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 * Permission is hereby granted, free of charge, to any person obtaining a copy
11 * of this software and associated documentation files (the "Software"), to deal
12 * in the Software without restriction, including without limitation the rights
13 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
14 * copies of the Software, and to permit persons to whom the Software is
15 * furnished to do so, subject to the following conditions:
17 * The above copyright notice and this permission notice shall be included in
18 * all copies or substantial portions of the Software.
20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
23 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
24 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
25 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
40 #include <sys/utsname.h>
47 #include "expression.h"
50 static struct ident_list
*macros
; // only needed for -dD
51 static int false_nesting
= 0;
52 static int counter_macro
= 0; // __COUNTER__ expansion
53 static int include_level
= 0;
54 static int expanding
= 0;
56 #define INCLUDEPATHS 300
57 const char *includepath
[INCLUDEPATHS
+1] = {
64 static const char **quote_includepath
= includepath
;
65 static const char **angle_includepath
= includepath
+ 1;
66 static const char **isys_includepath
= includepath
+ 1;
67 static const char **sys_includepath
= includepath
+ 1;
68 static const char **dirafter_includepath
= includepath
+ 3;
70 #define dirty_stream(stream) \
72 if (!stream->dirty) { \
74 if (!stream->ifndef) \
75 stream->protect = NULL; \
79 #define end_group(stream) \
81 if (stream->ifndef == stream->top_if) { \
82 stream->ifndef = NULL; \
84 stream->protect = NULL; \
85 else if (stream->protect) \
90 #define nesting_error(stream) \
93 stream->ifndef = NULL; \
94 stream->protect = NULL; \
97 static struct token
*alloc_token(struct position
*pos
)
99 struct token
*token
= __alloc_token(0);
101 token
->pos
.stream
= pos
->stream
;
102 token
->pos
.line
= pos
->line
;
103 token
->pos
.pos
= pos
->pos
;
104 token
->pos
.whitespace
= 1;
108 /* Expand symbol 'sym' at '*list' */
109 static int expand(struct token
**, struct symbol
*);
111 static void replace_with_string(struct token
*token
, const char *str
)
113 int size
= strlen(str
) + 1;
114 struct string
*s
= __alloc_string(size
);
117 memcpy(s
->data
, str
, size
);
118 token_type(token
) = TOKEN_STRING
;
122 static void replace_with_integer(struct token
*token
, unsigned int val
)
124 char *buf
= __alloc_bytes(11);
125 sprintf(buf
, "%u", val
);
126 token_type(token
) = TOKEN_NUMBER
;
130 static struct symbol
*lookup_macro(struct ident
*ident
)
132 struct symbol
*sym
= lookup_symbol(ident
, NS_MACRO
| NS_UNDEF
);
133 if (sym
&& sym
->namespace != NS_MACRO
)
138 static int token_defined(struct token
*token
)
140 if (token_type(token
) == TOKEN_IDENT
) {
141 struct symbol
*sym
= lookup_macro(token
->ident
);
143 sym
->used_in
= file_scope
;
149 sparse_error(token
->pos
, "expected preprocessor identifier");
153 static void replace_with_bool(struct token
*token
, bool val
)
155 static const char *string
[] = { "0", "1" };
157 token_type(token
) = TOKEN_NUMBER
;
158 token
->number
= string
[val
];
161 static void replace_with_defined(struct token
*token
)
163 replace_with_bool(token
, token_defined(token
));
166 static void expand_line(struct token
*token
)
168 replace_with_integer(token
, token
->pos
.line
);
171 static void expand_file(struct token
*token
)
173 replace_with_string(token
, stream_name(token
->pos
.stream
));
176 static void expand_basefile(struct token
*token
)
178 replace_with_string(token
, base_filename
);
182 static void expand_date(struct token
*token
)
184 static char buffer
[12]; /* __DATE__: 3 + ' ' + 2 + ' ' + 4 + '\0' */
188 strftime(buffer
, 12, "%b %e %Y", localtime(&t
));
189 replace_with_string(token
, buffer
);
192 static void expand_time(struct token
*token
)
194 static char buffer
[9]; /* __TIME__: 2 + ':' + 2 + ':' + 2 + '\0' */
198 strftime(buffer
, 9, "%T", localtime(&t
));
199 replace_with_string(token
, buffer
);
202 static void expand_counter(struct token
*token
)
204 replace_with_integer(token
, counter_macro
++);
207 static void expand_include_level(struct token
*token
)
209 replace_with_integer(token
, include_level
- 1);
212 static int expand_one_symbol(struct token
**list
)
214 struct token
*token
= *list
;
217 if (token
->pos
.noexpand
)
220 sym
= lookup_macro(token
->ident
);
223 store_macro_pos(token
);
224 if (sym
->expand_simple
) {
225 sym
->expand_simple(token
);
230 sym
->used_in
= file_scope
;
232 rc
= expand(list
, sym
);
238 static inline struct token
*scan_next(struct token
**where
)
240 struct token
*token
= *where
;
241 if (token_type(token
) != TOKEN_UNTAINT
)
244 token
->ident
->tainted
= 0;
246 } while (token_type(token
) == TOKEN_UNTAINT
);
251 static void expand_list(struct token
**list
)
254 while (!eof_token(next
= scan_next(list
))) {
255 if (token_type(next
) != TOKEN_IDENT
|| expand_one_symbol(list
))
260 static void preprocessor_line(struct stream
*stream
, struct token
**line
);
262 static struct token
*collect_arg(struct token
*prev
, int vararg
, struct position
*pos
, int count
)
264 struct stream
*stream
= input_streams
+ prev
->pos
.stream
;
265 struct token
**p
= &prev
->next
;
269 while (!eof_token(next
= scan_next(p
))) {
270 if (next
->pos
.newline
&& match_op(next
, '#')) {
271 if (!next
->pos
.noexpand
) {
272 preprocessor_line(stream
, p
);
273 __free_token(next
); /* Free the '#' token */
277 switch (token_type(next
)) {
278 case TOKEN_STREAMEND
:
279 case TOKEN_STREAMBEGIN
:
280 *p
= &eof_token_entry
;
283 case TOKEN_WIDE_STRING
:
285 next
->string
->immutable
= 1;
293 if (match_op(next
, '(')) {
295 } else if (match_op(next
, ')')) {
298 } else if (match_op(next
, ',') && !nesting
&& !vararg
) {
301 next
->pos
.stream
= pos
->stream
;
302 next
->pos
.line
= pos
->line
;
303 next
->pos
.pos
= pos
->pos
;
304 next
->pos
.newline
= 0;
307 *p
= &eof_token_entry
;
312 * We store arglist as <counter> [arg1] <number of uses for arg1> ... eof
317 struct token
*expanded
;
324 static int collect_arguments(struct token
*start
, struct token
*arglist
, struct arg
*args
, struct token
*what
)
326 int wanted
= arglist
->count
.normal
;
327 struct token
*next
= NULL
;
330 arglist
= arglist
->next
; /* skip counter */
333 next
= collect_arg(start
, 0, &what
->pos
, 0);
336 if (!eof_token(start
->next
) || !match_op(next
, ')')) {
341 for (count
= 0; count
< wanted
; count
++) {
342 struct argcount
*p
= &arglist
->next
->count
;
343 next
= collect_arg(start
, p
->vararg
, &what
->pos
, p
->normal
);
346 if (p
->vararg
&& wanted
== 1 && eof_token(start
->next
))
348 arglist
= arglist
->next
->next
;
349 args
[count
].arg
= start
->next
;
350 args
[count
].n_normal
= p
->normal
;
351 args
[count
].n_quoted
= p
->quoted
;
352 args
[count
].n_str
= p
->str
;
353 if (match_op(next
, ')')) {
359 if (count
== wanted
&& !match_op(next
, ')'))
361 if (count
== wanted
- 1) {
362 struct argcount
*p
= &arglist
->next
->count
;
365 args
[count
].arg
= NULL
;
366 args
[count
].n_normal
= p
->normal
;
367 args
[count
].n_quoted
= p
->quoted
;
368 args
[count
].n_str
= p
->str
;
370 if (count
< wanted
- 1)
373 what
->next
= next
->next
;
377 sparse_error(what
->pos
, "macro \"%s\" requires %d arguments, but only %d given",
378 show_token(what
), wanted
, count
);
381 while (match_op(next
, ',')) {
382 next
= collect_arg(next
, 0, &what
->pos
, 0);
387 sparse_error(what
->pos
, "macro \"%s\" passed %d arguments, but takes just %d",
388 show_token(what
), count
, wanted
);
391 sparse_error(what
->pos
, "unterminated argument list invoking macro \"%s\"",
394 what
->next
= next
->next
;
398 static struct token
*dup_list(struct token
*list
)
400 struct token
*res
= NULL
;
401 struct token
**p
= &res
;
403 while (!eof_token(list
)) {
404 struct token
*newtok
= __alloc_token(0);
413 static const char *show_token_sequence(struct token
*token
, int quote
)
415 static char buffer
[MAX_STRING
];
419 if (!token
&& !quote
)
421 while (!eof_token(token
)) {
422 const char *val
= quote
? quote_token(token
) : show_token(token
);
423 int len
= strlen(val
);
425 if (ptr
+ whitespace
+ len
>= buffer
+ sizeof(buffer
)) {
426 sparse_error(token
->pos
, "too long token expansion");
432 memcpy(ptr
, val
, len
);
435 whitespace
= token
->pos
.whitespace
;
441 static struct token
*stringify(struct token
*arg
)
443 const char *s
= show_token_sequence(arg
, 1);
444 int size
= strlen(s
)+1;
445 struct token
*token
= __alloc_token(0);
446 struct string
*string
= __alloc_string(size
);
448 memcpy(string
->data
, s
, size
);
449 string
->length
= size
;
450 token
->pos
= arg
->pos
;
451 token_type(token
) = TOKEN_STRING
;
452 token
->string
= string
;
453 token
->next
= &eof_token_entry
;
457 static void expand_arguments(int count
, struct arg
*args
)
460 for (i
= 0; i
< count
; i
++) {
461 struct token
*arg
= args
[i
].arg
;
463 arg
= &eof_token_entry
;
465 args
[i
].str
= stringify(arg
);
466 if (args
[i
].n_normal
) {
467 if (!args
[i
].n_quoted
) {
468 args
[i
].expanded
= arg
;
470 } else if (eof_token(arg
)) {
471 args
[i
].expanded
= arg
;
473 args
[i
].expanded
= dup_list(arg
);
475 expand_list(&args
[i
].expanded
);
481 * Possibly valid combinations:
482 * - ident + ident -> ident
483 * - ident + number -> ident unless number contains '.', '+' or '-'.
484 * - 'L' + char constant -> wide char constant
485 * - 'L' + string literal -> wide string literal
486 * - number + number -> number
487 * - number + ident -> number
488 * - number + '.' -> number
489 * - number + '+' or '-' -> number, if number used to end on [eEpP].
490 * - '.' + number -> number, if number used to start with a digit.
491 * - special + special -> either special or an error.
493 static enum token_type
combine(struct token
*left
, struct token
*right
, char *p
)
496 enum token_type t1
= token_type(left
), t2
= token_type(right
);
498 if (t1
!= TOKEN_IDENT
&& t1
!= TOKEN_NUMBER
&& t1
!= TOKEN_SPECIAL
)
501 if (t1
== TOKEN_IDENT
&& left
->ident
== &L_ident
) {
502 if (t2
>= TOKEN_CHAR
&& t2
< TOKEN_WIDE_CHAR
)
503 return t2
+ TOKEN_WIDE_CHAR
- TOKEN_CHAR
;
504 if (t2
== TOKEN_STRING
)
505 return TOKEN_WIDE_STRING
;
508 if (t2
!= TOKEN_IDENT
&& t2
!= TOKEN_NUMBER
&& t2
!= TOKEN_SPECIAL
)
511 strcpy(p
, show_token(left
));
512 strcat(p
, show_token(right
));
518 if (t1
== TOKEN_IDENT
) {
519 if (t2
== TOKEN_SPECIAL
)
521 if (t2
== TOKEN_NUMBER
&& strpbrk(p
, "+-."))
526 if (t1
== TOKEN_NUMBER
) {
527 if (t2
== TOKEN_SPECIAL
) {
528 switch (right
->special
) {
532 if (strchr("eEpP", p
[len
- 2]))
541 if (p
[0] == '.' && isdigit((unsigned char)p
[1]))
544 return TOKEN_SPECIAL
;
547 static int merge(struct token
*left
, struct token
*right
)
549 static char buffer
[512];
550 enum token_type res
= combine(left
, right
, buffer
);
555 left
->ident
= built_in_ident(buffer
);
556 left
->pos
.noexpand
= 0;
560 token_type(left
) = TOKEN_NUMBER
; /* could be . + num */
561 left
->number
= xstrdup(buffer
);
565 if (buffer
[2] && buffer
[3])
567 for (n
= SPECIAL_BASE
; n
< SPECIAL_ARG_SEPARATOR
; n
++) {
568 if (!memcmp(buffer
, combinations
[n
-SPECIAL_BASE
], 3)) {
575 case TOKEN_WIDE_CHAR
:
576 case TOKEN_WIDE_STRING
:
577 token_type(left
) = res
;
578 left
->pos
.noexpand
= 0;
579 left
->string
= right
->string
;
582 case TOKEN_WIDE_CHAR_EMBEDDED_0
... TOKEN_WIDE_CHAR_EMBEDDED_3
:
583 token_type(left
) = res
;
584 left
->pos
.noexpand
= 0;
585 memcpy(left
->embedded
, right
->embedded
, 4);
591 sparse_error(left
->pos
, "'##' failed: concatenation is not a valid token");
595 static struct token
*dup_token(struct token
*token
, struct position
*streampos
)
597 struct token
*alloc
= alloc_token(streampos
);
598 token_type(alloc
) = token_type(token
);
599 alloc
->pos
.newline
= token
->pos
.newline
;
600 alloc
->pos
.whitespace
= token
->pos
.whitespace
;
601 alloc
->number
= token
->number
;
602 alloc
->pos
.noexpand
= token
->pos
.noexpand
;
606 static struct token
**copy(struct token
**where
, struct token
*list
, int *count
)
608 int need_copy
= --*count
;
609 while (!eof_token(list
)) {
612 token
= dup_token(list
, &list
->pos
);
615 if (token_type(token
) == TOKEN_IDENT
&& token
->ident
->tainted
)
616 token
->pos
.noexpand
= 1;
618 where
= &token
->next
;
621 *where
= &eof_token_entry
;
625 static int handle_kludge(struct token
**p
, struct arg
*args
)
627 struct token
*t
= (*p
)->next
->next
;
629 struct arg
*v
= &args
[t
->argnum
];
630 if (token_type(t
->next
) != TOKEN_CONCAT
) {
632 /* ignore the first ## */
636 /* skip the entire thing */
640 if (v
->arg
&& !eof_token(v
->arg
))
641 return 0; /* no magic */
646 static struct token
**substitute(struct token
**list
, struct token
*body
, struct arg
*args
)
648 struct position
*base_pos
= &(*list
)->pos
;
650 enum {Normal
, Placeholder
, Concat
} state
= Normal
;
652 for (; !eof_token(body
); body
= body
->next
) {
653 struct token
*added
, *arg
;
657 switch (token_type(body
)) {
658 case TOKEN_GNU_KLUDGE
:
660 * GNU kludge: if we had <comma>##<vararg>, behaviour
661 * depends on whether we had enough arguments to have
662 * a vararg. If we did, ## is just ignored. Otherwise
663 * both , and ## are ignored. Worse, there can be
664 * an arbitrary number of ##<arg> in between; if all of
665 * those are empty, we act as if they hadn't been there,
666 * otherwise we act as if the kludge didn't exist.
669 if (handle_kludge(&body
, args
)) {
676 added
= dup_token(t
, base_pos
);
677 token_type(added
) = TOKEN_SPECIAL
;
681 case TOKEN_STR_ARGUMENT
:
682 arg
= args
[body
->argnum
].str
;
683 count
= &args
[body
->argnum
].n_str
;
686 case TOKEN_QUOTED_ARGUMENT
:
687 arg
= args
[body
->argnum
].arg
;
688 count
= &args
[body
->argnum
].n_quoted
;
689 if (!arg
|| eof_token(arg
)) {
698 case TOKEN_MACRO_ARGUMENT
:
699 arg
= args
[body
->argnum
].expanded
;
700 count
= &args
[body
->argnum
].n_normal
;
701 if (eof_token(arg
)) {
706 tail
= copy(&added
, arg
, count
);
707 added
->pos
.newline
= body
->pos
.newline
;
708 added
->pos
.whitespace
= body
->pos
.whitespace
;
712 if (state
== Placeholder
)
719 added
= dup_token(body
, base_pos
);
720 if (added
->ident
->tainted
)
721 added
->pos
.noexpand
= 1;
726 added
= dup_token(body
, base_pos
);
732 * if we got to doing real concatenation, we already have
733 * added something into the list, so containing_token() is OK.
735 if (state
== Concat
&& merge(containing_token(list
), added
)) {
737 if (tail
!= &added
->next
)
745 *list
= &eof_token_entry
;
749 static int expand(struct token
**list
, struct symbol
*sym
)
752 struct token
*token
= *list
;
753 struct ident
*expanding
= token
->ident
;
755 struct token
*expansion
= sym
->expansion
;
756 int nargs
= sym
->arglist
? sym
->arglist
->count
.normal
: 0;
757 struct arg args
[nargs
];
759 if (expanding
->tainted
) {
760 token
->pos
.noexpand
= 1;
765 if (!match_op(scan_next(&token
->next
), '('))
767 if (!collect_arguments(token
->next
, sym
->arglist
, args
, token
))
769 expand_arguments(nargs
, args
);
773 return sym
->expand(token
, args
) ? 0 : 1;
775 expanding
->tainted
= 1;
778 tail
= substitute(list
, expansion
, args
);
780 * Note that it won't be eof - at least TOKEN_UNTAINT will be there.
781 * We still can lose the newline flag if the sucker expands to nothing,
782 * but the price of dealing with that is probably too high (we'd need
783 * to collect the flags during scan_next())
785 (*list
)->pos
.newline
= token
->pos
.newline
;
786 (*list
)->pos
.whitespace
= token
->pos
.whitespace
;
792 static const char *token_name_sequence(struct token
*token
, int endop
, struct token
*start
)
794 static char buffer
[256];
797 while (!eof_token(token
) && !match_op(token
, endop
)) {
799 const char *val
= token
->string
->data
;
800 if (token_type(token
) != TOKEN_STRING
)
801 val
= show_token(token
);
803 memcpy(ptr
, val
, len
);
808 if (endop
&& !match_op(token
, endop
))
809 sparse_error(start
->pos
, "expected '>' at end of filename");
813 static int already_tokenized(const char *path
)
817 for (stream
= *hash_stream(path
); stream
>= 0 ; stream
= next
) {
818 struct stream
*s
= input_streams
+ stream
;
820 next
= s
->next_stream
;
822 if (strcmp(path
, s
->name
))
826 if (s
->constant
!= CONSTANT_FILE_YES
)
828 if (strcmp(path
, s
->name
))
830 if (s
->protect
&& !lookup_macro(s
->protect
))
837 /* Handle include of header files.
838 * The relevant options are made compatible with gcc. The only options that
839 * are not supported is -withprefix and friends.
841 * Three set of include paths are known:
842 * quote_includepath: Path to search when using #include "file.h"
843 * angle_includepath: Paths to search when using #include <file.h>
844 * isys_includepath: Paths specified with -isystem, come before the
845 * built-in system include paths. Gcc would suppress
846 * warnings from system headers. Here we separate
847 * them from the angle_ ones to keep search ordering.
849 * sys_includepath: Built-in include paths.
850 * dirafter_includepath Paths added with -dirafter.
852 * The above is implemented as one array with pointers
854 * quote_includepath ---> | |
858 * angle_includepath ---> | |
860 * isys_includepath ---> | |
862 * sys_includepath ---> | |
864 * dirafter_includepath -> | |
867 * -I dir insert dir just before isys_includepath and move the rest
868 * -I- makes all dirs specified with -I before to quote dirs only and
869 * angle_includepath is set equal to isys_includepath.
870 * -nostdinc removes all sys dirs by storing NULL in entry pointed
871 * to by * sys_includepath. Note that this will reset all dirs built-in
872 * and added before -nostdinc by -isystem and -idirafter.
873 * -isystem dir adds dir where isys_includepath points adding this dir as
875 * -idirafter dir adds dir to the end of the list
878 static void set_stream_include_path(struct stream
*stream
)
880 const char *path
= stream
->path
;
882 const char *p
= strrchr(stream
->name
, '/');
885 int len
= p
- stream
->name
+ 1;
886 char *m
= malloc(len
+1);
887 /* This includes the final "/" */
888 memcpy(m
, stream
->name
, len
);
891 /* normalize this path */
892 while (path
[0] == '.' && path
[1] == '/') {
894 while (path
[0] == '/')
900 includepath
[0] = path
;
903 static int try_include(struct position pos
, const char *path
, const char *filename
, int flen
, struct token
**where
, const char **next_path
)
906 int plen
= strlen(path
);
907 static char fullname
[PATH_MAX
];
909 memcpy(fullname
, path
, plen
);
910 if (plen
&& path
[plen
-1] != '/') {
911 fullname
[plen
] = '/';
914 memcpy(fullname
+plen
, filename
, flen
);
915 if (already_tokenized(fullname
))
917 fd
= open(fullname
, O_RDONLY
);
919 char *streamname
= xmemdup(fullname
, plen
+ flen
);
920 *where
= tokenize(&pos
, streamname
, fd
, *where
, next_path
);
927 static int do_include_path(const char **pptr
, struct token
**list
, struct token
*token
, const char *filename
, int flen
)
931 while ((path
= *pptr
++) != NULL
) {
932 if (!try_include(token
->pos
, path
, filename
, flen
, list
, pptr
))
939 static int free_preprocessor_line(struct token
*token
)
941 while (token_type(token
) != TOKEN_EOF
) {
942 struct token
*free
= token
;
949 const char *find_include(const char *skip
, const char *look_for
)
952 struct dirent
*entry
;
956 static char buf
[PATH_MAX
+ 1];
962 if (!getcwd(cwd
, sizeof(cwd
)))
965 while ((entry
= readdir(dp
))) {
966 lstat(entry
->d_name
, &statbuf
);
968 if (strcmp(entry
->d_name
, look_for
) == 0) {
969 snprintf(buf
, sizeof(buf
), "%s/%s", cwd
, entry
->d_name
);
974 if (S_ISDIR(statbuf
.st_mode
)) {
975 /* Found a directory, but ignore . and .. */
976 if (strcmp(".", entry
->d_name
) == 0 ||
977 strcmp("..", entry
->d_name
) == 0 ||
978 strcmp(skip
, entry
->d_name
) == 0)
981 chdir(entry
->d_name
);
982 ret
= find_include("", look_for
);
996 const char *search_dir(const char *stop
, const char *look_for
)
1003 if (!getcwd(cwd
, sizeof(cwd
)))
1008 ret
= find_include(cnt
++ ? cwd
+ len
+ 1 : "", look_for
);
1012 if (strcmp(cwd
, stop
) == 0 ||
1013 strcmp(cwd
, "/usr/include") == 0 ||
1014 strcmp(cwd
, "/usr/local/include") == 0 ||
1015 strlen(cwd
) <= 10 || /* heck... don't search /usr/lib/ */
1016 strcmp(cwd
, "/") == 0)
1019 while (--len
>= 0) {
1020 if (cwd
[len
] == '/') {
1031 static void use_best_guess_header_file(struct token
*token
, const char *filename
, struct token
**list
)
1034 char dir_part
[PATH_MAX
];
1035 const char *file_part
;
1036 const char *include_name
;
1040 /* Avoid guessing includes recursively. */
1044 if (!filename
|| filename
[0] == '\0')
1047 file_part
= filename
;
1048 while ((filename
= strchr(filename
, '/'))) {
1051 file_part
= filename
;
1054 snprintf(dir_part
, sizeof(dir_part
), "%s", stream_name(token
->pos
.stream
));
1055 len
= strlen(dir_part
);
1056 while (--len
>= 0) {
1057 if (dir_part
[len
] == '/') {
1058 dir_part
[len
] = '\0';
1063 sprintf(dir_part
, ".");
1065 if (!getcwd(cwd
, sizeof(cwd
)))
1069 include_name
= search_dir(cwd
, file_part
);
1073 sparse_error(token
->pos
, "using '%s'", include_name
);
1075 try_include(token
->pos
, "", include_name
, strlen(include_name
), list
, includepath
);
1078 static int handle_include_path(struct stream
*stream
, struct token
**list
, struct token
*token
, int how
)
1080 const char *filename
;
1088 if (!match_op(next
, '<')) {
1089 expand_list(&token
->next
);
1092 if (match_op(token
->next
, '<')) {
1099 filename
= token_name_sequence(token
, expect
, token
);
1100 flen
= strlen(filename
) + 1;
1102 /* Absolute path? */
1103 if (filename
[0] == '/') {
1104 if (try_include(token
->pos
, "", filename
, flen
, list
, includepath
))
1111 path
= stream
->next_path
;
1114 includepath
[0] = "";
1118 /* Dir of input file is first dir to search for quoted includes */
1119 set_stream_include_path(stream
);
1120 path
= expect
? angle_includepath
: quote_includepath
;
1123 /* Check the standard include paths.. */
1124 if (do_include_path(path
, list
, token
, filename
, flen
))
1127 sparse_error(token
->pos
, "unable to open '%s'", filename
);
1128 use_best_guess_header_file(token
, filename
, list
);
1132 static int handle_include(struct stream
*stream
, struct token
**list
, struct token
*token
)
1134 return handle_include_path(stream
, list
, token
, 0);
1137 static int handle_include_next(struct stream
*stream
, struct token
**list
, struct token
*token
)
1139 return handle_include_path(stream
, list
, token
, 1);
1142 static int handle_argv_include(struct stream
*stream
, struct token
**list
, struct token
*token
)
1144 return handle_include_path(stream
, list
, token
, 2);
1147 static int token_different(struct token
*t1
, struct token
*t2
)
1151 if (token_type(t1
) != token_type(t2
))
1154 switch (token_type(t1
)) {
1156 different
= t1
->ident
!= t2
->ident
;
1158 case TOKEN_ARG_COUNT
:
1161 case TOKEN_GNU_KLUDGE
:
1165 different
= strcmp(t1
->number
, t2
->number
);
1168 different
= t1
->special
!= t2
->special
;
1170 case TOKEN_MACRO_ARGUMENT
:
1171 case TOKEN_QUOTED_ARGUMENT
:
1172 case TOKEN_STR_ARGUMENT
:
1173 different
= t1
->argnum
!= t2
->argnum
;
1175 case TOKEN_CHAR_EMBEDDED_0
... TOKEN_CHAR_EMBEDDED_3
:
1176 case TOKEN_WIDE_CHAR_EMBEDDED_0
... TOKEN_WIDE_CHAR_EMBEDDED_3
:
1177 different
= memcmp(t1
->embedded
, t2
->embedded
, 4);
1180 case TOKEN_WIDE_CHAR
:
1182 case TOKEN_WIDE_STRING
: {
1183 struct string
*s1
, *s2
;
1188 if (s1
->length
!= s2
->length
)
1190 different
= memcmp(s1
->data
, s2
->data
, s1
->length
);
1200 static int token_list_different(struct token
*list1
, struct token
*list2
)
1205 if (!list1
|| !list2
)
1207 if (token_different(list1
, list2
))
1209 list1
= list1
->next
;
1210 list2
= list2
->next
;
1214 static inline void set_arg_count(struct token
*token
)
1216 token_type(token
) = TOKEN_ARG_COUNT
;
1217 token
->count
.normal
= token
->count
.quoted
=
1218 token
->count
.str
= token
->count
.vararg
= 0;
1221 static struct token
*parse_arguments(struct token
*list
)
1223 struct token
*arg
= list
->next
, *next
= list
;
1224 struct argcount
*count
= &list
->count
;
1226 set_arg_count(list
);
1228 if (match_op(arg
, ')')) {
1230 list
->next
= &eof_token_entry
;
1234 while (token_type(arg
) == TOKEN_IDENT
) {
1235 if (arg
->ident
== &__VA_ARGS___ident
)
1237 if (!++count
->normal
)
1241 if (match_op(next
, ',')) {
1242 set_arg_count(next
);
1247 if (match_op(next
, ')')) {
1248 set_arg_count(next
);
1250 arg
->next
->next
= &eof_token_entry
;
1254 /* normal cases are finished here */
1256 if (match_op(next
, SPECIAL_ELLIPSIS
)) {
1257 if (match_op(next
->next
, ')')) {
1258 set_arg_count(next
);
1259 next
->count
.vararg
= 1;
1261 arg
->next
->next
= &eof_token_entry
;
1269 if (eof_token(next
)) {
1277 if (match_op(arg
, SPECIAL_ELLIPSIS
)) {
1279 token_type(arg
) = TOKEN_IDENT
;
1280 arg
->ident
= &__VA_ARGS___ident
;
1281 if (!match_op(next
, ')'))
1283 if (!++count
->normal
)
1285 set_arg_count(next
);
1286 next
->count
.vararg
= 1;
1288 arg
->next
->next
= &eof_token_entry
;
1292 if (eof_token(arg
)) {
1296 if (match_op(arg
, ','))
1303 sparse_error(arg
->pos
, "parameter name missing");
1306 sparse_error(arg
->pos
, "\"%s\" may not appear in macro parameter list",
1310 sparse_error(arg
->pos
, "missing ')' in macro parameter list");
1313 sparse_error(arg
->pos
, "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
1316 sparse_error(arg
->pos
, "too many arguments in macro definition");
1320 static int try_arg(struct token
*token
, enum token_type type
, struct token
*arglist
)
1322 struct ident
*ident
= token
->ident
;
1325 if (!arglist
|| token_type(token
) != TOKEN_IDENT
)
1328 arglist
= arglist
->next
;
1330 for (nr
= 0; !eof_token(arglist
); nr
++, arglist
= arglist
->next
->next
) {
1331 if (arglist
->ident
== ident
) {
1332 struct argcount
*count
= &arglist
->next
->count
;
1336 token_type(token
) = type
;
1338 case TOKEN_MACRO_ARGUMENT
:
1339 n
= ++count
->normal
;
1341 case TOKEN_QUOTED_ARGUMENT
:
1342 n
= ++count
->quoted
;
1348 return count
->vararg
? 2 : 1;
1350 * XXX - need saner handling of that
1351 * (>= 1024 instances of argument)
1353 token_type(token
) = TOKEN_ERROR
;
1360 static struct token
*handle_hash(struct token
**p
, struct token
*arglist
)
1362 struct token
*token
= *p
;
1364 struct token
*next
= token
->next
;
1365 if (!try_arg(next
, TOKEN_STR_ARGUMENT
, arglist
))
1367 next
->pos
.whitespace
= token
->pos
.whitespace
;
1368 __free_token(token
);
1371 token
->pos
.noexpand
= 1;
1376 sparse_error(token
->pos
, "'#' is not followed by a macro parameter");
1380 /* token->next is ## */
1381 static struct token
*handle_hashhash(struct token
*token
, struct token
*arglist
)
1383 struct token
*last
= token
;
1384 struct token
*concat
;
1385 int state
= match_op(token
, ',');
1387 try_arg(token
, TOKEN_QUOTED_ARGUMENT
, arglist
);
1393 /* eat duplicate ## */
1394 concat
= token
->next
;
1395 while (match_op(t
= concat
->next
, SPECIAL_HASHHASH
)) {
1397 __free_token(concat
);
1400 token_type(concat
) = TOKEN_CONCAT
;
1405 if (match_op(t
, '#')) {
1406 t
= handle_hash(&concat
->next
, arglist
);
1411 is_arg
= try_arg(t
, TOKEN_QUOTED_ARGUMENT
, arglist
);
1413 if (state
== 1 && is_arg
) {
1417 state
= match_op(t
, ',');
1421 if (!match_op(token
->next
, SPECIAL_HASHHASH
))
1424 /* handle GNU ,##__VA_ARGS__ kludge, in all its weirdness */
1426 token_type(last
) = TOKEN_GNU_KLUDGE
;
1430 sparse_error(concat
->pos
, "'##' cannot appear at the ends of macro expansion");
1434 static struct token
*parse_expansion(struct token
*expansion
, struct token
*arglist
, struct ident
*name
)
1436 struct token
*token
= expansion
;
1439 if (match_op(token
, SPECIAL_HASHHASH
))
1442 for (p
= &expansion
; !eof_token(token
); p
= &token
->next
, token
= *p
) {
1443 if (match_op(token
, '#')) {
1444 token
= handle_hash(p
, arglist
);
1448 if (match_op(token
->next
, SPECIAL_HASHHASH
)) {
1449 token
= handle_hashhash(token
, arglist
);
1453 try_arg(token
, TOKEN_MACRO_ARGUMENT
, arglist
);
1455 switch (token_type(token
)) {
1460 case TOKEN_WIDE_STRING
:
1461 token
->string
->immutable
= 1;
1465 token
= alloc_token(&expansion
->pos
);
1466 token_type(token
) = TOKEN_UNTAINT
;
1467 token
->ident
= name
;
1473 sparse_error(token
->pos
, "'##' cannot appear at the ends of macro expansion");
1476 sparse_error(token
->pos
, "too many instances of argument in body");
1480 static int do_define(struct position pos
, struct token
*token
, struct ident
*name
,
1481 struct token
*arglist
, struct token
*expansion
, int attr
)
1486 expansion
= parse_expansion(expansion
, arglist
, name
);
1490 sym
= lookup_symbol(name
, NS_MACRO
| NS_UNDEF
);
1494 if (attr
< sym
->attr
)
1497 clean
= (attr
== sym
->attr
&& sym
->namespace == NS_MACRO
);
1499 if (token_list_different(sym
->expansion
, expansion
) ||
1500 token_list_different(sym
->arglist
, arglist
)) {
1502 if ((clean
&& attr
== SYM_ATTR_NORMAL
)
1503 || sym
->used_in
== file_scope
) {
1504 warning(pos
, "preprocessor token %.*s redefined",
1505 name
->len
, name
->name
);
1506 info(sym
->pos
, "this was the original definition");
1512 if (!sym
|| sym
->scope
!= file_scope
) {
1513 sym
= alloc_symbol(pos
, SYM_NODE
);
1514 bind_symbol(sym
, name
, NS_MACRO
);
1515 add_ident(¯os
, name
);
1520 sym
->expansion
= expansion
;
1521 sym
->arglist
= arglist
;
1522 if (token
) /* Free the "define" token, but not the rest of the line */
1523 __free_token(token
);
1526 sym
->namespace = NS_MACRO
;
1527 sym
->used_in
= NULL
;
1534 // predefine a macro with a printf-formatted value
1535 // @name: the name of the macro
1536 // @weak: 0/1 for a normal or a weak define
1537 // @fmt: the printf format followed by it's arguments.
1539 // The type of the value is automatically infered:
1540 // TOKEN_NUMBER if it starts by a digit, TOKEN_IDENT otherwise.
1541 // If @fmt is null or empty, the macro is defined with an empty definition.
1542 void predefine(const char *name
, int weak
, const char *fmt
, ...)
1544 struct ident
*ident
= built_in_ident(name
);
1545 struct token
*value
= &eof_token_entry
;
1546 int attr
= weak
? SYM_ATTR_WEAK
: SYM_ATTR_NORMAL
;
1548 if (fmt
&& fmt
[0]) {
1549 static char buf
[256];
1553 vsnprintf(buf
, sizeof(buf
), fmt
, ap
);
1556 value
= __alloc_token(0);
1557 if (isdigit((unsigned char)buf
[0])) {
1558 token_type(value
) = TOKEN_NUMBER
;
1559 value
->number
= xstrdup(buf
);
1561 token_type(value
) = TOKEN_IDENT
;
1562 value
->ident
= built_in_ident(buf
);
1564 value
->pos
.whitespace
= 1;
1565 value
->next
= &eof_token_entry
;
1568 do_define(value
->pos
, NULL
, ident
, NULL
, value
, attr
);
1572 // like predefine() but only if one of the non-standard dialect is chosen
1573 void predefine_nostd(const char *name
)
1575 if ((standard
& STANDARD_GNU
) || (standard
== STANDARD_NONE
))
1576 predefine(name
, 1, "1");
1579 static void predefine_fmt(const char *fmt
, int weak
, va_list ap
)
1581 static char buf
[256];
1583 vsnprintf(buf
, sizeof(buf
), fmt
, ap
);
1584 predefine(buf
, weak
, "1");
1587 void predefine_strong(const char *fmt
, ...)
1592 predefine_fmt(fmt
, 0, ap
);
1596 void predefine_weak(const char *fmt
, ...)
1601 predefine_fmt(fmt
, 1, ap
);
1605 static int do_handle_define(struct stream
*stream
, struct token
**line
, struct token
*token
, int attr
)
1607 struct token
*arglist
, *expansion
;
1608 struct token
*left
= token
->next
;
1611 if (token_type(left
) != TOKEN_IDENT
) {
1612 sparse_error(token
->pos
, "expected identifier to 'define'");
1619 expansion
= left
->next
;
1620 if (!expansion
->pos
.whitespace
) {
1621 if (match_op(expansion
, '(')) {
1622 arglist
= expansion
;
1623 expansion
= parse_arguments(expansion
);
1626 } else if (!eof_token(expansion
)) {
1627 warning(expansion
->pos
,
1628 "no whitespace before object-like macro body");
1632 return do_define(left
->pos
, token
, name
, arglist
, expansion
, attr
);
1635 static int handle_define(struct stream
*stream
, struct token
**line
, struct token
*token
)
1637 return do_handle_define(stream
, line
, token
, SYM_ATTR_NORMAL
);
1640 static int handle_weak_define(struct stream
*stream
, struct token
**line
, struct token
*token
)
1642 return do_handle_define(stream
, line
, token
, SYM_ATTR_WEAK
);
1645 static int handle_strong_define(struct stream
*stream
, struct token
**line
, struct token
*token
)
1647 return do_handle_define(stream
, line
, token
, SYM_ATTR_STRONG
);
1650 static int do_handle_undef(struct stream
*stream
, struct token
**line
, struct token
*token
, int attr
)
1652 struct token
*left
= token
->next
;
1655 if (token_type(left
) != TOKEN_IDENT
) {
1656 sparse_error(token
->pos
, "expected identifier to 'undef'");
1660 sym
= lookup_symbol(left
->ident
, NS_MACRO
| NS_UNDEF
);
1662 if (attr
< sym
->attr
)
1664 if (attr
== sym
->attr
&& sym
->namespace == NS_UNDEF
)
1666 } else if (attr
<= SYM_ATTR_NORMAL
)
1669 if (!sym
|| sym
->scope
!= file_scope
) {
1670 sym
= alloc_symbol(left
->pos
, SYM_NODE
);
1671 bind_symbol(sym
, left
->ident
, NS_MACRO
);
1674 sym
->namespace = NS_UNDEF
;
1675 sym
->used_in
= NULL
;
1681 static int handle_undef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1683 return do_handle_undef(stream
, line
, token
, SYM_ATTR_NORMAL
);
1686 static int handle_strong_undef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1688 return do_handle_undef(stream
, line
, token
, SYM_ATTR_STRONG
);
1691 static int preprocessor_if(struct stream
*stream
, struct token
*token
, int cond
)
1693 token_type(token
) = false_nesting
? TOKEN_SKIP_GROUPS
: TOKEN_IF
;
1694 free_preprocessor_line(token
->next
);
1695 token
->next
= stream
->top_if
;
1696 stream
->top_if
= token
;
1697 if (false_nesting
|| cond
!= 1)
1702 static int handle_ifdef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1704 struct token
*next
= token
->next
;
1706 if (token_type(next
) == TOKEN_IDENT
) {
1707 arg
= token_defined(next
);
1709 dirty_stream(stream
);
1711 sparse_error(token
->pos
, "expected preprocessor identifier");
1714 return preprocessor_if(stream
, token
, arg
);
1717 static int handle_ifndef(struct stream
*stream
, struct token
**line
, struct token
*token
)
1719 struct token
*next
= token
->next
;
1721 if (token_type(next
) == TOKEN_IDENT
) {
1722 if (!stream
->dirty
&& !stream
->ifndef
) {
1723 if (!stream
->protect
) {
1724 stream
->ifndef
= token
;
1725 stream
->protect
= next
->ident
;
1726 } else if (stream
->protect
== next
->ident
) {
1727 stream
->ifndef
= token
;
1731 arg
= !token_defined(next
);
1733 dirty_stream(stream
);
1735 sparse_error(token
->pos
, "expected preprocessor identifier");
1739 return preprocessor_if(stream
, token
, arg
);
1743 * Expression handling for #if and #elif; it differs from normal expansion
1744 * due to special treatment of "defined".
1746 static int expression_value(struct token
**where
)
1748 struct expression
*expr
;
1750 struct token
**list
= where
, **beginning
= NULL
;
1754 while (!eof_token(p
= scan_next(list
))) {
1757 if (token_type(p
) != TOKEN_IDENT
)
1759 if (p
->ident
== &defined_ident
) {
1764 if (!expand_one_symbol(list
))
1766 if (token_type(p
) != TOKEN_IDENT
)
1768 token_type(p
) = TOKEN_ZERO_IDENT
;
1771 if (match_op(p
, '(')) {
1775 replace_with_defined(p
);
1780 if (token_type(p
) == TOKEN_IDENT
)
1784 replace_with_defined(p
);
1789 if (!match_op(p
, ')'))
1790 sparse_error(p
->pos
, "missing ')' after \"defined\"");
1797 p
= constant_expression(*where
, &expr
);
1799 sparse_error(p
->pos
, "garbage at end: %s", show_token_sequence(p
, 0));
1800 value
= get_expression_value(expr
);
1804 static int handle_if(struct stream
*stream
, struct token
**line
, struct token
*token
)
1808 value
= expression_value(&token
->next
);
1810 dirty_stream(stream
);
1811 return preprocessor_if(stream
, token
, value
);
1814 static int handle_elif(struct stream
* stream
, struct token
**line
, struct token
*token
)
1816 struct token
*top_if
= stream
->top_if
;
1820 nesting_error(stream
);
1821 sparse_error(token
->pos
, "unmatched #elif within stream");
1825 if (token_type(top_if
) == TOKEN_ELSE
) {
1826 nesting_error(stream
);
1827 sparse_error(token
->pos
, "#elif after #else");
1833 dirty_stream(stream
);
1834 if (token_type(top_if
) != TOKEN_IF
)
1836 if (false_nesting
) {
1838 if (!expression_value(&token
->next
))
1842 token_type(top_if
) = TOKEN_SKIP_GROUPS
;
1847 static int handle_else(struct stream
*stream
, struct token
**line
, struct token
*token
)
1849 struct token
*top_if
= stream
->top_if
;
1853 nesting_error(stream
);
1854 sparse_error(token
->pos
, "unmatched #else within stream");
1858 if (token_type(top_if
) == TOKEN_ELSE
) {
1859 nesting_error(stream
);
1860 sparse_error(token
->pos
, "#else after #else");
1862 if (false_nesting
) {
1863 if (token_type(top_if
) == TOKEN_IF
)
1868 token_type(top_if
) = TOKEN_ELSE
;
1872 static int handle_endif(struct stream
*stream
, struct token
**line
, struct token
*token
)
1874 struct token
*top_if
= stream
->top_if
;
1877 nesting_error(stream
);
1878 sparse_error(token
->pos
, "unmatched #endif in stream");
1883 stream
->top_if
= top_if
->next
;
1884 __free_token(top_if
);
1888 static int handle_warning(struct stream
*stream
, struct token
**line
, struct token
*token
)
1890 warning(token
->pos
, "%s", show_token_sequence(token
->next
, 0));
1894 static int handle_error(struct stream
*stream
, struct token
**line
, struct token
*token
)
1896 sparse_error(token
->pos
, "%s", show_token_sequence(token
->next
, 0));
1900 static int handle_nostdinc(struct stream
*stream
, struct token
**line
, struct token
*token
)
1903 * Do we have any non-system includes?
1904 * Clear them out if so..
1906 *sys_includepath
= NULL
;
1910 static inline void update_inc_ptrs(const char ***where
)
1913 if (*where
<= dirafter_includepath
) {
1914 dirafter_includepath
++;
1915 /* If this was the entry that we prepend, don't
1916 * rise the lower entries, even if they are at
1917 * the same level. */
1918 if (where
== &dirafter_includepath
)
1921 if (*where
<= sys_includepath
) {
1923 if (where
== &sys_includepath
)
1926 if (*where
<= isys_includepath
) {
1928 if (where
== &isys_includepath
)
1932 /* angle_includepath is actually never updated, since we
1933 * don't suppport -iquote rught now. May change some day. */
1934 if (*where
<= angle_includepath
) {
1935 angle_includepath
++;
1936 if (where
== &angle_includepath
)
1941 /* Add a path before 'where' and update the pointers associated with the
1942 * includepath array */
1943 static void add_path_entry(struct token
*token
, const char *path
,
1944 const char ***where
)
1949 /* Need one free entry.. */
1950 if (includepath
[INCLUDEPATHS
-2])
1951 error_die(token
->pos
, "too many include path entries");
1953 /* check that this is not a duplicate */
1956 if (strcmp(*dst
, path
) == 0)
1963 update_inc_ptrs(where
);
1966 * Move them all up starting at dst,
1967 * insert the new entry..
1970 const char *tmp
= *dst
;
1977 static int handle_add_include(struct stream
*stream
, struct token
**line
, struct token
*token
)
1980 token
= token
->next
;
1981 if (eof_token(token
))
1983 if (token_type(token
) != TOKEN_STRING
) {
1984 warning(token
->pos
, "expected path string");
1987 add_path_entry(token
, token
->string
->data
, &isys_includepath
);
1991 static int handle_add_isystem(struct stream
*stream
, struct token
**line
, struct token
*token
)
1994 token
= token
->next
;
1995 if (eof_token(token
))
1997 if (token_type(token
) != TOKEN_STRING
) {
1998 sparse_error(token
->pos
, "expected path string");
2001 add_path_entry(token
, token
->string
->data
, &sys_includepath
);
2005 static int handle_add_system(struct stream
*stream
, struct token
**line
, struct token
*token
)
2008 token
= token
->next
;
2009 if (eof_token(token
))
2011 if (token_type(token
) != TOKEN_STRING
) {
2012 sparse_error(token
->pos
, "expected path string");
2015 add_path_entry(token
, token
->string
->data
, &dirafter_includepath
);
2019 /* Add to end on includepath list - no pointer updates */
2020 static void add_dirafter_entry(struct token
*token
, const char *path
)
2022 const char **dst
= includepath
;
2024 /* Need one free entry.. */
2025 if (includepath
[INCLUDEPATHS
-2])
2026 error_die(token
->pos
, "too many include path entries");
2028 /* Add to the end */
2036 static int handle_add_dirafter(struct stream
*stream
, struct token
**line
, struct token
*token
)
2039 token
= token
->next
;
2040 if (eof_token(token
))
2042 if (token_type(token
) != TOKEN_STRING
) {
2043 sparse_error(token
->pos
, "expected path string");
2046 add_dirafter_entry(token
, token
->string
->data
);
2050 static int handle_split_include(struct stream
*stream
, struct token
**line
, struct token
*token
)
2055 * Split the include path. Any directories specified with `-I'
2056 * options before `-I-' are searched only for headers requested with
2057 * `#include "FILE"'; they are not searched for `#include <FILE>'.
2058 * If additional directories are specified with `-I' options after
2059 * the `-I-', those directories are searched for all `#include'
2061 * In addition, `-I-' inhibits the use of the directory of the current
2062 * file directory as the first search directory for `#include "FILE"'.
2064 quote_includepath
= includepath
+1;
2065 angle_includepath
= sys_includepath
;
2070 * We replace "#pragma xxx" with "__pragma__" in the token
2071 * stream. Just as an example.
2073 * We'll just #define that away for now, but the theory here
2074 * is that we can use this to insert arbitrary token sequences
2075 * to turn the pragmas into internal front-end sequences for
2076 * when we actually start caring about them.
2078 * So eventually this will turn into some kind of extended
2079 * __attribute__() like thing, except called __pragma__(xxx).
2081 static int handle_pragma(struct stream
*stream
, struct token
**line
, struct token
*token
)
2083 struct token
*next
= *line
;
2085 if (match_ident(token
->next
, &once_ident
) && eof_token(token
->next
->next
)) {
2089 token
->ident
= &pragma_ident
;
2090 token
->pos
.newline
= 1;
2091 token
->pos
.whitespace
= 1;
2099 * We ignore #line for now.
2101 static int handle_line(struct stream
*stream
, struct token
**line
, struct token
*token
)
2106 static int handle_ident(struct stream
*stream
, struct token
**line
, struct token
*token
)
2111 static int handle_nondirective(struct stream
*stream
, struct token
**line
, struct token
*token
)
2113 sparse_error(token
->pos
, "unrecognized preprocessor line '%s'", show_token_sequence(token
, 0));
2117 static bool expand_has_attribute(struct token
*token
, struct arg
*args
)
2119 struct token
*arg
= args
[0].expanded
;
2122 if (token_type(arg
) != TOKEN_IDENT
) {
2123 sparse_error(arg
->pos
, "identifier expected");
2127 sym
= lookup_symbol(arg
->ident
, NS_KEYWORD
);
2128 replace_with_bool(token
, sym
&& sym
->op
&& sym
->op
->attribute
);
2132 static bool expand_has_builtin(struct token
*token
, struct arg
*args
)
2134 struct token
*arg
= args
[0].expanded
;
2137 if (token_type(arg
) != TOKEN_IDENT
) {
2138 sparse_error(arg
->pos
, "identifier expected");
2142 sym
= lookup_symbol(arg
->ident
, NS_SYMBOL
);
2143 replace_with_bool(token
, sym
&& sym
->builtin
);
2147 static bool expand_has_extension(struct token
*token
, struct arg
*args
)
2149 struct token
*arg
= args
[0].expanded
;
2150 struct ident
*ident
;
2153 if (token_type(arg
) != TOKEN_IDENT
) {
2154 sparse_error(arg
->pos
, "identifier expected");
2159 if (ident
== &c_alignas_ident
)
2161 else if (ident
== &c_alignof_ident
)
2163 else if (ident
== &c_generic_selections_ident
)
2165 else if (ident
== &c_static_assert_ident
)
2168 replace_with_bool(token
, val
);
2172 static bool expand_has_feature(struct token
*token
, struct arg
*args
)
2174 struct token
*arg
= args
[0].expanded
;
2175 struct ident
*ident
;
2178 if (token_type(arg
) != TOKEN_IDENT
) {
2179 sparse_error(arg
->pos
, "identifier expected");
2184 if (standard
>= STANDARD_C11
) {
2185 if (ident
== &c_alignas_ident
)
2187 else if (ident
== &c_alignof_ident
)
2189 else if (ident
== &c_generic_selections_ident
)
2191 else if (ident
== &c_static_assert_ident
)
2195 replace_with_bool(token
, val
);
2199 static void create_arglist(struct symbol
*sym
, int count
)
2201 struct token
*token
;
2202 struct token
**next
;
2207 token
= __alloc_token(0);
2208 token_type(token
) = TOKEN_ARG_COUNT
;
2209 token
->count
.normal
= count
;
2210 sym
->arglist
= token
;
2211 next
= &token
->next
;
2214 struct token
*id
, *uses
;
2215 id
= __alloc_token(0);
2216 token_type(id
) = TOKEN_IDENT
;
2217 uses
= __alloc_token(0);
2218 token_type(uses
) = TOKEN_ARG_COUNT
;
2219 uses
->count
.normal
= 1;
2225 *next
= &eof_token_entry
;
2228 static void init_preprocessor(void)
2231 int stream
= init_stream(NULL
, "preprocessor", -1, includepath
);
2234 int (*handler
)(struct stream
*, struct token
**, struct token
*);
2236 { "define", handle_define
},
2237 { "weak_define", handle_weak_define
},
2238 { "strong_define", handle_strong_define
},
2239 { "undef", handle_undef
},
2240 { "strong_undef", handle_strong_undef
},
2241 { "warning", handle_warning
},
2242 { "error", handle_error
},
2243 { "include", handle_include
},
2244 { "include_next", handle_include_next
},
2245 { "pragma", handle_pragma
},
2246 { "line", handle_line
},
2247 { "ident", handle_ident
},
2249 // our internal preprocessor tokens
2250 { "nostdinc", handle_nostdinc
},
2251 { "add_include", handle_add_include
},
2252 { "add_isystem", handle_add_isystem
},
2253 { "add_system", handle_add_system
},
2254 { "add_dirafter", handle_add_dirafter
},
2255 { "split_include", handle_split_include
},
2256 { "argv_include", handle_argv_include
},
2258 { "ifdef", handle_ifdef
},
2259 { "ifndef", handle_ifndef
},
2260 { "else", handle_else
},
2261 { "endif", handle_endif
},
2262 { "if", handle_if
},
2263 { "elif", handle_elif
},
2267 void (*expand_simple
)(struct token
*);
2268 bool (*expand
)(struct token
*, struct arg
*args
);
2270 { "__LINE__", expand_line
},
2271 { "__FILE__", expand_file
},
2272 { "__BASE_FILE__", expand_basefile
},
2273 { "__DATE__", expand_date
},
2274 { "__TIME__", expand_time
},
2275 { "__COUNTER__", expand_counter
},
2276 { "__INCLUDE_LEVEL__", expand_include_level
},
2277 { "__has_attribute", NULL
, expand_has_attribute
},
2278 { "__has_builtin", NULL
, expand_has_builtin
},
2279 { "__has_extension", NULL
, expand_has_extension
},
2280 { "__has_feature", NULL
, expand_has_feature
},
2283 for (i
= 0; i
< ARRAY_SIZE(normal
); i
++) {
2285 sym
= create_symbol(stream
, normal
[i
].name
, SYM_PREPROCESSOR
, NS_PREPROCESSOR
);
2286 sym
->handler
= normal
[i
].handler
;
2289 for (i
= 0; i
< ARRAY_SIZE(special
); i
++) {
2291 sym
= create_symbol(stream
, special
[i
].name
, SYM_PREPROCESSOR
, NS_PREPROCESSOR
);
2292 sym
->handler
= special
[i
].handler
;
2295 for (i
= 0; i
< ARRAY_SIZE(dynamic
); i
++) {
2297 sym
= create_symbol(stream
, dynamic
[i
].name
, SYM_NODE
, NS_MACRO
);
2298 sym
->expand_simple
= dynamic
[i
].expand_simple
;
2299 if ((sym
->expand
= dynamic
[i
].expand
) != NULL
)
2300 create_arglist(sym
, 1);
2306 static void handle_preprocessor_line(struct stream
*stream
, struct token
**line
, struct token
*start
)
2308 int (*handler
)(struct stream
*, struct token
**, struct token
*);
2309 struct token
*token
= start
->next
;
2311 int is_cond
= 0; // is one of {is,ifdef,ifndef,elif,else,endif}
2313 if (eof_token(token
))
2316 if (token_type(token
) == TOKEN_IDENT
) {
2317 struct symbol
*sym
= lookup_symbol(token
->ident
, NS_PREPROCESSOR
);
2319 handler
= sym
->handler
;
2320 is_normal
= sym
->normal
;
2321 is_cond
= !sym
->normal
;
2323 handler
= handle_nondirective
;
2325 } else if (token_type(token
) == TOKEN_NUMBER
) {
2326 handler
= handle_line
;
2328 handler
= handle_nondirective
;
2332 dirty_stream(stream
);
2338 if (!is_cond
|| Wpedantic
)
2339 warning(start
->pos
, "directive in macro's argument list");
2341 if (!handler(stream
, line
, token
)) /* all set */
2345 free_preprocessor_line(token
);
2348 static void preprocessor_line(struct stream
*stream
, struct token
**line
)
2350 struct token
*start
= *line
, *next
;
2351 struct token
**tp
= &start
->next
;
2355 if (next
->pos
.newline
)
2360 *tp
= &eof_token_entry
;
2361 handle_preprocessor_line(stream
, line
, start
);
2364 static void do_preprocess(struct token
**list
)
2368 while (!eof_token(next
= scan_next(list
))) {
2369 struct stream
*stream
= input_streams
+ next
->pos
.stream
;
2371 if (next
->pos
.newline
&& match_op(next
, '#')) {
2372 if (!next
->pos
.noexpand
) {
2373 preprocessor_line(stream
, list
);
2374 __free_token(next
); /* Free the '#' token */
2379 switch (token_type(next
)) {
2380 case TOKEN_STREAMEND
:
2381 if (stream
->top_if
) {
2382 nesting_error(stream
);
2383 sparse_error(stream
->top_if
->pos
, "unterminated preprocessor conditional");
2384 stream
->top_if
= NULL
;
2388 stream
->constant
= CONSTANT_FILE_YES
;
2392 case TOKEN_STREAMBEGIN
:
2398 dirty_stream(stream
);
2399 if (false_nesting
) {
2405 if (token_type(next
) != TOKEN_IDENT
||
2406 expand_one_symbol(list
))
2412 void init_include_path(void)
2417 struct utsname name
;
2419 error
= uname(&name
);
2423 if (strcmp(name
.sysname
, "Linux") != 0)
2425 strcpy(os
, "linux-gnu");
2427 snprintf(path
, sizeof(path
), "/usr/include/%s-%s/",
2429 add_pre_buffer("#add_system \"%s/\"\n", path
);
2432 struct token
* preprocess(struct token
*token
)
2435 init_preprocessor();
2436 do_preprocess(&token
);
2438 // Drop all expressions from preprocessing, they're not used any more.
2439 // This is not true when we have multiple files, though ;/
2440 // clear_expression_alloc();
2446 static int is_VA_ARGS_token(struct token
*token
)
2448 return (token_type(token
) == TOKEN_IDENT
) &&
2449 (token
->ident
== &__VA_ARGS___ident
);
2452 static void dump_macro(struct symbol
*sym
)
2454 int nargs
= sym
->arglist
? sym
->arglist
->count
.normal
: 0;
2455 struct token
*args
[nargs
];
2456 struct token
*token
;
2458 printf("#define %s", show_ident(sym
->ident
));
2459 token
= sym
->arglist
;
2461 const char *sep
= "";
2464 for (; !eof_token(token
); token
= token
->next
) {
2465 if (token_type(token
) == TOKEN_ARG_COUNT
)
2467 if (is_VA_ARGS_token(token
))
2468 printf("%s...", sep
);
2470 printf("%s%s", sep
, show_token(token
));
2471 args
[narg
++] = token
;
2477 token
= sym
->expansion
;
2478 while (token_type(token
) != TOKEN_UNTAINT
) {
2479 struct token
*next
= token
->next
;
2480 if (token
->pos
.whitespace
)
2482 switch (token_type(token
)) {
2486 case TOKEN_STR_ARGUMENT
:
2489 case TOKEN_QUOTED_ARGUMENT
:
2490 case TOKEN_MACRO_ARGUMENT
:
2491 token
= args
[token
->argnum
];
2494 printf("%s", show_token(token
));
2501 void dump_macro_definitions(void)
2505 FOR_EACH_PTR(macros
, name
) {
2506 struct symbol
*sym
= lookup_macro(name
);
2509 } END_FOR_EACH_PTR(name
);