 0;
31 static int false_nesting
= 0;
32 static struct token
*unmatched_if
= NULL
;
33 #define if_nesting (true_nesting + false_nesting)
35 #define MAX_NEST (256)
36 static unsigned char elif_ignore
[MAX_NEST
];
37 enum { ELIF_IGNORE
= 1, ELIF_SEEN_ELSE
= 2 };
39 #define INCLUDEPATHS 300
40 const char *includepath
[INCLUDEPATHS
+1] = {
48 static const char **quote_includepath
= includepath
;
49 static const char **angle_includepath
= includepath
+ 1;
50 static const char **sys_includepath
= includepath
+ 1;
52 #define MARK_STREAM_NONCONST(pos) do { \
53 if (stream->constant != CONSTANT_FILE_NOPE) { \
55 info(pos, "%s triggers non-const", __func__); \
56 stream->constant = CONSTANT_FILE_NOPE; \
61 static struct token
*alloc_token(struct position
*pos
)
63 struct token
*token
= __alloc_token(0);
65 token
->pos
.stream
= pos
->stream
;
66 token
->pos
.line
= pos
->line
;
67 token
->pos
.pos
= pos
->pos
;
68 token
->pos
.whitespace
= 1;
72 static const char *show_token_sequence(struct token
*token
);
74 /* Expand symbol 'sym' at '*list' */
75 static int expand(struct token
**, struct symbol
*);
77 static void replace_with_string(struct token
*token
, const char *str
)
79 int size
= strlen(str
) + 1;
80 struct string
*s
= __alloc_string(size
);
83 memcpy(s
->data
, str
, size
);
84 token_type(token
) = TOKEN_STRING
;
88 static void replace_with_integer(struct token
*token
, unsigned int val
)
90 char *buf
= __alloc_bytes(11);
91 sprintf(buf
, "%u", val
);
92 token_type(token
) = TOKEN_NUMBER
;
96 static int token_defined(struct token
*token
)
98 if (token_type(token
) == TOKEN_IDENT
) {
99 struct symbol
*sym
= lookup_symbol(token
->ident
, NS_MACRO
);
107 warning(token
->pos
, "expected preprocessor identifier");
111 static void replace_with_defined(struct token
*token
)
113 static const char *string
[] = { "0", "1" };
114 int defined
= token_defined(token
);
116 token_type(token
) = TOKEN_NUMBER
;
117 token
->number
= string
[defined
];
120 static int expand_one_symbol(struct token
**list
)
122 struct token
*token
= *list
;
125 if (token
->pos
.noexpand
)
128 sym
= lookup_symbol(token
->ident
, NS_MACRO
);
131 return expand(list
, sym
);
133 if (token
->ident
== &__LINE___ident
) {
134 replace_with_integer(token
, token
->pos
.line
);
135 } else if (token
->ident
== &__FILE___ident
) {
136 replace_with_string(token
, stream_name(token
->pos
.stream
));
141 static inline struct token
*scan_next(struct token
**where
)
143 struct token
*token
= *where
;
144 if (token_type(token
) != TOKEN_UNTAINT
)
147 token
->ident
->tainted
= 0;
149 } while (token_type(token
) == TOKEN_UNTAINT
);
154 static void expand_list(struct token
**list
)
157 while (!eof_token(next
= scan_next(list
))) {
158 if (token_type(next
) != TOKEN_IDENT
|| expand_one_symbol(list
))
163 static struct token
*collect_arg(struct token
*prev
, int vararg
, struct position
*pos
)
165 struct token
**p
= &prev
->next
;
169 while (!eof_token(next
= scan_next(p
))) {
170 if (match_op(next
, '(')) {
172 } else if (match_op(next
, ')')) {
175 } else if (match_op(next
, ',') && !nesting
&& !vararg
) {
178 next
->pos
.stream
= pos
->stream
;
179 next
->pos
.line
= pos
->line
;
180 next
->pos
.pos
= pos
->pos
;
183 *p
= &eof_token_entry
;
188 * We store arglist as <counter> [arg1] <number of uses for arg1> ... eof
193 struct token
*expanded
;
200 static int collect_arguments(struct token
*start
, struct token
*arglist
, struct arg
*args
, struct token
*what
)
202 int wanted
= arglist
->count
.normal
;
203 struct token
*next
= NULL
;
206 arglist
= arglist
->next
; /* skip counter */
209 next
= collect_arg(start
, 0, &what
->pos
);
212 if (!eof_token(start
->next
) || !match_op(next
, ')')) {
217 for (count
= 0; count
< wanted
; count
++) {
218 struct argcount
*p
= &arglist
->next
->count
;
219 next
= collect_arg(start
, p
->vararg
, &what
->pos
);
220 arglist
= arglist
->next
->next
;
223 args
[count
].arg
= start
->next
;
224 args
[count
].n_normal
= p
->normal
;
225 args
[count
].n_quoted
= p
->quoted
;
226 args
[count
].n_str
= p
->str
;
227 if (match_op(next
, ')')) {
233 if (count
== wanted
&& !match_op(next
, ')'))
235 if (count
== wanted
- 1) {
236 struct argcount
*p
= &arglist
->next
->count
;
239 args
[count
].arg
= NULL
;
240 args
[count
].n_normal
= p
->normal
;
241 args
[count
].n_quoted
= p
->quoted
;
242 args
[count
].n_str
= p
->str
;
244 if (count
< wanted
- 1)
247 what
->next
= next
->next
;
251 warning(what
->pos
, "macro \"%s\" requires %d arguments, but only %d given",
252 show_token(what
), wanted
, count
);
255 while (match_op(next
, ',')) {
256 next
= collect_arg(next
, 0, &what
->pos
);
261 warning(what
->pos
, "macro \"%s\" passed %d arguments, but takes just %d",
262 show_token(what
), count
, wanted
);
265 warning(what
->pos
, "unterminated argument list invoking macro \"%s\"",
268 what
->next
= next
->next
;
272 static struct token
*dup_list(struct token
*list
)
275 struct token
**p
= &res
;
277 while (!eof_token(list
)) {
278 struct token
*newtok
= __alloc_token(0);
287 static struct token
*stringify(struct token
*arg
)
289 const char *s
= show_token_sequence(arg
);
290 int size
= strlen(s
)+1;
291 struct token
*token
= __alloc_token(0);
292 struct string
*string
= __alloc_string(size
);
294 memcpy(string
->data
, s
, size
);
295 string
->length
= size
;
296 token
->pos
= arg
->pos
;
297 token_type(token
) = TOKEN_STRING
;
298 token
->string
= string
;
299 token
->next
= &eof_token_entry
;
303 static void expand_arguments(int count
, struct arg
*args
)
306 for (i
= 0; i
< count
; i
++) {
307 struct token
*arg
= args
[i
].arg
;
309 arg
= &eof_token_entry
;
311 args
[i
].str
= stringify(arg
);
312 if (args
[i
].n_normal
) {
313 if (!args
[i
].n_quoted
) {
314 args
[i
].expanded
= arg
;
316 } else if (eof_token(arg
)) {
317 args
[i
].expanded
= arg
;
319 args
[i
].expanded
= dup_list(arg
);
321 expand_list(&args
[i
].expanded
);
327 * Possibly valid combinations:
328 * - ident + ident -> ident
329 * - ident + number -> ident unless number contains '.', '+' or '-'.
330 * - number + number -> number
331 * - number + ident -> number
332 * - number + '.' -> number
333 * - number + '+' or '-' -> number, if number used to end on [eEpP].
334 * - '.' + number -> number, if number used to start with a digit.
335 * - special + special -> either special or an error.
337 static enum token_type
combine(struct token
*left
, struct token
*right
, char *p
)
340 enum token_type t1
= token_type(left
), t2
= token_type(right
);
342 if (t1
!= TOKEN_IDENT
&& t1
!= TOKEN_NUMBER
&& t1
!= TOKEN_SPECIAL
)
345 if (t2
!= TOKEN_IDENT
&& t2
!= TOKEN_NUMBER
&& t2
!= TOKEN_SPECIAL
)
348 strcpy(p
, show_token(left
));
349 strcat(p
, show_token(right
));
355 if (t1
== TOKEN_IDENT
) {
356 if (t2
== TOKEN_SPECIAL
)
358 if (t2
== TOKEN_NUMBER
&& strpbrk(p
, "+-."))
363 if (t1
== TOKEN_NUMBER
) {
364 if (t2
== TOKEN_SPECIAL
) {
365 switch (right
->special
) {
369 if (strchr("eEpP", p
[len
- 2]))
378 if (p
[0] == '.' && isdigit((unsigned char)p
[1]))
381 return TOKEN_SPECIAL
;
384 static int merge(struct token
*left
, struct token
*right
)
386 extern unsigned char combinations
[][3];
387 static char buffer
[512];
390 switch (combine(left
, right
, buffer
)) {
392 left
->ident
= built_in_ident(buffer
);
393 left
->pos
.noexpand
= 0;
397 char *number
= __alloc_bytes(strlen(buffer
) + 1);
398 memcpy(number
, buffer
, strlen(buffer
) + 1);
399 token_type(left
) = TOKEN_NUMBER
; /* could be . + num */
400 left
->number
= number
;
405 if (buffer
[2] && buffer
[3])
407 for (n
= SPECIAL_BASE
; n
< SPECIAL_ARG_SEPARATOR
; n
++) {
408 if (!memcmp(buffer
, combinations
[n
-SPECIAL_BASE
], 3)) {
416 warning(left
->pos
, "'##' failed: concatenation is not a valid token");
420 static struct token
*dup_token(struct token
*token
, struct position
*streampos
, struct position
*pos
)
422 struct token
*alloc
= alloc_token(streampos
);
423 token_type(alloc
) = token_type(token
);
424 alloc
->pos
.newline
= pos
->newline
;
425 alloc
->pos
.whitespace
= pos
->whitespace
;
426 alloc
->number
= token
->number
;
427 alloc
->pos
.noexpand
= token
->pos
.noexpand
;
431 static struct token
**copy(struct token
**where
, struct token
*list
, int *count
)
433 int need_copy
= --*count
;
434 while (!eof_token(list
)) {
437 token
= dup_token(list
, &list
->pos
, &list
->pos
);
440 if (token_type(token
) == TOKEN_IDENT
&& token
->ident
->tainted
)
441 token
->pos
.noexpand
= 1;
443 where
= &token
->next
;
446 *where
= &eof_token_entry
;
450 static struct token
**substitute(struct token
**list
, struct token
*body
, struct arg
*args
)
452 struct token
*token
= *list
;
453 struct position
*base_pos
= &token
->pos
;
454 struct position
*pos
= base_pos
;
456 enum {Normal
, Placeholder
, Concat
} state
= Normal
;
458 for (; !eof_token(body
); body
= body
->next
, pos
= &body
->pos
) {
459 struct token
*added
, *arg
;
462 switch (token_type(body
)) {
463 case TOKEN_GNU_KLUDGE
:
465 * GNU kludge: if we had <comma>##<vararg>, behaviour
466 * depends on whether we had enough arguments to have
467 * a vararg. If we did, ## is just ignored. Otherwise
468 * both , and ## are ignored. Comma should come from
469 * the body of macro and not be an argument of earlier
472 if (!args
[body
->next
->argnum
].arg
)
474 added
= dup_token(body
, base_pos
, pos
);
475 token_type(added
) = TOKEN_SPECIAL
;
479 case TOKEN_STR_ARGUMENT
:
480 arg
= args
[body
->argnum
].str
;
481 count
= &args
[body
->argnum
].n_str
;
484 case TOKEN_QUOTED_ARGUMENT
:
485 arg
= args
[body
->argnum
].arg
;
486 count
= &args
[body
->argnum
].n_quoted
;
487 if (!arg
|| eof_token(arg
)) {
496 case TOKEN_MACRO_ARGUMENT
:
497 arg
= args
[body
->argnum
].expanded
;
498 count
= &args
[body
->argnum
].n_normal
;
499 if (eof_token(arg
)) {
504 tail
= copy(&added
, arg
, count
);
505 added
->pos
.newline
= pos
->newline
;
506 added
->pos
.whitespace
= pos
->whitespace
;
510 if (state
== Placeholder
)
517 added
= dup_token(body
, base_pos
, pos
);
518 if (added
->ident
->tainted
)
519 added
->pos
.noexpand
= 1;
524 added
= dup_token(body
, base_pos
, pos
);
530 * if we got to doing real concatenation, we already have
531 * added something into the list, so containing_token() is OK.
533 if (state
== Concat
&& merge(containing_token(list
), added
)) {
535 if (tail
!= &added
->next
)
543 *list
= &eof_token_entry
;
547 static int expand(struct token
**list
, struct symbol
*sym
)
550 struct token
*token
= *list
;
551 struct ident
*expanding
= token
->ident
;
553 int nargs
= sym
->arglist
? sym
->arglist
->count
.normal
: 0;
554 struct arg args
[nargs
];
556 if (expanding
->tainted
) {
557 token
->pos
.noexpand
= 1;
562 if (!match_op(scan_next(&token
->next
), '('))
564 if (!collect_arguments(token
->next
, sym
->arglist
, args
, token
))
566 expand_arguments(nargs
, args
);
569 expanding
->tainted
= 1;
572 tail
= substitute(list
, sym
->expansion
, args
);
578 static const char *token_name_sequence(struct token
*token
, int endop
, struct token
*start
)
581 static char buffer
[256];
585 while (!eof_token(token
) && !match_op(token
, endop
)) {
587 const char *val
= token
->string
->data
;
588 if (token_type(token
) != TOKEN_STRING
)
589 val
= show_token(token
);
591 memcpy(ptr
, val
, len
);
596 if (endop
&& !match_op(token
, endop
))
597 warning(start
->pos
, "expected '>' at end of filename");
601 static int already_tokenized(const char *path
)
604 struct stream
*s
= input_streams
;
606 for (i
= input_stream_nr
; --i
>= 0; s
++) {
607 if (s
->constant
!= CONSTANT_FILE_YES
)
609 if (strcmp(path
, s
->name
))
611 if (!lookup_symbol(s
->protect
, NS_MACRO
))
618 /* Hande include of header files.
619 * The relevant options are made compatible with gcc. The only options that
620 * are not supported is -withprefix and friends.
622 * Three set of include paths are known:
623 * quote_includepath: Path to search when using #include "file.h"
624 * angle_includepath: Path to search when using #include <file.h>
625 * sys_includepath: Built-in include paths
627 * The above is implmented as one array with pointes
629 * quote_includepath ---> | |
633 * angle_includepath ---> | |
635 * sys_includepath ---> | |
640 * -I dir insert dir just before sys_includepath and move the rest
641 * -I- makes all dirs specified with -I before to quote dirs only and
642 * angle_includepath is set equal to sys_includepath.
643 * -nostdinc removes all sys dirs be storing NULL in entry pointed
644 * to by * sys_includepath. Note this will reset all dirs built-in and added
645 * before -nostdinc by -isystem and -dirafter
646 * -isystem dir adds dir where sys_includepath points adding this dir as
648 * -dirafter dir adds dir to the end of the list
651 static void set_stream_include_path(struct stream
*stream
)
653 const char *path
= stream
->path
;
655 const char *p
= strrchr(stream
->name
, '/');
658 int len
= p
- stream
->name
+ 1;
659 char *m
= malloc(len
+1);
660 /* This includes the final "/" */
661 memcpy(m
, stream
->name
, len
);
667 includepath
[0] = path
;
670 static int try_include(const char *path
, const char *filename
, int flen
, struct token
**where
, const char **next_path
)
673 int plen
= strlen(path
);
674 static char fullname
[PATH_MAX
];
676 memcpy(fullname
, path
, plen
);
677 if (plen
&& path
[plen
-1] != '/') {
678 fullname
[plen
] = '/';
681 memcpy(fullname
+plen
, filename
, flen
);
682 if (already_tokenized(fullname
))
684 fd
= open(fullname
, O_RDONLY
);
686 char * streamname
= __alloc_bytes(plen
+ flen
);
687 memcpy(streamname
, fullname
, plen
+ flen
);
688 *where
= tokenize(streamname
, fd
, *where
, next_path
);
695 static int do_include_path(const char **pptr
, struct token
**list
, struct token
*token
, const char *filename
, int flen
)
699 while ((path
= *pptr
++) != NULL
) {
700 if (!try_include(path
, filename
, flen
, list
, pptr
))
707 static void do_include(int local
, struct stream
*stream
, struct token
**list
, struct token
*token
, const char *filename
, const char **path
)
709 int flen
= strlen(filename
) + 1;
712 if (filename
[0] == '/') {
713 if (try_include("", filename
, flen
, list
, includepath
))
718 /* Dir of inputfile is first dir to search for quoted includes */
719 set_stream_include_path(stream
);
722 /* Do not search quote include if <> is in use */
723 path
= local
? quote_includepath
: angle_includepath
;
725 /* Check the standard include paths.. */
726 if (do_include_path(path
, list
, token
, filename
, flen
))
729 error_die(token
->pos
, "unable to open '%s'", filename
);
732 static int free_preprocessor_line(struct token
*token
)
735 struct token
*free
= token
;
738 } while (token_type(token
) != TOKEN_EOF
);
742 static int handle_include_path(struct stream
*stream
, struct token
**list
, struct token
*token
, const char **path
)
744 const char *filename
;
749 return free_preprocessor_line(token
);
751 if (stream
->constant
== CONSTANT_FILE_MAYBE
)
752 MARK_STREAM_NONCONST(token
->pos
);
756 if (!match_op(next
, '<')) {
757 expand_list(&token
->next
);
760 if (match_op(token
->next
, '<')) {
766 filename
= token_name_sequence(token
, expect
, token
);
767 do_include(!expect
, stream
, list
, token
, filename
, path
);
771 static int handle_include(struct stream
*stream
, struct token
**list
, struct token
*token
)
773 return handle_include_path(stream
, list
, token
, NULL
);
776 static int handle_include_next(struct stream
*stream
, struct token
**list
, struct token
*token
)
778 return handle_include_path(stream
, list
, token
, stream
->next_path
);
781 static int token_different(struct token
*t1
, struct token
*t2
)
785 if (token_type(t1
) != token_type(t2
))
788 switch (token_type(t1
)) {
790 different
= t1
->ident
!= t2
->ident
;
792 case TOKEN_ARG_COUNT
:
795 case TOKEN_GNU_KLUDGE
:
799 different
= strcmp(t1
->number
, t2
->number
);
802 different
= t1
->special
!= t2
->special
;
804 case TOKEN_MACRO_ARGUMENT
:
805 case TOKEN_QUOTED_ARGUMENT
:
806 case TOKEN_STR_ARGUMENT
:
807 different
= t1
->argnum
!= t2
->argnum
;
810 different
= t1
->character
!= t2
->character
;
813 struct string
*s1
, *s2
;
818 if (s1
->length
!= s2
->length
)
820 different
= memcmp(s1
->data
, s2
->data
, s1
->length
);
830 static int token_list_different(struct token
*list1
, struct token
*list2
)
835 if (!list1
|| !list2
)
837 if (token_different(list1
, list2
))
844 static inline void set_arg_count(struct token
*token
)
846 token_type(token
) = TOKEN_ARG_COUNT
;
847 token
->count
.normal
= token
->count
.quoted
=
848 token
->count
.str
= token
->count
.vararg
= 0;
851 static struct token
*parse_arguments(struct token
*list
)
853 struct token
*arg
= list
->next
, *next
= list
;
854 struct argcount
*count
= &list
->count
;
858 if (match_op(arg
, ')')) {
860 list
->next
= &eof_token_entry
;
864 while (token_type(arg
) == TOKEN_IDENT
) {
865 if (arg
->ident
== &__VA_ARGS___ident
)
867 if (!++count
->normal
)
871 if (match_op(next
, ',')) {
877 if (match_op(next
, ')')) {
880 arg
->next
->next
= &eof_token_entry
;
884 /* normal cases are finished here */
886 if (match_op(next
, SPECIAL_ELLIPSIS
)) {
887 if (match_op(next
->next
, ')')) {
889 next
->count
.vararg
= 1;
891 arg
->next
->next
= &eof_token_entry
;
899 if (eof_token(next
)) {
907 if (match_op(arg
, SPECIAL_ELLIPSIS
)) {
909 token_type(arg
) = TOKEN_IDENT
;
910 arg
->ident
= &__VA_ARGS___ident
;
911 if (!match_op(next
, ')'))
913 if (!++count
->normal
)
916 next
->count
.vararg
= 1;
918 arg
->next
->next
= &eof_token_entry
;
922 if (eof_token(arg
)) {
926 if (match_op(arg
, ','))
933 warning(arg
->pos
, "parameter name missing");
936 warning(arg
->pos
, "\"%s\" may not appear in macro parameter list",
940 warning(arg
->pos
, "missing ')' in macro parameter list");
943 warning(arg
->pos
, "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
946 warning(arg
->pos
, "too many arguments in macro definition");
950 static int try_arg(struct token
*token
, enum token_type type
, struct token
*arglist
)
952 struct ident
*ident
= token
->ident
;
955 if (!arglist
|| token_type(token
) != TOKEN_IDENT
)
958 arglist
= arglist
->next
;
960 for (nr
= 0; !eof_token(arglist
); nr
++, arglist
= arglist
->next
->next
) {
961 if (arglist
->ident
== ident
) {
962 struct argcount
*count
= &arglist
->next
->count
;
966 token_type(token
) = type
;
968 case TOKEN_MACRO_ARGUMENT
:
971 case TOKEN_QUOTED_ARGUMENT
:
978 return count
->vararg
? 2 : 1;
979 token_type(token
) = TOKEN_ERROR
;
986 static struct token
*parse_expansion(struct token
*expansion
, struct token
*arglist
, struct ident
*name
)
988 struct token
*token
= expansion
;
990 struct token
*last
= NULL
;
992 if (match_op(token
, SPECIAL_HASHHASH
))
995 for (p
= &expansion
; !eof_token(token
); p
= &token
->next
, token
= *p
) {
996 if (match_op(token
, '#')) {
998 struct token
*next
= token
->next
;
999 if (!try_arg(next
, TOKEN_STR_ARGUMENT
, arglist
))
1001 next
->pos
.whitespace
= token
->pos
.whitespace
;
1004 token
->pos
.noexpand
= 1;
1006 } else if (match_op(token
, SPECIAL_HASHHASH
)) {
1007 struct token
*next
= token
->next
;
1008 int arg
= try_arg(next
, TOKEN_QUOTED_ARGUMENT
, arglist
);
1009 token_type(token
) = TOKEN_CONCAT
;
1013 if (arg
== 2 && last
&& match_op(last
, ',')) {
1014 token_type(last
) = TOKEN_GNU_KLUDGE
;
1017 } else if (match_op(next
, SPECIAL_HASHHASH
))
1019 else if (match_op(next
, ','))
1021 else if (eof_token(next
))
1023 } else if (match_op(token
->next
, SPECIAL_HASHHASH
)) {
1024 try_arg(token
, TOKEN_QUOTED_ARGUMENT
, arglist
);
1026 try_arg(token
, TOKEN_MACRO_ARGUMENT
, arglist
);
1028 if (token_type(token
) == TOKEN_ERROR
)
1032 token
= alloc_token(&expansion
->pos
);
1033 token_type(token
) = TOKEN_UNTAINT
;
1034 token
->ident
= name
;
1040 warning(token
->pos
, "'#' is not followed by a macro parameter");
1044 warning(token
->pos
, "'##' cannot appear at the ends of macro expansion");
1047 warning(token
->pos
, "too many instances of argument in body");
1051 static int do_handle_define(struct stream
*stream
, struct token
**line
, struct token
*token
, int weak
)
1053 struct token
*arglist
, *expansion
;
1054 struct token
*left
= token
->next
;
1058 if (token_type(left
) != TOKEN_IDENT
) {
1059 warning(token
->pos
, "expected identifier to 'define'");
1063 return free_preprocessor_line(token
);
1065 if (stream
->constant
== CONSTANT_FILE_MAYBE
)
1066 MARK_STREAM_NONCONST(token
->pos
);
1068 __free_token(token
); /* Free the "define" token, but not the rest of the line */
1072 expansion
= left
->next
;
1073 if (!expansion
->pos
.whitespace
&& match_op(expansion
, '(')) {
1074 arglist
= expansion
;
1075 expansion
= parse_arguments(expansion
);
1080 expansion
= parse_expansion(expansion
, arglist
, name
);
1084 sym
= lookup_symbol(name
, NS_MACRO
);
1086 if (token_list_different(sym
->expansion
, expansion
) ||
1087 token_list_different(sym
->arglist
, arglist
)) {
1092 warning(left
->pos
, "preprocessor token %.*s redefined",
1093 name
->len
, name
->name
);
1094 info(sym
->pos
, "this was the original definition");
1095 sym
->expansion
= expansion
;
1096 sym
->arglist
= arglist
;
1100 sym
= alloc_symbol(left
->pos
, SYM_NODE
);
1101 bind_symbol(sym
, name
, NS_MACRO
);
1104 sym
->expansion
= expansion
;
1105 sym
->arglist
= arglist
;
1110 static int handle_define(struct stream
*stream
, struct token
**line
, struct token
*token
)
1112 return do_handle_define(stream
, line
, token
, 0);
1115 static int handle_weak_define(struct stream
*stream
, struct token
**line
, struct token
*token
)
1117 return do_handle_define(stream
, line
, token
, 1);
1120 static int handle_undef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1122 struct token
*left
= token
->next
;
1123 struct symbol
**sym
;
1125 if (token_type(left
) != TOKEN_IDENT
) {
1126 warning(token
->pos
, "expected identifier to 'undef'");
1130 return free_preprocessor_line(token
);
1132 if (stream
->constant
== CONSTANT_FILE_MAYBE
)
1133 MARK_STREAM_NONCONST(token
->pos
);
1135 sym
= &left
->ident
->symbols
;
1137 struct symbol
*t
= *sym
;
1138 if (t
->namespace == NS_MACRO
) {
1144 return free_preprocessor_line(token
);
1147 static int preprocessor_if(struct token
*token
, int true)
1149 if (if_nesting
== 0)
1150 unmatched_if
= token
;
1151 if (if_nesting
>= MAX_NEST
)
1152 error_die(token
->pos
, "Maximum preprocessor conditional level exhausted");
1153 elif_ignore
[if_nesting
] = (false_nesting
|| true) ? ELIF_IGNORE
: 0;
1154 if (false_nesting
|| !true) {
1156 return free_preprocessor_line(token
);
1159 return free_preprocessor_line(token
);
1162 static int handle_ifdef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1164 return preprocessor_if(token
, token_defined(token
->next
));
1167 static int handle_ifndef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1169 struct token
*next
= token
->next
;
1170 if (stream
->constant
== CONSTANT_FILE_MAYBE
) {
1171 if (token_type(next
) == TOKEN_IDENT
&&
1172 (!stream
->protect
|| stream
->protect
== next
->ident
)) {
1173 stream
->constant
= CONSTANT_FILE_IFNDEF
;
1174 stream
->protect
= next
->ident
;
1176 MARK_STREAM_NONCONST(token
->pos
);
1178 return preprocessor_if(token
, !token_defined(next
));
1182 * Expression handling for #if and #elif; it differs from normal expansion
1183 * due to special treatment of "defined".
1185 static int expression_value(struct token
**where
)
1187 struct expression
*expr
;
1189 struct token
**list
= where
, **beginning
= NULL
;
1193 while (!eof_token(p
= scan_next(list
))) {
1196 if (token_type(p
) != TOKEN_IDENT
)
1198 if (p
->ident
== &defined_ident
) {
1203 if (!expand_one_symbol(list
))
1205 if (token_type(p
) != TOKEN_IDENT
)
1207 if (Wundefined_preprocessor
)
1208 warning(p
->pos
, "undefined preprocessor identifier '%s'", show_ident(p
->ident
));
1209 replace_with_integer(p
, 0);
1212 if (match_op(p
, '(')) {
1216 replace_with_defined(p
);
1221 if (token_type(p
) == TOKEN_IDENT
)
1225 replace_with_defined(p
);
1230 if (!match_op(p
, ')'))
1231 warning(p
->pos
, "missing ')' after \"defined\"");
1238 p
= constant_expression(*where
, &expr
);
1240 warning(p
->pos
, "garbage at end: %s", show_token_sequence(p
));
1241 value
= get_expression_value(expr
);
1245 static int handle_if(struct stream
*stream
, struct token
**line
, struct token
*token
)
1249 value
= expression_value(&token
->next
);
1251 // This is an approximation. We really only need this if the
1252 // condition does depends on a pre-processor symbol. Note, that
1253 // the important #ifndef case has already changed ->constant.
1254 if (stream
->constant
== CONSTANT_FILE_MAYBE
)
1255 MARK_STREAM_NONCONST(token
->pos
);
1257 return preprocessor_if(token
, value
);
1260 static int handle_elif(struct stream
* stream
, struct token
**line
, struct token
*token
)
1262 if (stream
->nesting
== if_nesting
)
1263 MARK_STREAM_NONCONST(token
->pos
);
1265 if (stream
->nesting
> if_nesting
)
1266 warning(token
->pos
, "unmatched #elif within stream");
1268 if (elif_ignore
[if_nesting
-1] & ELIF_SEEN_ELSE
)
1269 warning(token
->pos
, "#elif after #else");
1271 if (false_nesting
) {
1272 /* If this whole if-thing is if'ed out, an elif cannot help */
1273 if (elif_ignore
[if_nesting
-1] & ELIF_IGNORE
)
1275 if (expression_value(&token
->next
)) {
1278 elif_ignore
[if_nesting
-1] |= ELIF_IGNORE
;
1284 return free_preprocessor_line(token
);
1287 static int handle_else(struct stream
*stream
, struct token
**line
, struct token
*token
)
1289 if (stream
->nesting
== if_nesting
)
1290 MARK_STREAM_NONCONST(token
->pos
);
1292 if (stream
->nesting
> if_nesting
)
1293 warning(token
->pos
, "unmatched #else within stream");
1295 if (elif_ignore
[if_nesting
-1] & ELIF_SEEN_ELSE
)
1296 warning(token
->pos
, "#else after #else");
1298 elif_ignore
[if_nesting
-1] |= ELIF_SEEN_ELSE
;
1300 if (false_nesting
) {
1301 /* If this whole if-thing is if'ed out, an else cannot help */
1302 if (elif_ignore
[if_nesting
-1] & ELIF_IGNORE
)
1306 elif_ignore
[if_nesting
-1] |= ELIF_IGNORE
;
1311 return free_preprocessor_line(token
);
1314 static int handle_endif(struct stream
*stream
, struct token
**line
, struct token
*token
)
1316 if (stream
->constant
== CONSTANT_FILE_IFNDEF
&& stream
->nesting
== if_nesting
)
1317 stream
->constant
= CONSTANT_FILE_MAYBE
;
1319 if (stream
->nesting
> if_nesting
)
1320 warning(token
->pos
, "unmatched #endif in stream");
1325 return free_preprocessor_line(token
);
1328 static const char *show_token_sequence(struct token
*token
)
1330 static char buffer
[1024];
1336 while (!eof_token(token
)) {
1337 const char *val
= show_token(token
);
1338 int len
= strlen(val
);
1340 if (ptr
+ whitespace
+ len
>= buffer
+ sizeof(buffer
)) {
1341 warning(token
->pos
, "too long token expansion");
1347 memcpy(ptr
, val
, len
);
1349 token
= token
->next
;
1350 whitespace
= token
->pos
.whitespace
;
1356 static int handle_warning(struct stream
*stream
, struct token
**line
, struct token
*token
)
1359 return free_preprocessor_line(token
);
1360 if (stream
->constant
== CONSTANT_FILE_MAYBE
)
1361 MARK_STREAM_NONCONST(token
->pos
);
1362 warning(token
->pos
, "%s", show_token_sequence(token
->next
));
1363 return free_preprocessor_line(token
);
1366 static int handle_error(struct stream
*stream
, struct token
**line
, struct token
*token
)
1369 return free_preprocessor_line(token
);
1370 if (stream
->constant
== CONSTANT_FILE_MAYBE
)
1371 MARK_STREAM_NONCONST(token
->pos
);
1372 warning(token
->pos
, "%s", show_token_sequence(token
->next
));
1373 return free_preprocessor_line(token
);
1376 static int handle_nostdinc(struct stream
*stream
, struct token
**line
, struct token
*token
)
1379 return free_preprocessor_line(token
);
1382 * Do we have any non-system includes?
1383 * Clear them out if so..
1385 *sys_includepath
= NULL
;
1386 return free_preprocessor_line(token
);
1389 static void add_path_entry(struct token
*token
, const char *path
, const char ***where
, const char **new_path
)
1394 /* Need one free entry.. */
1395 if (includepath
[INCLUDEPATHS
-2])
1396 error_die(token
->pos
, "too many include path entries");
1398 /* check that this is not a duplicate */
1401 if (strcmp(*dst
, path
) == 0)
1410 * Move them all up starting at dst,
1411 * insert the new entry..
1414 const char *tmp
= *dst
;
1423 static int handle_add_include(struct stream
*stream
, struct token
**line
, struct token
*token
)
1426 token
= token
->next
;
1427 if (eof_token(token
))
1429 if (token_type(token
) != TOKEN_STRING
) {
1430 warning(token
->pos
, "expected path string");
1433 add_path_entry(token
, token
->string
->data
, &sys_includepath
, sys_includepath
+ 1);
1437 static int handle_add_isystem(struct stream
*stream
, struct token
**line
, struct token
*token
)
1440 token
= token
->next
;
1441 if (eof_token(token
))
1443 if (token_type(token
) != TOKEN_STRING
) {
1444 warning(token
->pos
, "expected path string");
1447 add_path_entry(token
, token
->string
->data
, &sys_includepath
, sys_includepath
);
1451 /* Add to end on includepath list - no pointer updates */
1452 static void add_dirafter_entry(struct token
*token
, const char *path
)
1454 const char **dst
= includepath
;
1456 /* Need one free entry.. */
1457 if (includepath
[INCLUDEPATHS
-2])
1458 error_die(token
->pos
, "too many include path entries");
1460 /* Add to the end */
1468 static int handle_add_dirafter(struct stream
*stream
, struct token
**line
, struct token
*token
)
1471 token
= token
->next
;
1472 if (eof_token(token
))
1474 if (token_type(token
) != TOKEN_STRING
) {
1475 warning(token
->pos
, "expected path string");
1478 add_dirafter_entry(token
, token
->string
->data
);
1482 static int handle_split_include(struct stream
*stream
, struct token
**line
, struct token
*token
)
1485 return free_preprocessor_line(token
);
1490 * Split the include path. Any directories specified with `-I'
1491 * options before `-I-' are searched only for headers requested with
1492 * `#include "FILE"'; they are not searched for `#include <FILE>'.
1493 * If additional directories are specified with `-I' options after
1494 * the `-I-', those directories are searched for all `#include'
1496 * In addition, `-I-' inhibits the use of the directory of the current
1497 * file directory as the first search directory for `#include "FILE"'.
1499 quote_includepath
= includepath
+1;
1500 angle_includepath
= sys_includepath
;
1501 return free_preprocessor_line(token
);
1505 * We replace "#pragma xxx" with "__pragma__" in the token
1506 * stream. Just as an example.
1508 * We'll just #define that away for now, but the theory here
1509 * is that we can use this to insert arbitrary token sequences
1510 * to turn the pragma's into internal front-end sequences for
1511 * when we actually start caring about them.
1513 * So eventually this will turn into some kind of extended
1514 * __attribute__() like thing, except called __pragma__(xxx).
1516 static int handle_pragma(struct stream
*stream
, struct token
**line
, struct token
*token
)
1518 struct token
*next
= *line
;
1520 token
->ident
= &pragma_ident
;
1521 token
->pos
.newline
= 1;
1522 token
->pos
.whitespace
= 1;
1530 * We ignore #line for now.
1532 static int handle_line(struct stream
*stream
, struct token
**line
, struct token
*token
)
1538 void init_preprocessor(void)
1541 int stream
= init_stream("preprocessor", -1, includepath
);
1544 int (*handler
)(struct stream
*, struct token
**, struct token
*);
1546 { "define", handle_define
},
1547 { "weak_define",handle_weak_define
},
1548 { "undef", handle_undef
},
1549 { "ifdef", handle_ifdef
},
1550 { "ifndef", handle_ifndef
},
1551 { "else", handle_else
},
1552 { "endif", handle_endif
},
1553 { "if", handle_if
},
1554 { "elif", handle_elif
},
1555 { "warning", handle_warning
},
1556 { "error", handle_error
},
1557 { "include", handle_include
},
1558 { "include_next",handle_include_next
},
1559 { "pragma", handle_pragma
},
1560 { "line", handle_line
},
1562 // our internal preprocessor tokens
1563 { "nostdinc", handle_nostdinc
},
1564 { "add_include", handle_add_include
},
1565 { "add_isystem", handle_add_isystem
},
1566 { "add_dirafter", handle_add_dirafter
},
1567 { "split_include", handle_split_include
},
1570 for (i
= 0; i
< (sizeof (handlers
) / sizeof (handlers
[0])); i
++) {
1572 sym
= create_symbol(stream
, handlers
[i
].name
, SYM_PREPROCESSOR
, NS_PREPROCESSOR
);
1573 sym
->handler
= handlers
[i
].handler
;
1577 static void handle_preprocessor_line(struct stream
*stream
, struct token
**line
, struct token
*start
)
1579 struct token
*token
= start
->next
;
1584 if (token_type(token
) == TOKEN_NUMBER
)
1585 if (handle_line(stream
, line
, start
))
1588 if (token_type(token
) == TOKEN_IDENT
) {
1589 struct symbol
*sym
= lookup_symbol(token
->ident
, NS_PREPROCESSOR
);
1590 if (sym
&& sym
->handler(stream
, line
, token
))
1594 warning(token
->pos
, "unrecognized preprocessor line '%s'", show_token_sequence(token
));
1597 static void preprocessor_line(struct stream
*stream
, struct token
**line
)
1599 struct token
*start
= *line
, *next
;
1600 struct token
**tp
= &start
->next
;
1604 if (next
->pos
.newline
)
1609 *tp
= &eof_token_entry
;
1610 handle_preprocessor_line(stream
, line
, start
);
1613 static void do_preprocess(struct token
**list
)
1617 while (!eof_token(next
= scan_next(list
))) {
1618 struct stream
*stream
= input_streams
+ next
->pos
.stream
;
1620 if (next
->pos
.newline
&& match_op(next
, '#')) {
1621 if (!next
->pos
.noexpand
) {
1622 preprocessor_line(stream
, list
);
1623 __free_token(next
); /* Free the '#' token */
1628 switch (token_type(next
)) {
1629 case TOKEN_STREAMEND
:
1630 if (stream
->nesting
< if_nesting
+ 1) {
1631 warning(unmatched_if
->pos
, "unterminated preprocessor conditional");
1632 // Pretend to see a series of #endifs
1633 MARK_STREAM_NONCONST(next
->pos
);
1635 handle_endif (stream
, NULL
, NULL
);
1636 } while (stream
->nesting
< if_nesting
+ 1);
1638 if (stream
->constant
== CONSTANT_FILE_MAYBE
&& stream
->protect
) {
1639 stream
->constant
= CONSTANT_FILE_YES
;
1643 case TOKEN_STREAMBEGIN
:
1644 stream
->nesting
= if_nesting
+ 1;
1649 if (false_nesting
) {
1655 if (token_type(next
) != TOKEN_IDENT
||
1656 expand_one_symbol(list
))
1660 if (stream
->constant
== CONSTANT_FILE_MAYBE
) {
1662 * Any token expansion (even if it ended up being an
1663 * empty expansion) in this stream implies it can't
1666 MARK_STREAM_NONCONST(next
->pos
);
1671 struct token
* preprocess(struct token
*token
)
1674 init_preprocessor();
1675 do_preprocess(&token
);
1677 // Drop all expressions from pre-processing, they're not used any more.
1678 clear_expression_alloc();