Control symbol table of which to relocate symbols
[tinycc.git] / tccpp.c
blobd2b8e969c85fb40fc2d4098eca807b5ddd200b86
1 /*
2 * TCC - Tiny C Compiler
3 *
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
21 #include "tcc.h"
23 /********************************************************/
24 /* global variables */
26 ST_DATA int tok_flags;
27 ST_DATA int parse_flags;
29 ST_DATA struct BufferedFile *file;
30 ST_DATA int ch, tok;
31 ST_DATA CValue tokc;
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 TokenSym *hash_ident[TOK_HASH_SIZE];
44 static char token_buf[STRING_MAX_SIZE + 1];
45 static CString cstr_buf;
46 static TokenString tokstr_buf;
47 static unsigned char isidnum_table[256 - CH_EOF];
48 static int pp_debug_tok, pp_debug_symv;
49 static int pp_once;
50 static void tok_print(const char *msg, const int *str);
52 static struct TinyAlloc *toksym_alloc;
53 static struct TinyAlloc *tokstr_alloc;
54 static struct TinyAlloc *cstr_alloc;
56 /* isidnum_table flags: */
57 #define IS_SPC 1
58 #define IS_ID 2
59 #define IS_NUM 4
61 static TokenString *macro_stack;
63 static const char tcc_keywords[] =
64 #define DEF(id, str) str "\0"
65 #include "tcctok.h"
66 #undef DEF
69 /* WARNING: the content of this string encodes token numbers */
70 static const unsigned char tok_two_chars[] =
71 /* outdated -- gr
72 "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
73 "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
74 */{
75 '<','=', TOK_LE,
76 '>','=', TOK_GE,
77 '!','=', TOK_NE,
78 '&','&', TOK_LAND,
79 '|','|', TOK_LOR,
80 '+','+', TOK_INC,
81 '-','-', TOK_DEC,
82 '=','=', TOK_EQ,
83 '<','<', TOK_SHL,
84 '>','>', TOK_SAR,
85 '+','=', TOK_A_ADD,
86 '-','=', TOK_A_SUB,
87 '*','=', TOK_A_MUL,
88 '/','=', TOK_A_DIV,
89 '%','=', TOK_A_MOD,
90 '&','=', TOK_A_AND,
91 '^','=', TOK_A_XOR,
92 '|','=', TOK_A_OR,
93 '-','>', TOK_ARROW,
94 '.','.', 0xa8, // C++ token ?
95 '#','#', TOK_TWOSHARPS,
99 static void next_nomacro_spc(void);
101 ST_FUNC void skip(int c)
103 if (tok != c)
104 tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
105 next();
108 ST_FUNC void expect(const char *msg)
110 tcc_error("%s expected", msg);
113 /* ------------------------------------------------------------------------- */
114 /* Custom allocator for tiny objects */
116 #define USE_TAL
118 #ifndef USE_TAL
119 #define tal_free(al, p) tcc_free(p)
120 #define tal_realloc(al, p, size) tcc_realloc(p, size)
121 #define tal_new(a,b,c)
122 #define tal_delete(a)
123 #else
124 #if !defined(MEM_DEBUG)
125 #define tal_free(al, p) tal_free_impl(al, p)
126 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size)
127 #define TAL_DEBUG_PARAMS
128 #else
129 #define TAL_DEBUG 1
130 //#define TAL_INFO 1 /* collect and dump allocators stats */
131 #define tal_free(al, p) tal_free_impl(al, p, __FILE__, __LINE__)
132 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size, __FILE__, __LINE__)
133 #define TAL_DEBUG_PARAMS , const char *file, int line
134 #define TAL_DEBUG_FILE_LEN 15
135 #endif
137 #define TOKSYM_TAL_SIZE (768 * 1024) /* allocator for tiny TokenSym in table_ident */
138 #define TOKSTR_TAL_SIZE (768 * 1024) /* allocator for tiny TokenString instances */
139 #define CSTR_TAL_SIZE (256 * 1024) /* allocator for tiny CString instances */
140 #define TOKSYM_TAL_LIMIT 256 /* prefer unique limits to distinguish allocators debug msgs */
141 #define TOKSTR_TAL_LIMIT 128 /* 32 * sizeof(int) */
142 #define CSTR_TAL_LIMIT 1024
144 typedef struct TinyAlloc {
145 size_t limit;
146 size_t size;
147 uint8_t *buffer;
148 uint8_t *p;
149 size_t nb_allocs;
150 struct TinyAlloc *next, *top;
151 #ifdef TAL_INFO
152 size_t nb_peak;
153 size_t nb_total;
154 size_t nb_missed;
155 uint8_t *peak_p;
156 #endif
157 } TinyAlloc;
159 typedef struct tal_header_t {
160 size_t size;
161 #ifdef TAL_DEBUG
162 int line_num; /* negative line_num used for double free check */
163 char file_name[TAL_DEBUG_FILE_LEN + 1];
164 #endif
165 } tal_header_t;
167 /* ------------------------------------------------------------------------- */
169 ST_FUNC TinyAlloc *tal_new(TinyAlloc **pal, size_t limit, size_t size)
171 TinyAlloc *al = tcc_mallocz(sizeof(TinyAlloc));
172 al->p = al->buffer = tcc_malloc(size);
173 al->limit = limit;
174 al->size = size;
175 if (pal) *pal = al;
176 return al;
179 ST_FUNC void tal_delete(TinyAlloc *al)
181 TinyAlloc *next;
183 tail_call:
184 if (!al)
185 return;
186 #ifdef TAL_INFO
187 fprintf(stderr, "limit=%5d, size=%5g MB, nb_peak=%6d, nb_total=%8d, nb_missed=%6d, usage=%5.1f%%\n",
188 al->limit, al->size / 1024.0 / 1024.0, al->nb_peak, al->nb_total, al->nb_missed,
189 (al->peak_p - al->buffer) * 100.0 / al->size);
190 #endif
191 #ifdef TAL_DEBUG
192 if (al->nb_allocs > 0) {
193 uint8_t *p;
194 fprintf(stderr, "TAL_DEBUG: mem leak %d chunks (limit= %d)\n",
195 al->nb_allocs, al->limit);
196 p = al->buffer;
197 while (p < al->p) {
198 tal_header_t *header = (tal_header_t *)p;
199 if (header->line_num > 0) {
200 fprintf(stderr, " file %s, line %u: %u bytes\n",
201 header->file_name, header->line_num, header->size);
203 p += header->size + sizeof(tal_header_t);
206 #endif
207 next = al->next;
208 tcc_free(al->buffer);
209 tcc_free(al);
210 al = next;
211 goto tail_call;
214 ST_FUNC void tal_free_impl(TinyAlloc *al, void *p TAL_DEBUG_PARAMS)
216 if (!p)
217 return;
218 tail_call:
219 if (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size) {
220 #ifdef TAL_DEBUG
221 tal_header_t *header = (((tal_header_t *)p) - 1);
222 if (header->line_num < 0) {
223 fprintf(stderr, "TAL_DEBUG: file %s, line %u double frees chunk from\n",
224 file, line);
225 fprintf(stderr, " file %s, line %u: %u bytes\n",
226 header->file_name, -header->line_num, header->size);
227 } else
228 header->line_num = -header->line_num;
229 #endif
230 al->nb_allocs--;
231 if (!al->nb_allocs)
232 al->p = al->buffer;
233 } else if (al->next) {
234 al = al->next;
235 goto tail_call;
237 else
238 tcc_free(p);
241 ST_FUNC void *tal_realloc_impl(TinyAlloc **pal, void *p, size_t size TAL_DEBUG_PARAMS)
243 tal_header_t *header;
244 void *ret;
245 int is_own;
246 size_t adj_size = (size + 3) & -4;
247 TinyAlloc *al = *pal;
249 tail_call:
250 is_own = (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size);
251 if ((!p || is_own) && size <= al->limit) {
252 if (al->p + adj_size + sizeof(tal_header_t) < al->buffer + al->size) {
253 header = (tal_header_t *)al->p;
254 header->size = adj_size;
255 #ifdef TAL_DEBUG
256 { int ofs = strlen(file) - TAL_DEBUG_FILE_LEN;
257 strncpy(header->file_name, file + (ofs > 0 ? ofs : 0), TAL_DEBUG_FILE_LEN);
258 header->file_name[TAL_DEBUG_FILE_LEN] = 0;
259 header->line_num = line; }
260 #endif
261 ret = al->p + sizeof(tal_header_t);
262 al->p += adj_size + sizeof(tal_header_t);
263 if (is_own) {
264 header = (((tal_header_t *)p) - 1);
265 memcpy(ret, p, header->size);
266 #ifdef TAL_DEBUG
267 header->line_num = -header->line_num;
268 #endif
269 } else {
270 al->nb_allocs++;
272 #ifdef TAL_INFO
273 if (al->nb_peak < al->nb_allocs)
274 al->nb_peak = al->nb_allocs;
275 if (al->peak_p < al->p)
276 al->peak_p = al->p;
277 al->nb_total++;
278 #endif
279 return ret;
280 } else if (is_own) {
281 al->nb_allocs--;
282 ret = tal_realloc(*pal, 0, size);
283 header = (((tal_header_t *)p) - 1);
284 memcpy(ret, p, header->size);
285 #ifdef TAL_DEBUG
286 header->line_num = -header->line_num;
287 #endif
288 return ret;
290 if (al->next) {
291 al = al->next;
292 } else {
293 TinyAlloc *bottom = al, *next = al->top ? al->top : al;
295 al = tal_new(pal, next->limit, next->size * 2);
296 al->next = next;
297 bottom->top = al;
299 goto tail_call;
301 if (is_own) {
302 al->nb_allocs--;
303 ret = tcc_malloc(size);
304 header = (((tal_header_t *)p) - 1);
305 memcpy(ret, p, header->size);
306 #ifdef TAL_DEBUG
307 header->line_num = -header->line_num;
308 #endif
309 } else if (al->next) {
310 al = al->next;
311 goto tail_call;
312 } else
313 ret = tcc_realloc(p, size);
314 #ifdef TAL_INFO
315 al->nb_missed++;
316 #endif
317 return ret;
320 #endif /* USE_TAL */
322 /* ------------------------------------------------------------------------- */
323 /* CString handling */
324 static void cstr_realloc(CString *cstr, int new_size)
326 int size;
328 size = cstr->size_allocated;
329 if (size < 8)
330 size = 8; /* no need to allocate a too small first string */
331 while (size < new_size)
332 size = size * 2;
333 cstr->data = tal_realloc(cstr_alloc, cstr->data, size);
334 cstr->size_allocated = size;
337 /* add a byte */
338 ST_INLN void cstr_ccat(CString *cstr, int ch)
340 int size;
341 size = cstr->size + 1;
342 if (size > cstr->size_allocated)
343 cstr_realloc(cstr, size);
344 ((unsigned char *)cstr->data)[size - 1] = ch;
345 cstr->size = size;
348 ST_FUNC void cstr_cat(CString *cstr, const char *str, int len)
350 int size;
351 if (len <= 0)
352 len = strlen(str) + 1 + len;
353 size = cstr->size + len;
354 if (size > cstr->size_allocated)
355 cstr_realloc(cstr, size);
356 memmove(((unsigned char *)cstr->data) + cstr->size, str, len);
357 cstr->size = size;
360 /* add a wide char */
361 ST_FUNC void cstr_wccat(CString *cstr, int ch)
363 int size;
364 size = cstr->size + sizeof(nwchar_t);
365 if (size > cstr->size_allocated)
366 cstr_realloc(cstr, size);
367 *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
368 cstr->size = size;
371 ST_FUNC void cstr_new(CString *cstr)
373 memset(cstr, 0, sizeof(CString));
376 /* free string and reset it to NULL */
377 ST_FUNC void cstr_free(CString *cstr)
379 tal_free(cstr_alloc, cstr->data);
380 cstr_new(cstr);
383 /* reset string to empty */
384 ST_FUNC void cstr_reset(CString *cstr)
386 cstr->size = 0;
389 /* XXX: unicode ? */
390 static void add_char(CString *cstr, int c)
392 if (c == '\'' || c == '\"' || c == '\\') {
393 /* XXX: could be more precise if char or string */
394 cstr_ccat(cstr, '\\');
396 if (c >= 32 && c <= 126) {
397 cstr_ccat(cstr, c);
398 } else {
399 cstr_ccat(cstr, '\\');
400 if (c == '\n') {
401 cstr_ccat(cstr, 'n');
402 } else {
403 cstr_ccat(cstr, '0' + ((c >> 6) & 7));
404 cstr_ccat(cstr, '0' + ((c >> 3) & 7));
405 cstr_ccat(cstr, '0' + (c & 7));
410 /* ------------------------------------------------------------------------- */
411 /* allocate a new token */
412 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
414 TokenSym *ts, **ptable;
415 int i;
417 if (tok_ident >= SYM_FIRST_ANOM)
418 tcc_error("memory full (symbols)");
420 /* expand token table if needed */
421 i = tok_ident - TOK_IDENT;
422 if ((i % TOK_ALLOC_INCR) == 0) {
423 ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
424 table_ident = ptable;
427 ts = tal_realloc(toksym_alloc, 0, sizeof(TokenSym) + len);
428 table_ident[i] = ts;
429 ts->tok = tok_ident++;
430 ts->sym_define = NULL;
431 ts->sym_label = NULL;
432 ts->sym_struct = NULL;
433 ts->sym_identifier = NULL;
434 ts->len = len;
435 ts->hash_next = NULL;
436 memcpy(ts->str, str, len);
437 ts->str[len] = '\0';
438 *pts = ts;
439 return ts;
442 #define TOK_HASH_INIT 1
443 #define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c))
446 /* find a token and add it if not found */
447 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
449 TokenSym *ts, **pts;
450 int i;
451 unsigned int h;
453 h = TOK_HASH_INIT;
454 for(i=0;i<len;i++)
455 h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
456 h &= (TOK_HASH_SIZE - 1);
458 pts = &hash_ident[h];
459 for(;;) {
460 ts = *pts;
461 if (!ts)
462 break;
463 if (ts->len == len && !memcmp(ts->str, str, len))
464 return ts;
465 pts = &(ts->hash_next);
467 return tok_alloc_new(pts, str, len);
470 /* XXX: buffer overflow */
471 /* XXX: float tokens */
472 ST_FUNC const char *get_tok_str(int v, CValue *cv)
474 char *p;
475 int i, len;
477 cstr_reset(&cstr_buf);
478 p = cstr_buf.data;
480 switch(v) {
481 case TOK_CINT:
482 case TOK_CUINT:
483 case TOK_CLLONG:
484 case TOK_CULLONG:
485 /* XXX: not quite exact, but only useful for testing */
486 #ifdef _WIN32
487 sprintf(p, "%u", (unsigned)cv->i);
488 #else
489 sprintf(p, "%llu", (unsigned long long)cv->i);
490 #endif
491 break;
492 case TOK_LCHAR:
493 cstr_ccat(&cstr_buf, 'L');
494 case TOK_CCHAR:
495 cstr_ccat(&cstr_buf, '\'');
496 add_char(&cstr_buf, cv->i);
497 cstr_ccat(&cstr_buf, '\'');
498 cstr_ccat(&cstr_buf, '\0');
499 break;
500 case TOK_PPNUM:
501 case TOK_PPSTR:
502 return (char*)cv->str.data;
503 case TOK_LSTR:
504 cstr_ccat(&cstr_buf, 'L');
505 case TOK_STR:
506 cstr_ccat(&cstr_buf, '\"');
507 if (v == TOK_STR) {
508 len = cv->str.size - 1;
509 for(i=0;i<len;i++)
510 add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]);
511 } else {
512 len = (cv->str.size / sizeof(nwchar_t)) - 1;
513 for(i=0;i<len;i++)
514 add_char(&cstr_buf, ((nwchar_t *)cv->str.data)[i]);
516 cstr_ccat(&cstr_buf, '\"');
517 cstr_ccat(&cstr_buf, '\0');
518 break;
520 case TOK_CFLOAT:
521 cstr_cat(&cstr_buf, "<float>", 0);
522 break;
523 case TOK_CDOUBLE:
524 cstr_cat(&cstr_buf, "<double>", 0);
525 break;
526 case TOK_CLDOUBLE:
527 cstr_cat(&cstr_buf, "<long double>", 0);
528 break;
529 case TOK_LINENUM:
530 cstr_cat(&cstr_buf, "<linenumber>", 0);
531 break;
533 /* above tokens have value, the ones below don't */
535 case TOK_LT:
536 v = '<';
537 goto addv;
538 case TOK_GT:
539 v = '>';
540 goto addv;
541 case TOK_DOTS:
542 return strcpy(p, "...");
543 case TOK_A_SHL:
544 return strcpy(p, "<<=");
545 case TOK_A_SAR:
546 return strcpy(p, ">>=");
547 default:
548 if (v < TOK_IDENT) {
549 /* search in two bytes table */
550 const unsigned char *q = tok_two_chars;
551 while (*q) {
552 if (q[2] == v) {
553 *p++ = q[0];
554 *p++ = q[1];
555 *p = '\0';
556 return cstr_buf.data;
558 q += 3;
560 if (v >= 127) {
561 sprintf(cstr_buf.data, "<%02x>", v);
562 return cstr_buf.data;
564 addv:
565 *p++ = v;
566 *p = '\0';
567 } else if (v < tok_ident) {
568 return table_ident[v - TOK_IDENT]->str;
569 } else if (v >= SYM_FIRST_ANOM) {
570 /* special name for anonymous symbol */
571 sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
572 } else {
573 /* should never happen */
574 return NULL;
576 break;
578 return cstr_buf.data;
581 /* return the current character, handling end of block if necessary
582 (but not stray) */
583 ST_FUNC int handle_eob(void)
585 BufferedFile *bf = file;
586 int len;
588 /* only tries to read if really end of buffer */
589 if (bf->buf_ptr >= bf->buf_end) {
590 if (bf->fd != -1) {
591 #if defined(PARSE_DEBUG)
592 len = 1;
593 #else
594 len = IO_BUF_SIZE;
595 #endif
596 len = read(bf->fd, bf->buffer, len);
597 if (len < 0)
598 len = 0;
599 } else {
600 len = 0;
602 total_bytes += len;
603 bf->buf_ptr = bf->buffer;
604 bf->buf_end = bf->buffer + len;
605 *bf->buf_end = CH_EOB;
607 if (bf->buf_ptr < bf->buf_end) {
608 return bf->buf_ptr[0];
609 } else {
610 bf->buf_ptr = bf->buf_end;
611 return CH_EOF;
615 /* read next char from current input file and handle end of input buffer */
616 ST_INLN void inp(void)
618 ch = *(++(file->buf_ptr));
619 /* end of buffer/file handling */
620 if (ch == CH_EOB)
621 ch = handle_eob();
624 /* handle '\[\r]\n' */
625 static int handle_stray_noerror(void)
627 while (ch == '\\') {
628 inp();
629 if (ch == '\n') {
630 file->line_num++;
631 inp();
632 } else if (ch == '\r') {
633 inp();
634 if (ch != '\n')
635 goto fail;
636 file->line_num++;
637 inp();
638 } else {
639 fail:
640 return 1;
643 return 0;
646 static void handle_stray(void)
648 if (handle_stray_noerror())
649 tcc_error("stray '\\' in program");
652 /* skip the stray and handle the \\n case. Output an error if
653 incorrect char after the stray */
654 static int handle_stray1(uint8_t *p)
656 int c;
658 file->buf_ptr = p;
659 if (p >= file->buf_end) {
660 c = handle_eob();
661 if (c != '\\')
662 return c;
663 p = file->buf_ptr;
665 ch = *p;
666 if (handle_stray_noerror()) {
667 if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
668 tcc_error("stray '\\' in program");
669 *--file->buf_ptr = '\\';
671 p = file->buf_ptr;
672 c = *p;
673 return c;
676 /* handle just the EOB case, but not stray */
677 #define PEEKC_EOB(c, p)\
679 p++;\
680 c = *p;\
681 if (c == '\\') {\
682 file->buf_ptr = p;\
683 c = handle_eob();\
684 p = file->buf_ptr;\
688 /* handle the complicated stray case */
689 #define PEEKC(c, p)\
691 p++;\
692 c = *p;\
693 if (c == '\\') {\
694 c = handle_stray1(p);\
695 p = file->buf_ptr;\
699 /* input with '\[\r]\n' handling. Note that this function cannot
700 handle other characters after '\', so you cannot call it inside
701 strings or comments */
702 ST_FUNC void minp(void)
704 inp();
705 if (ch == '\\')
706 handle_stray();
709 /* single line C++ comments */
710 static uint8_t *parse_line_comment(uint8_t *p)
712 int c;
714 p++;
715 for(;;) {
716 c = *p;
717 redo:
718 if (c == '\n' || c == CH_EOF) {
719 break;
720 } else if (c == '\\') {
721 file->buf_ptr = p;
722 c = handle_eob();
723 p = file->buf_ptr;
724 if (c == '\\') {
725 PEEKC_EOB(c, p);
726 if (c == '\n') {
727 file->line_num++;
728 PEEKC_EOB(c, p);
729 } else if (c == '\r') {
730 PEEKC_EOB(c, p);
731 if (c == '\n') {
732 file->line_num++;
733 PEEKC_EOB(c, p);
736 } else {
737 goto redo;
739 } else {
740 p++;
743 return p;
746 /* C comments */
747 ST_FUNC uint8_t *parse_comment(uint8_t *p)
749 int c;
751 p++;
752 for(;;) {
753 /* fast skip loop */
754 for(;;) {
755 c = *p;
756 if (c == '\n' || c == '*' || c == '\\')
757 break;
758 p++;
759 c = *p;
760 if (c == '\n' || c == '*' || c == '\\')
761 break;
762 p++;
764 /* now we can handle all the cases */
765 if (c == '\n') {
766 file->line_num++;
767 p++;
768 } else if (c == '*') {
769 p++;
770 for(;;) {
771 c = *p;
772 if (c == '*') {
773 p++;
774 } else if (c == '/') {
775 goto end_of_comment;
776 } else if (c == '\\') {
777 file->buf_ptr = p;
778 c = handle_eob();
779 p = file->buf_ptr;
780 if (c == CH_EOF)
781 tcc_error("unexpected end of file in comment");
782 if (c == '\\') {
783 /* skip '\[\r]\n', otherwise just skip the stray */
784 while (c == '\\') {
785 PEEKC_EOB(c, p);
786 if (c == '\n') {
787 file->line_num++;
788 PEEKC_EOB(c, p);
789 } else if (c == '\r') {
790 PEEKC_EOB(c, p);
791 if (c == '\n') {
792 file->line_num++;
793 PEEKC_EOB(c, p);
795 } else {
796 goto after_star;
800 } else {
801 break;
804 after_star: ;
805 } else {
806 /* stray, eob or eof */
807 file->buf_ptr = p;
808 c = handle_eob();
809 p = file->buf_ptr;
810 if (c == CH_EOF) {
811 tcc_error("unexpected end of file in comment");
812 } else if (c == '\\') {
813 p++;
817 end_of_comment:
818 p++;
819 return p;
822 #define cinp minp
824 static inline void skip_spaces(void)
826 while (isidnum_table[ch - CH_EOF] & IS_SPC)
827 cinp();
830 static inline int check_space(int t, int *spc)
832 if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
833 if (*spc)
834 return 1;
835 *spc = 1;
836 } else
837 *spc = 0;
838 return 0;
841 /* parse a string without interpreting escapes */
842 static uint8_t *parse_pp_string(uint8_t *p,
843 int sep, CString *str)
845 int c;
846 p++;
847 for(;;) {
848 c = *p;
849 if (c == sep) {
850 break;
851 } else if (c == '\\') {
852 file->buf_ptr = p;
853 c = handle_eob();
854 p = file->buf_ptr;
855 if (c == CH_EOF) {
856 unterminated_string:
857 /* XXX: indicate line number of start of string */
858 tcc_error("missing terminating %c character", sep);
859 } else if (c == '\\') {
860 /* escape : just skip \[\r]\n */
861 PEEKC_EOB(c, p);
862 if (c == '\n') {
863 file->line_num++;
864 p++;
865 } else if (c == '\r') {
866 PEEKC_EOB(c, p);
867 if (c != '\n')
868 expect("'\n' after '\r'");
869 file->line_num++;
870 p++;
871 } else if (c == CH_EOF) {
872 goto unterminated_string;
873 } else {
874 if (str) {
875 cstr_ccat(str, '\\');
876 cstr_ccat(str, c);
878 p++;
881 } else if (c == '\n') {
882 file->line_num++;
883 goto add_char;
884 } else if (c == '\r') {
885 PEEKC_EOB(c, p);
886 if (c != '\n') {
887 if (str)
888 cstr_ccat(str, '\r');
889 } else {
890 file->line_num++;
891 goto add_char;
893 } else {
894 add_char:
895 if (str)
896 cstr_ccat(str, c);
897 p++;
900 p++;
901 return p;
904 /* skip block of text until #else, #elif or #endif. skip also pairs of
905 #if/#endif */
906 static void preprocess_skip(void)
908 int a, start_of_line, c, in_warn_or_error;
909 uint8_t *p;
911 p = file->buf_ptr;
912 a = 0;
913 redo_start:
914 start_of_line = 1;
915 in_warn_or_error = 0;
916 for(;;) {
917 redo_no_start:
918 c = *p;
919 switch(c) {
920 case ' ':
921 case '\t':
922 case '\f':
923 case '\v':
924 case '\r':
925 p++;
926 goto redo_no_start;
927 case '\n':
928 file->line_num++;
929 p++;
930 goto redo_start;
931 case '\\':
932 file->buf_ptr = p;
933 c = handle_eob();
934 if (c == CH_EOF) {
935 expect("#endif");
936 } else if (c == '\\') {
937 ch = file->buf_ptr[0];
938 handle_stray_noerror();
940 p = file->buf_ptr;
941 goto redo_no_start;
942 /* skip strings */
943 case '\"':
944 case '\'':
945 if (in_warn_or_error)
946 goto _default;
947 p = parse_pp_string(p, c, NULL);
948 break;
949 /* skip comments */
950 case '/':
951 if (in_warn_or_error)
952 goto _default;
953 file->buf_ptr = p;
954 ch = *p;
955 minp();
956 p = file->buf_ptr;
957 if (ch == '*') {
958 p = parse_comment(p);
959 } else if (ch == '/') {
960 p = parse_line_comment(p);
962 break;
963 case '#':
964 p++;
965 if (start_of_line) {
966 file->buf_ptr = p;
967 next_nomacro();
968 p = file->buf_ptr;
969 if (a == 0 &&
970 (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
971 goto the_end;
972 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
973 a++;
974 else if (tok == TOK_ENDIF)
975 a--;
976 else if( tok == TOK_ERROR || tok == TOK_WARNING)
977 in_warn_or_error = 1;
978 else if (tok == TOK_LINEFEED)
979 goto redo_start;
980 else if (parse_flags & PARSE_FLAG_ASM_FILE)
981 p = parse_line_comment(p - 1);
982 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
983 p = parse_line_comment(p - 1);
984 break;
985 _default:
986 default:
987 p++;
988 break;
990 start_of_line = 0;
992 the_end: ;
993 file->buf_ptr = p;
996 /* ParseState handling */
998 /* XXX: currently, no include file info is stored. Thus, we cannot display
999 accurate messages if the function or data definition spans multiple
1000 files */
1002 /* save current parse state in 's' */
1003 ST_FUNC void save_parse_state(ParseState *s)
1005 s->line_num = file->line_num;
1006 s->macro_ptr = macro_ptr;
1007 s->tok = tok;
1008 s->tokc = tokc;
1011 /* restore parse state from 's' */
1012 ST_FUNC void restore_parse_state(ParseState *s)
1014 file->line_num = s->line_num;
1015 macro_ptr = s->macro_ptr;
1016 tok = s->tok;
1017 tokc = s->tokc;
1020 /* return the number of additional 'ints' necessary to store the
1021 token */
1022 static inline int tok_size(const int *p)
1024 switch(*p) {
1025 /* 4 bytes */
1026 case TOK_CINT:
1027 case TOK_CUINT:
1028 case TOK_CCHAR:
1029 case TOK_LCHAR:
1030 case TOK_CFLOAT:
1031 case TOK_LINENUM:
1032 return 1 + 1;
1033 case TOK_STR:
1034 case TOK_LSTR:
1035 case TOK_PPNUM:
1036 case TOK_PPSTR:
1037 return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
1038 case TOK_CDOUBLE:
1039 case TOK_CLLONG:
1040 case TOK_CULLONG:
1041 return 1 + 2;
1042 case TOK_CLDOUBLE:
1043 return 1 + LDOUBLE_SIZE / 4;
1044 default:
1045 return 1 + 0;
1049 /* token string handling */
1051 ST_INLN void tok_str_new(TokenString *s)
1053 s->str = NULL;
1054 s->len = 0;
1055 s->allocated_len = 0;
1056 s->last_line_num = -1;
1059 ST_FUNC TokenString *tok_str_alloc(void)
1061 TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str);
1062 tok_str_new(str);
1063 return str;
1066 ST_FUNC int *tok_str_dup(TokenString *s)
1068 int *str;
1070 str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int));
1071 memcpy(str, s->str, s->len * sizeof(int));
1072 return str;
1075 ST_FUNC void tok_str_free_str(int *str)
1077 tal_free(tokstr_alloc, str);
1080 ST_FUNC void tok_str_free(TokenString *str)
1082 tok_str_free_str(str->str);
1083 tal_free(tokstr_alloc, str);
1086 ST_FUNC int *tok_str_realloc(TokenString *s, int new_size)
1088 int *str, size;
1090 size = s->allocated_len;
1091 if (size < 16)
1092 size = 16;
1093 while (size < new_size)
1094 size = size * 2;
1095 if (size > s->allocated_len) {
1096 str = tal_realloc(tokstr_alloc, s->str, size * sizeof(int));
1097 s->allocated_len = size;
1098 s->str = str;
1100 return s->str;
1103 ST_FUNC void tok_str_add(TokenString *s, int t)
1105 int len, *str;
1107 len = s->len;
1108 str = s->str;
1109 if (len >= s->allocated_len)
1110 str = tok_str_realloc(s, len + 1);
1111 str[len++] = t;
1112 s->len = len;
1115 ST_FUNC void begin_macro(TokenString *str, int alloc)
1117 str->alloc = alloc;
1118 str->prev = macro_stack;
1119 str->prev_ptr = macro_ptr;
1120 macro_ptr = str->str;
1121 macro_stack = str;
1124 ST_FUNC void end_macro(void)
1126 TokenString *str = macro_stack;
1127 macro_stack = str->prev;
1128 macro_ptr = str->prev_ptr;
1129 if (str->alloc == 2) {
1130 str->alloc = 3; /* just mark as finished */
1131 } else {
1132 tok_str_free(str);
1136 static void tok_str_add2(TokenString *s, int t, CValue *cv)
1138 int len, *str;
1140 len = s->len;
1141 str = s->str;
1143 /* allocate space for worst case */
1144 if (len + TOK_MAX_SIZE >= s->allocated_len)
1145 str = tok_str_realloc(s, len + TOK_MAX_SIZE + 1);
1146 str[len++] = t;
1147 switch(t) {
1148 case TOK_CINT:
1149 case TOK_CUINT:
1150 case TOK_CCHAR:
1151 case TOK_LCHAR:
1152 case TOK_CFLOAT:
1153 case TOK_LINENUM:
1154 str[len++] = cv->tab[0];
1155 break;
1156 case TOK_PPNUM:
1157 case TOK_PPSTR:
1158 case TOK_STR:
1159 case TOK_LSTR:
1161 /* Insert the string into the int array. */
1162 size_t nb_words =
1163 1 + (cv->str.size + sizeof(int) - 1) / sizeof(int);
1164 if (len + nb_words >= s->allocated_len)
1165 str = tok_str_realloc(s, len + nb_words + 1);
1166 str[len] = cv->str.size;
1167 memcpy(&str[len + 1], cv->str.data, cv->str.size);
1168 len += nb_words;
1170 break;
1171 case TOK_CDOUBLE:
1172 case TOK_CLLONG:
1173 case TOK_CULLONG:
1174 #if LDOUBLE_SIZE == 8
1175 case TOK_CLDOUBLE:
1176 #endif
1177 str[len++] = cv->tab[0];
1178 str[len++] = cv->tab[1];
1179 break;
1180 #if LDOUBLE_SIZE == 12
1181 case TOK_CLDOUBLE:
1182 str[len++] = cv->tab[0];
1183 str[len++] = cv->tab[1];
1184 str[len++] = cv->tab[2];
1185 #elif LDOUBLE_SIZE == 16
1186 case TOK_CLDOUBLE:
1187 str[len++] = cv->tab[0];
1188 str[len++] = cv->tab[1];
1189 str[len++] = cv->tab[2];
1190 str[len++] = cv->tab[3];
1191 #elif LDOUBLE_SIZE != 8
1192 #error add long double size support
1193 #endif
1194 break;
1195 default:
1196 break;
1198 s->len = len;
1201 /* add the current parse token in token string 's' */
1202 ST_FUNC void tok_str_add_tok(TokenString *s)
1204 CValue cval;
1206 /* save line number info */
1207 if (file->line_num != s->last_line_num) {
1208 s->last_line_num = file->line_num;
1209 cval.i = s->last_line_num;
1210 tok_str_add2(s, TOK_LINENUM, &cval);
1212 tok_str_add2(s, tok, &tokc);
1215 /* get a token from an integer array and increment pointer
1216 accordingly. we code it as a macro to avoid pointer aliasing. */
1217 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
1219 const int *p = *pp;
1220 int n, *tab;
1222 tab = cv->tab;
1223 switch(*t = *p++) {
1224 case TOK_CINT:
1225 case TOK_CUINT:
1226 case TOK_CCHAR:
1227 case TOK_LCHAR:
1228 case TOK_CFLOAT:
1229 case TOK_LINENUM:
1230 tab[0] = *p++;
1231 break;
1232 case TOK_STR:
1233 case TOK_LSTR:
1234 case TOK_PPNUM:
1235 case TOK_PPSTR:
1236 cv->str.size = *p++;
1237 cv->str.data = p;
1238 p += (cv->str.size + sizeof(int) - 1) / sizeof(int);
1239 break;
1240 case TOK_CDOUBLE:
1241 case TOK_CLLONG:
1242 case TOK_CULLONG:
1243 n = 2;
1244 goto copy;
1245 case TOK_CLDOUBLE:
1246 #if LDOUBLE_SIZE == 16
1247 n = 4;
1248 #elif LDOUBLE_SIZE == 12
1249 n = 3;
1250 #elif LDOUBLE_SIZE == 8
1251 n = 2;
1252 #else
1253 # error add long double size support
1254 #endif
1255 copy:
1257 *tab++ = *p++;
1258 while (--n);
1259 break;
1260 default:
1261 break;
1263 *pp = p;
1266 /* Calling this function is expensive, but it is not possible
1267 to read a token string backwards. */
1268 static int tok_last(const int *str0, const int *str1)
1270 const int *str = str0;
1271 int tok = 0;
1272 CValue cval;
1274 while (str < str1)
1275 TOK_GET(&tok, &str, &cval);
1276 return tok;
1279 static int macro_is_equal(const int *a, const int *b)
1281 CValue cv;
1282 int t;
1284 if (!a || !b)
1285 return 1;
1287 while (*a && *b) {
1288 /* first time preallocate static cstr_buf, next time only reset position to start */
1289 cstr_reset(&cstr_buf);
1290 TOK_GET(&t, &a, &cv);
1291 cstr_cat(&cstr_buf, get_tok_str(t, &cv), 0);
1292 TOK_GET(&t, &b, &cv);
1293 if (strcmp(cstr_buf.data, get_tok_str(t, &cv)))
1294 return 0;
1296 return !(*a || *b);
1299 /* defines handling */
1300 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1302 Sym *s, *o;
1304 o = define_find(v);
1305 s = sym_push2(&define_stack, v, macro_type, 0);
1306 s->d = str;
1307 s->next = first_arg;
1308 table_ident[v - TOK_IDENT]->sym_define = s;
1310 if (o && !macro_is_equal(o->d, s->d))
1311 tcc_warning("%s redefined", get_tok_str(v, NULL));
1314 /* undefined a define symbol. Its name is just set to zero */
1315 ST_FUNC void define_undef(Sym *s)
1317 int v = s->v;
1318 if (v >= TOK_IDENT && v < tok_ident)
1319 table_ident[v - TOK_IDENT]->sym_define = NULL;
1322 ST_INLN Sym *define_find(int v)
1324 v -= TOK_IDENT;
1325 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1326 return NULL;
1327 return table_ident[v]->sym_define;
1330 /* free define stack until top reaches 'b' */
1331 ST_FUNC void free_defines(Sym *b)
1333 while (define_stack != b) {
1334 Sym *top = define_stack;
1335 define_stack = top->prev;
1336 tok_str_free_str(top->d);
1337 define_undef(top);
1338 sym_free(top);
1341 /* restore remaining (-D or predefined) symbols */
1342 while (b) {
1343 int v = b->v;
1344 if (v >= TOK_IDENT && v < tok_ident) {
1345 Sym **d = &table_ident[v - TOK_IDENT]->sym_define;
1346 if (!*d)
1347 *d = b;
1349 b = b->prev;
1353 /* label lookup */
1354 ST_FUNC Sym *label_find(int v)
1356 v -= TOK_IDENT;
1357 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1358 return NULL;
1359 return table_ident[v]->sym_label;
1362 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1364 Sym *s, **ps;
1365 s = sym_push2(ptop, v, 0, 0);
1366 s->r = flags;
1367 ps = &table_ident[v - TOK_IDENT]->sym_label;
1368 if (ptop == &global_label_stack) {
1369 /* modify the top most local identifier, so that
1370 sym_identifier will point to 's' when popped */
1371 while (*ps != NULL)
1372 ps = &(*ps)->prev_tok;
1374 s->prev_tok = *ps;
1375 *ps = s;
1376 return s;
1379 /* pop labels until element last is reached. Look if any labels are
1380 undefined. Define symbols if '&&label' was used. */
1381 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1383 Sym *s, *s1;
1384 for(s = *ptop; s != slast; s = s1) {
1385 s1 = s->prev;
1386 if (s->r == LABEL_DECLARED) {
1387 tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1388 } else if (s->r == LABEL_FORWARD) {
1389 tcc_error("label '%s' used but not defined",
1390 get_tok_str(s->v, NULL));
1391 } else {
1392 if (s->c) {
1393 /* define corresponding symbol. A size of
1394 1 is put. */
1395 put_extern_sym(s, cur_text_section, s->jnext, 1);
1398 /* remove label */
1399 table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1400 sym_free(s);
1402 *ptop = slast;
1405 /* eval an expression for #if/#elif */
1406 static int expr_preprocess(void)
1408 int c, t;
1409 TokenString *str;
1411 str = tok_str_alloc();
1412 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1413 next(); /* do macro subst */
1414 if (tok == TOK_DEFINED) {
1415 next_nomacro();
1416 t = tok;
1417 if (t == '(')
1418 next_nomacro();
1419 c = define_find(tok) != 0;
1420 if (t == '(')
1421 next_nomacro();
1422 tok = TOK_CINT;
1423 tokc.i = c;
1424 } else if (tok >= TOK_IDENT) {
1425 /* if undefined macro */
1426 tok = TOK_CINT;
1427 tokc.i = 0;
1429 tok_str_add_tok(str);
1431 tok_str_add(str, -1); /* simulate end of file */
1432 tok_str_add(str, 0);
1433 /* now evaluate C constant expression */
1434 begin_macro(str, 1);
1435 next();
1436 c = expr_const();
1437 end_macro();
1438 return c != 0;
1442 /* parse after #define */
1443 ST_FUNC void parse_define(void)
1445 Sym *s, *first, **ps;
1446 int v, t, varg, is_vaargs, spc;
1447 int saved_parse_flags = parse_flags;
1449 v = tok;
1450 if (v < TOK_IDENT)
1451 tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1452 /* XXX: should check if same macro (ANSI) */
1453 first = NULL;
1454 t = MACRO_OBJ;
1455 /* '(' must be just after macro definition for MACRO_FUNC */
1456 parse_flags |= PARSE_FLAG_SPACES;
1457 next_nomacro_spc();
1458 if (tok == '(') {
1459 /* must be able to parse TOK_DOTS (in asm mode '.' can be part of identifier) */
1460 parse_flags &= ~PARSE_FLAG_ASM_FILE;
1461 isidnum_table['.' - CH_EOF] = 0;
1462 next_nomacro();
1463 ps = &first;
1464 if (tok != ')') for (;;) {
1465 varg = tok;
1466 next_nomacro();
1467 is_vaargs = 0;
1468 if (varg == TOK_DOTS) {
1469 varg = TOK___VA_ARGS__;
1470 is_vaargs = 1;
1471 } else if (tok == TOK_DOTS && gnu_ext) {
1472 is_vaargs = 1;
1473 next_nomacro();
1475 if (varg < TOK_IDENT)
1476 bad_list:
1477 tcc_error("bad macro parameter list");
1478 s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1479 *ps = s;
1480 ps = &s->next;
1481 if (tok == ')')
1482 break;
1483 if (tok != ',' || is_vaargs)
1484 goto bad_list;
1485 next_nomacro();
1487 next_nomacro_spc();
1488 t = MACRO_FUNC;
1489 parse_flags |= (saved_parse_flags & PARSE_FLAG_ASM_FILE);
1490 isidnum_table['.' - CH_EOF] =
1491 (parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0;
1494 tokstr_buf.len = 0;
1495 spc = 2;
1496 parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1497 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1498 /* remove spaces around ## and after '#' */
1499 if (TOK_TWOSHARPS == tok) {
1500 if (2 == spc)
1501 goto bad_twosharp;
1502 if (1 == spc)
1503 --tokstr_buf.len;
1504 spc = 3;
1505 } else if ('#' == tok) {
1506 spc = 4;
1507 } else if (check_space(tok, &spc)) {
1508 goto skip;
1510 tok_str_add2(&tokstr_buf, tok, &tokc);
1511 skip:
1512 next_nomacro_spc();
1515 parse_flags = saved_parse_flags;
1516 if (spc == 1)
1517 --tokstr_buf.len; /* remove trailing space */
1518 tok_str_add(&tokstr_buf, 0);
1519 if (3 == spc)
1520 bad_twosharp:
1521 tcc_error("'##' cannot appear at either end of macro");
1522 define_push(v, t, tok_str_dup(&tokstr_buf), first);
1525 static CachedInclude *search_cached_include(TCCState *s1, const char *filename, int add)
1527 const unsigned char *s;
1528 unsigned int h;
1529 CachedInclude *e;
1530 int i;
1532 h = TOK_HASH_INIT;
1533 s = (unsigned char *) filename;
1534 while (*s) {
1535 #ifdef _WIN32
1536 h = TOK_HASH_FUNC(h, toup(*s));
1537 #else
1538 h = TOK_HASH_FUNC(h, *s);
1539 #endif
1540 s++;
1542 h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1544 i = s1->cached_includes_hash[h];
1545 for(;;) {
1546 if (i == 0)
1547 break;
1548 e = s1->cached_includes[i - 1];
1549 if (0 == PATHCMP(e->filename, filename))
1550 return e;
1551 i = e->hash_next;
1553 if (!add)
1554 return NULL;
1556 e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1557 strcpy(e->filename, filename);
1558 e->ifndef_macro = e->once = 0;
1559 dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1560 /* add in hash table */
1561 e->hash_next = s1->cached_includes_hash[h];
1562 s1->cached_includes_hash[h] = s1->nb_cached_includes;
1563 #ifdef INC_DEBUG
1564 printf("adding cached '%s'\n", filename);
1565 #endif
1566 return e;
1569 static void pragma_parse(TCCState *s1)
1571 next_nomacro();
1572 if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1573 int t = tok, v;
1574 Sym *s;
1576 if (next(), tok != '(')
1577 goto pragma_err;
1578 if (next(), tok != TOK_STR)
1579 goto pragma_err;
1580 v = tok_alloc(tokc.str.data, tokc.str.size - 1)->tok;
1581 if (next(), tok != ')')
1582 goto pragma_err;
1583 if (t == TOK_push_macro) {
1584 while (NULL == (s = define_find(v)))
1585 define_push(v, 0, NULL, NULL);
1586 s->type.ref = s; /* set push boundary */
1587 } else {
1588 for (s = define_stack; s; s = s->prev)
1589 if (s->v == v && s->type.ref == s) {
1590 s->type.ref = NULL;
1591 break;
1594 if (s)
1595 table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1596 else
1597 tcc_warning("unbalanced #pragma pop_macro");
1598 pp_debug_tok = t, pp_debug_symv = v;
1600 } else if (tok == TOK_once) {
1601 search_cached_include(s1, file->filename, 1)->once = pp_once;
1603 } else if (s1->ppfp) {
1604 /* tcc -E: keep pragmas below unchanged */
1605 unget_tok(' ');
1606 unget_tok(TOK_PRAGMA);
1607 unget_tok('#');
1608 unget_tok(TOK_LINEFEED);
1610 } else if (tok == TOK_pack) {
1611 /* This may be:
1612 #pragma pack(1) // set
1613 #pragma pack() // reset to default
1614 #pragma pack(push,1) // push & set
1615 #pragma pack(pop) // restore previous */
1616 next();
1617 skip('(');
1618 if (tok == TOK_ASM_pop) {
1619 next();
1620 if (s1->pack_stack_ptr <= s1->pack_stack) {
1621 stk_error:
1622 tcc_error("out of pack stack");
1624 s1->pack_stack_ptr--;
1625 } else {
1626 int val = 0;
1627 if (tok != ')') {
1628 if (tok == TOK_ASM_push) {
1629 next();
1630 if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1631 goto stk_error;
1632 s1->pack_stack_ptr++;
1633 skip(',');
1635 if (tok != TOK_CINT)
1636 goto pragma_err;
1637 val = tokc.i;
1638 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1639 goto pragma_err;
1640 next();
1642 *s1->pack_stack_ptr = val;
1644 if (tok != ')')
1645 goto pragma_err;
1647 } else if (tok == TOK_comment) {
1648 char *file;
1649 next();
1650 skip('(');
1651 if (tok != TOK_lib)
1652 goto pragma_warn;
1653 next();
1654 skip(',');
1655 if (tok != TOK_STR)
1656 goto pragma_err;
1657 file = tcc_strdup((char *)tokc.str.data);
1658 dynarray_add((void ***)&s1->pragma_libs, &s1->nb_pragma_libs, file);
1659 next();
1660 if (tok != ')')
1661 goto pragma_err;
1662 } else {
1663 pragma_warn:
1664 if (s1->warn_unsupported)
1665 tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1667 return;
1669 pragma_err:
1670 tcc_error("malformed #pragma directive");
1671 return;
1674 /* is_bof is true if first non space token at beginning of file */
1675 ST_FUNC void preprocess(int is_bof)
1677 TCCState *s1 = tcc_state;
1678 int i, c, n, saved_parse_flags;
1679 char buf[1024], *q;
1680 Sym *s;
1682 saved_parse_flags = parse_flags;
1683 parse_flags = PARSE_FLAG_PREPROCESS
1684 | PARSE_FLAG_TOK_NUM
1685 | PARSE_FLAG_TOK_STR
1686 | PARSE_FLAG_LINEFEED
1687 | (parse_flags & PARSE_FLAG_ASM_FILE)
1690 next_nomacro();
1691 redo:
1692 switch(tok) {
1693 case TOK_DEFINE:
1694 pp_debug_tok = tok;
1695 next_nomacro();
1696 pp_debug_symv = tok;
1697 parse_define();
1698 break;
1699 case TOK_UNDEF:
1700 pp_debug_tok = tok;
1701 next_nomacro();
1702 pp_debug_symv = tok;
1703 s = define_find(tok);
1704 /* undefine symbol by putting an invalid name */
1705 if (s)
1706 define_undef(s);
1707 break;
1708 case TOK_INCLUDE:
1709 case TOK_INCLUDE_NEXT:
1710 ch = file->buf_ptr[0];
1711 /* XXX: incorrect if comments : use next_nomacro with a special mode */
1712 skip_spaces();
1713 if (ch == '<') {
1714 c = '>';
1715 goto read_name;
1716 } else if (ch == '\"') {
1717 c = ch;
1718 read_name:
1719 inp();
1720 q = buf;
1721 while (ch != c && ch != '\n' && ch != CH_EOF) {
1722 if ((q - buf) < sizeof(buf) - 1)
1723 *q++ = ch;
1724 if (ch == '\\') {
1725 if (handle_stray_noerror() == 0)
1726 --q;
1727 } else
1728 inp();
1730 *q = '\0';
1731 minp();
1732 #if 0
1733 /* eat all spaces and comments after include */
1734 /* XXX: slightly incorrect */
1735 while (ch1 != '\n' && ch1 != CH_EOF)
1736 inp();
1737 #endif
1738 } else {
1739 /* computed #include : either we have only strings or
1740 we have anything enclosed in '<>' */
1741 next();
1742 buf[0] = '\0';
1743 if (tok == TOK_STR) {
1744 while (tok != TOK_LINEFEED) {
1745 if (tok != TOK_STR) {
1746 include_syntax:
1747 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1749 pstrcat(buf, sizeof(buf), (char *)tokc.str.data);
1750 next();
1752 c = '\"';
1753 } else {
1754 int len;
1755 while (tok != TOK_LINEFEED) {
1756 pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1757 next();
1759 len = strlen(buf);
1760 /* check syntax and remove '<>' */
1761 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1762 goto include_syntax;
1763 memmove(buf, buf + 1, len - 2);
1764 buf[len - 2] = '\0';
1765 c = '>';
1769 if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1770 tcc_error("#include recursion too deep");
1771 /* store current file in stack, but increment stack later below */
1772 *s1->include_stack_ptr = file;
1773 i = tok == TOK_INCLUDE_NEXT ? file->include_next_index : 0;
1774 n = 2 + s1->nb_include_paths + s1->nb_sysinclude_paths;
1775 for (; i < n; ++i) {
1776 char buf1[sizeof file->filename];
1777 CachedInclude *e;
1778 const char *path;
1780 if (i == 0) {
1781 /* check absolute include path */
1782 if (!IS_ABSPATH(buf))
1783 continue;
1784 buf1[0] = 0;
1786 } else if (i == 1) {
1787 /* search in file's dir if "header.h" */
1788 if (c != '\"')
1789 continue;
1790 path = file->filename;
1791 pstrncpy(buf1, path, tcc_basename(path) - path);
1793 } else {
1794 /* search in all the include paths */
1795 int j = i - 2, k = j - s1->nb_include_paths;
1796 path = k < 0 ? s1->include_paths[j] : s1->sysinclude_paths[k];
1797 pstrcpy(buf1, sizeof(buf1), path);
1798 pstrcat(buf1, sizeof(buf1), "/");
1801 pstrcat(buf1, sizeof(buf1), buf);
1802 e = search_cached_include(s1, buf1, 0);
1803 if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) {
1804 /* no need to parse the include because the 'ifndef macro'
1805 is defined (or had #pragma once) */
1806 #ifdef INC_DEBUG
1807 printf("%s: skipping cached %s\n", file->filename, buf1);
1808 #endif
1809 goto include_done;
1812 if (tcc_open(s1, buf1) < 0)
1813 continue;
1815 file->include_next_index = i + 1;
1816 #ifdef INC_DEBUG
1817 printf("%s: including %s\n", file->prev->filename, file->filename);
1818 #endif
1819 /* update target deps */
1820 dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1821 tcc_strdup(buf1));
1822 /* push current file in stack */
1823 ++s1->include_stack_ptr;
1824 /* add include file debug info */
1825 if (s1->do_debug)
1826 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1827 tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1828 ch = file->buf_ptr[0];
1829 goto the_end;
1831 tcc_error("include file '%s' not found", buf);
1832 include_done:
1833 break;
1834 case TOK_IFNDEF:
1835 c = 1;
1836 goto do_ifdef;
1837 case TOK_IF:
1838 c = expr_preprocess();
1839 goto do_if;
1840 case TOK_IFDEF:
1841 c = 0;
1842 do_ifdef:
1843 next_nomacro();
1844 if (tok < TOK_IDENT)
1845 tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1846 if (is_bof) {
1847 if (c) {
1848 #ifdef INC_DEBUG
1849 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1850 #endif
1851 file->ifndef_macro = tok;
1854 c = (define_find(tok) != 0) ^ c;
1855 do_if:
1856 if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1857 tcc_error("memory full (ifdef)");
1858 *s1->ifdef_stack_ptr++ = c;
1859 goto test_skip;
1860 case TOK_ELSE:
1861 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1862 tcc_error("#else without matching #if");
1863 if (s1->ifdef_stack_ptr[-1] & 2)
1864 tcc_error("#else after #else");
1865 c = (s1->ifdef_stack_ptr[-1] ^= 3);
1866 goto test_else;
1867 case TOK_ELIF:
1868 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1869 tcc_error("#elif without matching #if");
1870 c = s1->ifdef_stack_ptr[-1];
1871 if (c > 1)
1872 tcc_error("#elif after #else");
1873 /* last #if/#elif expression was true: we skip */
1874 if (c == 1) {
1875 c = 0;
1876 } else {
1877 c = expr_preprocess();
1878 s1->ifdef_stack_ptr[-1] = c;
1880 test_else:
1881 if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1882 file->ifndef_macro = 0;
1883 test_skip:
1884 if (!(c & 1)) {
1885 preprocess_skip();
1886 is_bof = 0;
1887 goto redo;
1889 break;
1890 case TOK_ENDIF:
1891 if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1892 tcc_error("#endif without matching #if");
1893 s1->ifdef_stack_ptr--;
1894 /* '#ifndef macro' was at the start of file. Now we check if
1895 an '#endif' is exactly at the end of file */
1896 if (file->ifndef_macro &&
1897 s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1898 file->ifndef_macro_saved = file->ifndef_macro;
1899 /* need to set to zero to avoid false matches if another
1900 #ifndef at middle of file */
1901 file->ifndef_macro = 0;
1902 while (tok != TOK_LINEFEED)
1903 next_nomacro();
1904 tok_flags |= TOK_FLAG_ENDIF;
1905 goto the_end;
1907 break;
1908 case TOK_PPNUM:
1909 n = strtoul((char*)tokc.str.data, &q, 10);
1910 goto _line_num;
1911 case TOK_LINE:
1912 next();
1913 if (tok != TOK_CINT)
1914 _line_err:
1915 tcc_error("wrong #line format");
1916 n = tokc.i;
1917 _line_num:
1918 next();
1919 if (tok != TOK_LINEFEED) {
1920 if (tok == TOK_STR)
1921 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.str.data);
1922 else if (parse_flags & PARSE_FLAG_ASM_FILE)
1923 break;
1924 else
1925 goto _line_err;
1926 --n;
1928 if (file->fd > 0)
1929 total_lines += file->line_num - n;
1930 file->line_num = n;
1931 if (s1->do_debug)
1932 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1933 break;
1934 case TOK_ERROR:
1935 case TOK_WARNING:
1936 c = tok;
1937 ch = file->buf_ptr[0];
1938 skip_spaces();
1939 q = buf;
1940 while (ch != '\n' && ch != CH_EOF) {
1941 if ((q - buf) < sizeof(buf) - 1)
1942 *q++ = ch;
1943 if (ch == '\\') {
1944 if (handle_stray_noerror() == 0)
1945 --q;
1946 } else
1947 inp();
1949 *q = '\0';
1950 if (c == TOK_ERROR)
1951 tcc_error("#error %s", buf);
1952 else
1953 tcc_warning("#warning %s", buf);
1954 break;
1955 case TOK_PRAGMA:
1956 pragma_parse(s1);
1957 break;
1958 case TOK_LINEFEED:
1959 goto the_end;
1960 default:
1961 /* ignore gas line comment in an 'S' file. */
1962 if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
1963 goto ignore;
1964 if (tok == '!' && is_bof)
1965 /* '!' is ignored at beginning to allow C scripts. */
1966 goto ignore;
1967 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1968 ignore:
1969 file->buf_ptr = parse_line_comment(file->buf_ptr - 1);
1970 goto the_end;
1972 /* ignore other preprocess commands or #! for C scripts */
1973 while (tok != TOK_LINEFEED)
1974 next_nomacro();
1975 the_end:
1976 parse_flags = saved_parse_flags;
1979 /* evaluate escape codes in a string. */
1980 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1982 int c, n;
1983 const uint8_t *p;
1985 p = buf;
1986 for(;;) {
1987 c = *p;
1988 if (c == '\0')
1989 break;
1990 if (c == '\\') {
1991 p++;
1992 /* escape */
1993 c = *p;
1994 switch(c) {
1995 case '0': case '1': case '2': case '3':
1996 case '4': case '5': case '6': case '7':
1997 /* at most three octal digits */
1998 n = c - '0';
1999 p++;
2000 c = *p;
2001 if (isoct(c)) {
2002 n = n * 8 + c - '0';
2003 p++;
2004 c = *p;
2005 if (isoct(c)) {
2006 n = n * 8 + c - '0';
2007 p++;
2010 c = n;
2011 goto add_char_nonext;
2012 case 'x':
2013 case 'u':
2014 case 'U':
2015 p++;
2016 n = 0;
2017 for(;;) {
2018 c = *p;
2019 if (c >= 'a' && c <= 'f')
2020 c = c - 'a' + 10;
2021 else if (c >= 'A' && c <= 'F')
2022 c = c - 'A' + 10;
2023 else if (isnum(c))
2024 c = c - '0';
2025 else
2026 break;
2027 n = n * 16 + c;
2028 p++;
2030 c = n;
2031 goto add_char_nonext;
2032 case 'a':
2033 c = '\a';
2034 break;
2035 case 'b':
2036 c = '\b';
2037 break;
2038 case 'f':
2039 c = '\f';
2040 break;
2041 case 'n':
2042 c = '\n';
2043 break;
2044 case 'r':
2045 c = '\r';
2046 break;
2047 case 't':
2048 c = '\t';
2049 break;
2050 case 'v':
2051 c = '\v';
2052 break;
2053 case 'e':
2054 if (!gnu_ext)
2055 goto invalid_escape;
2056 c = 27;
2057 break;
2058 case '\'':
2059 case '\"':
2060 case '\\':
2061 case '?':
2062 break;
2063 default:
2064 invalid_escape:
2065 if (c >= '!' && c <= '~')
2066 tcc_warning("unknown escape sequence: \'\\%c\'", c);
2067 else
2068 tcc_warning("unknown escape sequence: \'\\x%x\'", c);
2069 break;
2072 p++;
2073 add_char_nonext:
2074 if (!is_long)
2075 cstr_ccat(outstr, c);
2076 else
2077 cstr_wccat(outstr, c);
2079 /* add a trailing '\0' */
2080 if (!is_long)
2081 cstr_ccat(outstr, '\0');
2082 else
2083 cstr_wccat(outstr, '\0');
2086 static void parse_string(const char *s, int len)
2088 uint8_t buf[1000], *p = buf;
2089 int is_long, sep;
2091 if ((is_long = *s == 'L'))
2092 ++s, --len;
2093 sep = *s++;
2094 len -= 2;
2095 if (len >= sizeof buf)
2096 p = tcc_malloc(len + 1);
2097 memcpy(p, s, len);
2098 p[len] = 0;
2100 cstr_reset(&tokcstr);
2101 parse_escape_string(&tokcstr, p, is_long);
2102 if (p != buf)
2103 tcc_free(p);
2105 if (sep == '\'') {
2106 int char_size;
2107 /* XXX: make it portable */
2108 if (!is_long)
2109 char_size = 1;
2110 else
2111 char_size = sizeof(nwchar_t);
2112 if (tokcstr.size <= char_size)
2113 tcc_error("empty character constant");
2114 if (tokcstr.size > 2 * char_size)
2115 tcc_warning("multi-character character constant");
2116 if (!is_long) {
2117 tokc.i = *(int8_t *)tokcstr.data;
2118 tok = TOK_CCHAR;
2119 } else {
2120 tokc.i = *(nwchar_t *)tokcstr.data;
2121 tok = TOK_LCHAR;
2123 } else {
2124 tokc.str.size = tokcstr.size;
2125 tokc.str.data = tokcstr.data;
2126 if (!is_long)
2127 tok = TOK_STR;
2128 else
2129 tok = TOK_LSTR;
2133 /* we use 64 bit numbers */
2134 #define BN_SIZE 2
2136 /* bn = (bn << shift) | or_val */
2137 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2139 int i;
2140 unsigned int v;
2141 for(i=0;i<BN_SIZE;i++) {
2142 v = bn[i];
2143 bn[i] = (v << shift) | or_val;
2144 or_val = v >> (32 - shift);
2148 static void bn_zero(unsigned int *bn)
2150 int i;
2151 for(i=0;i<BN_SIZE;i++) {
2152 bn[i] = 0;
2156 /* parse number in null terminated string 'p' and return it in the
2157 current token */
2158 static void parse_number(const char *p)
2160 int b, t, shift, frac_bits, s, exp_val, ch;
2161 char *q;
2162 unsigned int bn[BN_SIZE];
2163 double d;
2165 /* number */
2166 q = token_buf;
2167 ch = *p++;
2168 t = ch;
2169 ch = *p++;
2170 *q++ = t;
2171 b = 10;
2172 if (t == '.') {
2173 goto float_frac_parse;
2174 } else if (t == '0') {
2175 if (ch == 'x' || ch == 'X') {
2176 q--;
2177 ch = *p++;
2178 b = 16;
2179 } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2180 q--;
2181 ch = *p++;
2182 b = 2;
2185 /* parse all digits. cannot check octal numbers at this stage
2186 because of floating point constants */
2187 while (1) {
2188 if (ch >= 'a' && ch <= 'f')
2189 t = ch - 'a' + 10;
2190 else if (ch >= 'A' && ch <= 'F')
2191 t = ch - 'A' + 10;
2192 else if (isnum(ch))
2193 t = ch - '0';
2194 else
2195 break;
2196 if (t >= b)
2197 break;
2198 if (q >= token_buf + STRING_MAX_SIZE) {
2199 num_too_long:
2200 tcc_error("number too long");
2202 *q++ = ch;
2203 ch = *p++;
2205 if (ch == '.' ||
2206 ((ch == 'e' || ch == 'E') && b == 10) ||
2207 ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2208 if (b != 10) {
2209 /* NOTE: strtox should support that for hexa numbers, but
2210 non ISOC99 libcs do not support it, so we prefer to do
2211 it by hand */
2212 /* hexadecimal or binary floats */
2213 /* XXX: handle overflows */
2214 *q = '\0';
2215 if (b == 16)
2216 shift = 4;
2217 else
2218 shift = 1;
2219 bn_zero(bn);
2220 q = token_buf;
2221 while (1) {
2222 t = *q++;
2223 if (t == '\0') {
2224 break;
2225 } else if (t >= 'a') {
2226 t = t - 'a' + 10;
2227 } else if (t >= 'A') {
2228 t = t - 'A' + 10;
2229 } else {
2230 t = t - '0';
2232 bn_lshift(bn, shift, t);
2234 frac_bits = 0;
2235 if (ch == '.') {
2236 ch = *p++;
2237 while (1) {
2238 t = ch;
2239 if (t >= 'a' && t <= 'f') {
2240 t = t - 'a' + 10;
2241 } else if (t >= 'A' && t <= 'F') {
2242 t = t - 'A' + 10;
2243 } else if (t >= '0' && t <= '9') {
2244 t = t - '0';
2245 } else {
2246 break;
2248 if (t >= b)
2249 tcc_error("invalid digit");
2250 bn_lshift(bn, shift, t);
2251 frac_bits += shift;
2252 ch = *p++;
2255 if (ch != 'p' && ch != 'P')
2256 expect("exponent");
2257 ch = *p++;
2258 s = 1;
2259 exp_val = 0;
2260 if (ch == '+') {
2261 ch = *p++;
2262 } else if (ch == '-') {
2263 s = -1;
2264 ch = *p++;
2266 if (ch < '0' || ch > '9')
2267 expect("exponent digits");
2268 while (ch >= '0' && ch <= '9') {
2269 exp_val = exp_val * 10 + ch - '0';
2270 ch = *p++;
2272 exp_val = exp_val * s;
2274 /* now we can generate the number */
2275 /* XXX: should patch directly float number */
2276 d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2277 d = ldexp(d, exp_val - frac_bits);
2278 t = toup(ch);
2279 if (t == 'F') {
2280 ch = *p++;
2281 tok = TOK_CFLOAT;
2282 /* float : should handle overflow */
2283 tokc.f = (float)d;
2284 } else if (t == 'L') {
2285 ch = *p++;
2286 #ifdef TCC_TARGET_PE
2287 tok = TOK_CDOUBLE;
2288 tokc.d = d;
2289 #else
2290 tok = TOK_CLDOUBLE;
2291 /* XXX: not large enough */
2292 tokc.ld = (long double)d;
2293 #endif
2294 } else {
2295 tok = TOK_CDOUBLE;
2296 tokc.d = d;
2298 } else {
2299 /* decimal floats */
2300 if (ch == '.') {
2301 if (q >= token_buf + STRING_MAX_SIZE)
2302 goto num_too_long;
2303 *q++ = ch;
2304 ch = *p++;
2305 float_frac_parse:
2306 while (ch >= '0' && ch <= '9') {
2307 if (q >= token_buf + STRING_MAX_SIZE)
2308 goto num_too_long;
2309 *q++ = ch;
2310 ch = *p++;
2313 if (ch == 'e' || ch == 'E') {
2314 if (q >= token_buf + STRING_MAX_SIZE)
2315 goto num_too_long;
2316 *q++ = ch;
2317 ch = *p++;
2318 if (ch == '-' || ch == '+') {
2319 if (q >= token_buf + STRING_MAX_SIZE)
2320 goto num_too_long;
2321 *q++ = ch;
2322 ch = *p++;
2324 if (ch < '0' || ch > '9')
2325 expect("exponent digits");
2326 while (ch >= '0' && ch <= '9') {
2327 if (q >= token_buf + STRING_MAX_SIZE)
2328 goto num_too_long;
2329 *q++ = ch;
2330 ch = *p++;
2333 *q = '\0';
2334 t = toup(ch);
2335 errno = 0;
2336 if (t == 'F') {
2337 ch = *p++;
2338 tok = TOK_CFLOAT;
2339 tokc.f = strtof(token_buf, NULL);
2340 } else if (t == 'L') {
2341 ch = *p++;
2342 #ifdef TCC_TARGET_PE
2343 tok = TOK_CDOUBLE;
2344 tokc.d = strtod(token_buf, NULL);
2345 #else
2346 tok = TOK_CLDOUBLE;
2347 tokc.ld = strtold(token_buf, NULL);
2348 #endif
2349 } else {
2350 tok = TOK_CDOUBLE;
2351 tokc.d = strtod(token_buf, NULL);
2354 } else {
2355 unsigned long long n, n1;
2356 int lcount, ucount, must_64bit;
2357 const char *p1;
2359 /* integer number */
2360 *q = '\0';
2361 q = token_buf;
2362 if (b == 10 && *q == '0') {
2363 b = 8;
2364 q++;
2366 n = 0;
2367 while(1) {
2368 t = *q++;
2369 /* no need for checks except for base 10 / 8 errors */
2370 if (t == '\0')
2371 break;
2372 else if (t >= 'a')
2373 t = t - 'a' + 10;
2374 else if (t >= 'A')
2375 t = t - 'A' + 10;
2376 else
2377 t = t - '0';
2378 if (t >= b)
2379 tcc_error("invalid digit");
2380 n1 = n;
2381 n = n * b + t;
2382 /* detect overflow */
2383 /* XXX: this test is not reliable */
2384 if (n < n1)
2385 tcc_error("integer constant overflow");
2388 /* Determine the characteristics (unsigned and/or 64bit) the type of
2389 the constant must have according to the constant suffix(es) */
2390 lcount = ucount = must_64bit = 0;
2391 p1 = p;
2392 for(;;) {
2393 t = toup(ch);
2394 if (t == 'L') {
2395 if (lcount >= 2)
2396 tcc_error("three 'l's in integer constant");
2397 if (lcount && *(p - 1) != ch)
2398 tcc_error("incorrect integer suffix: %s", p1);
2399 lcount++;
2400 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2401 if (lcount == 2)
2402 #endif
2403 must_64bit = 1;
2404 ch = *p++;
2405 } else if (t == 'U') {
2406 if (ucount >= 1)
2407 tcc_error("two 'u's in integer constant");
2408 ucount++;
2409 ch = *p++;
2410 } else {
2411 break;
2415 /* Whether 64 bits are needed to hold the constant's value */
2416 if (n & 0xffffffff00000000LL || must_64bit) {
2417 tok = TOK_CLLONG;
2418 n1 = n >> 32;
2419 } else {
2420 tok = TOK_CINT;
2421 n1 = n;
2424 /* Whether type must be unsigned to hold the constant's value */
2425 if (ucount || ((n1 >> 31) && (b != 10))) {
2426 if (tok == TOK_CLLONG)
2427 tok = TOK_CULLONG;
2428 else
2429 tok = TOK_CUINT;
2430 /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2431 } else if (n1 >> 31) {
2432 if (tok == TOK_CINT)
2433 tok = TOK_CLLONG;
2434 else
2435 tcc_error("integer constant overflow");
2438 tokc.i = n;
2440 if (ch)
2441 tcc_error("invalid number\n");
2445 #define PARSE2(c1, tok1, c2, tok2) \
2446 case c1: \
2447 PEEKC(c, p); \
2448 if (c == c2) { \
2449 p++; \
2450 tok = tok2; \
2451 } else { \
2452 tok = tok1; \
2454 break;
2456 /* return next token without macro substitution */
2457 static inline void next_nomacro1(void)
2459 int t, c, is_long, len;
2460 TokenSym *ts;
2461 uint8_t *p, *p1;
2462 unsigned int h;
2464 p = file->buf_ptr;
2465 redo_no_start:
2466 c = *p;
2467 switch(c) {
2468 case ' ':
2469 case '\t':
2470 tok = c;
2471 p++;
2472 if (parse_flags & PARSE_FLAG_SPACES)
2473 goto keep_tok_flags;
2474 while (isidnum_table[*p - CH_EOF] & IS_SPC)
2475 ++p;
2476 goto redo_no_start;
2477 case '\f':
2478 case '\v':
2479 case '\r':
2480 p++;
2481 goto redo_no_start;
2482 case '\\':
2483 /* first look if it is in fact an end of buffer */
2484 c = handle_stray1(p);
2485 p = file->buf_ptr;
2486 if (c == '\\')
2487 goto parse_simple;
2488 if (c != CH_EOF)
2489 goto redo_no_start;
2491 TCCState *s1 = tcc_state;
2492 if ((parse_flags & PARSE_FLAG_LINEFEED)
2493 && !(tok_flags & TOK_FLAG_EOF)) {
2494 tok_flags |= TOK_FLAG_EOF;
2495 tok = TOK_LINEFEED;
2496 goto keep_tok_flags;
2497 } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2498 tok = TOK_EOF;
2499 } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2500 tcc_error("missing #endif");
2501 } else if (s1->include_stack_ptr == s1->include_stack) {
2502 /* no include left : end of file. */
2503 tok = TOK_EOF;
2504 } else {
2505 tok_flags &= ~TOK_FLAG_EOF;
2506 /* pop include file */
2508 /* test if previous '#endif' was after a #ifdef at
2509 start of file */
2510 if (tok_flags & TOK_FLAG_ENDIF) {
2511 #ifdef INC_DEBUG
2512 printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2513 #endif
2514 search_cached_include(s1, file->filename, 1)
2515 ->ifndef_macro = file->ifndef_macro_saved;
2516 tok_flags &= ~TOK_FLAG_ENDIF;
2519 /* add end of include file debug info */
2520 if (tcc_state->do_debug) {
2521 put_stabd(N_EINCL, 0, 0);
2523 /* pop include stack */
2524 tcc_close();
2525 s1->include_stack_ptr--;
2526 p = file->buf_ptr;
2527 goto redo_no_start;
2530 break;
2532 case '\n':
2533 file->line_num++;
2534 tok_flags |= TOK_FLAG_BOL;
2535 p++;
2536 maybe_newline:
2537 if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2538 goto redo_no_start;
2539 tok = TOK_LINEFEED;
2540 goto keep_tok_flags;
2542 case '#':
2543 /* XXX: simplify */
2544 PEEKC(c, p);
2545 if ((tok_flags & TOK_FLAG_BOL) &&
2546 (parse_flags & PARSE_FLAG_PREPROCESS)) {
2547 file->buf_ptr = p;
2548 preprocess(tok_flags & TOK_FLAG_BOF);
2549 p = file->buf_ptr;
2550 goto maybe_newline;
2551 } else {
2552 if (c == '#') {
2553 p++;
2554 tok = TOK_TWOSHARPS;
2555 } else {
2556 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2557 p = parse_line_comment(p - 1);
2558 goto redo_no_start;
2559 } else {
2560 tok = '#';
2564 break;
2566 /* dollar is allowed to start identifiers when not parsing asm */
2567 case '$':
2568 if (!(isidnum_table[c - CH_EOF] & IS_ID)
2569 || (parse_flags & PARSE_FLAG_ASM_FILE))
2570 goto parse_simple;
2572 case 'a': case 'b': case 'c': case 'd':
2573 case 'e': case 'f': case 'g': case 'h':
2574 case 'i': case 'j': case 'k': case 'l':
2575 case 'm': case 'n': case 'o': case 'p':
2576 case 'q': case 'r': case 's': case 't':
2577 case 'u': case 'v': case 'w': case 'x':
2578 case 'y': case 'z':
2579 case 'A': case 'B': case 'C': case 'D':
2580 case 'E': case 'F': case 'G': case 'H':
2581 case 'I': case 'J': case 'K':
2582 case 'M': case 'N': case 'O': case 'P':
2583 case 'Q': case 'R': case 'S': case 'T':
2584 case 'U': case 'V': case 'W': case 'X':
2585 case 'Y': case 'Z':
2586 case '_':
2587 parse_ident_fast:
2588 p1 = p;
2589 h = TOK_HASH_INIT;
2590 h = TOK_HASH_FUNC(h, c);
2591 while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2592 h = TOK_HASH_FUNC(h, c);
2593 len = p - p1;
2594 if (c != '\\') {
2595 TokenSym **pts;
2597 /* fast case : no stray found, so we have the full token
2598 and we have already hashed it */
2599 h &= (TOK_HASH_SIZE - 1);
2600 pts = &hash_ident[h];
2601 for(;;) {
2602 ts = *pts;
2603 if (!ts)
2604 break;
2605 if (ts->len == len && !memcmp(ts->str, p1, len))
2606 goto token_found;
2607 pts = &(ts->hash_next);
2609 ts = tok_alloc_new(pts, (char *) p1, len);
2610 token_found: ;
2611 } else {
2612 /* slower case */
2613 cstr_reset(&tokcstr);
2614 cstr_cat(&tokcstr, p1, len);
2615 p--;
2616 PEEKC(c, p);
2617 parse_ident_slow:
2618 while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2620 cstr_ccat(&tokcstr, c);
2621 PEEKC(c, p);
2623 ts = tok_alloc(tokcstr.data, tokcstr.size);
2625 tok = ts->tok;
2626 break;
2627 case 'L':
2628 t = p[1];
2629 if (t != '\\' && t != '\'' && t != '\"') {
2630 /* fast case */
2631 goto parse_ident_fast;
2632 } else {
2633 PEEKC(c, p);
2634 if (c == '\'' || c == '\"') {
2635 is_long = 1;
2636 goto str_const;
2637 } else {
2638 cstr_reset(&tokcstr);
2639 cstr_ccat(&tokcstr, 'L');
2640 goto parse_ident_slow;
2643 break;
2645 case '0': case '1': case '2': case '3':
2646 case '4': case '5': case '6': case '7':
2647 case '8': case '9':
2648 t = c;
2649 PEEKC(c, p);
2650 /* after the first digit, accept digits, alpha, '.' or sign if
2651 prefixed by 'eEpP' */
2652 parse_num:
2653 cstr_reset(&tokcstr);
2654 for(;;) {
2655 cstr_ccat(&tokcstr, t);
2656 if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2657 || c == '.'
2658 || ((c == '+' || c == '-')
2659 && (((t == 'e' || t == 'E')
2660 && !(parse_flags & PARSE_FLAG_ASM_FILE
2661 /* 0xe+1 is 3 tokens in asm */
2662 && ((char*)tokcstr.data)[0] == '0'
2663 && toup(((char*)tokcstr.data)[1]) == 'X'))
2664 || t == 'p' || t == 'P'))))
2665 break;
2666 t = c;
2667 PEEKC(c, p);
2669 /* We add a trailing '\0' to ease parsing */
2670 cstr_ccat(&tokcstr, '\0');
2671 tokc.str.size = tokcstr.size;
2672 tokc.str.data = tokcstr.data;
2673 tok = TOK_PPNUM;
2674 break;
2676 case '.':
2677 /* special dot handling because it can also start a number */
2678 PEEKC(c, p);
2679 if (isnum(c)) {
2680 t = '.';
2681 goto parse_num;
2682 } else if ((parse_flags & PARSE_FLAG_ASM_FILE)
2683 && (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))) {
2684 *--p = c = '.';
2685 goto parse_ident_fast;
2686 } else if (c == '.') {
2687 PEEKC(c, p);
2688 if (c == '.') {
2689 p++;
2690 tok = TOK_DOTS;
2691 } else {
2692 *--p = '.'; /* may underflow into file->unget[] */
2693 tok = '.';
2695 } else {
2696 tok = '.';
2698 break;
2699 case '\'':
2700 case '\"':
2701 is_long = 0;
2702 str_const:
2703 cstr_reset(&tokcstr);
2704 if (is_long)
2705 cstr_ccat(&tokcstr, 'L');
2706 cstr_ccat(&tokcstr, c);
2707 p = parse_pp_string(p, c, &tokcstr);
2708 cstr_ccat(&tokcstr, c);
2709 cstr_ccat(&tokcstr, '\0');
2710 tokc.str.size = tokcstr.size;
2711 tokc.str.data = tokcstr.data;
2712 tok = TOK_PPSTR;
2713 break;
2715 case '<':
2716 PEEKC(c, p);
2717 if (c == '=') {
2718 p++;
2719 tok = TOK_LE;
2720 } else if (c == '<') {
2721 PEEKC(c, p);
2722 if (c == '=') {
2723 p++;
2724 tok = TOK_A_SHL;
2725 } else {
2726 tok = TOK_SHL;
2728 } else {
2729 tok = TOK_LT;
2731 break;
2732 case '>':
2733 PEEKC(c, p);
2734 if (c == '=') {
2735 p++;
2736 tok = TOK_GE;
2737 } else if (c == '>') {
2738 PEEKC(c, p);
2739 if (c == '=') {
2740 p++;
2741 tok = TOK_A_SAR;
2742 } else {
2743 tok = TOK_SAR;
2745 } else {
2746 tok = TOK_GT;
2748 break;
2750 case '&':
2751 PEEKC(c, p);
2752 if (c == '&') {
2753 p++;
2754 tok = TOK_LAND;
2755 } else if (c == '=') {
2756 p++;
2757 tok = TOK_A_AND;
2758 } else {
2759 tok = '&';
2761 break;
2763 case '|':
2764 PEEKC(c, p);
2765 if (c == '|') {
2766 p++;
2767 tok = TOK_LOR;
2768 } else if (c == '=') {
2769 p++;
2770 tok = TOK_A_OR;
2771 } else {
2772 tok = '|';
2774 break;
2776 case '+':
2777 PEEKC(c, p);
2778 if (c == '+') {
2779 p++;
2780 tok = TOK_INC;
2781 } else if (c == '=') {
2782 p++;
2783 tok = TOK_A_ADD;
2784 } else {
2785 tok = '+';
2787 break;
2789 case '-':
2790 PEEKC(c, p);
2791 if (c == '-') {
2792 p++;
2793 tok = TOK_DEC;
2794 } else if (c == '=') {
2795 p++;
2796 tok = TOK_A_SUB;
2797 } else if (c == '>') {
2798 p++;
2799 tok = TOK_ARROW;
2800 } else {
2801 tok = '-';
2803 break;
2805 PARSE2('!', '!', '=', TOK_NE)
2806 PARSE2('=', '=', '=', TOK_EQ)
2807 PARSE2('*', '*', '=', TOK_A_MUL)
2808 PARSE2('%', '%', '=', TOK_A_MOD)
2809 PARSE2('^', '^', '=', TOK_A_XOR)
2811 /* comments or operator */
2812 case '/':
2813 PEEKC(c, p);
2814 if (c == '*') {
2815 p = parse_comment(p);
2816 /* comments replaced by a blank */
2817 tok = ' ';
2818 goto keep_tok_flags;
2819 } else if (c == '/') {
2820 p = parse_line_comment(p);
2821 tok = ' ';
2822 goto keep_tok_flags;
2823 } else if (c == '=') {
2824 p++;
2825 tok = TOK_A_DIV;
2826 } else {
2827 tok = '/';
2829 break;
2831 /* simple tokens */
2832 case '(':
2833 case ')':
2834 case '[':
2835 case ']':
2836 case '{':
2837 case '}':
2838 case ',':
2839 case ';':
2840 case ':':
2841 case '?':
2842 case '~':
2843 case '@': /* only used in assembler */
2844 parse_simple:
2845 tok = c;
2846 p++;
2847 break;
2848 default:
2849 if (c >= 0x80 && c <= 0xFF) /* utf8 identifiers */
2850 goto parse_ident_fast;
2851 if (parse_flags & PARSE_FLAG_ASM_FILE)
2852 goto parse_simple;
2853 tcc_error("unrecognized character \\x%02x", c);
2854 break;
2856 tok_flags = 0;
2857 keep_tok_flags:
2858 file->buf_ptr = p;
2859 #if defined(PARSE_DEBUG)
2860 printf("token = %s\n", get_tok_str(tok, &tokc));
2861 #endif
2864 /* return next token without macro substitution. Can read input from
2865 macro_ptr buffer */
2866 static void next_nomacro_spc(void)
2868 if (macro_ptr) {
2869 redo:
2870 tok = *macro_ptr;
2871 if (tok) {
2872 TOK_GET(&tok, &macro_ptr, &tokc);
2873 if (tok == TOK_LINENUM) {
2874 file->line_num = tokc.i;
2875 goto redo;
2878 } else {
2879 next_nomacro1();
2883 ST_FUNC void next_nomacro(void)
2885 do {
2886 next_nomacro_spc();
2887 } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
2891 static void macro_subst(
2892 TokenString *tok_str,
2893 Sym **nested_list,
2894 const int *macro_str,
2895 int can_read_stream
2898 /* substitute arguments in replacement lists in macro_str by the values in
2899 args (field d) and return allocated string */
2900 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2902 int t, t0, t1, spc;
2903 const int *st;
2904 Sym *s;
2905 CValue cval;
2906 TokenString str;
2907 CString cstr;
2909 tok_str_new(&str);
2910 t0 = t1 = 0;
2911 while(1) {
2912 TOK_GET(&t, &macro_str, &cval);
2913 if (!t)
2914 break;
2915 if (t == '#') {
2916 /* stringize */
2917 TOK_GET(&t, &macro_str, &cval);
2918 if (!t)
2919 goto bad_stringy;
2920 s = sym_find2(args, t);
2921 if (s) {
2922 cstr_new(&cstr);
2923 cstr_ccat(&cstr, '\"');
2924 st = s->d;
2925 spc = 0;
2926 while (*st) {
2927 TOK_GET(&t, &st, &cval);
2928 if (t != TOK_PLCHLDR
2929 && t != TOK_NOSUBST
2930 && 0 == check_space(t, &spc)) {
2931 const char *s = get_tok_str(t, &cval);
2932 while (*s) {
2933 if (t == TOK_PPSTR && *s != '\'')
2934 add_char(&cstr, *s);
2935 else
2936 cstr_ccat(&cstr, *s);
2937 ++s;
2941 cstr.size -= spc;
2942 cstr_ccat(&cstr, '\"');
2943 cstr_ccat(&cstr, '\0');
2944 #ifdef PP_DEBUG
2945 printf("\nstringize: <%s>\n", (char *)cstr.data);
2946 #endif
2947 /* add string */
2948 cval.str.size = cstr.size;
2949 cval.str.data = cstr.data;
2950 tok_str_add2(&str, TOK_PPSTR, &cval);
2951 cstr_free(&cstr);
2952 } else {
2953 bad_stringy:
2954 expect("macro parameter after '#'");
2956 } else if (t >= TOK_IDENT) {
2957 s = sym_find2(args, t);
2958 if (s) {
2959 int l0 = str.len;
2960 st = s->d;
2961 /* if '##' is present before or after, no arg substitution */
2962 if (*macro_str == TOK_TWOSHARPS || t1 == TOK_TWOSHARPS) {
2963 /* special case for var arg macros : ## eats the ','
2964 if empty VA_ARGS variable. */
2965 if (t1 == TOK_TWOSHARPS && t0 == ',' && gnu_ext && s->type.t) {
2966 if (*st == 0) {
2967 /* suppress ',' '##' */
2968 str.len -= 2;
2969 } else {
2970 /* suppress '##' and add variable */
2971 str.len--;
2972 goto add_var;
2974 } else {
2975 for(;;) {
2976 int t1;
2977 TOK_GET(&t1, &st, &cval);
2978 if (!t1)
2979 break;
2980 tok_str_add2(&str, t1, &cval);
2984 } else {
2985 add_var:
2986 /* NOTE: the stream cannot be read when macro
2987 substituing an argument */
2988 macro_subst(&str, nested_list, st, 0);
2990 if (str.len == l0) /* exanded to empty string */
2991 tok_str_add(&str, TOK_PLCHLDR);
2992 } else {
2993 tok_str_add(&str, t);
2995 } else {
2996 tok_str_add2(&str, t, &cval);
2998 t0 = t1, t1 = t;
3000 tok_str_add(&str, 0);
3001 return str.str;
3004 static char const ab_month_name[12][4] =
3006 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3007 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3010 /* peek or read [ws_str == NULL] next token from function macro call,
3011 walking up macro levels up to the file if necessary */
3012 static int next_argstream(Sym **nested_list, int can_read_stream, TokenString *ws_str)
3014 int t;
3015 const int *p;
3016 Sym *sa;
3018 for (;;) {
3019 if (macro_ptr) {
3020 p = macro_ptr, t = *p;
3021 if (ws_str) {
3022 while (is_space(t) || TOK_LINEFEED == t)
3023 tok_str_add(ws_str, t), t = *++p;
3025 if (t == 0 && can_read_stream) {
3026 end_macro();
3027 /* also, end of scope for nested defined symbol */
3028 sa = *nested_list;
3029 while (sa && sa->v == 0)
3030 sa = sa->prev;
3031 if (sa)
3032 sa->v = 0;
3033 continue;
3035 } else {
3036 ch = handle_eob();
3037 if (ws_str) {
3038 while (is_space(ch) || ch == '\n' || ch == '/') {
3039 if (ch == '/') {
3040 int c;
3041 uint8_t *p = file->buf_ptr;
3042 PEEKC(c, p);
3043 if (c == '*') {
3044 p = parse_comment(p);
3045 file->buf_ptr = p - 1;
3046 } else if (c == '/') {
3047 p = parse_line_comment(p);
3048 file->buf_ptr = p - 1;
3049 } else
3050 break;
3051 ch = ' ';
3053 if (!(ch == '\f' || ch == '\v' || ch == '\r'))
3054 tok_str_add(ws_str, ch);
3055 cinp();
3058 t = ch;
3061 if (ws_str)
3062 return t;
3063 next_nomacro_spc();
3064 return tok;
3068 /* do macro substitution of current token with macro 's' and add
3069 result to (tok_str,tok_len). 'nested_list' is the list of all
3070 macros we got inside to avoid recursing. Return non zero if no
3071 substitution needs to be done */
3072 static int macro_subst_tok(
3073 TokenString *tok_str,
3074 Sym **nested_list,
3075 Sym *s,
3076 int can_read_stream)
3078 Sym *args, *sa, *sa1;
3079 int parlevel, *mstr, t, t1, spc;
3080 TokenString str;
3081 char *cstrval;
3082 CValue cval;
3083 CString cstr;
3084 char buf[32];
3086 /* if symbol is a macro, prepare substitution */
3087 /* special macros */
3088 if (tok == TOK___LINE__) {
3089 snprintf(buf, sizeof(buf), "%d", file->line_num);
3090 cstrval = buf;
3091 t1 = TOK_PPNUM;
3092 goto add_cstr1;
3093 } else if (tok == TOK___FILE__) {
3094 cstrval = file->filename;
3095 goto add_cstr;
3096 } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
3097 time_t ti;
3098 struct tm *tm;
3100 time(&ti);
3101 tm = localtime(&ti);
3102 if (tok == TOK___DATE__) {
3103 snprintf(buf, sizeof(buf), "%s %2d %d",
3104 ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
3105 } else {
3106 snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
3107 tm->tm_hour, tm->tm_min, tm->tm_sec);
3109 cstrval = buf;
3110 add_cstr:
3111 t1 = TOK_STR;
3112 add_cstr1:
3113 cstr_new(&cstr);
3114 cstr_cat(&cstr, cstrval, 0);
3115 cval.str.size = cstr.size;
3116 cval.str.data = cstr.data;
3117 tok_str_add2(tok_str, t1, &cval);
3118 cstr_free(&cstr);
3119 } else {
3120 int saved_parse_flags = parse_flags;
3122 mstr = s->d;
3123 if (s->type.t == MACRO_FUNC) {
3124 /* whitespace between macro name and argument list */
3125 TokenString ws_str;
3126 tok_str_new(&ws_str);
3128 spc = 0;
3129 parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
3130 | PARSE_FLAG_ACCEPT_STRAYS;
3132 /* get next token from argument stream */
3133 t = next_argstream(nested_list, can_read_stream, &ws_str);
3134 if (t != '(') {
3135 /* not a macro substitution after all, restore the
3136 * macro token plus all whitespace we've read.
3137 * whitespace is intentionally not merged to preserve
3138 * newlines. */
3139 parse_flags = saved_parse_flags;
3140 tok_str_add(tok_str, tok);
3141 if (parse_flags & PARSE_FLAG_SPACES) {
3142 int i;
3143 for (i = 0; i < ws_str.len; i++)
3144 tok_str_add(tok_str, ws_str.str[i]);
3146 tok_str_free_str(ws_str.str);
3147 return 0;
3148 } else {
3149 tok_str_free_str(ws_str.str);
3151 next_nomacro(); /* eat '(' */
3153 /* argument macro */
3154 args = NULL;
3155 sa = s->next;
3156 /* NOTE: empty args are allowed, except if no args */
3157 for(;;) {
3158 do {
3159 next_argstream(nested_list, can_read_stream, NULL);
3160 } while (is_space(tok) || TOK_LINEFEED == tok);
3161 empty_arg:
3162 /* handle '()' case */
3163 if (!args && !sa && tok == ')')
3164 break;
3165 if (!sa)
3166 tcc_error("macro '%s' used with too many args",
3167 get_tok_str(s->v, 0));
3168 tok_str_new(&str);
3169 parlevel = spc = 0;
3170 /* NOTE: non zero sa->t indicates VA_ARGS */
3171 while ((parlevel > 0 ||
3172 (tok != ')' &&
3173 (tok != ',' || sa->type.t)))) {
3174 if (tok == TOK_EOF || tok == 0)
3175 break;
3176 if (tok == '(')
3177 parlevel++;
3178 else if (tok == ')')
3179 parlevel--;
3180 if (tok == TOK_LINEFEED)
3181 tok = ' ';
3182 if (!check_space(tok, &spc))
3183 tok_str_add2(&str, tok, &tokc);
3184 next_argstream(nested_list, can_read_stream, NULL);
3186 if (parlevel)
3187 expect(")");
3188 str.len -= spc;
3189 tok_str_add(&str, 0);
3190 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3191 sa1->d = str.str;
3192 sa = sa->next;
3193 if (tok == ')') {
3194 /* special case for gcc var args: add an empty
3195 var arg argument if it is omitted */
3196 if (sa && sa->type.t && gnu_ext)
3197 goto empty_arg;
3198 break;
3200 if (tok != ',')
3201 expect(",");
3203 if (sa) {
3204 tcc_error("macro '%s' used with too few args",
3205 get_tok_str(s->v, 0));
3208 parse_flags = saved_parse_flags;
3210 /* now subst each arg */
3211 mstr = macro_arg_subst(nested_list, mstr, args);
3212 /* free memory */
3213 sa = args;
3214 while (sa) {
3215 sa1 = sa->prev;
3216 tok_str_free_str(sa->d);
3217 sym_free(sa);
3218 sa = sa1;
3222 sym_push2(nested_list, s->v, 0, 0);
3223 parse_flags = saved_parse_flags;
3224 macro_subst(tok_str, nested_list, mstr, can_read_stream | 2);
3226 /* pop nested defined symbol */
3227 sa1 = *nested_list;
3228 *nested_list = sa1->prev;
3229 sym_free(sa1);
3230 if (mstr != s->d)
3231 tok_str_free_str(mstr);
3233 return 0;
3236 static int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3238 CString cstr;
3239 int n, ret = 1;
3241 cstr_new(&cstr);
3242 if (t1 != TOK_PLCHLDR)
3243 cstr_cat(&cstr, get_tok_str(t1, v1), -1);
3244 n = cstr.size;
3245 if (t2 != TOK_PLCHLDR)
3246 cstr_cat(&cstr, get_tok_str(t2, v2), -1);
3247 cstr_ccat(&cstr, '\0');
3249 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3250 memcpy(file->buffer, cstr.data, cstr.size);
3251 for (;;) {
3252 next_nomacro1();
3253 if (0 == *file->buf_ptr)
3254 break;
3255 if (is_space(tok))
3256 continue;
3257 tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid"
3258 " preprocessing token", n, cstr.data, (char*)cstr.data + n);
3259 ret = 0;
3260 break;
3262 tcc_close();
3263 //printf("paste <%s>\n", (char*)cstr.data);
3264 cstr_free(&cstr);
3265 return ret;
3268 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3269 return the resulting string (which must be freed). */
3270 static inline int *macro_twosharps(const int *ptr0)
3272 int t;
3273 CValue cval;
3274 TokenString macro_str1;
3275 int start_of_nosubsts = -1;
3276 const int *ptr;
3278 /* we search the first '##' */
3279 for (ptr = ptr0;;) {
3280 TOK_GET(&t, &ptr, &cval);
3281 if (t == TOK_TWOSHARPS)
3282 break;
3283 if (t == 0)
3284 return NULL;
3287 tok_str_new(&macro_str1);
3289 //tok_print(" $$$", ptr0);
3290 for (ptr = ptr0;;) {
3291 TOK_GET(&t, &ptr, &cval);
3292 if (t == 0)
3293 break;
3294 if (t == TOK_TWOSHARPS)
3295 continue;
3296 while (*ptr == TOK_TWOSHARPS) {
3297 int t1; CValue cv1;
3298 /* given 'a##b', remove nosubsts preceding 'a' */
3299 if (start_of_nosubsts >= 0)
3300 macro_str1.len = start_of_nosubsts;
3301 /* given 'a##b', remove nosubsts preceding 'b' */
3302 while ((t1 = *++ptr) == TOK_NOSUBST)
3304 if (t1 && t1 != TOK_TWOSHARPS) {
3305 TOK_GET(&t1, &ptr, &cv1);
3306 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3307 if (paste_tokens(t, &cval, t1, &cv1)) {
3308 t = tok, cval = tokc;
3309 } else {
3310 tok_str_add2(&macro_str1, t, &cval);
3311 t = t1, cval = cv1;
3316 if (t == TOK_NOSUBST) {
3317 if (start_of_nosubsts < 0)
3318 start_of_nosubsts = macro_str1.len;
3319 } else {
3320 start_of_nosubsts = -1;
3322 tok_str_add2(&macro_str1, t, &cval);
3324 tok_str_add(&macro_str1, 0);
3325 //tok_print(" ###", macro_str1.str);
3326 return macro_str1.str;
3329 /* do macro substitution of macro_str and add result to
3330 (tok_str,tok_len). 'nested_list' is the list of all macros we got
3331 inside to avoid recursing. */
3332 static void macro_subst(
3333 TokenString *tok_str,
3334 Sym **nested_list,
3335 const int *macro_str,
3336 int can_read_stream
3339 Sym *s;
3340 const int *ptr;
3341 int t, spc, nosubst;
3342 CValue cval;
3343 int *macro_str1 = NULL;
3345 /* first scan for '##' operator handling */
3346 ptr = macro_str;
3347 spc = nosubst = 0;
3349 /* first scan for '##' operator handling */
3350 if (can_read_stream & 1) {
3351 macro_str1 = macro_twosharps(ptr);
3352 if (macro_str1)
3353 ptr = macro_str1;
3356 while (1) {
3357 TOK_GET(&t, &ptr, &cval);
3358 if (t == 0)
3359 break;
3361 if (t >= TOK_IDENT && 0 == nosubst) {
3362 s = define_find(t);
3363 if (s == NULL)
3364 goto no_subst;
3366 /* if nested substitution, do nothing */
3367 if (sym_find2(*nested_list, t)) {
3368 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3369 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3370 goto no_subst;
3374 TokenString str;
3375 str.str = (int*)ptr;
3376 begin_macro(&str, 2);
3378 tok = t;
3379 macro_subst_tok(tok_str, nested_list, s, can_read_stream);
3381 if (str.alloc == 3) {
3382 /* already finished by reading function macro arguments */
3383 break;
3386 ptr = macro_ptr;
3387 end_macro ();
3390 spc = (tok_str->len &&
3391 is_space(tok_last(tok_str->str,
3392 tok_str->str + tok_str->len)));
3394 } else {
3396 if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3397 tcc_error("stray '\\' in program");
3399 no_subst:
3400 if (!check_space(t, &spc))
3401 tok_str_add2(tok_str, t, &cval);
3402 nosubst = 0;
3403 if (t == TOK_NOSUBST)
3404 nosubst = 1;
3407 if (macro_str1)
3408 tok_str_free_str(macro_str1);
3412 /* return next token with macro substitution */
3413 ST_FUNC void next(void)
3415 redo:
3416 if (parse_flags & PARSE_FLAG_SPACES)
3417 next_nomacro_spc();
3418 else
3419 next_nomacro();
3421 if (macro_ptr) {
3422 if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3423 /* discard preprocessor markers */
3424 goto redo;
3425 } else if (tok == 0) {
3426 /* end of macro or unget token string */
3427 end_macro();
3428 goto redo;
3430 } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3431 Sym *s;
3432 /* if reading from file, try to substitute macros */
3433 s = define_find(tok);
3434 if (s) {
3435 Sym *nested_list = NULL;
3436 tokstr_buf.len = 0;
3437 macro_subst_tok(&tokstr_buf, &nested_list, s, 1);
3438 tok_str_add(&tokstr_buf, 0);
3439 begin_macro(&tokstr_buf, 2);
3440 goto redo;
3443 /* convert preprocessor tokens into C tokens */
3444 if (tok == TOK_PPNUM) {
3445 if (parse_flags & PARSE_FLAG_TOK_NUM)
3446 parse_number((char *)tokc.str.data);
3447 } else if (tok == TOK_PPSTR) {
3448 if (parse_flags & PARSE_FLAG_TOK_STR)
3449 parse_string((char *)tokc.str.data, tokc.str.size - 1);
3453 /* push back current token and set current token to 'last_tok'. Only
3454 identifier case handled for labels. */
3455 ST_INLN void unget_tok(int last_tok)
3458 TokenString *str = tok_str_alloc();
3459 tok_str_add2(str, tok, &tokc);
3460 tok_str_add(str, 0);
3461 begin_macro(str, 1);
3462 tok = last_tok;
3465 ST_FUNC void preprocess_start(TCCState *s1)
3467 s1->include_stack_ptr = s1->include_stack;
3468 /* XXX: move that before to avoid having to initialize
3469 file->ifdef_stack_ptr ? */
3470 s1->ifdef_stack_ptr = s1->ifdef_stack;
3471 file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3472 pp_once++;
3474 pvtop = vtop = vstack - 1;
3475 s1->pack_stack[0] = 0;
3476 s1->pack_stack_ptr = s1->pack_stack;
3478 isidnum_table['$' - CH_EOF] =
3479 s1->dollars_in_identifiers ? IS_ID : 0;
3480 isidnum_table['.' - CH_EOF] =
3481 (parse_flags & PARSE_FLAG_ASM_FILE) ? IS_ID : 0;
3484 ST_FUNC void tccpp_new(TCCState *s)
3486 int i, c;
3487 const char *p, *r;
3489 /* might be used in error() before preprocess_start() */
3490 s->include_stack_ptr = s->include_stack;
3492 /* init isid table */
3493 for(i = CH_EOF; i<128; i++)
3494 isidnum_table[i - CH_EOF]
3495 = is_space(i) ? IS_SPC
3496 : isid(i) ? IS_ID
3497 : isnum(i) ? IS_NUM
3498 : 0;
3500 for(i = 128; i<256; i++)
3501 isidnum_table[i - CH_EOF] = IS_ID;
3503 /* init allocators */
3504 tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE);
3505 tal_new(&tokstr_alloc, TOKSTR_TAL_LIMIT, TOKSTR_TAL_SIZE);
3506 tal_new(&cstr_alloc, CSTR_TAL_LIMIT, CSTR_TAL_SIZE);
3508 memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3509 cstr_new(&cstr_buf);
3510 cstr_realloc(&cstr_buf, STRING_MAX_SIZE);
3511 tok_str_new(&tokstr_buf);
3512 tok_str_realloc(&tokstr_buf, TOKSTR_MAX_SIZE);
3514 tok_ident = TOK_IDENT;
3515 p = tcc_keywords;
3516 while (*p) {
3517 r = p;
3518 for(;;) {
3519 c = *r++;
3520 if (c == '\0')
3521 break;
3523 tok_alloc(p, r - p - 1);
3524 p = r;
3528 ST_FUNC void tccpp_delete(TCCState *s)
3530 int i, n;
3532 /* free -D and compiler defines */
3533 free_defines(NULL);
3535 /* cleanup from error/setjmp */
3536 while (macro_stack)
3537 end_macro();
3538 macro_ptr = NULL;
3540 /* free tokens */
3541 n = tok_ident - TOK_IDENT;
3542 for(i = 0; i < n; i++)
3543 tal_free(toksym_alloc, table_ident[i]);
3544 tcc_free(table_ident);
3545 table_ident = NULL;
3547 /* free static buffers */
3548 cstr_free(&tokcstr);
3549 cstr_free(&cstr_buf);
3550 tok_str_free_str(tokstr_buf.str);
3552 /* free allocators */
3553 tal_delete(toksym_alloc);
3554 toksym_alloc = NULL;
3555 tal_delete(tokstr_alloc);
3556 tokstr_alloc = NULL;
3557 tal_delete(cstr_alloc);
3558 cstr_alloc = NULL;
3561 /* ------------------------------------------------------------------------- */
3562 /* tcc -E [-P[1]] [-dD} support */
3564 static void tok_print(const char *msg, const int *str)
3566 FILE *fp;
3567 int t;
3568 CValue cval;
3570 fp = tcc_state->ppfp;
3571 if (!fp || !tcc_state->dflag)
3572 fp = stdout;
3574 fprintf(fp, "%s ", msg);
3575 while (str) {
3576 TOK_GET(&t, &str, &cval);
3577 if (!t)
3578 break;
3579 fprintf(fp,"%s", get_tok_str(t, &cval));
3581 fprintf(fp, "\n");
3584 static void pp_line(TCCState *s1, BufferedFile *f, int level)
3586 int d = f->line_num - f->line_ref;
3588 if (s1->dflag & 4)
3589 return;
3591 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE) {
3593 } else if (level == 0 && f->line_ref && d < 8) {
3594 while (d > 0)
3595 fputs("\n", s1->ppfp), --d;
3596 } else if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
3597 fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
3598 } else {
3599 fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
3600 level > 0 ? " 1" : level < 0 ? " 2" : "");
3602 f->line_ref = f->line_num;
3605 static void define_print(TCCState *s1, int v)
3607 FILE *fp;
3608 Sym *s;
3610 s = define_find(v);
3611 if (NULL == s || NULL == s->d)
3612 return;
3614 fp = s1->ppfp;
3615 fprintf(fp, "#define %s", get_tok_str(v, NULL));
3616 if (s->type.t == MACRO_FUNC) {
3617 Sym *a = s->next;
3618 fprintf(fp,"(");
3619 if (a)
3620 for (;;) {
3621 fprintf(fp,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
3622 if (!(a = a->next))
3623 break;
3624 fprintf(fp,",");
3626 fprintf(fp,")");
3628 tok_print("", s->d);
3631 static void pp_debug_defines(TCCState *s1)
3633 int v, t;
3634 const char *vs;
3635 FILE *fp;
3637 t = pp_debug_tok;
3638 if (t == 0)
3639 return;
3641 file->line_num--;
3642 pp_line(s1, file, 0);
3643 file->line_ref = ++file->line_num;
3645 fp = s1->ppfp;
3646 v = pp_debug_symv;
3647 vs = get_tok_str(v, NULL);
3648 if (t == TOK_DEFINE) {
3649 define_print(s1, v);
3650 } else if (t == TOK_UNDEF) {
3651 fprintf(fp, "#undef %s\n", vs);
3652 } else if (t == TOK_push_macro) {
3653 fprintf(fp, "#pragma push_macro(\"%s\")\n", vs);
3654 } else if (t == TOK_pop_macro) {
3655 fprintf(fp, "#pragma pop_macro(\"%s\")\n", vs);
3657 pp_debug_tok = 0;
3660 static void pp_debug_builtins(TCCState *s1)
3662 int v;
3663 for (v = TOK_IDENT; v < tok_ident; ++v)
3664 define_print(s1, v);
3667 /* Add a space between tokens a and b to avoid unwanted textual pasting */
3668 static int pp_need_space(int a, int b)
3670 return 'E' == a ? '+' == b || '-' == b
3671 : '+' == a ? TOK_INC == b || '+' == b
3672 : '-' == a ? TOK_DEC == b || '-' == b
3673 : a >= TOK_IDENT ? b >= TOK_IDENT
3674 : 0;
3677 /* maybe hex like 0x1e */
3678 static int pp_check_he0xE(int t, const char *p)
3680 if (t == TOK_PPNUM && toup(strchr(p, 0)[-1]) == 'E')
3681 return 'E';
3682 return t;
3685 /* Preprocess the current file */
3686 ST_FUNC int tcc_preprocess(TCCState *s1)
3688 BufferedFile **iptr;
3689 int token_seen, spcs, level;
3690 const char *p;
3691 Sym *define_start;
3693 preprocess_start(s1);
3694 ch = file->buf_ptr[0];
3695 tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3696 parse_flags = PARSE_FLAG_PREPROCESS
3697 | (parse_flags & PARSE_FLAG_ASM_FILE)
3698 | PARSE_FLAG_LINEFEED
3699 | PARSE_FLAG_SPACES
3700 | PARSE_FLAG_ACCEPT_STRAYS
3702 define_start = define_stack;
3704 /* Credits to Fabrice Bellard's initial revision to demonstrate its
3705 capability to compile and run itself, provided all numbers are
3706 given as decimals. tcc -E -P10 will do. */
3707 if (s1->Pflag == 1 + 10)
3708 parse_flags |= PARSE_FLAG_TOK_NUM, s1->Pflag = 1;
3710 #ifdef PP_BENCH
3711 /* for PP benchmarks */
3712 do next(); while (tok != TOK_EOF); return 0;
3713 #endif
3715 if (s1->dflag & 1) {
3716 pp_debug_builtins(s1);
3717 s1->dflag &= ~1;
3720 token_seen = TOK_LINEFEED, spcs = 0;
3721 pp_line(s1, file, 0);
3723 for (;;) {
3724 iptr = s1->include_stack_ptr;
3725 next();
3726 if (tok == TOK_EOF)
3727 break;
3728 level = s1->include_stack_ptr - iptr;
3729 if (level) {
3730 if (level > 0)
3731 pp_line(s1, *iptr, 0);
3732 pp_line(s1, file, level);
3735 if (s1->dflag) {
3736 pp_debug_defines(s1);
3737 if (s1->dflag & 4)
3738 continue;
3741 if (token_seen == TOK_LINEFEED) {
3742 if (tok == ' ') {
3743 ++spcs;
3744 continue;
3746 if (tok == TOK_LINEFEED) {
3747 spcs = 0;
3748 continue;
3750 pp_line(s1, file, 0);
3751 } else if (tok == TOK_LINEFEED) {
3752 ++file->line_ref;
3753 } else {
3754 spcs = pp_need_space(token_seen, tok);
3757 while (spcs)
3758 fputs(" ", s1->ppfp), --spcs;
3759 fputs(p = get_tok_str(tok, &tokc), s1->ppfp);
3760 token_seen = pp_check_he0xE(tok, p);;
3762 /* reset define stack, but keep -D and built-ins */
3763 free_defines(define_start);
3764 return 0;
3767 /* ------------------------------------------------------------------------- */