?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Fix Windows++ compilation of previous (YX Hao, Joel Bodenmann)
[tinycc.git] / tccpp.c
blob3d11dc3d43eded48fbd8c2fde58ad7d00672ec7e
1 /*
2 * TCC - Tiny C Compiler
3 *
4 * Copyright (c) 2001-2004 Fabrice Bellard
5 *
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.
10 *
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.
15 *
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
19 */
20
21 #include "tcc.h"
22
23 /********************************************************/
24 /* global variables */
25
26 ST_DATA int tok_flags;
27 ST_DATA int parse_flags;
28
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 */
34
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;
40
41 /* ------------------------------------------------------------------------- */
42
43 static TokenSym *hash_ident[TOK_HASH_SIZE];
44 static char token_buf[STRING_MAX_SIZE + 1];
45 static CString cstr_buf;
46 static CString macro_equal_buf;
47 static TokenString tokstr_buf;
48 static unsigned char isidnum_table[256 - CH_EOF];
49 static int pp_debug_tok, pp_debug_symv;
50 static int pp_once;
51 static int pp_expr;
52 static int pp_counter;
53 static void tok_print(const char *msg, const int *str);
54
55 static struct TinyAlloc *toksym_alloc;
56 static struct TinyAlloc *tokstr_alloc;
57 static struct TinyAlloc *cstr_alloc;
58
59 static TokenString *macro_stack;
60
61 static const char tcc_keywords[] =
62 #define DEF(id, str) str "\0"
63 #include "tcctok.h"
64 #undef DEF
65 ;
66
67 /* WARNING: the content of this string encodes token numbers */
68 static const unsigned char tok_two_chars[] =
69 /* outdated -- gr
70 "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
71 "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
72 */{
73 '<','=', TOK_LE,
74 '>','=', TOK_GE,
75 '!','=', TOK_NE,
76 '&','&', TOK_LAND,
77 '|','|', TOK_LOR,
78 '+','+', TOK_INC,
79 '-','-', TOK_DEC,
80 '=','=', TOK_EQ,
81 '<','<', TOK_SHL,
82 '>','>', TOK_SAR,
83 '+','=', TOK_A_ADD,
84 '-','=', TOK_A_SUB,
85 '*','=', TOK_A_MUL,
86 '/','=', TOK_A_DIV,
87 '%','=', TOK_A_MOD,
88 '&','=', TOK_A_AND,
89 '^','=', TOK_A_XOR,
90 '|','=', TOK_A_OR,
91 '-','>', TOK_ARROW,
92 '.','.', TOK_TWODOTS,
93 '#','#', TOK_TWOSHARPS,
94 0
95 };
96
97 static void next_nomacro_spc(void);
98
99 ST_FUNC void skip(int c)
100 {
101 if (tok != c)
102 tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
103 next();
104 }
105
106 ST_FUNC void expect(const char *msg)
107 {
108 tcc_error("%s expected", msg);
109 }
110
111 /* ------------------------------------------------------------------------- */
112 /* Custom allocator for tiny objects */
113
114 #define USE_TAL
115
116 #ifndef USE_TAL
117 #define tal_free(al, p) tcc_free(p)
118 #define tal_realloc(al, p, size) tcc_realloc(p, size)
119 #define tal_new(a,b,c)
120 #define tal_delete(a)
121 #else
122 #if !defined(MEM_DEBUG)
123 #define tal_free(al, p) tal_free_impl(al, p)
124 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size)
125 #define TAL_DEBUG_PARAMS
126 #else
127 #define TAL_DEBUG 1
128 //#define TAL_INFO 1 /* collect and dump allocators stats */
129 #define tal_free(al, p) tal_free_impl(al, p, __FILE__, __LINE__)
130 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size, __FILE__, __LINE__)
131 #define TAL_DEBUG_PARAMS , const char *file, int line
132 #define TAL_DEBUG_FILE_LEN 40
133 #endif
134
135 #define TOKSYM_TAL_SIZE (768 * 1024) /* allocator for tiny TokenSym in table_ident */
136 #define TOKSTR_TAL_SIZE (768 * 1024) /* allocator for tiny TokenString instances */
137 #define CSTR_TAL_SIZE (256 * 1024) /* allocator for tiny CString instances */
138 #define TOKSYM_TAL_LIMIT 256 /* prefer unique limits to distinguish allocators debug msgs */
139 #define TOKSTR_TAL_LIMIT 128 /* 32 * sizeof(int) */
140 #define CSTR_TAL_LIMIT 1024
141
142 typedef struct TinyAlloc {
143 unsigned limit;
144 unsigned size;
145 uint8_t *buffer;
146 uint8_t *p;
147 unsigned nb_allocs;
148 struct TinyAlloc *next, *top;
149 #ifdef TAL_INFO
150 unsigned nb_peak;
151 unsigned nb_total;
152 unsigned nb_missed;
153 uint8_t *peak_p;
154 #endif
155 } TinyAlloc;
156
157 typedef struct tal_header_t {
158 unsigned size;
159 #ifdef TAL_DEBUG
160 int line_num; /* negative line_num used for double free check */
161 char file_name[TAL_DEBUG_FILE_LEN + 1];
162 #endif
163 } tal_header_t;
164
165 /* ------------------------------------------------------------------------- */
166
167 static TinyAlloc *tal_new(TinyAlloc **pal, unsigned limit, unsigned size)
168 {
169 TinyAlloc *al = tcc_mallocz(sizeof(TinyAlloc));
170 al->p = al->buffer = tcc_malloc(size);
171 al->limit = limit;
172 al->size = size;
173 if (pal) *pal = al;
174 return al;
175 }
176
177 static void tal_delete(TinyAlloc *al)
178 {
179 TinyAlloc *next;
180
181 tail_call:
182 if (!al)
183 return;
184 #ifdef TAL_INFO
185 fprintf(stderr, "limit=%5d, size=%5g MB, nb_peak=%6d, nb_total=%8d, nb_missed=%6d, usage=%5.1f%%\n",
186 al->limit, al->size / 1024.0 / 1024.0, al->nb_peak, al->nb_total, al->nb_missed,
187 (al->peak_p - al->buffer) * 100.0 / al->size);
188 #endif
189 #ifdef TAL_DEBUG
190 if (al->nb_allocs > 0) {
191 uint8_t *p;
192 fprintf(stderr, "TAL_DEBUG: memory leak %d chunk(s) (limit= %d)\n",
193 al->nb_allocs, al->limit);
194 p = al->buffer;
195 while (p < al->p) {
196 tal_header_t *header = (tal_header_t *)p;
197 if (header->line_num > 0) {
198 fprintf(stderr, "%s:%d: chunk of %d bytes leaked\n",
199 header->file_name, header->line_num, header->size);
200 }
201 p += header->size + sizeof(tal_header_t);
202 }
203 #if MEM_DEBUG-0 == 2
204 exit(2);
205 #endif
206 }
207 #endif
208 next = al->next;
209 tcc_free(al->buffer);
210 tcc_free(al);
211 al = next;
212 goto tail_call;
213 }
214
215 static void tal_free_impl(TinyAlloc *al, void *p TAL_DEBUG_PARAMS)
216 {
217 if (!p)
218 return;
219 tail_call:
220 if (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size) {
221 #ifdef TAL_DEBUG
222 tal_header_t *header = (((tal_header_t *)p) - 1);
223 if (header->line_num < 0) {
224 fprintf(stderr, "%s:%d: TAL_DEBUG: double frees chunk from\n",
225 file, line);
226 fprintf(stderr, "%s:%d: %d bytes\n",
227 header->file_name, (int)-header->line_num, (int)header->size);
228 } else
229 header->line_num = -header->line_num;
230 #endif
231 al->nb_allocs--;
232 if (!al->nb_allocs)
233 al->p = al->buffer;
234 } else if (al->next) {
235 al = al->next;
236 goto tail_call;
237 }
238 else
239 tcc_free(p);
240 }
241
242 static void *tal_realloc_impl(TinyAlloc **pal, void *p, unsigned size TAL_DEBUG_PARAMS)
243 {
244 tal_header_t *header;
245 void *ret;
246 int is_own;
247 unsigned adj_size = (size + 3) & -4;
248 TinyAlloc *al = *pal;
249
250 tail_call:
251 is_own = (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size);
252 if ((!p || is_own) && size <= al->limit) {
253 if (al->p + adj_size + sizeof(tal_header_t) < al->buffer + al->size) {
254 header = (tal_header_t *)al->p;
255 header->size = adj_size;
256 #ifdef TAL_DEBUG
257 { int ofs = strlen(file) - TAL_DEBUG_FILE_LEN;
258 strncpy(header->file_name, file + (ofs > 0 ? ofs : 0), TAL_DEBUG_FILE_LEN);
259 header->file_name[TAL_DEBUG_FILE_LEN] = 0;
260 header->line_num = line; }
261 #endif
262 ret = al->p + sizeof(tal_header_t);
263 al->p += adj_size + sizeof(tal_header_t);
264 if (is_own) {
265 header = (((tal_header_t *)p) - 1);
266 memcpy(ret, p, header->size);
267 #ifdef TAL_DEBUG
268 header->line_num = -header->line_num;
269 #endif
270 } else {
271 al->nb_allocs++;
272 }
273 #ifdef TAL_INFO
274 if (al->nb_peak < al->nb_allocs)
275 al->nb_peak = al->nb_allocs;
276 if (al->peak_p < al->p)
277 al->peak_p = al->p;
278 al->nb_total++;
279 #endif
280 return ret;
281 } else if (is_own) {
282 al->nb_allocs--;
283 ret = tal_realloc(*pal, 0, size);
284 header = (((tal_header_t *)p) - 1);
285 memcpy(ret, p, header->size);
286 #ifdef TAL_DEBUG
287 header->line_num = -header->line_num;
288 #endif
289 return ret;
290 }
291 if (al->next) {
292 al = al->next;
293 } else {
294 TinyAlloc *bottom = al, *next = al->top ? al->top : al;
295
296 al = tal_new(pal, next->limit, next->size * 2);
297 al->next = next;
298 bottom->top = al;
299 }
300 goto tail_call;
301 }
302 if (is_own) {
303 al->nb_allocs--;
304 ret = tcc_malloc(size);
305 header = (((tal_header_t *)p) - 1);
306 memcpy(ret, p, header->size);
307 #ifdef TAL_DEBUG
308 header->line_num = -header->line_num;
309 #endif
310 } else if (al->next) {
311 al = al->next;
312 goto tail_call;
313 } else
314 ret = tcc_realloc(p, size);
315 #ifdef TAL_INFO
316 al->nb_missed++;
317 #endif
318 return ret;
319 }
320
321 #endif /* USE_TAL */
322
323 /* ------------------------------------------------------------------------- */
324 /* CString handling */
325 static void cstr_realloc(CString *cstr, int new_size)
326 {
327 int size;
328
329 size = cstr->size_allocated;
330 if (size < 8)
331 size = 8; /* no need to allocate a too small first string */
332 while (size < new_size)
333 size = size * 2;
334 cstr->data = tal_realloc(cstr_alloc, cstr->data, size);
335 cstr->size_allocated = size;
336 }
337
338 /* add a byte */
339 ST_INLN void cstr_ccat(CString *cstr, int ch)
340 {
341 int size;
342 size = cstr->size + 1;
343 if (size > cstr->size_allocated)
344 cstr_realloc(cstr, size);
345 ((unsigned char *)cstr->data)[size - 1] = ch;
346 cstr->size = size;
347 }
348
349 ST_FUNC void cstr_cat(CString *cstr, const char *str, int len)
350 {
351 int size;
352 if (len <= 0)
353 len = strlen(str) + 1 + len;
354 size = cstr->size + len;
355 if (size > cstr->size_allocated)
356 cstr_realloc(cstr, size);
357 memmove(((unsigned char *)cstr->data) + cstr->size, str, len);
358 cstr->size = size;
359 }
360
361 /* add a wide char */
362 ST_FUNC void cstr_wccat(CString *cstr, int ch)
363 {
364 int size;
365 size = cstr->size + sizeof(nwchar_t);
366 if (size > cstr->size_allocated)
367 cstr_realloc(cstr, size);
368 *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
369 cstr->size = size;
370 }
371
372 ST_FUNC void cstr_new(CString *cstr)
373 {
374 memset(cstr, 0, sizeof(CString));
375 }
376
377 /* free string and reset it to NULL */
378 ST_FUNC void cstr_free(CString *cstr)
379 {
380 tal_free(cstr_alloc, cstr->data);
381 cstr_new(cstr);
382 }
383
384 /* reset string to empty */
385 ST_FUNC void cstr_reset(CString *cstr)
386 {
387 cstr->size = 0;
388 }
389
390 /* XXX: unicode ? */
391 static void add_char(CString *cstr, int c)
392 {
393 if (c == '\'' || c == '\"' || c == '\\') {
394 /* XXX: could be more precise if char or string */
395 cstr_ccat(cstr, '\\');
396 }
397 if (c >= 32 && c <= 126) {
398 cstr_ccat(cstr, c);
399 } else {
400 cstr_ccat(cstr, '\\');
401 if (c == '\n') {
402 cstr_ccat(cstr, 'n');
403 } else {
404 cstr_ccat(cstr, '0' + ((c >> 6) & 7));
405 cstr_ccat(cstr, '0' + ((c >> 3) & 7));
406 cstr_ccat(cstr, '0' + (c & 7));
407 }
408 }
409 }
410
411 /* ------------------------------------------------------------------------- */
412 /* allocate a new token */
413 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
414 {
415 TokenSym *ts, **ptable;
416 int i;
417
418 if (tok_ident >= SYM_FIRST_ANOM)
419 tcc_error("memory full (symbols)");
420
421 /* expand token table if needed */
422 i = tok_ident - TOK_IDENT;
423 if ((i % TOK_ALLOC_INCR) == 0) {
424 ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
425 table_ident = ptable;
426 }
427
428 ts = tal_realloc(toksym_alloc, 0, sizeof(TokenSym) + len);
429 table_ident[i] = ts;
430 ts->tok = tok_ident++;
431 ts->sym_define = NULL;
432 ts->sym_label = NULL;
433 ts->sym_struct = NULL;
434 ts->sym_identifier = NULL;
435 ts->len = len;
436 ts->hash_next = NULL;
437 memcpy(ts->str, str, len);
438 ts->str[len] = '\0';
439 *pts = ts;
440 return ts;
441 }
442
443 #define TOK_HASH_INIT 1
444 #define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c))
445
446
447 /* find a token and add it if not found */
448 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
449 {
450 TokenSym *ts, **pts;
451 int i;
452 unsigned int h;
453
454 h = TOK_HASH_INIT;
455 for(i=0;i<len;i++)
456 h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
457 h &= (TOK_HASH_SIZE - 1);
458
459 pts = &hash_ident[h];
460 for(;;) {
461 ts = *pts;
462 if (!ts)
463 break;
464 if (ts->len == len && !memcmp(ts->str, str, len))
465 return ts;
466 pts = &(ts->hash_next);
467 }
468 return tok_alloc_new(pts, str, len);
469 }
470
471 /* XXX: buffer overflow */
472 /* XXX: float tokens */
473 ST_FUNC const char *get_tok_str(int v, CValue *cv)
474 {
475 char *p;
476 int i, len;
477
478 cstr_reset(&cstr_buf);
479 p = cstr_buf.data;
480
481 switch(v) {
482 case TOK_CINT:
483 case TOK_CUINT:
484 case TOK_CLONG:
485 case TOK_CULONG:
486 case TOK_CLLONG:
487 case TOK_CULLONG:
488 /* XXX: not quite exact, but only useful for testing */
489 #ifdef _WIN32
490 sprintf(p, "%u", (unsigned)cv->i);
491 #else
492 sprintf(p, "%llu", (unsigned long long)cv->i);
493 #endif
494 break;
495 case TOK_LCHAR:
496 cstr_ccat(&cstr_buf, 'L');
497 case TOK_CCHAR:
498 cstr_ccat(&cstr_buf, '\'');
499 add_char(&cstr_buf, cv->i);
500 cstr_ccat(&cstr_buf, '\'');
501 cstr_ccat(&cstr_buf, '\0');
502 break;
503 case TOK_PPNUM:
504 case TOK_PPSTR:
505 return (char*)cv->str.data;
506 case TOK_LSTR:
507 cstr_ccat(&cstr_buf, 'L');
508 case TOK_STR:
509 cstr_ccat(&cstr_buf, '\"');
510 if (v == TOK_STR) {
511 len = cv->str.size - 1;
512 for(i=0;i<len;i++)
513 add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]);
514 } else {
515 len = (cv->str.size / sizeof(nwchar_t)) - 1;
516 for(i=0;i<len;i++)
517 add_char(&cstr_buf, ((nwchar_t *)cv->str.data)[i]);
518 }
519 cstr_ccat(&cstr_buf, '\"');
520 cstr_ccat(&cstr_buf, '\0');
521 break;
522
523 case TOK_CFLOAT:
524 cstr_cat(&cstr_buf, "<float>", 0);
525 break;
526 case TOK_CDOUBLE:
527 cstr_cat(&cstr_buf, "<double>", 0);
528 break;
529 case TOK_CLDOUBLE:
530 cstr_cat(&cstr_buf, "<long double>", 0);
531 break;
532 case TOK_LINENUM:
533 cstr_cat(&cstr_buf, "<linenumber>", 0);
534 break;
535
536 /* above tokens have value, the ones below don't */
537 case TOK_LT:
538 v = '<';
539 goto addv;
540 case TOK_GT:
541 v = '>';
542 goto addv;
543 case TOK_DOTS:
544 return strcpy(p, "...");
545 case TOK_A_SHL:
546 return strcpy(p, "<<=");
547 case TOK_A_SAR:
548 return strcpy(p, ">>=");
549 case TOK_EOF:
550 return strcpy(p, "<eof>");
551 default:
552 if (v < TOK_IDENT) {
553 /* search in two bytes table */
554 const unsigned char *q = tok_two_chars;
555 while (*q) {
556 if (q[2] == v) {
557 *p++ = q[0];
558 *p++ = q[1];
559 *p = '\0';
560 return cstr_buf.data;
561 }
562 q += 3;
563 }
564 if (v >= 127) {
565 sprintf(cstr_buf.data, "<%02x>", v);
566 return cstr_buf.data;
567 }
568 addv:
569 *p++ = v;
570 *p = '\0';
571 } else if (v < tok_ident) {
572 return table_ident[v - TOK_IDENT]->str;
573 } else if (v >= SYM_FIRST_ANOM) {
574 /* special name for anonymous symbol */
575 sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
576 } else {
577 /* should never happen */
578 return NULL;
579 }
580 break;
581 }
582 return cstr_buf.data;
583 }
584
585 /* return the current character, handling end of block if necessary
586 (but not stray) */
587 ST_FUNC int handle_eob(void)
588 {
589 BufferedFile *bf = file;
590 int len;
591
592 /* only tries to read if really end of buffer */
593 if (bf->buf_ptr >= bf->buf_end) {
594 if (bf->fd != -1) {
595 #if defined(PARSE_DEBUG)
596 len = 1;
597 #else
598 len = IO_BUF_SIZE;
599 #endif
600 len = read(bf->fd, bf->buffer, len);
601 if (len < 0)
602 len = 0;
603 } else {
604 len = 0;
605 }
606 total_bytes += len;
607 bf->buf_ptr = bf->buffer;
608 bf->buf_end = bf->buffer + len;
609 *bf->buf_end = CH_EOB;
610 }
611 if (bf->buf_ptr < bf->buf_end) {
612 return bf->buf_ptr[0];
613 } else {
614 bf->buf_ptr = bf->buf_end;
615 return CH_EOF;
616 }
617 }
618
619 /* read next char from current input file and handle end of input buffer */
620 ST_INLN void inp(void)
621 {
622 ch = *(++(file->buf_ptr));
623 /* end of buffer/file handling */
624 if (ch == CH_EOB)
625 ch = handle_eob();
626 }
627
628 /* handle '\[\r]\n' */
629 static int handle_stray_noerror(void)
630 {
631 while (ch == '\\') {
632 inp();
633 if (ch == '\n') {
634 file->line_num++;
635 inp();
636 } else if (ch == '\r') {
637 inp();
638 if (ch != '\n')
639 goto fail;
640 file->line_num++;
641 inp();
642 } else {
643 fail:
644 return 1;
645 }
646 }
647 return 0;
648 }
649
650 static void handle_stray(void)
651 {
652 if (handle_stray_noerror())
653 tcc_error("stray '\\' in program");
654 }
655
656 /* skip the stray and handle the \\n case. Output an error if
657 incorrect char after the stray */
658 static int handle_stray1(uint8_t *p)
659 {
660 int c;
661
662 file->buf_ptr = p;
663 if (p >= file->buf_end) {
664 c = handle_eob();
665 if (c != '\\')
666 return c;
667 p = file->buf_ptr;
668 }
669 ch = *p;
670 if (handle_stray_noerror()) {
671 if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
672 tcc_error("stray '\\' in program");
673 *--file->buf_ptr = '\\';
674 }
675 p = file->buf_ptr;
676 c = *p;
677 return c;
678 }
679
680 /* handle just the EOB case, but not stray */
681 #define PEEKC_EOB(c, p)\
682 {\
683 p++;\
684 c = *p;\
685 if (c == '\\') {\
686 file->buf_ptr = p;\
687 c = handle_eob();\
688 p = file->buf_ptr;\
689 }\
690 }
691
692 /* handle the complicated stray case */
693 #define PEEKC(c, p)\
694 {\
695 p++;\
696 c = *p;\
697 if (c == '\\') {\
698 c = handle_stray1(p);\
699 p = file->buf_ptr;\
700 }\
701 }
702
703 /* input with '\[\r]\n' handling. Note that this function cannot
704 handle other characters after '\', so you cannot call it inside
705 strings or comments */
706 ST_FUNC void minp(void)
707 {
708 inp();
709 if (ch == '\\')
710 handle_stray();
711 }
712
713 /* single line C++ comments */
714 static uint8_t *parse_line_comment(uint8_t *p)
715 {
716 int c;
717
718 p++;
719 for(;;) {
720 c = *p;
721 redo:
722 if (c == '\n' || c == CH_EOF) {
723 break;
724 } else if (c == '\\') {
725 file->buf_ptr = p;
726 c = handle_eob();
727 p = file->buf_ptr;
728 if (c == '\\') {
729 PEEKC_EOB(c, p);
730 if (c == '\n') {
731 file->line_num++;
732 PEEKC_EOB(c, p);
733 } else if (c == '\r') {
734 PEEKC_EOB(c, p);
735 if (c == '\n') {
736 file->line_num++;
737 PEEKC_EOB(c, p);
738 }
739 }
740 } else {
741 goto redo;
742 }
743 } else {
744 p++;
745 }
746 }
747 return p;
748 }
749
750 /* C comments */
751 ST_FUNC uint8_t *parse_comment(uint8_t *p)
752 {
753 int c;
754
755 p++;
756 for(;;) {
757 /* fast skip loop */
758 for(;;) {
759 c = *p;
760 if (c == '\n' || c == '*' || c == '\\')
761 break;
762 p++;
763 c = *p;
764 if (c == '\n' || c == '*' || c == '\\')
765 break;
766 p++;
767 }
768 /* now we can handle all the cases */
769 if (c == '\n') {
770 file->line_num++;
771 p++;
772 } else if (c == '*') {
773 p++;
774 for(;;) {
775 c = *p;
776 if (c == '*') {
777 p++;
778 } else if (c == '/') {
779 goto end_of_comment;
780 } else if (c == '\\') {
781 file->buf_ptr = p;
782 c = handle_eob();
783 p = file->buf_ptr;
784 if (c == CH_EOF)
785 tcc_error("unexpected end of file in comment");
786 if (c == '\\') {
787 /* skip '\[\r]\n', otherwise just skip the stray */
788 while (c == '\\') {
789 PEEKC_EOB(c, p);
790 if (c == '\n') {
791 file->line_num++;
792 PEEKC_EOB(c, p);
793 } else if (c == '\r') {
794 PEEKC_EOB(c, p);
795 if (c == '\n') {
796 file->line_num++;
797 PEEKC_EOB(c, p);
798 }
799 } else {
800 goto after_star;
801 }
802 }
803 }
804 } else {
805 break;
806 }
807 }
808 after_star: ;
809 } else {
810 /* stray, eob or eof */
811 file->buf_ptr = p;
812 c = handle_eob();
813 p = file->buf_ptr;
814 if (c == CH_EOF) {
815 tcc_error("unexpected end of file in comment");
816 } else if (c == '\\') {
817 p++;
818 }
819 }
820 }
821 end_of_comment:
822 p++;
823 return p;
824 }
825
826 ST_FUNC int set_idnum(int c, int val)
827 {
828 int prev = isidnum_table[c - CH_EOF];
829 isidnum_table[c - CH_EOF] = val;
830 return prev;
831 }
832
833 #define cinp minp
834
835 static inline void skip_spaces(void)
836 {
837 while (isidnum_table[ch - CH_EOF] & IS_SPC)
838 cinp();
839 }
840
841 static inline int check_space(int t, int *spc)
842 {
843 if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
844 if (*spc)
845 return 1;
846 *spc = 1;
847 } else
848 *spc = 0;
849 return 0;
850 }
851
852 /* parse a string without interpreting escapes */
853 static uint8_t *parse_pp_string(uint8_t *p,
854 int sep, CString *str)
855 {
856 int c;
857 p++;
858 for(;;) {
859 c = *p;
860 if (c == sep) {
861 break;
862 } else if (c == '\\') {
863 file->buf_ptr = p;
864 c = handle_eob();
865 p = file->buf_ptr;
866 if (c == CH_EOF) {
867 unterminated_string:
868 /* XXX: indicate line number of start of string */
869 tcc_error("missing terminating %c character", sep);
870 } else if (c == '\\') {
871 /* escape : just skip \[\r]\n */
872 PEEKC_EOB(c, p);
873 if (c == '\n') {
874 file->line_num++;
875 p++;
876 } else if (c == '\r') {
877 PEEKC_EOB(c, p);
878 if (c != '\n')
879 expect("'\n' after '\r'");
880 file->line_num++;
881 p++;
882 } else if (c == CH_EOF) {
883 goto unterminated_string;
884 } else {
885 if (str) {
886 cstr_ccat(str, '\\');
887 cstr_ccat(str, c);
888 }
889 p++;
890 }
891 }
892 } else if (c == '\n') {
893 file->line_num++;
894 goto add_char;
895 } else if (c == '\r') {
896 PEEKC_EOB(c, p);
897 if (c != '\n') {
898 if (str)
899 cstr_ccat(str, '\r');
900 } else {
901 file->line_num++;
902 goto add_char;
903 }
904 } else {
905 add_char:
906 if (str)
907 cstr_ccat(str, c);
908 p++;
909 }
910 }
911 p++;
912 return p;
913 }
914
915 /* skip block of text until #else, #elif or #endif. skip also pairs of
916 #if/#endif */
917 static void preprocess_skip(void)
918 {
919 int a, start_of_line, c, in_warn_or_error;
920 uint8_t *p;
921
922 p = file->buf_ptr;
923 a = 0;
924 redo_start:
925 start_of_line = 1;
926 in_warn_or_error = 0;
927 for(;;) {
928 redo_no_start:
929 c = *p;
930 switch(c) {
931 case ' ':
932 case '\t':
933 case '\f':
934 case '\v':
935 case '\r':
936 p++;
937 goto redo_no_start;
938 case '\n':
939 file->line_num++;
940 p++;
941 goto redo_start;
942 case '\\':
943 file->buf_ptr = p;
944 c = handle_eob();
945 if (c == CH_EOF) {
946 expect("#endif");
947 } else if (c == '\\') {
948 ch = file->buf_ptr[0];
949 handle_stray_noerror();
950 }
951 p = file->buf_ptr;
952 goto redo_no_start;
953 /* skip strings */
954 case '\"':
955 case '\'':
956 if (in_warn_or_error)
957 goto _default;
958 p = parse_pp_string(p, c, NULL);
959 break;
960 /* skip comments */
961 case '/':
962 if (in_warn_or_error)
963 goto _default;
964 file->buf_ptr = p;
965 ch = *p;
966 minp();
967 p = file->buf_ptr;
968 if (ch == '*') {
969 p = parse_comment(p);
970 } else if (ch == '/') {
971 p = parse_line_comment(p);
972 }
973 break;
974 case '#':
975 p++;
976 if (start_of_line) {
977 file->buf_ptr = p;
978 next_nomacro();
979 p = file->buf_ptr;
980 if (a == 0 &&
981 (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
982 goto the_end;
983 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
984 a++;
985 else if (tok == TOK_ENDIF)
986 a--;
987 else if( tok == TOK_ERROR || tok == TOK_WARNING)
988 in_warn_or_error = 1;
989 else if (tok == TOK_LINEFEED)
990 goto redo_start;
991 else if (parse_flags & PARSE_FLAG_ASM_FILE)
992 p = parse_line_comment(p - 1);
993 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
994 p = parse_line_comment(p - 1);
995 break;
996 _default:
997 default:
998 p++;
999 break;
1000 }
1001 start_of_line = 0;
1002 }
1003 the_end: ;
1004 file->buf_ptr = p;
1005 }
1006
1007 #if 0
1008 /* return the number of additional 'ints' necessary to store the
1009 token */
1010 static inline int tok_size(const int *p)
1011 {
1012 switch(*p) {
1013 /* 4 bytes */
1014 case TOK_CINT:
1015 case TOK_CUINT:
1016 case TOK_CCHAR:
1017 case TOK_LCHAR:
1018 case TOK_CFLOAT:
1019 case TOK_LINENUM:
1020 return 1 + 1;
1021 case TOK_STR:
1022 case TOK_LSTR:
1023 case TOK_PPNUM:
1024 case TOK_PPSTR:
1025 return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
1026 case TOK_CLONG:
1027 case TOK_CULONG:
1028 return 1 + LONG_SIZE / 4;
1029 case TOK_CDOUBLE:
1030 case TOK_CLLONG:
1031 case TOK_CULLONG:
1032 return 1 + 2;
1033 case TOK_CLDOUBLE:
1034 return 1 + LDOUBLE_SIZE / 4;
1035 default:
1036 return 1 + 0;
1037 }
1038 }
1039 #endif
1040
1041 /* token string handling */
1042 ST_INLN void tok_str_new(TokenString *s)
1043 {
1044 s->str = NULL;
1045 s->len = s->lastlen = 0;
1046 s->allocated_len = 0;
1047 s->last_line_num = -1;
1048 }
1049
1050 ST_FUNC TokenString *tok_str_alloc(void)
1051 {
1052 TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str);
1053 tok_str_new(str);
1054 return str;
1055 }
1056
1057 ST_FUNC int *tok_str_dup(TokenString *s)
1058 {
1059 int *str;
1060
1061 str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int));
1062 memcpy(str, s->str, s->len * sizeof(int));
1063 return str;
1064 }
1065
1066 ST_FUNC void tok_str_free_str(int *str)
1067 {
1068 tal_free(tokstr_alloc, str);
1069 }
1070
1071 ST_FUNC void tok_str_free(TokenString *str)
1072 {
1073 tok_str_free_str(str->str);
1074 tal_free(tokstr_alloc, str);
1075 }
1076
1077 ST_FUNC int *tok_str_realloc(TokenString *s, int new_size)
1078 {
1079 int *str, size;
1080
1081 size = s->allocated_len;
1082 if (size < 16)
1083 size = 16;
1084 while (size < new_size)
1085 size = size * 2;
1086 if (size > s->allocated_len) {
1087 str = tal_realloc(tokstr_alloc, s->str, size * sizeof(int));
1088 s->allocated_len = size;
1089 s->str = str;
1090 }
1091 return s->str;
1092 }
1093
1094 ST_FUNC void tok_str_add(TokenString *s, int t)
1095 {
1096 int len, *str;
1097
1098 len = s->len;
1099 str = s->str;
1100 if (len >= s->allocated_len)
1101 str = tok_str_realloc(s, len + 1);
1102 str[len++] = t;
1103 s->len = len;
1104 }
1105
1106 ST_FUNC void begin_macro(TokenString *str, int alloc)
1107 {
1108 str->alloc = alloc;
1109 str->prev = macro_stack;
1110 str->prev_ptr = macro_ptr;
1111 str->save_line_num = file->line_num;
1112 macro_ptr = str->str;
1113 macro_stack = str;
1114 }
1115
1116 ST_FUNC void end_macro(void)
1117 {
1118 TokenString *str = macro_stack;
1119 macro_stack = str->prev;
1120 macro_ptr = str->prev_ptr;
1121 file->line_num = str->save_line_num;
1122 if (str->alloc == 2) {
1123 str->alloc = 3; /* just mark as finished */
1124 } else {
1125 tok_str_free(str);
1126 }
1127 }
1128
1129 static void tok_str_add2(TokenString *s, int t, CValue *cv)
1130 {
1131 int len, *str;
1132
1133 len = s->lastlen = s->len;
1134 str = s->str;
1135
1136 /* allocate space for worst case */
1137 if (len + TOK_MAX_SIZE >= s->allocated_len)
1138 str = tok_str_realloc(s, len + TOK_MAX_SIZE + 1);
1139 str[len++] = t;
1140 switch(t) {
1141 case TOK_CINT:
1142 case TOK_CUINT:
1143 case TOK_CCHAR:
1144 case TOK_LCHAR:
1145 case TOK_CFLOAT:
1146 case TOK_LINENUM:
1147 #if LONG_SIZE == 4
1148 case TOK_CLONG:
1149 case TOK_CULONG:
1150 #endif
1151 str[len++] = cv->tab[0];
1152 break;
1153 case TOK_PPNUM:
1154 case TOK_PPSTR:
1155 case TOK_STR:
1156 case TOK_LSTR:
1157 {
1158 /* Insert the string into the int array. */
1159 size_t nb_words =
1160 1 + (cv->str.size + sizeof(int) - 1) / sizeof(int);
1161 if (len + nb_words >= s->allocated_len)
1162 str = tok_str_realloc(s, len + nb_words + 1);
1163 str[len] = cv->str.size;
1164 memcpy(&str[len + 1], cv->str.data, cv->str.size);
1165 len += nb_words;
1166 }
1167 break;
1168 case TOK_CDOUBLE:
1169 case TOK_CLLONG:
1170 case TOK_CULLONG:
1171 #if LONG_SIZE == 8
1172 case TOK_CLONG:
1173 case TOK_CULONG:
1174 #endif
1175 #if LDOUBLE_SIZE == 8
1176 case TOK_CLDOUBLE:
1177 #endif
1178 str[len++] = cv->tab[0];
1179 str[len++] = cv->tab[1];
1180 break;
1181 #if LDOUBLE_SIZE == 12
1182 case TOK_CLDOUBLE:
1183 str[len++] = cv->tab[0];
1184 str[len++] = cv->tab[1];
1185 str[len++] = cv->tab[2];
1186 #elif LDOUBLE_SIZE == 16
1187 case TOK_CLDOUBLE:
1188 str[len++] = cv->tab[0];
1189 str[len++] = cv->tab[1];
1190 str[len++] = cv->tab[2];
1191 str[len++] = cv->tab[3];
1192 #elif LDOUBLE_SIZE != 8
1193 #error add long double size support
1194 #endif
1195 break;
1196 default:
1197 break;
1198 }
1199 s->len = len;
1200 }
1201
1202 /* add the current parse token in token string 's' */
1203 ST_FUNC void tok_str_add_tok(TokenString *s)
1204 {
1205 CValue cval;
1206
1207 /* save line number info */
1208 if (file->line_num != s->last_line_num) {
1209 s->last_line_num = file->line_num;
1210 cval.i = s->last_line_num;
1211 tok_str_add2(s, TOK_LINENUM, &cval);
1212 }
1213 tok_str_add2(s, tok, &tokc);
1214 }
1215
1216 /* get a token from an integer array and increment pointer
1217 accordingly. we code it as a macro to avoid pointer aliasing. */
1218 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
1219 {
1220 const int *p = *pp;
1221 int n, *tab;
1222
1223 tab = cv->tab;
1224 switch(*t = *p++) {
1225 #if LONG_SIZE == 4
1226 case TOK_CLONG:
1227 #endif
1228 case TOK_CINT:
1229 case TOK_CCHAR:
1230 case TOK_LCHAR:
1231 case TOK_LINENUM:
1232 cv->i = *p++;
1233 break;
1234 #if LONG_SIZE == 4
1235 case TOK_CULONG:
1236 #endif
1237 case TOK_CUINT:
1238 cv->i = (unsigned)*p++;
1239 break;
1240 case TOK_CFLOAT:
1241 tab[0] = *p++;
1242 break;
1243 case TOK_STR:
1244 case TOK_LSTR:
1245 case TOK_PPNUM:
1246 case TOK_PPSTR:
1247 cv->str.size = *p++;
1248 cv->str.data = p;
1249 p += (cv->str.size + sizeof(int) - 1) / sizeof(int);
1250 break;
1251 case TOK_CDOUBLE:
1252 case TOK_CLLONG:
1253 case TOK_CULLONG:
1254 #if LONG_SIZE == 8
1255 case TOK_CLONG:
1256 case TOK_CULONG:
1257 #endif
1258 n = 2;
1259 goto copy;
1260 case TOK_CLDOUBLE:
1261 #if LDOUBLE_SIZE == 16
1262 n = 4;
1263 #elif LDOUBLE_SIZE == 12
1264 n = 3;
1265 #elif LDOUBLE_SIZE == 8
1266 n = 2;
1267 #else
1268 # error add long double size support
1269 #endif
1270 copy:
1271 do
1272 *tab++ = *p++;
1273 while (--n);
1274 break;
1275 default:
1276 break;
1277 }
1278 *pp = p;
1279 }
1280
1281 static int macro_is_equal(const int *a, const int *b)
1282 {
1283 CValue cv;
1284 int t;
1285
1286 if (!a || !b)
1287 return 1;
1288
1289 while (*a && *b) {
1290 /* first time preallocate macro_equal_buf, next time only reset position to start */
1291 cstr_reset(&macro_equal_buf);
1292 TOK_GET(&t, &a, &cv);
1293 cstr_cat(&macro_equal_buf, get_tok_str(t, &cv), 0);
1294 TOK_GET(&t, &b, &cv);
1295 if (strcmp(macro_equal_buf.data, get_tok_str(t, &cv)))
1296 return 0;
1297 }
1298 return !(*a || *b);
1299 }
1300
1301 /* defines handling */
1302 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1303 {
1304 Sym *s, *o;
1305
1306 o = define_find(v);
1307 s = sym_push2(&define_stack, v, macro_type, 0);
1308 s->d = str;
1309 s->next = first_arg;
1310 table_ident[v - TOK_IDENT]->sym_define = s;
1311
1312 if (o && !macro_is_equal(o->d, s->d))
1313 tcc_warning("%s redefined", get_tok_str(v, NULL));
1314 }
1315
1316 /* undefined a define symbol. Its name is just set to zero */
1317 ST_FUNC void define_undef(Sym *s)
1318 {
1319 int v = s->v;
1320 if (v >= TOK_IDENT && v < tok_ident)
1321 table_ident[v - TOK_IDENT]->sym_define = NULL;
1322 }
1323
1324 ST_INLN Sym *define_find(int v)
1325 {
1326 v -= TOK_IDENT;
1327 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1328 return NULL;
1329 return table_ident[v]->sym_define;
1330 }
1331
1332 /* free define stack until top reaches 'b' */
1333 ST_FUNC void free_defines(Sym *b)
1334 {
1335 while (define_stack != b) {
1336 Sym *top = define_stack;
1337 define_stack = top->prev;
1338 tok_str_free_str(top->d);
1339 define_undef(top);
1340 sym_free(top);
1341 }
1342
1343 /* restore remaining (-D or predefined) symbols if they were
1344 #undef'd in the file */
1345 while (b) {
1346 int v = b->v;
1347 if (v >= TOK_IDENT && v < tok_ident) {
1348 Sym **d = &table_ident[v - TOK_IDENT]->sym_define;
1349 if (!*d)
1350 *d = b;
1351 }
1352 b = b->prev;
1353 }
1354 }
1355
1356 /* label lookup */
1357 ST_FUNC Sym *label_find(int v)
1358 {
1359 v -= TOK_IDENT;
1360 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1361 return NULL;
1362 return table_ident[v]->sym_label;
1363 }
1364
1365 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1366 {
1367 Sym *s, **ps;
1368 s = sym_push2(ptop, v, 0, 0);
1369 s->r = flags;
1370 ps = &table_ident[v - TOK_IDENT]->sym_label;
1371 if (ptop == &global_label_stack) {
1372 /* modify the top most local identifier, so that
1373 sym_identifier will point to 's' when popped */
1374 while (*ps != NULL)
1375 ps = &(*ps)->prev_tok;
1376 }
1377 s->prev_tok = *ps;
1378 *ps = s;
1379 return s;
1380 }
1381
1382 /* pop labels until element last is reached. Look if any labels are
1383 undefined. Define symbols if '&&label' was used. */
1384 ST_FUNC void label_pop(Sym **ptop, Sym *slast, int keep)
1385 {
1386 Sym *s, *s1;
1387 for(s = *ptop; s != slast; s = s1) {
1388 s1 = s->prev;
1389 if (s->r == LABEL_DECLARED) {
1390 tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1391 } else if (s->r == LABEL_FORWARD) {
1392 tcc_error("label '%s' used but not defined",
1393 get_tok_str(s->v, NULL));
1394 } else {
1395 if (s->c) {
1396 /* define corresponding symbol. A size of
1397 1 is put. */
1398 put_extern_sym(s, cur_text_section, s->jnext, 1);
1399 }
1400 }
1401 /* remove label */
1402 table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1403 if (!keep)
1404 sym_free(s);
1405 }
1406 if (!keep)
1407 *ptop = slast;
1408 }
1409
1410 /* fake the nth "#if defined test_..." for tcc -dt -run */
1411 static void maybe_run_test(TCCState *s)
1412 {
1413 const char *p;
1414 if (s->include_stack_ptr != s->include_stack)
1415 return;
1416 p = get_tok_str(tok, NULL);
1417 if (0 != memcmp(p, "test_", 5))
1418 return;
1419 if (0 != --s->run_test)
1420 return;
1421 fprintf(s->ppfp, "\n[%s]\n" + !(s->dflag & 32), p), fflush(s->ppfp);
1422 define_push(tok, MACRO_OBJ, NULL, NULL);
1423 }
1424
1425 /* eval an expression for #if/#elif */
1426 static int expr_preprocess(void)
1427 {
1428 int c, t;
1429 TokenString *str;
1430
1431 str = tok_str_alloc();
1432 pp_expr = 1;
1433 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1434 next(); /* do macro subst */
1435 if (tok == TOK_DEFINED) {
1436 next_nomacro();
1437 t = tok;
1438 if (t == '(')
1439 next_nomacro();
1440 if (tok < TOK_IDENT)
1441 expect("identifier");
1442 if (tcc_state->run_test)
1443 maybe_run_test(tcc_state);
1444 c = define_find(tok) != 0;
1445 if (t == '(') {
1446 next_nomacro();
1447 if (tok != ')')
1448 expect("')'");
1449 }
1450 tok = TOK_CINT;
1451 tokc.i = c;
1452 } else if (tok >= TOK_IDENT) {
1453 /* if undefined macro */
1454 tok = TOK_CINT;
1455 tokc.i = 0;
1456 }
1457 tok_str_add_tok(str);
1458 }
1459 pp_expr = 0;
1460 tok_str_add(str, -1); /* simulate end of file */
1461 tok_str_add(str, 0);
1462 /* now evaluate C constant expression */
1463 begin_macro(str, 1);
1464 next();
1465 c = expr_const();
1466 end_macro();
1467 return c != 0;
1468 }
1469
1470
1471 /* parse after #define */
1472 ST_FUNC void parse_define(void)
1473 {
1474 Sym *s, *first, **ps;
1475 int v, t, varg, is_vaargs, spc;
1476 int saved_parse_flags = parse_flags;
1477
1478 v = tok;
1479 if (v < TOK_IDENT || v == TOK_DEFINED)
1480 tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1481 /* XXX: should check if same macro (ANSI) */
1482 first = NULL;
1483 t = MACRO_OBJ;
1484 /* We have to parse the whole define as if not in asm mode, in particular
1485 no line comment with '#' must be ignored. Also for function
1486 macros the argument list must be parsed without '.' being an ID
1487 character. */
1488 parse_flags = ((parse_flags & ~PARSE_FLAG_ASM_FILE) | PARSE_FLAG_SPACES);
1489 /* '(' must be just after macro definition for MACRO_FUNC */
1490 next_nomacro_spc();
1491 if (tok == '(') {
1492 int dotid = set_idnum('.', 0);
1493 next_nomacro();
1494 ps = &first;
1495 if (tok != ')') for (;;) {
1496 varg = tok;
1497 next_nomacro();
1498 is_vaargs = 0;
1499 if (varg == TOK_DOTS) {
1500 varg = TOK___VA_ARGS__;
1501 is_vaargs = 1;
1502 } else if (tok == TOK_DOTS && gnu_ext) {
1503 is_vaargs = 1;
1504 next_nomacro();
1505 }
1506 if (varg < TOK_IDENT)
1507 bad_list:
1508 tcc_error("bad macro parameter list");
1509 s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1510 *ps = s;
1511 ps = &s->next;
1512 if (tok == ')')
1513 break;
1514 if (tok != ',' || is_vaargs)
1515 goto bad_list;
1516 next_nomacro();
1517 }
1518 next_nomacro_spc();
1519 t = MACRO_FUNC;
1520 set_idnum('.', dotid);
1521 }
1522
1523 tokstr_buf.len = 0;
1524 spc = 2;
1525 parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1526 /* The body of a macro definition should be parsed such that identifiers
1527 are parsed like the file mode determines (i.e. with '.' being an
1528 ID character in asm mode). But '#' should be retained instead of
1529 regarded as line comment leader, so still don't set ASM_FILE
1530 in parse_flags. */
1531 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1532 /* remove spaces around ## and after '#' */
1533 if (TOK_TWOSHARPS == tok) {
1534 if (2 == spc)
1535 goto bad_twosharp;
1536 if (1 == spc)
1537 --tokstr_buf.len;
1538 spc = 3;
1539 tok = TOK_PPJOIN;
1540 } else if ('#' == tok) {
1541 spc = 4;
1542 } else if (check_space(tok, &spc)) {
1543 goto skip;
1544 }
1545 tok_str_add2(&tokstr_buf, tok, &tokc);
1546 skip:
1547 next_nomacro_spc();
1548 }
1549
1550 parse_flags = saved_parse_flags;
1551 if (spc == 1)
1552 --tokstr_buf.len; /* remove trailing space */
1553 tok_str_add(&tokstr_buf, 0);
1554 if (3 == spc)
1555 bad_twosharp:
1556 tcc_error("'##' cannot appear at either end of macro");
1557 define_push(v, t, tok_str_dup(&tokstr_buf), first);
1558 }
1559
1560 static CachedInclude *search_cached_include(TCCState *s1, const char *filename, int add)
1561 {
1562 const unsigned char *s;
1563 unsigned int h;
1564 CachedInclude *e;
1565 int i;
1566
1567 h = TOK_HASH_INIT;
1568 s = (unsigned char *) filename;
1569 while (*s) {
1570 #ifdef _WIN32
1571 h = TOK_HASH_FUNC(h, toup(*s));
1572 #else
1573 h = TOK_HASH_FUNC(h, *s);
1574 #endif
1575 s++;
1576 }
1577 h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1578
1579 i = s1->cached_includes_hash[h];
1580 for(;;) {
1581 if (i == 0)
1582 break;
1583 e = s1->cached_includes[i - 1];
1584 if (0 == PATHCMP(e->filename, filename))
1585 return e;
1586 i = e->hash_next;
1587 }
1588 if (!add)
1589 return NULL;
1590
1591 e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1592 strcpy(e->filename, filename);
1593 e->ifndef_macro = e->once = 0;
1594 dynarray_add(&s1->cached_includes, &s1->nb_cached_includes, e);
1595 /* add in hash table */
1596 e->hash_next = s1->cached_includes_hash[h];
1597 s1->cached_includes_hash[h] = s1->nb_cached_includes;
1598 #ifdef INC_DEBUG
1599 printf("adding cached '%s'\n", filename);
1600 #endif
1601 return e;
1602 }
1603
1604 static void pragma_parse(TCCState *s1)
1605 {
1606 next_nomacro();
1607 if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1608 int t = tok, v;
1609 Sym *s;
1610
1611 if (next(), tok != '(')
1612 goto pragma_err;
1613 if (next(), tok != TOK_STR)
1614 goto pragma_err;
1615 v = tok_alloc(tokc.str.data, tokc.str.size - 1)->tok;
1616 if (next(), tok != ')')
1617 goto pragma_err;
1618 if (t == TOK_push_macro) {
1619 while (NULL == (s = define_find(v)))
1620 define_push(v, 0, NULL, NULL);
1621 s->type.ref = s; /* set push boundary */
1622 } else {
1623 for (s = define_stack; s; s = s->prev)
1624 if (s->v == v && s->type.ref == s) {
1625 s->type.ref = NULL;
1626 break;
1627 }
1628 }
1629 if (s)
1630 table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1631 else
1632 tcc_warning("unbalanced #pragma pop_macro");
1633 pp_debug_tok = t, pp_debug_symv = v;
1634
1635 } else if (tok == TOK_once) {
1636 search_cached_include(s1, file->filename, 1)->once = pp_once;
1637
1638 } else if (s1->output_type == TCC_OUTPUT_PREPROCESS) {
1639 /* tcc -E: keep pragmas below unchanged */
1640 unget_tok(' ');
1641 unget_tok(TOK_PRAGMA);
1642 unget_tok('#');
1643 unget_tok(TOK_LINEFEED);
1644
1645 } else if (tok == TOK_pack) {
1646 /* This may be:
1647 #pragma pack(1) // set
1648 #pragma pack() // reset to default
1649 #pragma pack(push,1) // push & set
1650 #pragma pack(pop) // restore previous */
1651 next();
1652 skip('(');
1653 if (tok == TOK_ASM_pop) {
1654 next();
1655 if (s1->pack_stack_ptr <= s1->pack_stack) {
1656 stk_error:
1657 tcc_error("out of pack stack");
1658 }
1659 s1->pack_stack_ptr--;
1660 } else {
1661 int val = 0;
1662 if (tok != ')') {
1663 if (tok == TOK_ASM_push) {
1664 next();
1665 if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1666 goto stk_error;
1667 s1->pack_stack_ptr++;
1668 skip(',');
1669 }
1670 if (tok != TOK_CINT)
1671 goto pragma_err;
1672 val = tokc.i;
1673 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1674 goto pragma_err;
1675 next();
1676 }
1677 *s1->pack_stack_ptr = val;
1678 }
1679 if (tok != ')')
1680 goto pragma_err;
1681
1682 } else if (tok == TOK_comment) {
1683 char *p; int t;
1684 next();
1685 skip('(');
1686 t = tok;
1687 next();
1688 skip(',');
1689 if (tok != TOK_STR)
1690 goto pragma_err;
1691 p = tcc_strdup((char *)tokc.str.data);
1692 next();
1693 if (tok != ')')
1694 goto pragma_err;
1695 if (t == TOK_lib) {
1696 dynarray_add(&s1->pragma_libs, &s1->nb_pragma_libs, p);
1697 } else {
1698 if (t == TOK_option)
1699 tcc_set_options(s1, p);
1700 tcc_free(p);
1701 }
1702
1703 } else if (s1->warn_unsupported) {
1704 tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1705 }
1706 return;
1707
1708 pragma_err:
1709 tcc_error("malformed #pragma directive");
1710 return;
1711 }
1712
1713 /* is_bof is true if first non space token at beginning of file */
1714 ST_FUNC void preprocess(int is_bof)
1715 {
1716 TCCState *s1 = tcc_state;
1717 int i, c, n, saved_parse_flags;
1718 char buf[1024], *q;
1719 Sym *s;
1720
1721 saved_parse_flags = parse_flags;
1722 parse_flags = PARSE_FLAG_PREPROCESS
1723 | PARSE_FLAG_TOK_NUM
1724 | PARSE_FLAG_TOK_STR
1725 | PARSE_FLAG_LINEFEED
1726 | (parse_flags & PARSE_FLAG_ASM_FILE)
1727 ;
1728
1729 next_nomacro();
1730 redo:
1731 switch(tok) {
1732 case TOK_DEFINE:
1733 pp_debug_tok = tok;
1734 next_nomacro();
1735 pp_debug_symv = tok;
1736 parse_define();
1737 break;
1738 case TOK_UNDEF:
1739 pp_debug_tok = tok;
1740 next_nomacro();
1741 pp_debug_symv = tok;
1742 s = define_find(tok);
1743 /* undefine symbol by putting an invalid name */
1744 if (s)
1745 define_undef(s);
1746 break;
1747 case TOK_INCLUDE:
1748 case TOK_INCLUDE_NEXT:
1749 ch = file->buf_ptr[0];
1750 /* XXX: incorrect if comments : use next_nomacro with a special mode */
1751 skip_spaces();
1752 if (ch == '<') {
1753 c = '>';
1754 goto read_name;
1755 } else if (ch == '\"') {
1756 c = ch;
1757 read_name:
1758 inp();
1759 q = buf;
1760 while (ch != c && ch != '\n' && ch != CH_EOF) {
1761 if ((q - buf) < sizeof(buf) - 1)
1762 *q++ = ch;
1763 if (ch == '\\') {
1764 if (handle_stray_noerror() == 0)
1765 --q;
1766 } else
1767 inp();
1768 }
1769 *q = '\0';
1770 minp();
1771 #if 0
1772 /* eat all spaces and comments after include */
1773 /* XXX: slightly incorrect */
1774 while (ch1 != '\n' && ch1 != CH_EOF)
1775 inp();
1776 #endif
1777 } else {
1778 int len;
1779 /* computed #include : concatenate everything up to linefeed,
1780 the result must be one of the two accepted forms.
1781 Don't convert pp-tokens to tokens here. */
1782 parse_flags = (PARSE_FLAG_PREPROCESS
1783 | PARSE_FLAG_LINEFEED
1784 | (parse_flags & PARSE_FLAG_ASM_FILE));
1785 next();
1786 buf[0] = '\0';
1787 while (tok != TOK_LINEFEED) {
1788 pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1789 next();
1790 }
1791 len = strlen(buf);
1792 /* check syntax and remove '<>|""' */
1793 if ((len < 2 || ((buf[0] != '"' || buf[len-1] != '"') &&
1794 (buf[0] != '<' || buf[len-1] != '>'))))
1795 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1796 c = buf[len-1];
1797 memmove(buf, buf + 1, len - 2);
1798 buf[len - 2] = '\0';
1799 }
1800
1801 if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1802 tcc_error("#include recursion too deep");
1803 /* store current file in stack, but increment stack later below */
1804 *s1->include_stack_ptr = file;
1805 i = tok == TOK_INCLUDE_NEXT ? file->include_next_index : 0;
1806 n = 2 + s1->nb_tccinclude_paths + s1->nb_include_paths +
1807 s1->nb_sysinclude_paths;
1808 for (; i < n; ++i) {
1809 char buf1[sizeof file->filename];
1810 CachedInclude *e;
1811 const char *path;
1812
1813 if (i == 0) {
1814 /* check absolute include path */
1815 if (!IS_ABSPATH(buf))
1816 continue;
1817 buf1[0] = 0;
1818
1819 } else if (i == 1) {
1820 /* search in file's dir if "header.h" */
1821 if (c != '\"')
1822 continue;
1823 /* https://savannah.nongnu.org/bugs/index.php?50847 */
1824 path = file->true_filename;
1825 pstrncpy(buf1, path, tcc_basename(path) - path);
1826
1827 } else {
1828 /* search in all the include paths */
1829 int k, j = i - 2;
1830
1831 if (j < (k = s1->nb_tccinclude_paths))
1832 path = s1->tccinclude_paths[j];
1833 else if ((j -= k) < s1->nb_include_paths)
1834 path = s1->include_paths[j];
1835 else if ((j -= s1->nb_include_paths) < s1->nb_sysinclude_paths)
1836 path = s1->sysinclude_paths[j];
1837 pstrcpy(buf1, sizeof(buf1), path);
1838 pstrcat(buf1, sizeof(buf1), "/");
1839 }
1840
1841 pstrcat(buf1, sizeof(buf1), buf);
1842 e = search_cached_include(s1, buf1, 0);
1843 if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) {
1844 /* no need to parse the include because the 'ifndef macro'
1845 is defined (or had #pragma once) */
1846 #ifdef INC_DEBUG
1847 printf("%s: skipping cached %s\n", file->filename, buf1);
1848 #endif
1849 goto include_done;
1850 }
1851
1852 if (tcc_open(s1, buf1) < 0)
1853 continue;
1854
1855 file->include_next_index = i + 1;
1856 #ifdef INC_DEBUG
1857 printf("%s: including %s\n", file->prev->filename, file->filename);
1858 #endif
1859 /* update target deps */
1860 dynarray_add(&s1->target_deps, &s1->nb_target_deps,
1861 tcc_strdup(buf1));
1862 /* push current file in stack */
1863 ++s1->include_stack_ptr;
1864 /* add include file debug info */
1865 if (s1->do_debug)
1866 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1867 tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1868 ch = file->buf_ptr[0];
1869 goto the_end;
1870 }
1871 tcc_error("include file '%s' not found", buf);
1872 include_done:
1873 break;
1874 case TOK_IFNDEF:
1875 c = 1;
1876 goto do_ifdef;
1877 case TOK_IF:
1878 c = expr_preprocess();
1879 goto do_if;
1880 case TOK_IFDEF:
1881 c = 0;
1882 do_ifdef:
1883 next_nomacro();
1884 if (tok < TOK_IDENT)
1885 tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1886 if (is_bof) {
1887 if (c) {
1888 #ifdef INC_DEBUG
1889 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1890 #endif
1891 file->ifndef_macro = tok;
1892 }
1893 }
1894 c = (define_find(tok) != 0) ^ c;
1895 do_if:
1896 if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1897 tcc_error("memory full (ifdef)");
1898 *s1->ifdef_stack_ptr++ = c;
1899 goto test_skip;
1900 case TOK_ELSE:
1901 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1902 tcc_error("#else without matching #if");
1903 if (s1->ifdef_stack_ptr[-1] & 2)
1904 tcc_error("#else after #else");
1905 c = (s1->ifdef_stack_ptr[-1] ^= 3);
1906 goto test_else;
1907 case TOK_ELIF:
1908 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1909 tcc_error("#elif without matching #if");
1910 c = s1->ifdef_stack_ptr[-1];
1911 if (c > 1)
1912 tcc_error("#elif after #else");
1913 /* last #if/#elif expression was true: we skip */
1914 if (c == 1) {
1915 c = 0;
1916 } else {
1917 c = expr_preprocess();
1918 s1->ifdef_stack_ptr[-1] = c;
1919 }
1920 test_else:
1921 if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1922 file->ifndef_macro = 0;
1923 test_skip:
1924 if (!(c & 1)) {
1925 preprocess_skip();
1926 is_bof = 0;
1927 goto redo;
1928 }
1929 break;
1930 case TOK_ENDIF:
1931 if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1932 tcc_error("#endif without matching #if");
1933 s1->ifdef_stack_ptr--;
1934 /* '#ifndef macro' was at the start of file. Now we check if
1935 an '#endif' is exactly at the end of file */
1936 if (file->ifndef_macro &&
1937 s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1938 file->ifndef_macro_saved = file->ifndef_macro;
1939 /* need to set to zero to avoid false matches if another
1940 #ifndef at middle of file */
1941 file->ifndef_macro = 0;
1942 while (tok != TOK_LINEFEED)
1943 next_nomacro();
1944 tok_flags |= TOK_FLAG_ENDIF;
1945 goto the_end;
1946 }
1947 break;
1948 case TOK_PPNUM:
1949 n = strtoul((char*)tokc.str.data, &q, 10);
1950 goto _line_num;
1951 case TOK_LINE:
1952 next();
1953 if (tok != TOK_CINT)
1954 _line_err:
1955 tcc_error("wrong #line format");
1956 n = tokc.i;
1957 _line_num:
1958 next();
1959 if (tok != TOK_LINEFEED) {
1960 if (tok == TOK_STR) {
1961 if (file->true_filename == file->filename)
1962 file->true_filename = tcc_strdup(file->filename);
1963 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.str.data);
1964 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
1965 break;
1966 else
1967 goto _line_err;
1968 --n;
1969 }
1970 if (file->fd > 0)
1971 total_lines += file->line_num - n;
1972 file->line_num = n;
1973 if (s1->do_debug)
1974 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1975 break;
1976 case TOK_ERROR:
1977 case TOK_WARNING:
1978 c = tok;
1979 ch = file->buf_ptr[0];
1980 skip_spaces();
1981 q = buf;
1982 while (ch != '\n' && ch != CH_EOF) {
1983 if ((q - buf) < sizeof(buf) - 1)
1984 *q++ = ch;
1985 if (ch == '\\') {
1986 if (handle_stray_noerror() == 0)
1987 --q;
1988 } else
1989 inp();
1990 }
1991 *q = '\0';
1992 if (c == TOK_ERROR)
1993 tcc_error("#error %s", buf);
1994 else
1995 tcc_warning("#warning %s", buf);
1996 break;
1997 case TOK_PRAGMA:
1998 pragma_parse(s1);
1999 break;
2000 case TOK_LINEFEED:
2001 goto the_end;
2002 default:
2003 /* ignore gas line comment in an 'S' file. */
2004 if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
2005 goto ignore;
2006 if (tok == '!' && is_bof)
2007 /* '!' is ignored at beginning to allow C scripts. */
2008 goto ignore;
2009 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
2010 ignore:
2011 file->buf_ptr = parse_line_comment(file->buf_ptr - 1);
2012 goto the_end;
2013 }
2014 /* ignore other preprocess commands or #! for C scripts */
2015 while (tok != TOK_LINEFEED)
2016 next_nomacro();
2017 the_end:
2018 parse_flags = saved_parse_flags;
2019 }
2020
2021 /* evaluate escape codes in a string. */
2022 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
2023 {
2024 int c, n;
2025 const uint8_t *p;
2026
2027 p = buf;
2028 for(;;) {
2029 c = *p;
2030 if (c == '\0')
2031 break;
2032 if (c == '\\') {
2033 p++;
2034 /* escape */
2035 c = *p;
2036 switch(c) {
2037 case '0': case '1': case '2': case '3':
2038 case '4': case '5': case '6': case '7':
2039 /* at most three octal digits */
2040 n = c - '0';
2041 p++;
2042 c = *p;
2043 if (isoct(c)) {
2044 n = n * 8 + c - '0';
2045 p++;
2046 c = *p;
2047 if (isoct(c)) {
2048 n = n * 8 + c - '0';
2049 p++;
2050 }
2051 }
2052 c = n;
2053 goto add_char_nonext;
2054 case 'x':
2055 case 'u':
2056 case 'U':
2057 p++;
2058 n = 0;
2059 for(;;) {
2060 c = *p;
2061 if (c >= 'a' && c <= 'f')
2062 c = c - 'a' + 10;
2063 else if (c >= 'A' && c <= 'F')
2064 c = c - 'A' + 10;
2065 else if (isnum(c))
2066 c = c - '0';
2067 else
2068 break;
2069 n = n * 16 + c;
2070 p++;
2071 }
2072 c = n;
2073 goto add_char_nonext;
2074 case 'a':
2075 c = '\a';
2076 break;
2077 case 'b':
2078 c = '\b';
2079 break;
2080 case 'f':
2081 c = '\f';
2082 break;
2083 case 'n':
2084 c = '\n';
2085 break;
2086 case 'r':
2087 c = '\r';
2088 break;
2089 case 't':
2090 c = '\t';
2091 break;
2092 case 'v':
2093 c = '\v';
2094 break;
2095 case 'e':
2096 if (!gnu_ext)
2097 goto invalid_escape;
2098 c = 27;
2099 break;
2100 case '\'':
2101 case '\"':
2102 case '\\':
2103 case '?':
2104 break;
2105 default:
2106 invalid_escape:
2107 if (c >= '!' && c <= '~')
2108 tcc_warning("unknown escape sequence: \'\\%c\'", c);
2109 else
2110 tcc_warning("unknown escape sequence: \'\\x%x\'", c);
2111 break;
2112 }
2113 } else if (is_long && c >= 0x80) {
2114 /* assume we are processing UTF-8 sequence */
2115 /* reference: The Unicode Standard, Version 10.0, ch3.9 */
2116
2117 int cont; /* count of continuation bytes */
2118 int skip; /* how many bytes should skip when error occurred */
2119 int i;
2120
2121 /* decode leading byte */
2122 if (c < 0xC2) {
2123 skip = 1; goto invalid_utf8_sequence;
2124 } else if (c <= 0xDF) {
2125 cont = 1; n = c & 0x1f;
2126 } else if (c <= 0xEF) {
2127 cont = 2; n = c & 0xf;
2128 } else if (c <= 0xF4) {
2129 cont = 3; n = c & 0x7;
2130 } else {
2131 skip = 1; goto invalid_utf8_sequence;
2132 }
2133
2134 /* decode continuation bytes */
2135 for (i = 1; i <= cont; i++) {
2136 int l = 0x80, h = 0xBF;
2137
2138 /* adjust limit for second byte */
2139 if (i == 1) {
2140 switch (c) {
2141 case 0xE0: l = 0xA0; break;
2142 case 0xED: h = 0x9F; break;
2143 case 0xF0: l = 0x90; break;
2144 case 0xF4: h = 0x8F; break;
2145 }
2146 }
2147
2148 if (p[i] < l || p[i] > h) {
2149 skip = i; goto invalid_utf8_sequence;
2150 }
2151
2152 n = (n << 6) | (p[i] & 0x3f);
2153 }
2154
2155 /* advance pointer */
2156 p += 1 + cont;
2157 c = n;
2158 goto add_char_nonext;
2159
2160 /* error handling */
2161 invalid_utf8_sequence:
2162 tcc_warning("ill-formed UTF-8 subsequence starting with: \'\\x%x\'", c);
2163 c = 0xFFFD;
2164 p += skip;
2165 goto add_char_nonext;
2166
2167 }
2168 p++;
2169 add_char_nonext:
2170 if (!is_long)
2171 cstr_ccat(outstr, c);
2172 else {
2173 #ifdef TCC_TARGET_PE
2174 /* store as UTF-16 */
2175 if (c < 0x10000) {
2176 cstr_wccat(outstr, c);
2177 } else {
2178 c -= 0x10000;
2179 cstr_wccat(outstr, (c >> 10) + 0xD800);
2180 cstr_wccat(outstr, (c & 0x3FF) + 0xDC00);
2181 }
2182 #else
2183 cstr_wccat(outstr, c);
2184 #endif
2185 }
2186 }
2187 /* add a trailing '\0' */
2188 if (!is_long)
2189 cstr_ccat(outstr, '\0');
2190 else
2191 cstr_wccat(outstr, '\0');
2192 }
2193
2194 static void parse_string(const char *s, int len)
2195 {
2196 uint8_t buf[1000], *p = buf;
2197 int is_long, sep;
2198
2199 if ((is_long = *s == 'L'))
2200 ++s, --len;
2201 sep = *s++;
2202 len -= 2;
2203 if (len >= sizeof buf)
2204 p = tcc_malloc(len + 1);
2205 memcpy(p, s, len);
2206 p[len] = 0;
2207
2208 cstr_reset(&tokcstr);
2209 parse_escape_string(&tokcstr, p, is_long);
2210 if (p != buf)
2211 tcc_free(p);
2212
2213 if (sep == '\'') {
2214 int char_size, i, n, c;
2215 /* XXX: make it portable */
2216 if (!is_long)
2217 tok = TOK_CCHAR, char_size = 1;
2218 else
2219 tok = TOK_LCHAR, char_size = sizeof(nwchar_t);
2220 n = tokcstr.size / char_size - 1;
2221 if (n < 1)
2222 tcc_error("empty character constant");
2223 if (n > 1)
2224 tcc_warning("multi-character character constant");
2225 for (c = i = 0; i < n; ++i) {
2226 if (is_long)
2227 c = ((nwchar_t *)tokcstr.data)[i];
2228 else
2229 c = (c << 8) | ((char *)tokcstr.data)[i];
2230 }
2231 tokc.i = c;
2232 } else {
2233 tokc.str.size = tokcstr.size;
2234 tokc.str.data = tokcstr.data;
2235 if (!is_long)
2236 tok = TOK_STR;
2237 else
2238 tok = TOK_LSTR;
2239 }
2240 }
2241
2242 /* we use 64 bit numbers */
2243 #define BN_SIZE 2
2244
2245 /* bn = (bn << shift) | or_val */
2246 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2247 {
2248 int i;
2249 unsigned int v;
2250 for(i=0;i<BN_SIZE;i++) {
2251 v = bn[i];
2252 bn[i] = (v << shift) | or_val;
2253 or_val = v >> (32 - shift);
2254 }
2255 }
2256
2257 static void bn_zero(unsigned int *bn)
2258 {
2259 int i;
2260 for(i=0;i<BN_SIZE;i++) {
2261 bn[i] = 0;
2262 }
2263 }
2264
2265 /* parse number in null terminated string 'p' and return it in the
2266 current token */
2267 static void parse_number(const char *p)
2268 {
2269 int b, t, shift, frac_bits, s, exp_val, ch;
2270 char *q;
2271 unsigned int bn[BN_SIZE];
2272 double d;
2273
2274 /* number */
2275 q = token_buf;
2276 ch = *p++;
2277 t = ch;
2278 ch = *p++;
2279 *q++ = t;
2280 b = 10;
2281 if (t == '.') {
2282 goto float_frac_parse;
2283 } else if (t == '0') {
2284 if (ch == 'x' || ch == 'X') {
2285 q--;
2286 ch = *p++;
2287 b = 16;
2288 } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2289 q--;
2290 ch = *p++;
2291 b = 2;
2292 }
2293 }
2294 /* parse all digits. cannot check octal numbers at this stage
2295 because of floating point constants */
2296 while (1) {
2297 if (ch >= 'a' && ch <= 'f')
2298 t = ch - 'a' + 10;
2299 else if (ch >= 'A' && ch <= 'F')
2300 t = ch - 'A' + 10;
2301 else if (isnum(ch))
2302 t = ch - '0';
2303 else
2304 break;
2305 if (t >= b)
2306 break;
2307 if (q >= token_buf + STRING_MAX_SIZE) {
2308 num_too_long:
2309 tcc_error("number too long");
2310 }
2311 *q++ = ch;
2312 ch = *p++;
2313 }
2314 if (ch == '.' ||
2315 ((ch == 'e' || ch == 'E') && b == 10) ||
2316 ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2317 if (b != 10) {
2318 /* NOTE: strtox should support that for hexa numbers, but
2319 non ISOC99 libcs do not support it, so we prefer to do
2320 it by hand */
2321 /* hexadecimal or binary floats */
2322 /* XXX: handle overflows */
2323 *q = '\0';
2324 if (b == 16)
2325 shift = 4;
2326 else
2327 shift = 1;
2328 bn_zero(bn);
2329 q = token_buf;
2330 while (1) {
2331 t = *q++;
2332 if (t == '\0') {
2333 break;
2334 } else if (t >= 'a') {
2335 t = t - 'a' + 10;
2336 } else if (t >= 'A') {
2337 t = t - 'A' + 10;
2338 } else {
2339 t = t - '0';
2340 }
2341 bn_lshift(bn, shift, t);
2342 }
2343 frac_bits = 0;
2344 if (ch == '.') {
2345 ch = *p++;
2346 while (1) {
2347 t = ch;
2348 if (t >= 'a' && t <= 'f') {
2349 t = t - 'a' + 10;
2350 } else if (t >= 'A' && t <= 'F') {
2351 t = t - 'A' + 10;
2352 } else if (t >= '0' && t <= '9') {
2353 t = t - '0';
2354 } else {
2355 break;
2356 }
2357 if (t >= b)
2358 tcc_error("invalid digit");
2359 bn_lshift(bn, shift, t);
2360 frac_bits += shift;
2361 ch = *p++;
2362 }
2363 }
2364 if (ch != 'p' && ch != 'P')
2365 expect("exponent");
2366 ch = *p++;
2367 s = 1;
2368 exp_val = 0;
2369 if (ch == '+') {
2370 ch = *p++;
2371 } else if (ch == '-') {
2372 s = -1;
2373 ch = *p++;
2374 }
2375 if (ch < '0' || ch > '9')
2376 expect("exponent digits");
2377 while (ch >= '0' && ch <= '9') {
2378 exp_val = exp_val * 10 + ch - '0';
2379 ch = *p++;
2380 }
2381 exp_val = exp_val * s;
2382
2383 /* now we can generate the number */
2384 /* XXX: should patch directly float number */
2385 d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2386 d = ldexp(d, exp_val - frac_bits);
2387 t = toup(ch);
2388 if (t == 'F') {
2389 ch = *p++;
2390 tok = TOK_CFLOAT;
2391 /* float : should handle overflow */
2392 tokc.f = (float)d;
2393 } else if (t == 'L') {
2394 ch = *p++;
2395 #ifdef TCC_TARGET_PE
2396 tok = TOK_CDOUBLE;
2397 tokc.d = d;
2398 #else
2399 tok = TOK_CLDOUBLE;
2400 /* XXX: not large enough */
2401 tokc.ld = (long double)d;
2402 #endif
2403 } else {
2404 tok = TOK_CDOUBLE;
2405 tokc.d = d;
2406 }
2407 } else {
2408 /* decimal floats */
2409 if (ch == '.') {
2410 if (q >= token_buf + STRING_MAX_SIZE)
2411 goto num_too_long;
2412 *q++ = ch;
2413 ch = *p++;
2414 float_frac_parse:
2415 while (ch >= '0' && ch <= '9') {
2416 if (q >= token_buf + STRING_MAX_SIZE)
2417 goto num_too_long;
2418 *q++ = ch;
2419 ch = *p++;
2420 }
2421 }
2422 if (ch == 'e' || ch == 'E') {
2423 if (q >= token_buf + STRING_MAX_SIZE)
2424 goto num_too_long;
2425 *q++ = ch;
2426 ch = *p++;
2427 if (ch == '-' || ch == '+') {
2428 if (q >= token_buf + STRING_MAX_SIZE)
2429 goto num_too_long;
2430 *q++ = ch;
2431 ch = *p++;
2432 }
2433 if (ch < '0' || ch > '9')
2434 expect("exponent digits");
2435 while (ch >= '0' && ch <= '9') {
2436 if (q >= token_buf + STRING_MAX_SIZE)
2437 goto num_too_long;
2438 *q++ = ch;
2439 ch = *p++;
2440 }
2441 }
2442 *q = '\0';
2443 t = toup(ch);
2444 errno = 0;
2445 if (t == 'F') {
2446 ch = *p++;
2447 tok = TOK_CFLOAT;
2448 tokc.f = strtof(token_buf, NULL);
2449 } else if (t == 'L') {
2450 ch = *p++;
2451 #ifdef TCC_TARGET_PE
2452 tok = TOK_CDOUBLE;
2453 tokc.d = strtod(token_buf, NULL);
2454 #else
2455 tok = TOK_CLDOUBLE;
2456 tokc.ld = strtold(token_buf, NULL);
2457 #endif
2458 } else {
2459 tok = TOK_CDOUBLE;
2460 tokc.d = strtod(token_buf, NULL);
2461 }
2462 }
2463 } else {
2464 unsigned long long n, n1;
2465 int lcount, ucount, ov = 0;
2466 const char *p1;
2467
2468 /* integer number */
2469 *q = '\0';
2470 q = token_buf;
2471 if (b == 10 && *q == '0') {
2472 b = 8;
2473 q++;
2474 }
2475 n = 0;
2476 while(1) {
2477 t = *q++;
2478 /* no need for checks except for base 10 / 8 errors */
2479 if (t == '\0')
2480 break;
2481 else if (t >= 'a')
2482 t = t - 'a' + 10;
2483 else if (t >= 'A')
2484 t = t - 'A' + 10;
2485 else
2486 t = t - '0';
2487 if (t >= b)
2488 tcc_error("invalid digit");
2489 n1 = n;
2490 n = n * b + t;
2491 /* detect overflow */
2492 if (n1 >= 0x1000000000000000ULL && n / b != n1)
2493 ov = 1;
2494 }
2495
2496 /* Determine the characteristics (unsigned and/or 64bit) the type of
2497 the constant must have according to the constant suffix(es) */
2498 lcount = ucount = 0;
2499 p1 = p;
2500 for(;;) {
2501 t = toup(ch);
2502 if (t == 'L') {
2503 if (lcount >= 2)
2504 tcc_error("three 'l's in integer constant");
2505 if (lcount && *(p - 1) != ch)
2506 tcc_error("incorrect integer suffix: %s", p1);
2507 lcount++;
2508 ch = *p++;
2509 } else if (t == 'U') {
2510 if (ucount >= 1)
2511 tcc_error("two 'u's in integer constant");
2512 ucount++;
2513 ch = *p++;
2514 } else {
2515 break;
2516 }
2517 }
2518
2519 /* Determine if it needs 64 bits and/or unsigned in order to fit */
2520 if (ucount == 0 && b == 10) {
2521 if (lcount <= (LONG_SIZE == 4)) {
2522 if (n >= 0x80000000U)
2523 lcount = (LONG_SIZE == 4) + 1;
2524 }
2525 if (n >= 0x8000000000000000ULL)
2526 ov = 1, ucount = 1;
2527 } else {
2528 if (lcount <= (LONG_SIZE == 4)) {
2529 if (n >= 0x100000000ULL)
2530 lcount = (LONG_SIZE == 4) + 1;
2531 else if (n >= 0x80000000U)
2532 ucount = 1;
2533 }
2534 if (n >= 0x8000000000000000ULL)
2535 ucount = 1;
2536 }
2537
2538 if (ov)
2539 tcc_warning("integer constant overflow");
2540
2541 tok = TOK_CINT;
2542 if (lcount) {
2543 tok = TOK_CLONG;
2544 if (lcount == 2)
2545 tok = TOK_CLLONG;
2546 }
2547 if (ucount)
2548 ++tok; /* TOK_CU... */
2549 tokc.i = n;
2550 }
2551 if (ch)
2552 tcc_error("invalid number\n");
2553 }
2554
2555
2556 #define PARSE2(c1, tok1, c2, tok2) \
2557 case c1: \
2558 PEEKC(c, p); \
2559 if (c == c2) { \
2560 p++; \
2561 tok = tok2; \
2562 } else { \
2563 tok = tok1; \
2564 } \
2565 break;
2566
2567 /* return next token without macro substitution */
2568 static inline void next_nomacro1(void)
2569 {
2570 int t, c, is_long, len;
2571 TokenSym *ts;
2572 uint8_t *p, *p1;
2573 unsigned int h;
2574
2575 p = file->buf_ptr;
2576 redo_no_start:
2577 c = *p;
2578 switch(c) {
2579 case ' ':
2580 case '\t':
2581 tok = c;
2582 p++;
2583 if (parse_flags & PARSE_FLAG_SPACES)
2584 goto keep_tok_flags;
2585 while (isidnum_table[*p - CH_EOF] & IS_SPC)
2586 ++p;
2587 goto redo_no_start;
2588 case '\f':
2589 case '\v':
2590 case '\r':
2591 p++;
2592 goto redo_no_start;
2593 case '\\':
2594 /* first look if it is in fact an end of buffer */
2595 c = handle_stray1(p);
2596 p = file->buf_ptr;
2597 if (c == '\\')
2598 goto parse_simple;
2599 if (c != CH_EOF)
2600 goto redo_no_start;
2601 {
2602 TCCState *s1 = tcc_state;
2603 if ((parse_flags & PARSE_FLAG_LINEFEED)
2604 && !(tok_flags & TOK_FLAG_EOF)) {
2605 tok_flags |= TOK_FLAG_EOF;
2606 tok = TOK_LINEFEED;
2607 goto keep_tok_flags;
2608 } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2609 tok = TOK_EOF;
2610 } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2611 tcc_error("missing #endif");
2612 } else if (s1->include_stack_ptr == s1->include_stack) {
2613 /* no include left : end of file. */
2614 tok = TOK_EOF;
2615 } else {
2616 tok_flags &= ~TOK_FLAG_EOF;
2617 /* pop include file */
2618
2619 /* test if previous '#endif' was after a #ifdef at
2620 start of file */
2621 if (tok_flags & TOK_FLAG_ENDIF) {
2622 #ifdef INC_DEBUG
2623 printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2624 #endif
2625 search_cached_include(s1, file->filename, 1)
2626 ->ifndef_macro = file->ifndef_macro_saved;
2627 tok_flags &= ~TOK_FLAG_ENDIF;
2628 }
2629
2630 /* add end of include file debug info */
2631 if (tcc_state->do_debug) {
2632 put_stabd(N_EINCL, 0, 0);
2633 }
2634 /* pop include stack */
2635 tcc_close();
2636 s1->include_stack_ptr--;
2637 p = file->buf_ptr;
2638 goto redo_no_start;
2639 }
2640 }
2641 break;
2642
2643 case '\n':
2644 file->line_num++;
2645 tok_flags |= TOK_FLAG_BOL;
2646 p++;
2647 maybe_newline:
2648 if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2649 goto redo_no_start;
2650 tok = TOK_LINEFEED;
2651 goto keep_tok_flags;
2652
2653 case '#':
2654 /* XXX: simplify */
2655 PEEKC(c, p);
2656 if ((tok_flags & TOK_FLAG_BOL) &&
2657 (parse_flags & PARSE_FLAG_PREPROCESS)) {
2658 file->buf_ptr = p;
2659 preprocess(tok_flags & TOK_FLAG_BOF);
2660 p = file->buf_ptr;
2661 goto maybe_newline;
2662 } else {
2663 if (c == '#') {
2664 p++;
2665 tok = TOK_TWOSHARPS;
2666 } else {
2667 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2668 p = parse_line_comment(p - 1);
2669 goto redo_no_start;
2670 } else {
2671 tok = '#';
2672 }
2673 }
2674 }
2675 break;
2676
2677 /* dollar is allowed to start identifiers when not parsing asm */
2678 case '$':
2679 if (!(isidnum_table[c - CH_EOF] & IS_ID)
2680 || (parse_flags & PARSE_FLAG_ASM_FILE))
2681 goto parse_simple;
2682
2683 case 'a': case 'b': case 'c': case 'd':
2684 case 'e': case 'f': case 'g': case 'h':
2685 case 'i': case 'j': case 'k': case 'l':
2686 case 'm': case 'n': case 'o': case 'p':
2687 case 'q': case 'r': case 's': case 't':
2688 case 'u': case 'v': case 'w': case 'x':
2689 case 'y': case 'z':
2690 case 'A': case 'B': case 'C': case 'D':
2691 case 'E': case 'F': case 'G': case 'H':
2692 case 'I': case 'J': case 'K':
2693 case 'M': case 'N': case 'O': case 'P':
2694 case 'Q': case 'R': case 'S': case 'T':
2695 case 'U': case 'V': case 'W': case 'X':
2696 case 'Y': case 'Z':
2697 case '_':
2698 parse_ident_fast:
2699 p1 = p;
2700 h = TOK_HASH_INIT;
2701 h = TOK_HASH_FUNC(h, c);
2702 while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2703 h = TOK_HASH_FUNC(h, c);
2704 len = p - p1;
2705 if (c != '\\') {
2706 TokenSym **pts;
2707
2708 /* fast case : no stray found, so we have the full token
2709 and we have already hashed it */
2710 h &= (TOK_HASH_SIZE - 1);
2711 pts = &hash_ident[h];
2712 for(;;) {
2713 ts = *pts;
2714 if (!ts)
2715 break;
2716 if (ts->len == len && !memcmp(ts->str, p1, len))
2717 goto token_found;
2718 pts = &(ts->hash_next);
2719 }
2720 ts = tok_alloc_new(pts, (char *) p1, len);
2721 token_found: ;
2722 } else {
2723 /* slower case */
2724 cstr_reset(&tokcstr);
2725 cstr_cat(&tokcstr, (char *) p1, len);
2726 p--;
2727 PEEKC(c, p);
2728 parse_ident_slow:
2729 while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2730 {
2731 cstr_ccat(&tokcstr, c);
2732 PEEKC(c, p);
2733 }
2734 ts = tok_alloc(tokcstr.data, tokcstr.size);
2735 }
2736 tok = ts->tok;
2737 break;
2738 case 'L':
2739 t = p[1];
2740 if (t != '\\' && t != '\'' && t != '\"') {
2741 /* fast case */
2742 goto parse_ident_fast;
2743 } else {
2744 PEEKC(c, p);
2745 if (c == '\'' || c == '\"') {
2746 is_long = 1;
2747 goto str_const;
2748 } else {
2749 cstr_reset(&tokcstr);
2750 cstr_ccat(&tokcstr, 'L');
2751 goto parse_ident_slow;
2752 }
2753 }
2754 break;
2755
2756 case '0': case '1': case '2': case '3':
2757 case '4': case '5': case '6': case '7':
2758 case '8': case '9':
2759 t = c;
2760 PEEKC(c, p);
2761 /* after the first digit, accept digits, alpha, '.' or sign if
2762 prefixed by 'eEpP' */
2763 parse_num:
2764 cstr_reset(&tokcstr);
2765 for(;;) {
2766 cstr_ccat(&tokcstr, t);
2767 if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2768 || c == '.'
2769 || ((c == '+' || c == '-')
2770 && (((t == 'e' || t == 'E')
2771 && !(parse_flags & PARSE_FLAG_ASM_FILE
2772 /* 0xe+1 is 3 tokens in asm */
2773 && ((char*)tokcstr.data)[0] == '0'
2774 && toup(((char*)tokcstr.data)[1]) == 'X'))
2775 || t == 'p' || t == 'P'))))
2776 break;
2777 t = c;
2778 PEEKC(c, p);
2779 }
2780 /* We add a trailing '\0' to ease parsing */
2781 cstr_ccat(&tokcstr, '\0');
2782 tokc.str.size = tokcstr.size;
2783 tokc.str.data = tokcstr.data;
2784 tok = TOK_PPNUM;
2785 break;
2786
2787 case '.':
2788 /* special dot handling because it can also start a number */
2789 PEEKC(c, p);
2790 if (isnum(c)) {
2791 t = '.';
2792 goto parse_num;
2793 } else if ((isidnum_table['.' - CH_EOF] & IS_ID)
2794 && (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))) {
2795 *--p = c = '.';
2796 goto parse_ident_fast;
2797 } else if (c == '.') {
2798 PEEKC(c, p);
2799 if (c == '.') {
2800 p++;
2801 tok = TOK_DOTS;
2802 } else {
2803 *--p = '.'; /* may underflow into file->unget[] */
2804 tok = '.';
2805 }
2806 } else {
2807 tok = '.';
2808 }
2809 break;
2810 case '\'':
2811 case '\"':
2812 is_long = 0;
2813 str_const:
2814 cstr_reset(&tokcstr);
2815 if (is_long)
2816 cstr_ccat(&tokcstr, 'L');
2817 cstr_ccat(&tokcstr, c);
2818 p = parse_pp_string(p, c, &tokcstr);
2819 cstr_ccat(&tokcstr, c);
2820 cstr_ccat(&tokcstr, '\0');
2821 tokc.str.size = tokcstr.size;
2822 tokc.str.data = tokcstr.data;
2823 tok = TOK_PPSTR;
2824 break;
2825
2826 case '<':
2827 PEEKC(c, p);
2828 if (c == '=') {
2829 p++;
2830 tok = TOK_LE;
2831 } else if (c == '<') {
2832 PEEKC(c, p);
2833 if (c == '=') {
2834 p++;
2835 tok = TOK_A_SHL;
2836 } else {
2837 tok = TOK_SHL;
2838 }
2839 } else {
2840 tok = TOK_LT;
2841 }
2842 break;
2843 case '>':
2844 PEEKC(c, p);
2845 if (c == '=') {
2846 p++;
2847 tok = TOK_GE;
2848 } else if (c == '>') {
2849 PEEKC(c, p);
2850 if (c == '=') {
2851 p++;
2852 tok = TOK_A_SAR;
2853 } else {
2854 tok = TOK_SAR;
2855 }
2856 } else {
2857 tok = TOK_GT;
2858 }
2859 break;
2860
2861 case '&':
2862 PEEKC(c, p);
2863 if (c == '&') {
2864 p++;
2865 tok = TOK_LAND;
2866 } else if (c == '=') {
2867 p++;
2868 tok = TOK_A_AND;
2869 } else {
2870 tok = '&';
2871 }
2872 break;
2873
2874 case '|':
2875 PEEKC(c, p);
2876 if (c == '|') {
2877 p++;
2878 tok = TOK_LOR;
2879 } else if (c == '=') {
2880 p++;
2881 tok = TOK_A_OR;
2882 } else {
2883 tok = '|';
2884 }
2885 break;
2886
2887 case '+':
2888 PEEKC(c, p);
2889 if (c == '+') {
2890 p++;
2891 tok = TOK_INC;
2892 } else if (c == '=') {
2893 p++;
2894 tok = TOK_A_ADD;
2895 } else {
2896 tok = '+';
2897 }
2898 break;
2899
2900 case '-':
2901 PEEKC(c, p);
2902 if (c == '-') {
2903 p++;
2904 tok = TOK_DEC;
2905 } else if (c == '=') {
2906 p++;
2907 tok = TOK_A_SUB;
2908 } else if (c == '>') {
2909 p++;
2910 tok = TOK_ARROW;
2911 } else {
2912 tok = '-';
2913 }
2914 break;
2915
2916 PARSE2('!', '!', '=', TOK_NE)
2917 PARSE2('=', '=', '=', TOK_EQ)
2918 PARSE2('*', '*', '=', TOK_A_MUL)
2919 PARSE2('%', '%', '=', TOK_A_MOD)
2920 PARSE2('^', '^', '=', TOK_A_XOR)
2921
2922 /* comments or operator */
2923 case '/':
2924 PEEKC(c, p);
2925 if (c == '*') {
2926 p = parse_comment(p);
2927 /* comments replaced by a blank */
2928 tok = ' ';
2929 goto keep_tok_flags;
2930 } else if (c == '/') {
2931 p = parse_line_comment(p);
2932 tok = ' ';
2933 goto keep_tok_flags;
2934 } else if (c == '=') {
2935 p++;
2936 tok = TOK_A_DIV;
2937 } else {
2938 tok = '/';
2939 }
2940 break;
2941
2942 /* simple tokens */
2943 case '(':
2944 case ')':
2945 case '[':
2946 case ']':
2947 case '{':
2948 case '}':
2949 case ',':
2950 case ';':
2951 case ':':
2952 case '?':
2953 case '~':
2954 case '@': /* only used in assembler */
2955 parse_simple:
2956 tok = c;
2957 p++;
2958 break;
2959 default:
2960 if (c >= 0x80 && c <= 0xFF) /* utf8 identifiers */
2961 goto parse_ident_fast;
2962 if (parse_flags & PARSE_FLAG_ASM_FILE)
2963 goto parse_simple;
2964 tcc_error("unrecognized character \\x%02x", c);
2965 break;
2966 }
2967 tok_flags = 0;
2968 keep_tok_flags:
2969 file->buf_ptr = p;
2970 #if defined(PARSE_DEBUG)
2971 printf("token = %s\n", get_tok_str(tok, &tokc));
2972 #endif
2973 }
2974
2975 /* return next token without macro substitution. Can read input from
2976 macro_ptr buffer */
2977 static void next_nomacro_spc(void)
2978 {
2979 if (macro_ptr) {
2980 redo:
2981 tok = *macro_ptr;
2982 if (tok) {
2983 TOK_GET(&tok, &macro_ptr, &tokc);
2984 if (tok == TOK_LINENUM) {
2985 file->line_num = tokc.i;
2986 goto redo;
2987 }
2988 }
2989 } else {
2990 next_nomacro1();
2991 }
2992 //printf("token = %s\n", get_tok_str(tok, &tokc));
2993 }
2994
2995 ST_FUNC void next_nomacro(void)
2996 {
2997 do {
2998 next_nomacro_spc();
2999 } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
3000 }
3001
3002
3003 static void macro_subst(
3004 TokenString *tok_str,
3005 Sym **nested_list,
3006 const int *macro_str
3007 );
3008
3009 /* substitute arguments in replacement lists in macro_str by the values in
3010 args (field d) and return allocated string */
3011 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
3012 {
3013 int t, t0, t1, spc;
3014 const int *st;
3015 Sym *s;
3016 CValue cval;
3017 TokenString str;
3018 CString cstr;
3019
3020 tok_str_new(&str);
3021 t0 = t1 = 0;
3022 while(1) {
3023 TOK_GET(&t, &macro_str, &cval);
3024 if (!t)
3025 break;
3026 if (t == '#') {
3027 /* stringize */
3028 TOK_GET(&t, &macro_str, &cval);
3029 if (!t)
3030 goto bad_stringy;
3031 s = sym_find2(args, t);
3032 if (s) {
3033 cstr_new(&cstr);
3034 cstr_ccat(&cstr, '\"');
3035 st = s->d;
3036 spc = 0;
3037 while (*st >= 0) {
3038 TOK_GET(&t, &st, &cval);
3039 if (t != TOK_PLCHLDR
3040 && t != TOK_NOSUBST
3041 && 0 == check_space(t, &spc)) {
3042 const char *s = get_tok_str(t, &cval);
3043 while (*s) {
3044 if (t == TOK_PPSTR && *s != '\'')
3045 add_char(&cstr, *s);
3046 else
3047 cstr_ccat(&cstr, *s);
3048 ++s;
3049 }
3050 }
3051 }
3052 cstr.size -= spc;
3053 cstr_ccat(&cstr, '\"');
3054 cstr_ccat(&cstr, '\0');
3055 #ifdef PP_DEBUG
3056 printf("\nstringize: <%s>\n", (char *)cstr.data);
3057 #endif
3058 /* add string */
3059 cval.str.size = cstr.size;
3060 cval.str.data = cstr.data;
3061 tok_str_add2(&str, TOK_PPSTR, &cval);
3062 cstr_free(&cstr);
3063 } else {
3064 bad_stringy:
3065 expect("macro parameter after '#'");
3066 }
3067 } else if (t >= TOK_IDENT) {
3068 s = sym_find2(args, t);
3069 if (s) {
3070 int l0 = str.len;
3071 st = s->d;
3072 /* if '##' is present before or after, no arg substitution */
3073 if (*macro_str == TOK_PPJOIN || t1 == TOK_PPJOIN) {
3074 /* special case for var arg macros : ## eats the ','
3075 if empty VA_ARGS variable. */
3076 if (t1 == TOK_PPJOIN && t0 == ',' && gnu_ext && s->type.t) {
3077 if (*st <= 0) {
3078 /* suppress ',' '##' */
3079 str.len -= 2;
3080 } else {
3081 /* suppress '##' and add variable */
3082 str.len--;
3083 goto add_var;
3084 }
3085 }
3086 } else {
3087 add_var:
3088 if (!s->next) {
3089 /* Expand arguments tokens and store them. In most
3090 cases we could also re-expand each argument if
3091 used multiple times, but not if the argument
3092 contains the __COUNTER__ macro. */
3093 TokenString str2;
3094 sym_push2(&s->next, s->v, s->type.t, 0);
3095 tok_str_new(&str2);
3096 macro_subst(&str2, nested_list, st);
3097 tok_str_add(&str2, 0);
3098 s->next->d = str2.str;
3099 }
3100 st = s->next->d;
3101 }
3102 for(;;) {
3103 int t2;
3104 TOK_GET(&t2, &st, &cval);
3105 if (t2 <= 0)
3106 break;
3107 tok_str_add2(&str, t2, &cval);
3108 }
3109 if (str.len == l0) /* expanded to empty string */
3110 tok_str_add(&str, TOK_PLCHLDR);
3111 } else {
3112 tok_str_add(&str, t);
3113 }
3114 } else {
3115 tok_str_add2(&str, t, &cval);
3116 }
3117 t0 = t1, t1 = t;
3118 }
3119 tok_str_add(&str, 0);
3120 return str.str;
3121 }
3122
3123 static char const ab_month_name[12][4] =
3124 {
3125 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3126 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3127 };
3128
3129 static int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3130 {
3131 CString cstr;
3132 int n, ret = 1;
3133
3134 cstr_new(&cstr);
3135 if (t1 != TOK_PLCHLDR)
3136 cstr_cat(&cstr, get_tok_str(t1, v1), -1);
3137 n = cstr.size;
3138 if (t2 != TOK_PLCHLDR)
3139 cstr_cat(&cstr, get_tok_str(t2, v2), -1);
3140 cstr_ccat(&cstr, '\0');
3141
3142 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3143 memcpy(file->buffer, cstr.data, cstr.size);
3144 tok_flags = 0;
3145 for (;;) {
3146 next_nomacro1();
3147 if (0 == *file->buf_ptr)
3148 break;
3149 if (is_space(tok))
3150 continue;
3151 tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid"
3152 " preprocessing token", n, cstr.data, (char*)cstr.data + n);
3153 ret = 0;
3154 break;
3155 }
3156 tcc_close();
3157 //printf("paste <%s>\n", (char*)cstr.data);
3158 cstr_free(&cstr);
3159 return ret;
3160 }
3161
3162 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3163 return the resulting string (which must be freed). */
3164 static inline int *macro_twosharps(const int *ptr0)
3165 {
3166 int t;
3167 CValue cval;
3168 TokenString macro_str1;
3169 int start_of_nosubsts = -1;
3170 const int *ptr;
3171
3172 /* we search the first '##' */
3173 for (ptr = ptr0;;) {
3174 TOK_GET(&t, &ptr, &cval);
3175 if (t == TOK_PPJOIN)
3176 break;
3177 if (t == 0)
3178 return NULL;
3179 }
3180
3181 tok_str_new(&macro_str1);
3182
3183 //tok_print(" $$$", ptr0);
3184 for (ptr = ptr0;;) {
3185 TOK_GET(&t, &ptr, &cval);
3186 if (t == 0)
3187 break;
3188 if (t == TOK_PPJOIN)
3189 continue;
3190 while (*ptr == TOK_PPJOIN) {
3191 int t1; CValue cv1;
3192 /* given 'a##b', remove nosubsts preceding 'a' */
3193 if (start_of_nosubsts >= 0)
3194 macro_str1.len = start_of_nosubsts;
3195 /* given 'a##b', remove nosubsts preceding 'b' */
3196 while ((t1 = *++ptr) == TOK_NOSUBST)
3197 ;
3198 if (t1 && t1 != TOK_PPJOIN) {
3199 TOK_GET(&t1, &ptr, &cv1);
3200 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3201 if (paste_tokens(t, &cval, t1, &cv1)) {
3202 t = tok, cval = tokc;
3203 } else {
3204 tok_str_add2(&macro_str1, t, &cval);
3205 t = t1, cval = cv1;
3206 }
3207 }
3208 }
3209 }
3210 if (t == TOK_NOSUBST) {
3211 if (start_of_nosubsts < 0)
3212 start_of_nosubsts = macro_str1.len;
3213 } else {
3214 start_of_nosubsts = -1;
3215 }
3216 tok_str_add2(&macro_str1, t, &cval);
3217 }
3218 tok_str_add(&macro_str1, 0);
3219 //tok_print(" ###", macro_str1.str);
3220 return macro_str1.str;
3221 }
3222
3223 /* peek or read [ws_str == NULL] next token from function macro call,
3224 walking up macro levels up to the file if necessary */
3225 static int next_argstream(Sym **nested_list, TokenString *ws_str)
3226 {
3227 int t;
3228 const int *p;
3229 Sym *sa;
3230
3231 for (;;) {
3232 if (macro_ptr) {
3233 p = macro_ptr, t = *p;
3234 if (ws_str) {
3235 while (is_space(t) || TOK_LINEFEED == t || TOK_PLCHLDR == t)
3236 tok_str_add(ws_str, t), t = *++p;
3237 }
3238 if (t == 0) {
3239 end_macro();
3240 /* also, end of scope for nested defined symbol */
3241 sa = *nested_list;
3242 while (sa && sa->v == 0)
3243 sa = sa->prev;
3244 if (sa)
3245 sa->v = 0;
3246 continue;
3247 }
3248 } else {
3249 ch = handle_eob();
3250 if (ws_str) {
3251 while (is_space(ch) || ch == '\n' || ch == '/') {
3252 if (ch == '/') {
3253 int c;
3254 uint8_t *p = file->buf_ptr;
3255 PEEKC(c, p);
3256 if (c == '*') {
3257 p = parse_comment(p);
3258 file->buf_ptr = p - 1;
3259 } else if (c == '/') {
3260 p = parse_line_comment(p);
3261 file->buf_ptr = p - 1;
3262 } else
3263 break;
3264 ch = ' ';
3265 }
3266 if (ch == '\n')
3267 file->line_num++;
3268 if (!(ch == '\f' || ch == '\v' || ch == '\r'))
3269 tok_str_add(ws_str, ch);
3270 cinp();
3271 }
3272 }
3273 t = ch;
3274 }
3275
3276 if (ws_str)
3277 return t;
3278 next_nomacro_spc();
3279 return tok;
3280 }
3281 }
3282
3283 /* do macro substitution of current token with macro 's' and add
3284 result to (tok_str,tok_len). 'nested_list' is the list of all
3285 macros we got inside to avoid recursing. Return non zero if no
3286 substitution needs to be done */
3287 static int macro_subst_tok(
3288 TokenString *tok_str,
3289 Sym **nested_list,
3290 Sym *s)
3291 {
3292 Sym *args, *sa, *sa1;
3293 int parlevel, t, t1, spc;
3294 TokenString str;
3295 char *cstrval;
3296 CValue cval;
3297 CString cstr;
3298 char buf[32];
3299
3300 /* if symbol is a macro, prepare substitution */
3301 /* special macros */
3302 if (tok == TOK___LINE__ || tok == TOK___COUNTER__) {
3303 t = tok == TOK___LINE__ ? file->line_num : pp_counter++;
3304 snprintf(buf, sizeof(buf), "%d", t);
3305 cstrval = buf;
3306 t1 = TOK_PPNUM;
3307 goto add_cstr1;
3308 } else if (tok == TOK___FILE__) {
3309 cstrval = file->filename;
3310 goto add_cstr;
3311 } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
3312 time_t ti;
3313 struct tm *tm;
3314
3315 time(&ti);
3316 tm = localtime(&ti);
3317 if (tok == TOK___DATE__) {
3318 snprintf(buf, sizeof(buf), "%s %2d %d",
3319 ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
3320 } else {
3321 snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
3322 tm->tm_hour, tm->tm_min, tm->tm_sec);
3323 }
3324 cstrval = buf;
3325 add_cstr:
3326 t1 = TOK_STR;
3327 add_cstr1:
3328 cstr_new(&cstr);
3329 cstr_cat(&cstr, cstrval, 0);
3330 cval.str.size = cstr.size;
3331 cval.str.data = cstr.data;
3332 tok_str_add2(tok_str, t1, &cval);
3333 cstr_free(&cstr);
3334 } else if (s->d) {
3335 int saved_parse_flags = parse_flags;
3336 int *joined_str = NULL;
3337 int *mstr = s->d;
3338
3339 if (s->type.t == MACRO_FUNC) {
3340 /* whitespace between macro name and argument list */
3341 TokenString ws_str;
3342 tok_str_new(&ws_str);
3343
3344 spc = 0;
3345 parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
3346 | PARSE_FLAG_ACCEPT_STRAYS;
3347
3348 /* get next token from argument stream */
3349 t = next_argstream(nested_list, &ws_str);
3350 if (t != '(') {
3351 /* not a macro substitution after all, restore the
3352 * macro token plus all whitespace we've read.
3353 * whitespace is intentionally not merged to preserve
3354 * newlines. */
3355 parse_flags = saved_parse_flags;
3356 tok_str_add(tok_str, tok);
3357 if (parse_flags & PARSE_FLAG_SPACES) {
3358 int i;
3359 for (i = 0; i < ws_str.len; i++)
3360 tok_str_add(tok_str, ws_str.str[i]);
3361 }
3362 tok_str_free_str(ws_str.str);
3363 return 0;
3364 } else {
3365 tok_str_free_str(ws_str.str);
3366 }
3367 do {
3368 next_nomacro(); /* eat '(' */
3369 } while (tok == TOK_PLCHLDR);
3370
3371 /* argument macro */
3372 args = NULL;
3373 sa = s->next;
3374 /* NOTE: empty args are allowed, except if no args */
3375 for(;;) {
3376 do {
3377 next_argstream(nested_list, NULL);
3378 } while (is_space(tok) || TOK_LINEFEED == tok);
3379 empty_arg:
3380 /* handle '()' case */
3381 if (!args && !sa && tok == ')')
3382 break;
3383 if (!sa)
3384 tcc_error("macro '%s' used with too many args",
3385 get_tok_str(s->v, 0));
3386 tok_str_new(&str);
3387 parlevel = spc = 0;
3388 /* NOTE: non zero sa->t indicates VA_ARGS */
3389 while ((parlevel > 0 ||
3390 (tok != ')' &&
3391 (tok != ',' || sa->type.t)))) {
3392 if (tok == TOK_EOF || tok == 0)
3393 break;
3394 if (tok == '(')
3395 parlevel++;
3396 else if (tok == ')')
3397 parlevel--;
3398 if (tok == TOK_LINEFEED)
3399 tok = ' ';
3400 if (!check_space(tok, &spc))
3401 tok_str_add2(&str, tok, &tokc);
3402 next_argstream(nested_list, NULL);
3403 }
3404 if (parlevel)
3405 expect(")");
3406 str.len -= spc;
3407 tok_str_add(&str, -1);
3408 tok_str_add(&str, 0);
3409 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3410 sa1->d = str.str;
3411 sa = sa->next;
3412 if (tok == ')') {
3413 /* special case for gcc var args: add an empty
3414 var arg argument if it is omitted */
3415 if (sa && sa->type.t && gnu_ext)
3416 goto empty_arg;
3417 break;
3418 }
3419 if (tok != ',')
3420 expect(",");
3421 }
3422 if (sa) {
3423 tcc_error("macro '%s' used with too few args",
3424 get_tok_str(s->v, 0));
3425 }
3426
3427 parse_flags = saved_parse_flags;
3428
3429 /* now subst each arg */
3430 mstr = macro_arg_subst(nested_list, mstr, args);
3431 /* free memory */
3432 sa = args;
3433 while (sa) {
3434 sa1 = sa->prev;
3435 tok_str_free_str(sa->d);
3436 if (sa->next) {
3437 tok_str_free_str(sa->next->d);
3438 sym_free(sa->next);
3439 }
3440 sym_free(sa);
3441 sa = sa1;
3442 }
3443 }
3444
3445 sym_push2(nested_list, s->v, 0, 0);
3446 parse_flags = saved_parse_flags;
3447 joined_str = macro_twosharps(mstr);
3448 macro_subst(tok_str, nested_list, joined_str ? joined_str : mstr);
3449
3450 /* pop nested defined symbol */
3451 sa1 = *nested_list;
3452 *nested_list = sa1->prev;
3453 sym_free(sa1);
3454 if (joined_str)
3455 tok_str_free_str(joined_str);
3456 if (mstr != s->d)
3457 tok_str_free_str(mstr);
3458 }
3459 return 0;
3460 }
3461
3462 /* do macro substitution of macro_str and add result to
3463 (tok_str,tok_len). 'nested_list' is the list of all macros we got
3464 inside to avoid recursing. */
3465 static void macro_subst(
3466 TokenString *tok_str,
3467 Sym **nested_list,
3468 const int *macro_str
3469 )
3470 {
3471 Sym *s;
3472 int t, spc, nosubst;
3473 CValue cval;
3474
3475 spc = nosubst = 0;
3476
3477 while (1) {
3478 TOK_GET(&t, &macro_str, &cval);
3479 if (t <= 0)
3480 break;
3481
3482 if (t >= TOK_IDENT && 0 == nosubst) {
3483 s = define_find(t);
3484 if (s == NULL)
3485 goto no_subst;
3486
3487 /* if nested substitution, do nothing */
3488 if (sym_find2(*nested_list, t)) {
3489 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3490 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3491 goto no_subst;
3492 }
3493
3494 {
3495 TokenString str;
3496 str.str = (int*)macro_str;
3497 begin_macro(&str, 2);
3498
3499 tok = t;
3500 macro_subst_tok(tok_str, nested_list, s);
3501
3502 if (str.alloc == 3) {
3503 /* already finished by reading function macro arguments */
3504 break;
3505 }
3506
3507 macro_str = macro_ptr;
3508 end_macro ();
3509 }
3510 if (tok_str->len)
3511 spc = is_space(t = tok_str->str[tok_str->lastlen]);
3512 } else {
3513 if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3514 tcc_error("stray '\\' in program");
3515 no_subst:
3516 if (!check_space(t, &spc))
3517 tok_str_add2(tok_str, t, &cval);
3518
3519 if (nosubst) {
3520 if (nosubst > 1 && (spc || (++nosubst == 3 && t == '(')))
3521 continue;
3522 nosubst = 0;
3523 }
3524 if (t == TOK_NOSUBST)
3525 nosubst = 1;
3526 }
3527 /* GCC supports 'defined' as result of a macro substitution */
3528 if (t == TOK_DEFINED && pp_expr)
3529 nosubst = 2;
3530 }
3531 }
3532
3533 /* return next token with macro substitution */
3534 ST_FUNC void next(void)
3535 {
3536 redo:
3537 if (parse_flags & PARSE_FLAG_SPACES)
3538 next_nomacro_spc();
3539 else
3540 next_nomacro();
3541
3542 if (macro_ptr) {
3543 if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3544 /* discard preprocessor markers */
3545 goto redo;
3546 } else if (tok == 0) {
3547 /* end of macro or unget token string */
3548 end_macro();
3549 goto redo;
3550 }
3551 } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3552 Sym *s;
3553 /* if reading from file, try to substitute macros */
3554 s = define_find(tok);
3555 if (s) {
3556 Sym *nested_list = NULL;
3557 tokstr_buf.len = 0;
3558 macro_subst_tok(&tokstr_buf, &nested_list, s);
3559 tok_str_add(&tokstr_buf, 0);
3560 begin_macro(&tokstr_buf, 2);
3561 goto redo;
3562 }
3563 }
3564 /* convert preprocessor tokens into C tokens */
3565 if (tok == TOK_PPNUM) {
3566 if (parse_flags & PARSE_FLAG_TOK_NUM)
3567 parse_number((char *)tokc.str.data);
3568 } else if (tok == TOK_PPSTR) {
3569 if (parse_flags & PARSE_FLAG_TOK_STR)
3570 parse_string((char *)tokc.str.data, tokc.str.size - 1);
3571 }
3572 }
3573
3574 /* push back current token and set current token to 'last_tok'. Only
3575 identifier case handled for labels. */
3576 ST_INLN void unget_tok(int last_tok)
3577 {
3578
3579 TokenString *str = tok_str_alloc();
3580 tok_str_add2(str, tok, &tokc);
3581 tok_str_add(str, 0);
3582 begin_macro(str, 1);
3583 tok = last_tok;
3584 }
3585
3586 ST_FUNC void preprocess_start(TCCState *s1, int is_asm)
3587 {
3588 char *buf;
3589
3590 s1->include_stack_ptr = s1->include_stack;
3591 s1->ifdef_stack_ptr = s1->ifdef_stack;
3592 file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3593 pp_expr = 0;
3594 pp_counter = 0;
3595 pp_debug_tok = pp_debug_symv = 0;
3596 pp_once++;
3597 pvtop = vtop = vstack - 1;
3598 s1->pack_stack[0] = 0;
3599 s1->pack_stack_ptr = s1->pack_stack;
3600
3601 set_idnum('$', s1->dollars_in_identifiers ? IS_ID : 0);
3602 set_idnum('.', is_asm ? IS_ID : 0);
3603
3604 buf = tcc_malloc(3 + strlen(file->filename));
3605 sprintf(buf, "\"%s\"", file->filename);
3606 tcc_define_symbol(s1, "__BASE_FILE__", buf);
3607 tcc_free(buf);
3608
3609 if (s1->nb_cmd_include_files) {
3610 CString cstr;
3611 int i;
3612 cstr_new(&cstr);
3613 for (i = 0; i < s1->nb_cmd_include_files; i++) {
3614 cstr_cat(&cstr, "#include \"", -1);
3615 cstr_cat(&cstr, s1->cmd_include_files[i], -1);
3616 cstr_cat(&cstr, "\"\n", -1);
3617 }
3618 *s1->include_stack_ptr++ = file;
3619 tcc_open_bf(s1, "<command line>", cstr.size);
3620 memcpy(file->buffer, cstr.data, cstr.size);
3621 cstr_free(&cstr);
3622 }
3623
3624 if (is_asm)
3625 tcc_define_symbol(s1, "__ASSEMBLER__", NULL);
3626
3627 parse_flags = is_asm ? PARSE_FLAG_ASM_FILE : 0;
3628 tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3629 }
3630
3631 /* cleanup from error/setjmp */
3632 ST_FUNC void preprocess_end(TCCState *s1)
3633 {
3634 while (macro_stack)
3635 end_macro();
3636 macro_ptr = NULL;
3637 }
3638
3639 ST_FUNC void tccpp_new(TCCState *s)
3640 {
3641 int i, c;
3642 const char *p, *r;
3643
3644 /* might be used in error() before preprocess_start() */
3645 s->include_stack_ptr = s->include_stack;
3646 s->ppfp = stdout;
3647
3648 /* init isid table */
3649 for(i = CH_EOF; i<128; i++)
3650 set_idnum(i,
3651 is_space(i) ? IS_SPC
3652 : isid(i) ? IS_ID
3653 : isnum(i) ? IS_NUM
3654 : 0);
3655
3656 for(i = 128; i<256; i++)
3657 set_idnum(i, IS_ID);
3658
3659 /* init allocators */
3660 tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE);
3661 tal_new(&tokstr_alloc, TOKSTR_TAL_LIMIT, TOKSTR_TAL_SIZE);
3662 tal_new(&cstr_alloc, CSTR_TAL_LIMIT, CSTR_TAL_SIZE);
3663
3664 memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3665 cstr_new(&cstr_buf);
3666 cstr_realloc(&cstr_buf, STRING_MAX_SIZE);
3667 tok_str_new(&tokstr_buf);
3668 tok_str_realloc(&tokstr_buf, TOKSTR_MAX_SIZE);
3669
3670 tok_ident = TOK_IDENT;
3671 p = tcc_keywords;
3672 while (*p) {
3673 r = p;
3674 for(;;) {
3675 c = *r++;
3676 if (c == '\0')
3677 break;
3678 }
3679 tok_alloc(p, r - p - 1);
3680 p = r;
3681 }
3682 }
3683
3684 ST_FUNC void tccpp_delete(TCCState *s)
3685 {
3686 int i, n;
3687
3688 /* free -D and compiler defines */
3689 free_defines(NULL);
3690
3691 /* free tokens */
3692 n = tok_ident - TOK_IDENT;
3693 for(i = 0; i < n; i++)
3694 tal_free(toksym_alloc, table_ident[i]);
3695 tcc_free(table_ident);
3696 table_ident = NULL;
3697
3698 /* free static buffers */
3699 cstr_free(&tokcstr);
3700 cstr_free(&cstr_buf);
3701 cstr_free(&macro_equal_buf);
3702 tok_str_free_str(tokstr_buf.str);
3703
3704 /* free allocators */
3705 tal_delete(toksym_alloc);
3706 toksym_alloc = NULL;
3707 tal_delete(tokstr_alloc);
3708 tokstr_alloc = NULL;
3709 tal_delete(cstr_alloc);
3710 cstr_alloc = NULL;
3711 }
3712
3713 /* ------------------------------------------------------------------------- */
3714 /* tcc -E [-P[1]] [-dD} support */
3715
3716 static void tok_print(const char *msg, const int *str)
3717 {
3718 FILE *fp;
3719 int t, s = 0;
3720 CValue cval;
3721
3722 fp = tcc_state->ppfp;
3723 fprintf(fp, "%s", msg);
3724 while (str) {
3725 TOK_GET(&t, &str, &cval);
3726 if (!t)
3727 break;
3728 fprintf(fp, " %s" + s, get_tok_str(t, &cval)), s = 1;
3729 }
3730 fprintf(fp, "\n");
3731 }
3732
3733 static void pp_line(TCCState *s1, BufferedFile *f, int level)
3734 {
3735 int d = f->line_num - f->line_ref;
3736
3737 if (s1->dflag & 4)
3738 return;
3739
3740 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE) {
3741 ;
3742 } else if (level == 0 && f->line_ref && d < 8) {
3743 while (d > 0)
3744 fputs("\n", s1->ppfp), --d;
3745 } else if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
3746 fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
3747 } else {
3748 fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
3749 level > 0 ? " 1" : level < 0 ? " 2" : "");
3750 }
3751 f->line_ref = f->line_num;
3752 }
3753
3754 static void define_print(TCCState *s1, int v)
3755 {
3756 FILE *fp;
3757 Sym *s;
3758
3759 s = define_find(v);
3760 if (NULL == s || NULL == s->d)
3761 return;
3762
3763 fp = s1->ppfp;
3764 fprintf(fp, "#define %s", get_tok_str(v, NULL));
3765 if (s->type.t == MACRO_FUNC) {
3766 Sym *a = s->next;
3767 fprintf(fp,"(");
3768 if (a)
3769 for (;;) {
3770 fprintf(fp,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
3771 if (!(a = a->next))
3772 break;
3773 fprintf(fp,",");
3774 }
3775 fprintf(fp,")");
3776 }
3777 tok_print("", s->d);
3778 }
3779
3780 static void pp_debug_defines(TCCState *s1)
3781 {
3782 int v, t;
3783 const char *vs;
3784 FILE *fp;
3785
3786 t = pp_debug_tok;
3787 if (t == 0)
3788 return;
3789
3790 file->line_num--;
3791 pp_line(s1, file, 0);
3792 file->line_ref = ++file->line_num;
3793
3794 fp = s1->ppfp;
3795 v = pp_debug_symv;
3796 vs = get_tok_str(v, NULL);
3797 if (t == TOK_DEFINE) {
3798 define_print(s1, v);
3799 } else if (t == TOK_UNDEF) {
3800 fprintf(fp, "#undef %s\n", vs);
3801 } else if (t == TOK_push_macro) {
3802 fprintf(fp, "#pragma push_macro(\"%s\")\n", vs);
3803 } else if (t == TOK_pop_macro) {
3804 fprintf(fp, "#pragma pop_macro(\"%s\")\n", vs);
3805 }
3806 pp_debug_tok = 0;
3807 }
3808
3809 static void pp_debug_builtins(TCCState *s1)
3810 {
3811 int v;
3812 for (v = TOK_IDENT; v < tok_ident; ++v)
3813 define_print(s1, v);
3814 }
3815
3816 /* Add a space between tokens a and b to avoid unwanted textual pasting */
3817 static int pp_need_space(int a, int b)
3818 {
3819 return 'E' == a ? '+' == b || '-' == b
3820 : '+' == a ? TOK_INC == b || '+' == b
3821 : '-' == a ? TOK_DEC == b || '-' == b
3822 : a >= TOK_IDENT ? b >= TOK_IDENT
3823 : a == TOK_PPNUM ? b >= TOK_IDENT
3824 : 0;
3825 }
3826
3827 /* maybe hex like 0x1e */
3828 static int pp_check_he0xE(int t, const char *p)
3829 {
3830 if (t == TOK_PPNUM && toup(strchr(p, 0)[-1]) == 'E')
3831 return 'E';
3832 return t;
3833 }
3834
3835 /* Preprocess the current file */
3836 ST_FUNC int tcc_preprocess(TCCState *s1)
3837 {
3838 BufferedFile **iptr;
3839 int token_seen, spcs, level;
3840 const char *p;
3841 char white[400];
3842
3843 parse_flags = PARSE_FLAG_PREPROCESS
3844 | (parse_flags & PARSE_FLAG_ASM_FILE)
3845 | PARSE_FLAG_LINEFEED
3846 | PARSE_FLAG_SPACES
3847 | PARSE_FLAG_ACCEPT_STRAYS
3848 ;
3849 /* Credits to Fabrice Bellard's initial revision to demonstrate its
3850 capability to compile and run itself, provided all numbers are
3851 given as decimals. tcc -E -P10 will do. */
3852 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_P10)
3853 parse_flags |= PARSE_FLAG_TOK_NUM, s1->Pflag = 1;
3854
3855 #ifdef PP_BENCH
3856 /* for PP benchmarks */
3857 do next(); while (tok != TOK_EOF);
3858 return 0;
3859 #endif
3860
3861 if (s1->dflag & 1) {
3862 pp_debug_builtins(s1);
3863 s1->dflag &= ~1;
3864 }
3865
3866 token_seen = TOK_LINEFEED, spcs = 0;
3867 pp_line(s1, file, 0);
3868 for (;;) {
3869 iptr = s1->include_stack_ptr;
3870 next();
3871 if (tok == TOK_EOF)
3872 break;
3873
3874 level = s1->include_stack_ptr - iptr;
3875 if (level) {
3876 if (level > 0)
3877 pp_line(s1, *iptr, 0);
3878 pp_line(s1, file, level);
3879 }
3880 if (s1->dflag & 7) {
3881 pp_debug_defines(s1);
3882 if (s1->dflag & 4)
3883 continue;
3884 }
3885
3886 if (is_space(tok)) {
3887 if (spcs < sizeof white - 1)
3888 white[spcs++] = tok;
3889 continue;
3890 } else if (tok == TOK_LINEFEED) {
3891 spcs = 0;
3892 if (token_seen == TOK_LINEFEED)
3893 continue;
3894 ++file->line_ref;
3895 } else if (token_seen == TOK_LINEFEED) {
3896 pp_line(s1, file, 0);
3897 } else if (spcs == 0 && pp_need_space(token_seen, tok)) {
3898 white[spcs++] = ' ';
3899 }
3900
3901 white[spcs] = 0, fputs(white, s1->ppfp), spcs = 0;
3902 fputs(p = get_tok_str(tok, &tokc), s1->ppfp);
3903 token_seen = pp_check_he0xE(tok, p);
3904 }
3905 return 0;
3906 }
3907
3908 /* ------------------------------------------------------------------------- */
Tiny C Compiler main repository