2 * TCC - Tiny C Compiler
4 * Copyright (c) 2001-2004 Fabrice Bellard
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 /********************************************************/
24 /* global variables */
26 ST_DATA
int tok_flags
;
27 ST_DATA
int parse_flags
;
29 ST_DATA
struct BufferedFile
*file
;
32 ST_DATA
const int *macro_ptr
;
33 ST_DATA CString tokcstr
; /* current parsed string, if any */
35 /* display benchmark infos */
36 ST_DATA
int total_lines
;
37 ST_DATA
int total_bytes
;
38 ST_DATA
int tok_ident
;
39 ST_DATA TokenSym
**table_ident
;
41 /* ------------------------------------------------------------------------- */
43 static int *macro_ptr_allocated
;
44 static const int *unget_saved_macro_ptr
;
45 static int unget_saved_buffer
[TOK_MAX_SIZE
+ 1];
46 static int unget_buffer_enabled
;
47 static TokenSym
*hash_ident
[TOK_HASH_SIZE
];
48 static char token_buf
[STRING_MAX_SIZE
+ 1];
49 /* true if isid(c) || isnum(c) */
50 static unsigned char isidnum_table
[256-CH_EOF
];
52 static const char tcc_keywords
[] =
53 #define DEF(id, str) str "\0"
58 /* WARNING: the content of this string encodes token numbers */
59 static const unsigned char tok_two_chars
[] =
61 "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
62 "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
83 '.','.', 0xa8, // C++ token ?
84 '#','#', TOK_TWOSHARPS
,
89 struct macro_level
*prev
;
93 static void next_nomacro_spc(void);
94 static void macro_subst(
98 struct macro_level
**can_read_stream
101 ST_FUNC
void skip(int c
)
104 tcc_error("'%c' expected (got \"%s\")", c
, get_tok_str(tok
, &tokc
));
108 ST_FUNC
void expect(const char *msg
)
110 tcc_error("%s expected", msg
);
113 /* ------------------------------------------------------------------------- */
114 /* CString handling */
115 static void cstr_realloc(CString
*cstr
, int new_size
)
120 size
= cstr
->size_allocated
;
122 size
= 8; /* no need to allocate a too small first string */
123 while (size
< new_size
)
125 data
= tcc_realloc(cstr
->data_allocated
, size
);
126 cstr
->data_allocated
= data
;
127 cstr
->size_allocated
= size
;
132 ST_FUNC
void cstr_ccat(CString
*cstr
, int ch
)
135 size
= cstr
->size
+ 1;
136 if (size
> cstr
->size_allocated
)
137 cstr_realloc(cstr
, size
);
138 ((unsigned char *)cstr
->data
)[size
- 1] = ch
;
142 ST_FUNC
void cstr_cat(CString
*cstr
, const char *str
)
154 /* add a wide char */
155 ST_FUNC
void cstr_wccat(CString
*cstr
, int ch
)
158 size
= cstr
->size
+ sizeof(nwchar_t
);
159 if (size
> cstr
->size_allocated
)
160 cstr_realloc(cstr
, size
);
161 *(nwchar_t
*)(((unsigned char *)cstr
->data
) + size
- sizeof(nwchar_t
)) = ch
;
165 ST_FUNC
void cstr_new(CString
*cstr
)
167 memset(cstr
, 0, sizeof(CString
));
170 /* free string and reset it to NULL */
171 ST_FUNC
void cstr_free(CString
*cstr
)
173 tcc_free(cstr
->data_allocated
);
177 /* reset string to empty */
178 ST_FUNC
void cstr_reset(CString
*cstr
)
184 static void add_char(CString
*cstr
, int c
)
186 if (c
== '\'' || c
== '\"' || c
== '\\') {
187 /* XXX: could be more precise if char or string */
188 cstr_ccat(cstr
, '\\');
190 if (c
>= 32 && c
<= 126) {
193 cstr_ccat(cstr
, '\\');
195 cstr_ccat(cstr
, 'n');
197 cstr_ccat(cstr
, '0' + ((c
>> 6) & 7));
198 cstr_ccat(cstr
, '0' + ((c
>> 3) & 7));
199 cstr_ccat(cstr
, '0' + (c
& 7));
204 /* ------------------------------------------------------------------------- */
205 /* allocate a new token */
206 static TokenSym
*tok_alloc_new(TokenSym
**pts
, const char *str
, int len
)
208 TokenSym
*ts
, **ptable
;
211 if (tok_ident
>= SYM_FIRST_ANOM
)
212 tcc_error("memory full (symbols)");
214 /* expand token table if needed */
215 i
= tok_ident
- TOK_IDENT
;
216 if ((i
% TOK_ALLOC_INCR
) == 0) {
217 ptable
= tcc_realloc(table_ident
, (i
+ TOK_ALLOC_INCR
) * sizeof(TokenSym
*));
218 table_ident
= ptable
;
221 ts
= tcc_malloc(sizeof(TokenSym
) + len
);
223 ts
->tok
= tok_ident
++;
224 ts
->sym_define
= NULL
;
225 ts
->sym_label
= NULL
;
226 ts
->sym_struct
= NULL
;
227 ts
->sym_identifier
= NULL
;
229 ts
->hash_next
= NULL
;
230 memcpy(ts
->str
, str
, len
);
236 #define TOK_HASH_INIT 1
237 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
239 /* find a token and add it if not found */
240 ST_FUNC TokenSym
*tok_alloc(const char *str
, int len
)
248 h
= TOK_HASH_FUNC(h
, ((unsigned char *)str
)[i
]);
249 h
&= (TOK_HASH_SIZE
- 1);
251 pts
= &hash_ident
[h
];
256 if (ts
->len
== len
&& !memcmp(ts
->str
, str
, len
))
258 pts
= &(ts
->hash_next
);
260 return tok_alloc_new(pts
, str
, len
);
263 /* XXX: buffer overflow */
264 /* XXX: float tokens */
265 ST_FUNC
char *get_tok_str(int v
, CValue
*cv
)
267 static char buf
[STRING_MAX_SIZE
+ 1];
268 static CString cstr_buf
;
273 /* NOTE: to go faster, we give a fixed buffer for small strings */
274 cstr_reset(&cstr_buf
);
276 cstr_buf
.size_allocated
= sizeof(buf
);
279 /* just an explanation, should never happen:
280 if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
281 tcc_error("internal error: get_tok_str"); */
286 /* XXX: not quite exact, but only useful for testing */
287 sprintf(p
, "%u", cv
->ui
);
291 /* XXX: not quite exact, but only useful for testing */
293 sprintf(p
, "%u", (unsigned)cv
->ull
);
295 sprintf(p
, "%llu", cv
->ull
);
299 cstr_ccat(&cstr_buf
, 'L');
301 cstr_ccat(&cstr_buf
, '\'');
302 add_char(&cstr_buf
, cv
->i
);
303 cstr_ccat(&cstr_buf
, '\'');
304 cstr_ccat(&cstr_buf
, '\0');
308 len
= cstr
->size
- 1;
310 add_char(&cstr_buf
, ((unsigned char *)cstr
->data
)[i
]);
311 cstr_ccat(&cstr_buf
, '\0');
314 cstr_ccat(&cstr_buf
, 'L');
317 cstr_ccat(&cstr_buf
, '\"');
319 len
= cstr
->size
- 1;
321 add_char(&cstr_buf
, ((unsigned char *)cstr
->data
)[i
]);
323 len
= (cstr
->size
/ sizeof(nwchar_t
)) - 1;
325 add_char(&cstr_buf
, ((nwchar_t
*)cstr
->data
)[i
]);
327 cstr_ccat(&cstr_buf
, '\"');
328 cstr_ccat(&cstr_buf
, '\0');
335 return NULL
; /* should not happen */
337 /* above tokens have value, the ones below don't */
346 return strcpy(p
, "...");
348 return strcpy(p
, "<<=");
350 return strcpy(p
, ">>=");
353 /* search in two bytes table */
354 const unsigned char *q
= tok_two_chars
;
367 } else if (v
< tok_ident
) {
368 return table_ident
[v
- TOK_IDENT
]->str
;
369 } else if (v
>= SYM_FIRST_ANOM
) {
370 /* special name for anonymous symbol */
371 sprintf(p
, "L.%u", v
- SYM_FIRST_ANOM
);
373 /* should never happen */
378 return cstr_buf
.data
;
381 /* fill input buffer and peek next char */
382 static int tcc_peekc_slow(BufferedFile
*bf
)
385 /* only tries to read if really end of buffer */
386 if (bf
->buf_ptr
>= bf
->buf_end
) {
388 #if defined(PARSE_DEBUG)
393 len
= read(bf
->fd
, bf
->buffer
, len
);
400 bf
->buf_ptr
= bf
->buffer
;
401 bf
->buf_end
= bf
->buffer
+ len
;
402 *bf
->buf_end
= CH_EOB
;
404 if (bf
->buf_ptr
< bf
->buf_end
) {
405 return bf
->buf_ptr
[0];
407 bf
->buf_ptr
= bf
->buf_end
;
412 /* return the current character, handling end of block if necessary
414 ST_FUNC
int handle_eob(void)
416 return tcc_peekc_slow(file
);
419 /* read next char from current input file and handle end of input buffer */
420 ST_INLN
void inp(void)
422 ch
= *(++(file
->buf_ptr
));
423 /* end of buffer/file handling */
428 /* handle '\[\r]\n' */
429 static int handle_stray_noerror(void)
436 } else if (ch
== '\r') {
450 static void handle_stray(void)
452 if (handle_stray_noerror())
453 tcc_error("stray '\\' in program");
456 /* skip the stray and handle the \\n case. Output an error if
457 incorrect char after the stray */
458 static int handle_stray1(uint8_t *p
)
462 if (p
>= file
->buf_end
) {
479 /* handle just the EOB case, but not stray */
480 #define PEEKC_EOB(c, p)\
491 /* handle the complicated stray case */
497 c = handle_stray1(p);\
502 /* input with '\[\r]\n' handling. Note that this function cannot
503 handle other characters after '\', so you cannot call it inside
504 strings or comments */
505 ST_FUNC
void minp(void)
513 /* single line C++ comments */
514 static uint8_t *parse_line_comment(uint8_t *p
)
522 if (c
== '\n' || c
== CH_EOF
) {
524 } else if (c
== '\\') {
533 } else if (c
== '\r') {
551 ST_FUNC
uint8_t *parse_comment(uint8_t *p
)
560 if (c
== '\n' || c
== '*' || c
== '\\')
564 if (c
== '\n' || c
== '*' || c
== '\\')
568 /* now we can handle all the cases */
572 } else if (c
== '*') {
578 } else if (c
== '/') {
580 } else if (c
== '\\') {
585 /* skip '\[\r]\n', otherwise just skip the stray */
591 } else if (c
== '\r') {
608 /* stray, eob or eof */
613 tcc_error("unexpected end of file in comment");
614 } else if (c
== '\\') {
626 static inline void skip_spaces(void)
632 static inline int check_space(int t
, int *spc
)
643 /* parse a string without interpreting escapes */
644 static uint8_t *parse_pp_string(uint8_t *p
,
645 int sep
, CString
*str
)
653 } else if (c
== '\\') {
659 /* XXX: indicate line number of start of string */
660 tcc_error("missing terminating %c character", sep
);
661 } else if (c
== '\\') {
662 /* escape : just skip \[\r]\n */
667 } else if (c
== '\r') {
670 expect("'\n' after '\r'");
673 } else if (c
== CH_EOF
) {
674 goto unterminated_string
;
677 cstr_ccat(str
, '\\');
683 } else if (c
== '\n') {
686 } else if (c
== '\r') {
690 cstr_ccat(str
, '\r');
706 /* skip block of text until #else, #elif or #endif. skip also pairs of
708 static void preprocess_skip(void)
710 int a
, start_of_line
, c
, in_warn_or_error
;
717 in_warn_or_error
= 0;
738 } else if (c
== '\\') {
739 ch
= file
->buf_ptr
[0];
740 handle_stray_noerror();
747 if (in_warn_or_error
)
749 p
= parse_pp_string(p
, c
, NULL
);
753 if (in_warn_or_error
)
760 p
= parse_comment(p
);
761 } else if (ch
== '/') {
762 p
= parse_line_comment(p
);
772 (tok
== TOK_ELSE
|| tok
== TOK_ELIF
|| tok
== TOK_ENDIF
))
774 if (tok
== TOK_IF
|| tok
== TOK_IFDEF
|| tok
== TOK_IFNDEF
)
776 else if (tok
== TOK_ENDIF
)
778 else if( tok
== TOK_ERROR
|| tok
== TOK_WARNING
)
779 in_warn_or_error
= 1;
780 else if (tok
== TOK_LINEFEED
)
782 else if (parse_flags
& PARSE_FLAG_ASM_COMMENTS
)
783 p
= parse_line_comment(p
);
785 else if (parse_flags
& PARSE_FLAG_ASM_FILE
)
786 p
= parse_line_comment(p
);
799 /* ParseState handling */
801 /* XXX: currently, no include file info is stored. Thus, we cannot display
802 accurate messages if the function or data definition spans multiple
805 /* save current parse state in 's' */
806 ST_FUNC
void save_parse_state(ParseState
*s
)
808 s
->line_num
= file
->line_num
;
809 s
->macro_ptr
= macro_ptr
;
814 /* restore parse state from 's' */
815 ST_FUNC
void restore_parse_state(ParseState
*s
)
817 file
->line_num
= s
->line_num
;
818 macro_ptr
= s
->macro_ptr
;
823 /* return the number of additional 'ints' necessary to store the
825 static inline int tok_ext_size(int t
)
839 tcc_error("unsupported token");
846 return LDOUBLE_SIZE
/ 4;
852 /* token string handling */
854 ST_INLN
void tok_str_new(TokenString
*s
)
858 s
->allocated_len
= 0;
859 s
->last_line_num
= -1;
862 ST_FUNC
void tok_str_free(int *str
)
867 static int *tok_str_realloc(TokenString
*s
)
871 if (s
->allocated_len
== 0) {
874 len
= s
->allocated_len
* 2;
876 str
= tcc_realloc(s
->str
, len
* sizeof(int));
877 s
->allocated_len
= len
;
882 ST_FUNC
void tok_str_add(TokenString
*s
, int t
)
888 if (len
>= s
->allocated_len
)
889 str
= tok_str_realloc(s
);
894 static void tok_str_add2(TokenString
*s
, int t
, CValue
*cv
)
901 /* allocate space for worst case */
902 if (len
+ TOK_MAX_SIZE
> s
->allocated_len
)
903 str
= tok_str_realloc(s
);
912 str
[len
++] = cv
->tab
[0];
921 nb_words
= (sizeof(CString
) + cv
->cstr
->size
+ 3) >> 2;
922 while ((len
+ nb_words
) > s
->allocated_len
)
923 str
= tok_str_realloc(s
);
924 cstr
= (CString
*)(str
+ len
);
926 cstr
->size
= cv
->cstr
->size
;
927 cstr
->data_allocated
= NULL
;
928 cstr
->size_allocated
= cstr
->size
;
929 memcpy((char *)cstr
+ sizeof(CString
),
930 cv
->cstr
->data
, cstr
->size
);
937 #if LDOUBLE_SIZE == 8
940 str
[len
++] = cv
->tab
[0];
941 str
[len
++] = cv
->tab
[1];
943 #if LDOUBLE_SIZE == 12
945 str
[len
++] = cv
->tab
[0];
946 str
[len
++] = cv
->tab
[1];
947 str
[len
++] = cv
->tab
[2];
948 #elif LDOUBLE_SIZE == 16
950 str
[len
++] = cv
->tab
[0];
951 str
[len
++] = cv
->tab
[1];
952 str
[len
++] = cv
->tab
[2];
953 str
[len
++] = cv
->tab
[3];
954 #elif LDOUBLE_SIZE != 8
955 #error add long double size support
964 /* add the current parse token in token string 's' */
965 ST_FUNC
void tok_str_add_tok(TokenString
*s
)
969 /* save line number info */
970 if (file
->line_num
!= s
->last_line_num
) {
971 s
->last_line_num
= file
->line_num
;
972 cval
.i
= s
->last_line_num
;
973 tok_str_add2(s
, TOK_LINENUM
, &cval
);
975 tok_str_add2(s
, tok
, &tokc
);
978 /* get a token from an integer array and increment pointer
979 accordingly. we code it as a macro to avoid pointer aliasing. */
980 static inline void TOK_GET(int *t
, const int **pp
, CValue
*cv
)
998 cv
->cstr
= (CString
*)p
;
999 cv
->cstr
->data
= (char *)p
+ sizeof(CString
);
1000 p
+= (sizeof(CString
) + cv
->cstr
->size
+ 3) >> 2;
1008 #if LDOUBLE_SIZE == 16
1010 #elif LDOUBLE_SIZE == 12
1012 #elif LDOUBLE_SIZE == 8
1015 # error add long double size support
1028 static int macro_is_equal(const int *a
, const int *b
)
1030 char buf
[STRING_MAX_SIZE
+ 1];
1034 TOK_GET(&t
, &a
, &cv
);
1035 pstrcpy(buf
, sizeof buf
, get_tok_str(t
, &cv
));
1036 TOK_GET(&t
, &b
, &cv
);
1037 if (strcmp(buf
, get_tok_str(t
, &cv
)))
1043 static void define_print(Sym
*s
, int is_undef
)
1050 if (tcc_state
->dflag
== 0 || !s
|| !tcc_state
->ppfp
)
1054 c
= file
->line_num
- file
->line_ref
- 1;
1057 fputs("\n", tcc_state
->ppfp
);
1058 file
->line_ref
= file
->line_num
;
1063 fprintf(tcc_state
->ppfp
, "// #undef %s\n", get_tok_str(s
->v
, NULL
));
1067 fprintf(tcc_state
->ppfp
, "// #define %s", get_tok_str(s
->v
, NULL
));
1072 fprintf(tcc_state
->ppfp
, "%s%s", sep
, get_tok_str(arg
->v
& ~SYM_FIELD
, NULL
));
1076 fprintf(tcc_state
->ppfp
, ")");
1081 fprintf(tcc_state
->ppfp
, " ");
1084 TOK_GET(&t
, &str
, &cval
);
1087 fprintf(tcc_state
->ppfp
, "%s", get_tok_str(t
, &cval
));
1089 fprintf(tcc_state
->ppfp
, "\n");
1092 /* defines handling */
1093 ST_INLN
void define_push(int v
, int macro_type
, int *str
, Sym
*first_arg
)
1098 if (s
&& !macro_is_equal(s
->d
, str
))
1099 tcc_warning("%s redefined", get_tok_str(v
, NULL
));
1101 s
= sym_push2(&define_stack
, v
, macro_type
, 0);
1103 s
->next
= first_arg
;
1104 table_ident
[v
- TOK_IDENT
]->sym_define
= s
;
1108 /* undefined a define symbol. Its name is just set to zero */
1109 ST_FUNC
void define_undef(Sym
*s
)
1113 if (v
>= TOK_IDENT
&& v
< tok_ident
) {
1115 table_ident
[v
- TOK_IDENT
]->sym_define
= NULL
;
1120 ST_INLN Sym
*define_find(int v
)
1123 if ((unsigned)v
>= (unsigned)(tok_ident
- TOK_IDENT
))
1125 return table_ident
[v
]->sym_define
;
1128 /* free define stack until top reaches 'b' */
1129 ST_FUNC
void free_defines(Sym
*b
)
1137 /* do not free args or predefined defines */
1139 tok_str_free(top
->d
);
1141 if (v
>= TOK_IDENT
&& v
< tok_ident
)
1142 table_ident
[v
- TOK_IDENT
]->sym_define
= NULL
;
1149 ST_FUNC
void print_defines(void)
1157 if (v
>= TOK_IDENT
&& v
< tok_ident
) {
1158 s
= table_ident
[v
- TOK_IDENT
]->sym_define
;
1166 ST_FUNC Sym
*label_find(int v
)
1169 if ((unsigned)v
>= (unsigned)(tok_ident
- TOK_IDENT
))
1171 return table_ident
[v
]->sym_label
;
1174 ST_FUNC Sym
*label_push(Sym
**ptop
, int v
, int flags
)
1177 s
= sym_push2(ptop
, v
, 0, 0);
1179 ps
= &table_ident
[v
- TOK_IDENT
]->sym_label
;
1180 if (ptop
== &global_label_stack
) {
1181 /* modify the top most local identifier, so that
1182 sym_identifier will point to 's' when popped */
1184 ps
= &(*ps
)->prev_tok
;
1191 /* pop labels until element last is reached. Look if any labels are
1192 undefined. Define symbols if '&&label' was used. */
1193 ST_FUNC
void label_pop(Sym
**ptop
, Sym
*slast
)
1196 for(s
= *ptop
; s
!= slast
; s
= s1
) {
1198 if (s
->r
== LABEL_DECLARED
) {
1199 tcc_warning("label '%s' declared but not used", get_tok_str(s
->v
, NULL
));
1200 } else if (s
->r
== LABEL_FORWARD
) {
1201 tcc_error("label '%s' used but not defined",
1202 get_tok_str(s
->v
, NULL
));
1205 /* define corresponding symbol. A size of
1207 put_extern_sym(s
, cur_text_section
, s
->jnext
, 1);
1211 table_ident
[s
->v
- TOK_IDENT
]->sym_label
= s
->prev_tok
;
1217 /* eval an expression for #if/#elif */
1218 static int expr_preprocess(void)
1224 while (tok
!= TOK_LINEFEED
&& tok
!= TOK_EOF
) {
1225 next(); /* do macro subst */
1226 if (tok
== TOK_DEFINED
) {
1231 c
= define_find(tok
) != 0;
1236 } else if (tok
>= TOK_IDENT
) {
1237 /* if undefined macro */
1241 tok_str_add_tok(&str
);
1243 tok_str_add(&str
, -1); /* simulate end of file */
1244 tok_str_add(&str
, 0);
1245 /* now evaluate C constant expression */
1246 macro_ptr
= str
.str
;
1250 tok_str_free(str
.str
);
1254 /* parse after #define */
1255 ST_FUNC
void parse_define(void)
1257 Sym
*s
, *first
, **ps
;
1258 int v
, t
, varg
, is_vaargs
, spc
, ptok
, macro_list_start
;
1263 tcc_error("invalid macro name '%s'", get_tok_str(tok
, &tokc
));
1264 /* XXX: should check if same macro (ANSI) */
1267 /* '(' must be just after macro definition for MACRO_FUNC */
1272 while (tok
!= ')') {
1276 if (varg
== TOK_DOTS
) {
1277 varg
= TOK___VA_ARGS__
;
1279 } else if (tok
== TOK_DOTS
&& gnu_ext
) {
1283 if (varg
< TOK_IDENT
)
1284 tcc_error( "\'%s\' may not appear in parameter list", get_tok_str(varg
, NULL
));
1285 s
= sym_push2(&define_stack
, varg
| SYM_FIELD
, is_vaargs
, 0);
1297 /* EOF testing necessary for '-D' handling */
1299 macro_list_start
= 1;
1300 while (tok
!= TOK_LINEFEED
&& tok
!= TOK_EOF
) {
1301 if (!macro_list_start
&& spc
== 2 && tok
== TOK_TWOSHARPS
)
1302 tcc_error("'##' invalid at start of macro");
1304 /* remove spaces around ## and after '#' */
1305 if (TOK_TWOSHARPS
== tok
) {
1309 } else if ('#' == tok
) {
1311 } else if (check_space(tok
, &spc
)) {
1314 tok_str_add2(&str
, tok
, &tokc
);
1317 macro_list_start
= 0;
1319 if (ptok
== TOK_TWOSHARPS
)
1320 tcc_error("'##' invalid at end of macro");
1322 --str
.len
; /* remove trailing space */
1323 tok_str_add(&str
, 0);
1324 define_push(v
, t
, str
.str
, first
);
1327 static inline int hash_cached_include(const char *filename
)
1329 const unsigned char *s
;
1333 s
= (unsigned char *) filename
;
1335 h
= TOK_HASH_FUNC(h
, *s
);
1338 h
&= (CACHED_INCLUDES_HASH_SIZE
- 1);
1342 static CachedInclude
*search_cached_include(TCCState
*s1
, const char *filename
)
1346 h
= hash_cached_include(filename
);
1347 i
= s1
->cached_includes_hash
[h
];
1351 e
= s1
->cached_includes
[i
- 1];
1352 if (0 == PATHCMP(e
->filename
, filename
))
1359 static inline void add_cached_include(TCCState
*s1
, const char *filename
, int ifndef_macro
)
1364 if (search_cached_include(s1
, filename
))
1367 printf("adding cached '%s' %s\n", filename
, get_tok_str(ifndef_macro
, NULL
));
1369 e
= tcc_malloc(sizeof(CachedInclude
) + strlen(filename
));
1370 strcpy(e
->filename
, filename
);
1371 e
->ifndef_macro
= ifndef_macro
;
1372 dynarray_add((void ***)&s1
->cached_includes
, &s1
->nb_cached_includes
, e
);
1373 /* add in hash table */
1374 h
= hash_cached_include(filename
);
1375 e
->hash_next
= s1
->cached_includes_hash
[h
];
1376 s1
->cached_includes_hash
[h
] = s1
->nb_cached_includes
;
1379 static void pragma_parse(TCCState
*s1
)
1384 if (tok
== TOK_pack
) {
1387 #pragma pack(1) // set
1388 #pragma pack() // reset to default
1389 #pragma pack(push,1) // push & set
1390 #pragma pack(pop) // restore previous
1394 if (tok
== TOK_ASM_pop
) {
1396 if (s1
->pack_stack_ptr
<= s1
->pack_stack
) {
1398 tcc_error("out of pack stack");
1400 s1
->pack_stack_ptr
--;
1404 if (tok
== TOK_ASM_push
) {
1406 if (s1
->pack_stack_ptr
>= s1
->pack_stack
+ PACK_STACK_SIZE
- 1)
1408 s1
->pack_stack_ptr
++;
1411 if (tok
!= TOK_CINT
) {
1413 tcc_error("invalid pack pragma");
1416 if (val
< 1 || val
> 16 || (val
& (val
- 1)) != 0)
1420 *s1
->pack_stack_ptr
= val
;
1423 } else if (tok
== TOK_comment
) {
1424 if (s1
->ms_extensions
) {
1427 if (tok
== TOK_lib
) {
1430 if (tok
!= TOK_STR
) {
1431 tcc_error("invalid library specification");
1433 int len
= strlen((char *)tokc
.cstr
->data
);
1434 char *file
= tcc_malloc(len
+ 4); /* filetype, "-l", and \0 at the end */
1435 file
[0] = TCC_FILETYPE_BINARY
;
1438 strcpy(&file
[3],(char *)tokc
.cstr
->data
);
1439 dynarray_add((void ***)&s1
->files
, &s1
->nb_files
, file
);
1444 tcc_warning("unknown specifier '%s' in #pragma comment", get_tok_str(tok
, &tokc
));
1447 tcc_warning("#pragma comment(lib) is ignored");
1449 } else if (tok
== TOK_once
) {
1450 add_cached_include(s1
, file
->filename
, TOK_once
);
1452 tcc_warning("unknown #pragma %s", get_tok_str(tok
, &tokc
));
1456 /* is_bof is true if first non space token at beginning of file */
1457 ST_FUNC
void preprocess(int is_bof
)
1459 TCCState
*s1
= tcc_state
;
1460 int i
, c
, n
, saved_parse_flags
;
1464 saved_parse_flags
= parse_flags
;
1465 parse_flags
= PARSE_FLAG_PREPROCESS
| PARSE_FLAG_TOK_NUM
|
1466 PARSE_FLAG_LINEFEED
;
1467 parse_flags
|= (saved_parse_flags
& (PARSE_FLAG_ASM_FILE
| PARSE_FLAG_ASM_COMMENTS
));
1477 s
= define_find(tok
);
1478 /* undefine symbol by putting an invalid name */
1483 case TOK_INCLUDE_NEXT
:
1484 ch
= file
->buf_ptr
[0];
1485 /* XXX: incorrect if comments : use next_nomacro with a special mode */
1490 } else if (ch
== '\"') {
1495 while (ch
!= c
&& ch
!= '\n' && ch
!= CH_EOF
) {
1496 if ((q
- buf
) < sizeof(buf
) - 1)
1499 if (handle_stray_noerror() == 0)
1507 /* eat all spaces and comments after include */
1508 /* XXX: slightly incorrect */
1509 while (ch1
!= '\n' && ch1
!= CH_EOF
)
1513 /* computed #include : either we have only strings or
1514 we have anything enclosed in '<>' */
1517 if (tok
== TOK_STR
) {
1518 while (tok
!= TOK_LINEFEED
) {
1519 if (tok
!= TOK_STR
) {
1521 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1523 pstrcat(buf
, sizeof(buf
), (char *)tokc
.cstr
->data
);
1529 while (tok
!= TOK_LINEFEED
) {
1530 pstrcat(buf
, sizeof(buf
), get_tok_str(tok
, &tokc
));
1534 /* check syntax and remove '<>' */
1535 if (len
< 2 || buf
[0] != '<' || buf
[len
- 1] != '>')
1536 goto include_syntax
;
1537 memmove(buf
, buf
+ 1, len
- 2);
1538 buf
[len
- 2] = '\0';
1543 if (s1
->include_stack_ptr
>= s1
->include_stack
+ INCLUDE_STACK_SIZE
)
1544 tcc_error("#include recursion too deep");
1545 /* store current file in stack, but increment stack later below */
1546 *s1
->include_stack_ptr
= file
;
1548 n
= s1
->nb_include_paths
+ s1
->nb_sysinclude_paths
;
1549 for (i
= -2; i
< n
; ++i
) {
1550 char buf1
[sizeof file
->filename
];
1556 /* check absolute include path */
1557 if (!IS_ABSPATH(buf
))
1560 i
= n
; /* force end loop */
1562 } else if (i
== -1) {
1563 /* search in current dir if "header.h" */
1566 path
= file
->filename
;
1567 pstrncpy(buf1
, path
, tcc_basename(path
) - path
);
1570 /* search in all the include paths */
1571 if (i
< s1
->nb_include_paths
)
1572 path
= s1
->include_paths
[i
];
1574 path
= s1
->sysinclude_paths
[i
- s1
->nb_include_paths
];
1575 pstrcpy(buf1
, sizeof(buf1
), path
);
1576 pstrcat(buf1
, sizeof(buf1
), "/");
1579 pstrcat(buf1
, sizeof(buf1
), buf
);
1581 if (tok
== TOK_INCLUDE_NEXT
)
1582 for (f
= s1
->include_stack_ptr
; f
>= s1
->include_stack
; --f
)
1583 if (0 == PATHCMP((*f
)->filename
, buf1
)) {
1585 printf("%s: #include_next skipping %s\n", file
->filename
, buf1
);
1587 goto include_trynext
;
1590 e
= search_cached_include(s1
, buf1
);
1591 if (e
&& (define_find(e
->ifndef_macro
) || e
->ifndef_macro
== TOK_once
)) {
1592 /* no need to parse the include because the 'ifndef macro'
1595 printf("%s: skipping cached %s\n", file
->filename
, buf1
);
1600 if (tcc_open(s1
, buf1
) < 0)
1605 printf("%s: including %s\n", file
->prev
->filename
, file
->filename
);
1607 /* update target deps */
1608 dynarray_add((void ***)&s1
->target_deps
, &s1
->nb_target_deps
,
1610 /* push current file in stack */
1611 ++s1
->include_stack_ptr
;
1612 /* add include file debug info */
1614 put_stabs(file
->filename
, N_BINCL
, 0, 0, 0);
1615 tok_flags
|= TOK_FLAG_BOF
| TOK_FLAG_BOL
;
1616 ch
= file
->buf_ptr
[0];
1619 tcc_error("include file '%s' not found", buf
);
1626 c
= expr_preprocess();
1632 if (tok
< TOK_IDENT
)
1633 tcc_error("invalid argument for '#if%sdef'", c
? "n" : "");
1637 printf("#ifndef %s\n", get_tok_str(tok
, NULL
));
1639 file
->ifndef_macro
= tok
;
1642 c
= (define_find(tok
) != 0) ^ c
;
1644 if (s1
->ifdef_stack_ptr
>= s1
->ifdef_stack
+ IFDEF_STACK_SIZE
)
1645 tcc_error("memory full (ifdef)");
1646 *s1
->ifdef_stack_ptr
++ = c
;
1649 if (s1
->ifdef_stack_ptr
== s1
->ifdef_stack
)
1650 tcc_error("#else without matching #if");
1651 if (s1
->ifdef_stack_ptr
[-1] & 2)
1652 tcc_error("#else after #else");
1653 c
= (s1
->ifdef_stack_ptr
[-1] ^= 3);
1656 if (s1
->ifdef_stack_ptr
== s1
->ifdef_stack
)
1657 tcc_error("#elif without matching #if");
1658 c
= s1
->ifdef_stack_ptr
[-1];
1660 tcc_error("#elif after #else");
1661 /* last #if/#elif expression was true: we skip */
1664 c
= expr_preprocess();
1665 s1
->ifdef_stack_ptr
[-1] = c
;
1667 if (s1
->ifdef_stack_ptr
== file
->ifdef_stack_ptr
+ 1)
1668 file
->ifndef_macro
= 0;
1678 if (s1
->ifdef_stack_ptr
<= file
->ifdef_stack_ptr
)
1679 tcc_error("#endif without matching #if");
1680 s1
->ifdef_stack_ptr
--;
1681 /* '#ifndef macro' was at the start of file. Now we check if
1682 an '#endif' is exactly at the end of file */
1683 if (file
->ifndef_macro
&&
1684 s1
->ifdef_stack_ptr
== file
->ifdef_stack_ptr
) {
1685 file
->ifndef_macro_saved
= file
->ifndef_macro
;
1686 /* need to set to zero to avoid false matches if another
1687 #ifndef at middle of file */
1688 file
->ifndef_macro
= 0;
1689 while (tok
!= TOK_LINEFEED
)
1691 tok_flags
|= TOK_FLAG_ENDIF
;
1697 if (tok
!= TOK_CINT
)
1698 tcc_error("A #line format is wrong");
1700 if (tok
!= TOK_CINT
) {
1701 char *p
= tokc
.cstr
->data
;
1702 tokc
.i
= strtoul(p
, (char **)&p
, 10);
1705 file
->line_num
= tokc
.i
- 1;
1707 if (tok
!= TOK_LINEFEED
) {
1708 if (tok
!= TOK_STR
) {
1709 if ((parse_flags
& PARSE_FLAG_ASM_COMMENTS
) == 0) {
1711 tcc_error("#line format is wrong");
1715 pstrcpy(file
->filename
, sizeof(file
->filename
),
1716 (char *)tokc
.cstr
->data
);
1719 put_stabs(file
->filename
, N_BINCL
, 0, 0, 0);
1724 ch
= file
->buf_ptr
[0];
1727 while (ch
!= '\n' && ch
!= CH_EOF
) {
1728 if ((q
- buf
) < sizeof(buf
) - 1)
1731 if (handle_stray_noerror() == 0)
1738 tcc_error("#error %s", buf
);
1740 tcc_warning("#warning %s", buf
);
1746 if (tok
== TOK_LINEFEED
|| tok
== '!' || tok
== TOK_PPNUM
) {
1747 /* '!' is ignored to allow C scripts. numbers are ignored
1748 to emulate cpp behaviour */
1750 if (!(parse_flags
& PARSE_FLAG_ASM_COMMENTS
))
1751 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok
, &tokc
));
1753 /* this is a gas line comment in an 'S' file. */
1754 file
->buf_ptr
= parse_line_comment(file
->buf_ptr
);
1760 /* ignore other preprocess commands or #! for C scripts */
1761 while (tok
!= TOK_LINEFEED
)
1764 parse_flags
= saved_parse_flags
;
1767 /* evaluate escape codes in a string. */
1768 static void parse_escape_string(CString
*outstr
, const uint8_t *buf
, int is_long
)
1783 case '0': case '1': case '2': case '3':
1784 case '4': case '5': case '6': case '7':
1785 /* at most three octal digits */
1790 n
= n
* 8 + c
- '0';
1794 n
= n
* 8 + c
- '0';
1799 goto add_char_nonext
;
1807 if (c
>= 'a' && c
<= 'f')
1809 else if (c
>= 'A' && c
<= 'F')
1819 goto add_char_nonext
;
1843 goto invalid_escape
;
1853 if (c
>= '!' && c
<= '~')
1854 tcc_warning("unknown escape sequence: \'\\%c\'", c
);
1856 tcc_warning("unknown escape sequence: \'\\x%x\'", c
);
1863 cstr_ccat(outstr
, c
);
1865 cstr_wccat(outstr
, c
);
1867 /* add a trailing '\0' */
1869 cstr_ccat(outstr
, '\0');
1871 cstr_wccat(outstr
, '\0');
1874 /* we use 64 bit numbers */
1877 /* bn = (bn << shift) | or_val */
1878 static void bn_lshift(unsigned int *bn
, int shift
, int or_val
)
1882 for(i
=0;i
<BN_SIZE
;i
++) {
1884 bn
[i
] = (v
<< shift
) | or_val
;
1885 or_val
= v
>> (32 - shift
);
1889 static void bn_zero(unsigned int *bn
)
1892 for(i
=0;i
<BN_SIZE
;i
++) {
1897 /* parse number in null terminated string 'p' and return it in the
1899 static void parse_number(const char *p
)
1901 int b
, t
, shift
, frac_bits
, s
, exp_val
, ch
;
1903 unsigned int bn
[BN_SIZE
];
1914 goto float_frac_parse
;
1915 } else if (t
== '0') {
1916 if (ch
== 'x' || ch
== 'X') {
1920 } else if (tcc_ext
&& (ch
== 'b' || ch
== 'B')) {
1926 /* parse all digits. cannot check octal numbers at this stage
1927 because of floating point constants */
1929 if (ch
>= 'a' && ch
<= 'f')
1931 else if (ch
>= 'A' && ch
<= 'F')
1939 if (q
>= token_buf
+ STRING_MAX_SIZE
) {
1941 tcc_error("number too long");
1947 ((ch
== 'e' || ch
== 'E') && b
== 10) ||
1948 ((ch
== 'p' || ch
== 'P') && (b
== 16 || b
== 2))) {
1950 /* NOTE: strtox should support that for hexa numbers, but
1951 non ISOC99 libcs do not support it, so we prefer to do
1953 /* hexadecimal or binary floats */
1954 /* XXX: handle overflows */
1966 } else if (t
>= 'a') {
1968 } else if (t
>= 'A') {
1973 bn_lshift(bn
, shift
, t
);
1980 if (t
>= 'a' && t
<= 'f') {
1982 } else if (t
>= 'A' && t
<= 'F') {
1984 } else if (t
>= '0' && t
<= '9') {
1990 tcc_error("invalid digit");
1991 bn_lshift(bn
, shift
, t
);
1996 if (ch
!= 'p' && ch
!= 'P')
2003 } else if (ch
== '-') {
2007 if (ch
< '0' || ch
> '9')
2008 expect("exponent digits");
2009 while (ch
>= '0' && ch
<= '9') {
2010 exp_val
= exp_val
* 10 + ch
- '0';
2013 exp_val
= exp_val
* s
;
2015 /* now we can generate the number */
2016 /* XXX: should patch directly float number */
2017 d
= (double)bn
[1] * 4294967296.0 + (double)bn
[0];
2018 d
= ldexp(d
, exp_val
- frac_bits
);
2023 /* float : should handle overflow */
2025 } else if (t
== 'L') {
2027 #ifdef TCC_TARGET_PE
2032 /* XXX: not large enough */
2033 tokc
.ld
= (long double)d
;
2040 /* decimal floats */
2042 if (q
>= token_buf
+ STRING_MAX_SIZE
)
2047 while (ch
>= '0' && ch
<= '9') {
2048 if (q
>= token_buf
+ STRING_MAX_SIZE
)
2054 if (ch
== 'e' || ch
== 'E') {
2055 if (q
>= token_buf
+ STRING_MAX_SIZE
)
2059 if (ch
== '-' || ch
== '+') {
2060 if (q
>= token_buf
+ STRING_MAX_SIZE
)
2065 if (ch
< '0' || ch
> '9')
2066 expect("exponent digits");
2067 while (ch
>= '0' && ch
<= '9') {
2068 if (q
>= token_buf
+ STRING_MAX_SIZE
)
2080 tokc
.f
= strtof(token_buf
, NULL
);
2081 } else if (t
== 'L') {
2083 #ifdef TCC_TARGET_PE
2085 tokc
.d
= strtod(token_buf
, NULL
);
2088 tokc
.ld
= strtold(token_buf
, NULL
);
2092 tokc
.d
= strtod(token_buf
, NULL
);
2096 unsigned long long n
, n1
;
2097 int lcount
, ucount
, must_64bit
;
2100 /* integer number */
2103 if (b
== 10 && *q
== '0') {
2110 /* no need for checks except for base 10 / 8 errors */
2120 tcc_error("invalid digit");
2123 /* detect overflow */
2124 /* XXX: this test is not reliable */
2126 tcc_error("integer constant overflow");
2129 /* Determine the characteristics (unsigned and/or 64bit) the type of
2130 the constant must have according to the constant suffix(es) */
2131 lcount
= ucount
= must_64bit
= 0;
2137 tcc_error("three 'l's in integer constant");
2138 if (lcount
&& *(p
- 1) != ch
)
2139 tcc_error("incorrect integer suffix: %s", p1
);
2141 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2146 } else if (t
== 'U') {
2148 tcc_error("two 'u's in integer constant");
2156 /* Whether 64 bits are needed to hold the constant's value */
2157 if (n
& 0xffffffff00000000LL
|| must_64bit
) {
2165 /* Whether type must be unsigned to hold the constant's value */
2166 if (ucount
|| ((n1
>> 31) && (b
!= 10))) {
2167 if (tok
== TOK_CLLONG
)
2171 /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2172 } else if (n1
>> 31) {
2173 if (tok
== TOK_CINT
)
2176 tcc_error("integer constant overflow");
2179 if (tok
== TOK_CINT
|| tok
== TOK_CUINT
)
2185 tcc_error("invalid number\n");
2189 #define PARSE2(c1, tok1, c2, tok2) \
2200 /* return next token without macro substitution */
2201 static inline void next_nomacro1(void)
2216 goto keep_tok_flags
;
2223 /* first look if it is in fact an end of buffer */
2224 if (p
>= file
->buf_end
) {
2228 if (p
>= file
->buf_end
)
2241 TCCState
*s1
= tcc_state
;
2242 if ((parse_flags
& PARSE_FLAG_LINEFEED
)
2243 && !(tok_flags
& TOK_FLAG_EOF
)) {
2244 tok_flags
|= TOK_FLAG_EOF
;
2246 goto keep_tok_flags
;
2247 } else if (!(parse_flags
& PARSE_FLAG_PREPROCESS
)) {
2249 } else if (s1
->ifdef_stack_ptr
!= file
->ifdef_stack_ptr
) {
2250 tcc_error("missing #endif");
2251 } else if (s1
->include_stack_ptr
== s1
->include_stack
) {
2252 /* no include left : end of file. */
2255 tok_flags
&= ~TOK_FLAG_EOF
;
2256 /* pop include file */
2258 /* test if previous '#endif' was after a #ifdef at
2260 if (tok_flags
& TOK_FLAG_ENDIF
) {
2262 printf("#endif %s\n", get_tok_str(file
->ifndef_macro_saved
, NULL
));
2264 add_cached_include(s1
, file
->filename
, file
->ifndef_macro_saved
);
2265 tok_flags
&= ~TOK_FLAG_ENDIF
;
2268 /* add end of include file debug info */
2269 if (tcc_state
->do_debug
) {
2270 put_stabd(N_EINCL
, 0, 0);
2272 /* pop include stack */
2274 s1
->include_stack_ptr
--;
2283 tok_flags
|= TOK_FLAG_BOL
;
2286 if (0 == (parse_flags
& PARSE_FLAG_LINEFEED
))
2289 goto keep_tok_flags
;
2294 if (is_space(c
) && (parse_flags
& PARSE_FLAG_ASM_FILE
)) {
2295 p
= parse_line_comment(p
);
2299 if ((tok_flags
& TOK_FLAG_BOL
) &&
2300 (parse_flags
& PARSE_FLAG_PREPROCESS
)) {
2302 preprocess(tok_flags
& TOK_FLAG_BOF
);
2308 tok
= TOK_TWOSHARPS
;
2310 if (parse_flags
& PARSE_FLAG_ASM_COMMENTS
) {
2311 p
= parse_line_comment(p
- 1);
2320 /* dollar is allowed to start identifiers when not parsing asm */
2322 if (!tcc_state
->dollars_in_identifiers
2323 || (parse_flags
& PARSE_FLAG_ASM_FILE
)) goto parse_simple
;
2325 case 'a': case 'b': case 'c': case 'd':
2326 case 'e': case 'f': case 'g': case 'h':
2327 case 'i': case 'j': case 'k': case 'l':
2328 case 'm': case 'n': case 'o': case 'p':
2329 case 'q': case 'r': case 's': case 't':
2330 case 'u': case 'v': case 'w': case 'x':
2332 case 'A': case 'B': case 'C': case 'D':
2333 case 'E': case 'F': case 'G': case 'H':
2334 case 'I': case 'J': case 'K':
2335 case 'M': case 'N': case 'O': case 'P':
2336 case 'Q': case 'R': case 'S': case 'T':
2337 case 'U': case 'V': case 'W': case 'X':
2343 h
= TOK_HASH_FUNC(h
, c
);
2347 if (!isidnum_table
[c
-CH_EOF
]
2348 && (tcc_state
->dollars_in_identifiers
? (c
!= '$') : 1))
2350 h
= TOK_HASH_FUNC(h
, c
);
2357 /* fast case : no stray found, so we have the full token
2358 and we have already hashed it */
2360 h
&= (TOK_HASH_SIZE
- 1);
2361 pts
= &hash_ident
[h
];
2366 if (ts
->len
== len
&& !memcmp(ts
->str
, p1
, len
))
2368 pts
= &(ts
->hash_next
);
2370 ts
= tok_alloc_new(pts
, (char *) p1
, len
);
2374 cstr_reset(&tokcstr
);
2377 cstr_ccat(&tokcstr
, *p1
);
2383 while (isidnum_table
[c
-CH_EOF
]
2384 || (tcc_state
->dollars_in_identifiers
? (c
== '$') : 0)) {
2385 cstr_ccat(&tokcstr
, c
);
2388 ts
= tok_alloc(tokcstr
.data
, tokcstr
.size
);
2394 if (t
!= '\\' && t
!= '\'' && t
!= '\"') {
2396 goto parse_ident_fast
;
2399 if (c
== '\'' || c
== '\"') {
2403 cstr_reset(&tokcstr
);
2404 cstr_ccat(&tokcstr
, 'L');
2405 goto parse_ident_slow
;
2409 case '0': case '1': case '2': case '3':
2410 case '4': case '5': case '6': case '7':
2413 cstr_reset(&tokcstr
);
2414 /* after the first digit, accept digits, alpha, '.' or sign if
2415 prefixed by 'eEpP' */
2419 cstr_ccat(&tokcstr
, c
);
2421 if (!(isnum(c
) || isid(c
) || c
== '.' ||
2422 ((c
== '+' || c
== '-') &&
2423 (t
== 'e' || t
== 'E' || t
== 'p' || t
== 'P'))))
2426 /* We add a trailing '\0' to ease parsing */
2427 cstr_ccat(&tokcstr
, '\0');
2428 tokc
.cstr
= &tokcstr
;
2432 /* special dot handling because it can also start a number */
2435 cstr_reset(&tokcstr
);
2436 cstr_ccat(&tokcstr
, '.');
2438 } else if (c
== '.') {
2441 if ((parse_flags
& PARSE_FLAG_ASM_COMMENTS
) == 0)
2462 /* parse the string */
2464 p
= parse_pp_string(p
, sep
, &str
);
2465 cstr_ccat(&str
, '\0');
2467 /* eval the escape (should be done as TOK_PPNUM) */
2468 cstr_reset(&tokcstr
);
2469 parse_escape_string(&tokcstr
, str
.data
, is_long
);
2474 /* XXX: make it portable */
2478 char_size
= sizeof(nwchar_t
);
2479 if (tokcstr
.size
<= char_size
)
2480 tcc_error("empty character constant");
2481 if (tokcstr
.size
> 2 * char_size
)
2482 tcc_warning("multi-character character constant");
2484 tokc
.i
= *(int8_t *)tokcstr
.data
;
2487 tokc
.i
= *(nwchar_t
*)tokcstr
.data
;
2491 tokc
.cstr
= &tokcstr
;
2505 } else if (c
== '<') {
2523 } else if (c
== '>') {
2541 } else if (c
== '=') {
2554 } else if (c
== '=') {
2567 } else if (c
== '=') {
2580 } else if (c
== '=') {
2583 } else if (c
== '>') {
2591 PARSE2('!', '!', '=', TOK_NE
)
2592 PARSE2('=', '=', '=', TOK_EQ
)
2593 PARSE2('*', '*', '=', TOK_A_MUL
)
2594 PARSE2('%', '%', '=', TOK_A_MOD
)
2595 PARSE2('^', '^', '=', TOK_A_XOR
)
2597 /* comments or operator */
2601 p
= parse_comment(p
);
2602 /* comments replaced by a blank */
2604 goto keep_tok_flags
;
2605 } else if (c
== '/') {
2606 p
= parse_line_comment(p
);
2608 goto keep_tok_flags
;
2609 } else if (c
== '=') {
2629 case '@': /* only used in assembler */
2635 if ((parse_flags
& PARSE_FLAG_ASM_FILE
) == 0)
2636 tcc_error("unrecognized character \\x%02x", c
);
2646 #if defined(PARSE_DEBUG)
2647 printf("token = %s\n", get_tok_str(tok
, &tokc
));
2651 /* return next token without macro substitution. Can read input from
2653 static void next_nomacro_spc(void)
2659 TOK_GET(&tok
, ¯o_ptr
, &tokc
);
2660 if (tok
== TOK_LINENUM
) {
2661 file
->line_num
= tokc
.i
;
2670 ST_FUNC
void next_nomacro(void)
2674 } while (is_space(tok
));
2677 /* substitute arguments in replacement lists in macro_str by the values in
2678 args (field d) and return allocated string */
2679 static int *macro_arg_subst(Sym
**nested_list
, const int *macro_str
, Sym
*args
)
2681 int last_tok
, t
, spc
;
2691 TOK_GET(&t
, ¯o_str
, &cval
);
2696 TOK_GET(&t
, ¯o_str
, &cval
);
2699 s
= sym_find2(args
, t
);
2705 TOK_GET(&t
, &st
, &cval
);
2706 if (!check_space(t
, &spc
))
2707 cstr_cat(&cstr
, get_tok_str(t
, &cval
));
2710 cstr_ccat(&cstr
, '\0');
2712 printf("stringize: %s\n", (char *)cstr
.data
);
2716 tok_str_add2(&str
, TOK_STR
, &cval
);
2719 tok_str_add2(&str
, t
, &cval
);
2721 } else if (t
>= TOK_IDENT
) {
2722 s
= sym_find2(args
, t
);
2725 /* if '##' is present before or after, no arg substitution */
2726 if (*macro_str
== TOK_TWOSHARPS
|| last_tok
== TOK_TWOSHARPS
) {
2727 /* special case for var arg macros : ## eats the
2728 ',' if empty VA_ARGS variable. */
2729 /* XXX: test of the ',' is not 100%
2730 reliable. should fix it to avoid security
2732 if (gnu_ext
&& s
->type
.t
&&
2733 last_tok
== TOK_TWOSHARPS
&&
2734 str
.len
>= 2 && str
.str
[str
.len
- 2] == ',') {
2735 if (*st
== TOK_PLCHLDR
) {
2736 /* suppress ',' '##' */
2739 /* suppress '##' and add variable */
2747 TOK_GET(&t1
, &st
, &cval
);
2750 tok_str_add2(&str
, t1
, &cval
);
2753 } else if (*st
!= TOK_PLCHLDR
) {
2754 /* NOTE: the stream cannot be read when macro
2755 substituing an argument */
2756 macro_subst(&str
, nested_list
, st
, NULL
);
2759 tok_str_add(&str
, t
);
2762 tok_str_add2(&str
, t
, &cval
);
2766 tok_str_add(&str
, 0);
2770 static char const ab_month_name
[12][4] =
2772 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2773 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2776 /* do macro substitution of current token with macro 's' and add
2777 result to (tok_str,tok_len). 'nested_list' is the list of all
2778 macros we got inside to avoid recursing. Return non zero if no
2779 substitution needs to be done */
2780 static int macro_subst_tok(TokenString
*tok_str
,
2781 Sym
**nested_list
, Sym
*s
, struct macro_level
**can_read_stream
)
2783 Sym
*args
, *sa
, *sa1
;
2784 int mstr_allocated
, parlevel
, *mstr
, t
, t1
, spc
;
2792 /* if symbol is a macro, prepare substitution */
2793 /* special macros */
2794 if (tok
== TOK___LINE__
) {
2795 snprintf(buf
, sizeof(buf
), "%d", file
->line_num
);
2799 } else if (tok
== TOK___FILE__
) {
2800 cstrval
= file
->filename
;
2802 } else if (tok
== TOK___DATE__
|| tok
== TOK___TIME__
) {
2807 tm
= localtime(&ti
);
2808 if (tok
== TOK___DATE__
) {
2809 snprintf(buf
, sizeof(buf
), "%s %2d %d",
2810 ab_month_name
[tm
->tm_mon
], tm
->tm_mday
, tm
->tm_year
+ 1900);
2812 snprintf(buf
, sizeof(buf
), "%02d:%02d:%02d",
2813 tm
->tm_hour
, tm
->tm_min
, tm
->tm_sec
);
2820 cstr_cat(&cstr
, cstrval
);
2821 cstr_ccat(&cstr
, '\0');
2823 tok_str_add2(tok_str
, t1
, &cval
);
2828 if (s
->type
.t
== MACRO_FUNC
) {
2829 /* NOTE: we do not use next_nomacro to avoid eating the
2830 next token. XXX: find better solution */
2834 while (is_space(t
= *p
) || TOK_LINEFEED
== t
)
2836 if (t
== 0 && can_read_stream
) {
2837 /* end of macro stream: we must look at the token
2838 after in the file */
2839 struct macro_level
*ml
= *can_read_stream
;
2845 *can_read_stream
= ml
-> prev
;
2847 /* also, end of scope for nested defined symbol */
2848 (*nested_list
)->v
= -1;
2852 ch
= file
->buf_ptr
[0];
2853 while (is_space(ch
) || ch
== '\n' || ch
== '/')
2858 uint8_t *p
= file
->buf_ptr
;
2861 p
= parse_comment(p
);
2862 file
->buf_ptr
= p
- 1;
2863 } else if (c
== '/') {
2864 p
= parse_line_comment(p
);
2865 file
->buf_ptr
= p
- 1;
2873 if (t
!= '(') /* no macro subst */
2876 /* argument macro */
2881 /* NOTE: empty args are allowed, except if no args */
2883 /* handle '()' case */
2884 if (!args
&& !sa
&& tok
== ')')
2887 tcc_error("macro '%s' used with too many args",
2888 get_tok_str(s
->v
, 0));
2891 /* NOTE: non zero sa->t indicates VA_ARGS */
2892 while ((parlevel
> 0 ||
2894 (tok
!= ',' || sa
->type
.t
))) &&
2898 else if (tok
== ')')
2900 if (tok
== TOK_LINEFEED
)
2902 if (!check_space(tok
, &spc
))
2903 tok_str_add2(&str
, tok
, &tokc
);
2907 tok_str_add(&str
, TOK_PLCHLDR
);
2909 tok_str_add(&str
, 0);
2910 sa1
= sym_push2(&args
, sa
->v
& ~SYM_FIELD
, sa
->type
.t
, 0);
2914 /* special case for gcc var args: add an empty
2915 var arg argument if it is omitted */
2916 if (sa
&& sa
->type
.t
&& gnu_ext
)
2926 tcc_error("macro '%s' used with too few args",
2927 get_tok_str(s
->v
, 0));
2930 /* now subst each arg */
2931 mstr
= macro_arg_subst(nested_list
, mstr
, args
);
2936 tok_str_free(sa
->d
);
2942 sym_push2(nested_list
, s
->v
, 0, 0);
2943 macro_subst(tok_str
, nested_list
, mstr
, can_read_stream
);
2944 /* pop nested defined symbol */
2946 *nested_list
= sa1
->prev
;
2954 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2955 return the resulting string (which must be freed). */
2956 static inline int *macro_twosharps(const int *macro_str
)
2960 TokenString macro_str1
;
2962 int n
, start_of_nosubsts
;
2964 /* we search the first '##' */
2965 for(ptr
= macro_str
;;) {
2967 TOK_GET(&t
, &ptr
, &cval
);
2968 if (t
== TOK_TWOSHARPS
)
2970 /* nothing more to do if end of string */
2975 /* we saw '##', so we need more processing to handle it */
2976 start_of_nosubsts
= -1;
2977 tok_str_new(¯o_str1
);
2978 for(ptr
= macro_str
;;) {
2979 TOK_GET(&tok
, &ptr
, &tokc
);
2982 if (tok
== TOK_TWOSHARPS
)
2984 if (tok
== TOK_NOSUBST
&& start_of_nosubsts
< 0)
2985 start_of_nosubsts
= macro_str1
.len
;
2986 while (*ptr
== TOK_TWOSHARPS
) {
2987 /* given 'a##b', remove nosubsts preceding 'a' */
2988 if (start_of_nosubsts
>= 0)
2989 macro_str1
.len
= start_of_nosubsts
;
2990 /* given 'a##b', skip '##' */
2992 /* given 'a##b', remove nosubsts preceding 'b' */
2993 while (t
== TOK_NOSUBST
)
2995 if (t
&& t
!= TOK_TWOSHARPS
) {
2997 TOK_GET(&t
, &ptr
, &cval
);
2998 /* We concatenate the two tokens */
3000 if (tok
!= TOK_PLCHLDR
)
3001 cstr_cat(&cstr
, get_tok_str(tok
, &tokc
));
3003 if (t
!= TOK_PLCHLDR
|| tok
== TOK_PLCHLDR
)
3004 cstr_cat(&cstr
, get_tok_str(t
, &cval
));
3005 cstr_ccat(&cstr
, '\0');
3007 tcc_open_bf(tcc_state
, ":paste:", cstr
.size
);
3008 memcpy(file
->buffer
, cstr
.data
, cstr
.size
);
3011 if (0 == *file
->buf_ptr
)
3013 tok_str_add2(¯o_str1
, tok
, &tokc
);
3014 tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
3015 n
, cstr
.data
, (char*)cstr
.data
+ n
);
3021 if (tok
!= TOK_NOSUBST
) {
3022 tok_str_add2(¯o_str1
, tok
, &tokc
);
3024 start_of_nosubsts
= -1;
3026 tok_str_add2(¯o_str1
, tok
, &tokc
);
3028 tok_str_add(¯o_str1
, 0);
3029 return macro_str1
.str
;
3033 /* do macro substitution of macro_str and add result to
3034 (tok_str,tok_len). 'nested_list' is the list of all macros we got
3035 inside to avoid recursing. */
3036 static void macro_subst(TokenString
*tok_str
, Sym
**nested_list
,
3037 const int *macro_str
, struct macro_level
** can_read_stream
)
3044 struct macro_level ml
;
3047 /* first scan for '##' operator handling */
3049 macro_str1
= macro_twosharps(ptr
);
3057 /* NOTE: ptr == NULL can only happen if tokens are read from
3058 file stream due to a macro function call */
3061 TOK_GET(&t
, &ptr
, &cval
);
3064 if (t
== TOK_NOSUBST
) {
3065 /* following token has already been subst'd. just copy it on */
3066 tok_str_add2(tok_str
, TOK_NOSUBST
, NULL
);
3067 TOK_GET(&t
, &ptr
, &cval
);
3072 /* if nested substitution, do nothing */
3073 if (sym_find2(*nested_list
, t
)) {
3074 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3075 tok_str_add2(tok_str
, TOK_NOSUBST
, NULL
);
3079 if (can_read_stream
)
3080 ml
.prev
= *can_read_stream
, *can_read_stream
= &ml
;
3081 macro_ptr
= (int *)ptr
;
3083 ret
= macro_subst_tok(tok_str
, nested_list
, s
, can_read_stream
);
3084 ptr
= (int *)macro_ptr
;
3086 if (can_read_stream
&& *can_read_stream
== &ml
)
3087 *can_read_stream
= ml
.prev
;
3090 if (parse_flags
& PARSE_FLAG_SPACES
)
3095 tok_str_add(tok_str
, ' ');
3099 if (!check_space(t
, &spc
))
3100 tok_str_add2(tok_str
, t
, &cval
);
3104 tok_str_free(macro_str1
);
3107 /* return next token with macro substitution */
3108 ST_FUNC
void next(void)
3110 Sym
*nested_list
, *s
;
3112 struct macro_level
*ml
;
3115 if (parse_flags
& PARSE_FLAG_SPACES
)
3120 /* if not reading from macro substituted string, then try
3121 to substitute macros */
3122 if (tok
>= TOK_IDENT
&&
3123 (parse_flags
& PARSE_FLAG_PREPROCESS
)) {
3124 s
= define_find(tok
);
3126 /* we have a macro: we try to substitute */
3130 if (macro_subst_tok(&str
, &nested_list
, s
, &ml
) == 0) {
3131 /* substitution done, NOTE: maybe empty */
3132 tok_str_add(&str
, 0);
3133 macro_ptr
= str
.str
;
3134 macro_ptr_allocated
= str
.str
;
3141 /* end of macro or end of unget buffer */
3142 if (unget_buffer_enabled
) {
3143 macro_ptr
= unget_saved_macro_ptr
;
3144 unget_buffer_enabled
= 0;
3146 /* end of macro string: free it */
3147 tok_str_free(macro_ptr_allocated
);
3148 macro_ptr_allocated
= NULL
;
3152 } else if (tok
== TOK_NOSUBST
) {
3153 /* discard preprocessor's nosubst markers */
3158 /* convert preprocessor tokens into C tokens */
3159 if (tok
== TOK_PPNUM
&&
3160 (parse_flags
& PARSE_FLAG_TOK_NUM
)) {
3161 parse_number((char *)tokc
.cstr
->data
);
3165 /* push back current token and set current token to 'last_tok'. Only
3166 identifier case handled for labels. */
3167 ST_INLN
void unget_tok(int last_tok
)
3171 if (unget_buffer_enabled
)
3173 /* assert(macro_ptr == unget_saved_buffer + 1);
3174 assert(*macro_ptr == 0); */
3178 unget_saved_macro_ptr
= macro_ptr
;
3179 unget_buffer_enabled
= 1;
3181 q
= unget_saved_buffer
;
3184 n
= tok_ext_size(tok
) - 1;
3187 *q
= 0; /* end of token string */
3192 /* better than nothing, but needs extension to handle '-E' option
3194 ST_FUNC
void preprocess_init(TCCState
*s1
)
3196 s1
->include_stack_ptr
= s1
->include_stack
;
3197 /* XXX: move that before to avoid having to initialize
3198 file->ifdef_stack_ptr ? */
3199 s1
->ifdef_stack_ptr
= s1
->ifdef_stack
;
3200 file
->ifdef_stack_ptr
= s1
->ifdef_stack_ptr
;
3203 s1
->pack_stack
[0] = 0;
3204 s1
->pack_stack_ptr
= s1
->pack_stack
;
3207 ST_FUNC
void preprocess_new(void)
3212 /* init isid table */
3214 for(i
=CH_EOF
;i
<256;i
++)
3215 isidnum_table
[i
-CH_EOF
] = isid(i
) || isnum(i
);
3217 /* add all tokens */
3219 tcc_free (table_ident
);
3222 memset(hash_ident
, 0, TOK_HASH_SIZE
* sizeof(TokenSym
*));
3224 tok_ident
= TOK_IDENT
;
3233 tok_alloc(p
, r
- p
- 1);
3238 static void line_macro_output(BufferedFile
*f
, const char *s
, TCCState
*s1
)
3240 switch (s1
->Pflag
) {
3241 case LINE_MACRO_OUTPUT_FORMAT_STD
:
3242 /* "tcc -E -P1" case */
3243 fprintf(s1
->ppfp
, "# line %d \"%s\"\n", f
->line_num
, f
->filename
);
3246 case LINE_MACRO_OUTPUT_FORMAT_NONE
:
3247 /* "tcc -E -P" case: don't output a line directive */
3250 case LINE_MACRO_OUTPUT_FORMAT_GCC
:
3252 /* "tcc -E" case: a gcc standard by default */
3253 fprintf(s1
->ppfp
, "# %d \"%s\"%s\n", f
->line_num
, f
->filename
, s
);
3258 /* Preprocess the current file */
3259 ST_FUNC
int tcc_preprocess(TCCState
*s1
)
3261 BufferedFile
*file_ref
, **iptr
, **iptr_new
;
3265 preprocess_init(s1
);
3266 ch
= file
->buf_ptr
[0];
3267 tok_flags
= TOK_FLAG_BOL
| TOK_FLAG_BOF
;
3268 parse_flags
= (parse_flags
& PARSE_FLAG_ASM_FILE
);
3269 parse_flags
|= PARSE_FLAG_ASM_COMMENTS
| PARSE_FLAG_PREPROCESS
|
3270 PARSE_FLAG_LINEFEED
| PARSE_FLAG_SPACES
;
3274 iptr
= s1
->include_stack_ptr
;
3278 if (tok
== TOK_EOF
) {
3280 } else if (file
!= file_ref
) {
3282 line_macro_output(file_ref
, "", s1
);
3284 } else if (tok
== TOK_LINEFEED
) {
3289 } else if (!token_seen
) {
3290 d
= file
->line_num
- file
->line_ref
;
3291 if (file
!= file_ref
|| d
>= 8) {
3294 if (tcc_state
->Pflag
== LINE_MACRO_OUTPUT_FORMAT_GCC
) {
3295 iptr_new
= s1
->include_stack_ptr
;
3296 s
= iptr_new
> iptr
? " 1"
3297 : iptr_new
< iptr
? " 2"
3298 : iptr_new
> s1
->include_stack
? " 3"
3302 line_macro_output(file
, s
, s1
);
3305 fputs("\n", s1
->ppfp
), --d
;
3307 file
->line_ref
= (file_ref
= file
)->line_num
;
3308 token_seen
= tok
!= TOK_LINEFEED
;
3312 fputs(get_tok_str(tok
, &tokc
), s1
->ppfp
);