?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tccasm: synch C and asm symtab tighter
[tinycc.git] / tccpp.c
blobd8e7f53169d5439664302a9a0a5cf36af2b270f5
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->sym_asm_label = NULL;
436 ts->len = len;
437 ts->hash_next = NULL;
438 memcpy(ts->str, str, len);
439 ts->str[len] = '\0';
440 *pts = ts;
441 return ts;
442 }
443
444 #define TOK_HASH_INIT 1
445 #define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c))
446
447
448 /* find a token and add it if not found */
449 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
450 {
451 TokenSym *ts, **pts;
452 int i;
453 unsigned int h;
454
455 h = TOK_HASH_INIT;
456 for(i=0;i<len;i++)
457 h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
458 h &= (TOK_HASH_SIZE - 1);
459
460 pts = &hash_ident[h];
461 for(;;) {
462 ts = *pts;
463 if (!ts)
464 break;
465 if (ts->len == len && !memcmp(ts->str, str, len))
466 return ts;
467 pts = &(ts->hash_next);
468 }
469 return tok_alloc_new(pts, str, len);
470 }
471
472 /* XXX: buffer overflow */
473 /* XXX: float tokens */
474 ST_FUNC const char *get_tok_str(int v, CValue *cv)
475 {
476 char *p;
477 int i, len;
478
479 cstr_reset(&cstr_buf);
480 p = cstr_buf.data;
481
482 switch(v) {
483 case TOK_CINT:
484 case TOK_CUINT:
485 case TOK_CLONG:
486 case TOK_CULONG:
487 case TOK_CLLONG:
488 case TOK_CULLONG:
489 /* XXX: not quite exact, but only useful for testing */
490 #ifdef _WIN32
491 sprintf(p, "%u", (unsigned)cv->i);
492 #else
493 sprintf(p, "%llu", (unsigned long long)cv->i);
494 #endif
495 break;
496 case TOK_LCHAR:
497 cstr_ccat(&cstr_buf, 'L');
498 case TOK_CCHAR:
499 cstr_ccat(&cstr_buf, '\'');
500 add_char(&cstr_buf, cv->i);
501 cstr_ccat(&cstr_buf, '\'');
502 cstr_ccat(&cstr_buf, '\0');
503 break;
504 case TOK_PPNUM:
505 case TOK_PPSTR:
506 return (char*)cv->str.data;
507 case TOK_LSTR:
508 cstr_ccat(&cstr_buf, 'L');
509 case TOK_STR:
510 cstr_ccat(&cstr_buf, '\"');
511 if (v == TOK_STR) {
512 len = cv->str.size - 1;
513 for(i=0;i<len;i++)
514 add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]);
515 } else {
516 len = (cv->str.size / sizeof(nwchar_t)) - 1;
517 for(i=0;i<len;i++)
518 add_char(&cstr_buf, ((nwchar_t *)cv->str.data)[i]);
519 }
520 cstr_ccat(&cstr_buf, '\"');
521 cstr_ccat(&cstr_buf, '\0');
522 break;
523
524 case TOK_CFLOAT:
525 cstr_cat(&cstr_buf, "<float>", 0);
526 break;
527 case TOK_CDOUBLE:
528 cstr_cat(&cstr_buf, "<double>", 0);
529 break;
530 case TOK_CLDOUBLE:
531 cstr_cat(&cstr_buf, "<long double>", 0);
532 break;
533 case TOK_LINENUM:
534 cstr_cat(&cstr_buf, "<linenumber>", 0);
535 break;
536
537 /* above tokens have value, the ones below don't */
538 case TOK_LT:
539 v = '<';
540 goto addv;
541 case TOK_GT:
542 v = '>';
543 goto addv;
544 case TOK_DOTS:
545 return strcpy(p, "...");
546 case TOK_A_SHL:
547 return strcpy(p, "<<=");
548 case TOK_A_SAR:
549 return strcpy(p, ">>=");
550 case TOK_EOF:
551 return strcpy(p, "<eof>");
552 default:
553 if (v < TOK_IDENT) {
554 /* search in two bytes table */
555 const unsigned char *q = tok_two_chars;
556 while (*q) {
557 if (q[2] == v) {
558 *p++ = q[0];
559 *p++ = q[1];
560 *p = '\0';
561 return cstr_buf.data;
562 }
563 q += 3;
564 }
565 if (v >= 127) {
566 sprintf(cstr_buf.data, "<%02x>", v);
567 return cstr_buf.data;
568 }
569 addv:
570 *p++ = v;
571 *p = '\0';
572 } else if (v < tok_ident) {
573 return table_ident[v - TOK_IDENT]->str;
574 } else if (v >= SYM_FIRST_ANOM) {
575 /* special name for anonymous symbol */
576 sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
577 } else {
578 /* should never happen */
579 return NULL;
580 }
581 break;
582 }
583 return cstr_buf.data;
584 }
585
586 /* return the current character, handling end of block if necessary
587 (but not stray) */
588 ST_FUNC int handle_eob(void)
589 {
590 BufferedFile *bf = file;
591 int len;
592
593 /* only tries to read if really end of buffer */
594 if (bf->buf_ptr >= bf->buf_end) {
595 if (bf->fd >= 0) {
596 #if defined(PARSE_DEBUG)
597 len = 1;
598 #else
599 len = IO_BUF_SIZE;
600 #endif
601 len = read(bf->fd, bf->buffer, len);
602 if (len < 0)
603 len = 0;
604 } else {
605 len = 0;
606 }
607 total_bytes += len;
608 bf->buf_ptr = bf->buffer;
609 bf->buf_end = bf->buffer + len;
610 *bf->buf_end = CH_EOB;
611 }
612 if (bf->buf_ptr < bf->buf_end) {
613 return bf->buf_ptr[0];
614 } else {
615 bf->buf_ptr = bf->buf_end;
616 return CH_EOF;
617 }
618 }
619
620 /* read next char from current input file and handle end of input buffer */
621 ST_INLN void inp(void)
622 {
623 ch = *(++(file->buf_ptr));
624 /* end of buffer/file handling */
625 if (ch == CH_EOB)
626 ch = handle_eob();
627 }
628
629 /* handle '\[\r]\n' */
630 static int handle_stray_noerror(void)
631 {
632 while (ch == '\\') {
633 inp();
634 if (ch == '\n') {
635 file->line_num++;
636 inp();
637 } else if (ch == '\r') {
638 inp();
639 if (ch != '\n')
640 goto fail;
641 file->line_num++;
642 inp();
643 } else {
644 fail:
645 return 1;
646 }
647 }
648 return 0;
649 }
650
651 static void handle_stray(void)
652 {
653 if (handle_stray_noerror())
654 tcc_error("stray '\\' in program");
655 }
656
657 /* skip the stray and handle the \\n case. Output an error if
658 incorrect char after the stray */
659 static int handle_stray1(uint8_t *p)
660 {
661 int c;
662
663 file->buf_ptr = p;
664 if (p >= file->buf_end) {
665 c = handle_eob();
666 if (c != '\\')
667 return c;
668 p = file->buf_ptr;
669 }
670 ch = *p;
671 if (handle_stray_noerror()) {
672 if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
673 tcc_error("stray '\\' in program");
674 *--file->buf_ptr = '\\';
675 }
676 p = file->buf_ptr;
677 c = *p;
678 return c;
679 }
680
681 /* handle just the EOB case, but not stray */
682 #define PEEKC_EOB(c, p)\
683 {\
684 p++;\
685 c = *p;\
686 if (c == '\\') {\
687 file->buf_ptr = p;\
688 c = handle_eob();\
689 p = file->buf_ptr;\
690 }\
691 }
692
693 /* handle the complicated stray case */
694 #define PEEKC(c, p)\
695 {\
696 p++;\
697 c = *p;\
698 if (c == '\\') {\
699 c = handle_stray1(p);\
700 p = file->buf_ptr;\
701 }\
702 }
703
704 /* input with '\[\r]\n' handling. Note that this function cannot
705 handle other characters after '\', so you cannot call it inside
706 strings or comments */
707 ST_FUNC void minp(void)
708 {
709 inp();
710 if (ch == '\\')
711 handle_stray();
712 }
713
714 /* single line C++ comments */
715 static uint8_t *parse_line_comment(uint8_t *p)
716 {
717 int c;
718
719 p++;
720 for(;;) {
721 c = *p;
722 redo:
723 if (c == '\n' || c == CH_EOF) {
724 break;
725 } else if (c == '\\') {
726 file->buf_ptr = p;
727 c = handle_eob();
728 p = file->buf_ptr;
729 if (c == '\\') {
730 PEEKC_EOB(c, p);
731 if (c == '\n') {
732 file->line_num++;
733 PEEKC_EOB(c, p);
734 } else if (c == '\r') {
735 PEEKC_EOB(c, p);
736 if (c == '\n') {
737 file->line_num++;
738 PEEKC_EOB(c, p);
739 }
740 }
741 } else {
742 goto redo;
743 }
744 } else {
745 p++;
746 }
747 }
748 return p;
749 }
750
751 /* C comments */
752 ST_FUNC uint8_t *parse_comment(uint8_t *p)
753 {
754 int c;
755
756 p++;
757 for(;;) {
758 /* fast skip loop */
759 for(;;) {
760 c = *p;
761 if (c == '\n' || c == '*' || c == '\\')
762 break;
763 p++;
764 c = *p;
765 if (c == '\n' || c == '*' || c == '\\')
766 break;
767 p++;
768 }
769 /* now we can handle all the cases */
770 if (c == '\n') {
771 file->line_num++;
772 p++;
773 } else if (c == '*') {
774 p++;
775 for(;;) {
776 c = *p;
777 if (c == '*') {
778 p++;
779 } else if (c == '/') {
780 goto end_of_comment;
781 } else if (c == '\\') {
782 file->buf_ptr = p;
783 c = handle_eob();
784 p = file->buf_ptr;
785 if (c == CH_EOF)
786 tcc_error("unexpected end of file in comment");
787 if (c == '\\') {
788 /* skip '\[\r]\n', otherwise just skip the stray */
789 while (c == '\\') {
790 PEEKC_EOB(c, p);
791 if (c == '\n') {
792 file->line_num++;
793 PEEKC_EOB(c, p);
794 } else if (c == '\r') {
795 PEEKC_EOB(c, p);
796 if (c == '\n') {
797 file->line_num++;
798 PEEKC_EOB(c, p);
799 }
800 } else {
801 goto after_star;
802 }
803 }
804 }
805 } else {
806 break;
807 }
808 }
809 after_star: ;
810 } else {
811 /* stray, eob or eof */
812 file->buf_ptr = p;
813 c = handle_eob();
814 p = file->buf_ptr;
815 if (c == CH_EOF) {
816 tcc_error("unexpected end of file in comment");
817 } else if (c == '\\') {
818 p++;
819 }
820 }
821 }
822 end_of_comment:
823 p++;
824 return p;
825 }
826
827 ST_FUNC int set_idnum(int c, int val)
828 {
829 int prev = isidnum_table[c - CH_EOF];
830 isidnum_table[c - CH_EOF] = val;
831 return prev;
832 }
833
834 #define cinp minp
835
836 static inline void skip_spaces(void)
837 {
838 while (isidnum_table[ch - CH_EOF] & IS_SPC)
839 cinp();
840 }
841
842 static inline int check_space(int t, int *spc)
843 {
844 if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
845 if (*spc)
846 return 1;
847 *spc = 1;
848 } else
849 *spc = 0;
850 return 0;
851 }
852
853 /* parse a string without interpreting escapes */
854 static uint8_t *parse_pp_string(uint8_t *p,
855 int sep, CString *str)
856 {
857 int c;
858 p++;
859 for(;;) {
860 c = *p;
861 if (c == sep) {
862 break;
863 } else if (c == '\\') {
864 file->buf_ptr = p;
865 c = handle_eob();
866 p = file->buf_ptr;
867 if (c == CH_EOF) {
868 unterminated_string:
869 /* XXX: indicate line number of start of string */
870 tcc_error("missing terminating %c character", sep);
871 } else if (c == '\\') {
872 /* escape : just skip \[\r]\n */
873 PEEKC_EOB(c, p);
874 if (c == '\n') {
875 file->line_num++;
876 p++;
877 } else if (c == '\r') {
878 PEEKC_EOB(c, p);
879 if (c != '\n')
880 expect("'\n' after '\r'");
881 file->line_num++;
882 p++;
883 } else if (c == CH_EOF) {
884 goto unterminated_string;
885 } else {
886 if (str) {
887 cstr_ccat(str, '\\');
888 cstr_ccat(str, c);
889 }
890 p++;
891 }
892 }
893 } else if (c == '\n') {
894 file->line_num++;
895 goto add_char;
896 } else if (c == '\r') {
897 PEEKC_EOB(c, p);
898 if (c != '\n') {
899 if (str)
900 cstr_ccat(str, '\r');
901 } else {
902 file->line_num++;
903 goto add_char;
904 }
905 } else {
906 add_char:
907 if (str)
908 cstr_ccat(str, c);
909 p++;
910 }
911 }
912 p++;
913 return p;
914 }
915
916 /* skip block of text until #else, #elif or #endif. skip also pairs of
917 #if/#endif */
918 static void preprocess_skip(void)
919 {
920 int a, start_of_line, c, in_warn_or_error;
921 uint8_t *p;
922
923 p = file->buf_ptr;
924 a = 0;
925 redo_start:
926 start_of_line = 1;
927 in_warn_or_error = 0;
928 for(;;) {
929 redo_no_start:
930 c = *p;
931 switch(c) {
932 case ' ':
933 case '\t':
934 case '\f':
935 case '\v':
936 case '\r':
937 p++;
938 goto redo_no_start;
939 case '\n':
940 file->line_num++;
941 p++;
942 goto redo_start;
943 case '\\':
944 file->buf_ptr = p;
945 c = handle_eob();
946 if (c == CH_EOF) {
947 expect("#endif");
948 } else if (c == '\\') {
949 ch = file->buf_ptr[0];
950 handle_stray_noerror();
951 }
952 p = file->buf_ptr;
953 goto redo_no_start;
954 /* skip strings */
955 case '\"':
956 case '\'':
957 if (in_warn_or_error)
958 goto _default;
959 p = parse_pp_string(p, c, NULL);
960 break;
961 /* skip comments */
962 case '/':
963 if (in_warn_or_error)
964 goto _default;
965 file->buf_ptr = p;
966 ch = *p;
967 minp();
968 p = file->buf_ptr;
969 if (ch == '*') {
970 p = parse_comment(p);
971 } else if (ch == '/') {
972 p = parse_line_comment(p);
973 }
974 break;
975 case '#':
976 p++;
977 if (start_of_line) {
978 file->buf_ptr = p;
979 next_nomacro();
980 p = file->buf_ptr;
981 if (a == 0 &&
982 (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
983 goto the_end;
984 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
985 a++;
986 else if (tok == TOK_ENDIF)
987 a--;
988 else if( tok == TOK_ERROR || tok == TOK_WARNING)
989 in_warn_or_error = 1;
990 else if (tok == TOK_LINEFEED)
991 goto redo_start;
992 else if (parse_flags & PARSE_FLAG_ASM_FILE)
993 p = parse_line_comment(p - 1);
994 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
995 p = parse_line_comment(p - 1);
996 break;
997 _default:
998 default:
999 p++;
1000 break;
1001 }
1002 start_of_line = 0;
1003 }
1004 the_end: ;
1005 file->buf_ptr = p;
1006 }
1007
1008 #if 0
1009 /* return the number of additional 'ints' necessary to store the
1010 token */
1011 static inline int tok_size(const int *p)
1012 {
1013 switch(*p) {
1014 /* 4 bytes */
1015 case TOK_CINT:
1016 case TOK_CUINT:
1017 case TOK_CCHAR:
1018 case TOK_LCHAR:
1019 case TOK_CFLOAT:
1020 case TOK_LINENUM:
1021 return 1 + 1;
1022 case TOK_STR:
1023 case TOK_LSTR:
1024 case TOK_PPNUM:
1025 case TOK_PPSTR:
1026 return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
1027 case TOK_CLONG:
1028 case TOK_CULONG:
1029 return 1 + LONG_SIZE / 4;
1030 case TOK_CDOUBLE:
1031 case TOK_CLLONG:
1032 case TOK_CULLONG:
1033 return 1 + 2;
1034 case TOK_CLDOUBLE:
1035 return 1 + LDOUBLE_SIZE / 4;
1036 default:
1037 return 1 + 0;
1038 }
1039 }
1040 #endif
1041
1042 /* token string handling */
1043 ST_INLN void tok_str_new(TokenString *s)
1044 {
1045 s->str = NULL;
1046 s->len = s->lastlen = 0;
1047 s->allocated_len = 0;
1048 s->last_line_num = -1;
1049 }
1050
1051 ST_FUNC TokenString *tok_str_alloc(void)
1052 {
1053 TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str);
1054 tok_str_new(str);
1055 return str;
1056 }
1057
1058 ST_FUNC int *tok_str_dup(TokenString *s)
1059 {
1060 int *str;
1061
1062 str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int));
1063 memcpy(str, s->str, s->len * sizeof(int));
1064 return str;
1065 }
1066
1067 ST_FUNC void tok_str_free_str(int *str)
1068 {
1069 tal_free(tokstr_alloc, str);
1070 }
1071
1072 ST_FUNC void tok_str_free(TokenString *str)
1073 {
1074 tok_str_free_str(str->str);
1075 tal_free(tokstr_alloc, str);
1076 }
1077
1078 ST_FUNC int *tok_str_realloc(TokenString *s, int new_size)
1079 {
1080 int *str, size;
1081
1082 size = s->allocated_len;
1083 if (size < 16)
1084 size = 16;
1085 while (size < new_size)
1086 size = size * 2;
1087 if (size > s->allocated_len) {
1088 str = tal_realloc(tokstr_alloc, s->str, size * sizeof(int));
1089 s->allocated_len = size;
1090 s->str = str;
1091 }
1092 return s->str;
1093 }
1094
1095 ST_FUNC void tok_str_add(TokenString *s, int t)
1096 {
1097 int len, *str;
1098
1099 len = s->len;
1100 str = s->str;
1101 if (len >= s->allocated_len)
1102 str = tok_str_realloc(s, len + 1);
1103 str[len++] = t;
1104 s->len = len;
1105 }
1106
1107 ST_FUNC void begin_macro(TokenString *str, int alloc)
1108 {
1109 str->alloc = alloc;
1110 str->prev = macro_stack;
1111 str->prev_ptr = macro_ptr;
1112 str->save_line_num = file->line_num;
1113 macro_ptr = str->str;
1114 macro_stack = str;
1115 }
1116
1117 ST_FUNC void end_macro(void)
1118 {
1119 TokenString *str = macro_stack;
1120 macro_stack = str->prev;
1121 macro_ptr = str->prev_ptr;
1122 file->line_num = str->save_line_num;
1123 if (str->alloc == 2) {
1124 str->alloc = 3; /* just mark as finished */
1125 } else {
1126 tok_str_free(str);
1127 }
1128 }
1129
1130 static void tok_str_add2(TokenString *s, int t, CValue *cv)
1131 {
1132 int len, *str;
1133
1134 len = s->lastlen = s->len;
1135 str = s->str;
1136
1137 /* allocate space for worst case */
1138 if (len + TOK_MAX_SIZE >= s->allocated_len)
1139 str = tok_str_realloc(s, len + TOK_MAX_SIZE + 1);
1140 str[len++] = t;
1141 switch(t) {
1142 case TOK_CINT:
1143 case TOK_CUINT:
1144 case TOK_CCHAR:
1145 case TOK_LCHAR:
1146 case TOK_CFLOAT:
1147 case TOK_LINENUM:
1148 #if LONG_SIZE == 4
1149 case TOK_CLONG:
1150 case TOK_CULONG:
1151 #endif
1152 str[len++] = cv->tab[0];
1153 break;
1154 case TOK_PPNUM:
1155 case TOK_PPSTR:
1156 case TOK_STR:
1157 case TOK_LSTR:
1158 {
1159 /* Insert the string into the int array. */
1160 size_t nb_words =
1161 1 + (cv->str.size + sizeof(int) - 1) / sizeof(int);
1162 if (len + nb_words >= s->allocated_len)
1163 str = tok_str_realloc(s, len + nb_words + 1);
1164 str[len] = cv->str.size;
1165 memcpy(&str[len + 1], cv->str.data, cv->str.size);
1166 len += nb_words;
1167 }
1168 break;
1169 case TOK_CDOUBLE:
1170 case TOK_CLLONG:
1171 case TOK_CULLONG:
1172 #if LONG_SIZE == 8
1173 case TOK_CLONG:
1174 case TOK_CULONG:
1175 #endif
1176 #if LDOUBLE_SIZE == 8
1177 case TOK_CLDOUBLE:
1178 #endif
1179 str[len++] = cv->tab[0];
1180 str[len++] = cv->tab[1];
1181 break;
1182 #if LDOUBLE_SIZE == 12
1183 case TOK_CLDOUBLE:
1184 str[len++] = cv->tab[0];
1185 str[len++] = cv->tab[1];
1186 str[len++] = cv->tab[2];
1187 #elif LDOUBLE_SIZE == 16
1188 case TOK_CLDOUBLE:
1189 str[len++] = cv->tab[0];
1190 str[len++] = cv->tab[1];
1191 str[len++] = cv->tab[2];
1192 str[len++] = cv->tab[3];
1193 #elif LDOUBLE_SIZE != 8
1194 #error add long double size support
1195 #endif
1196 break;
1197 default:
1198 break;
1199 }
1200 s->len = len;
1201 }
1202
1203 /* add the current parse token in token string 's' */
1204 ST_FUNC void tok_str_add_tok(TokenString *s)
1205 {
1206 CValue cval;
1207
1208 /* save line number info */
1209 if (file->line_num != s->last_line_num) {
1210 s->last_line_num = file->line_num;
1211 cval.i = s->last_line_num;
1212 tok_str_add2(s, TOK_LINENUM, &cval);
1213 }
1214 tok_str_add2(s, tok, &tokc);
1215 }
1216
1217 /* get a token from an integer array and increment pointer
1218 accordingly. we code it as a macro to avoid pointer aliasing. */
1219 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
1220 {
1221 const int *p = *pp;
1222 int n, *tab;
1223
1224 tab = cv->tab;
1225 switch(*t = *p++) {
1226 #if LONG_SIZE == 4
1227 case TOK_CLONG:
1228 #endif
1229 case TOK_CINT:
1230 case TOK_CCHAR:
1231 case TOK_LCHAR:
1232 case TOK_LINENUM:
1233 cv->i = *p++;
1234 break;
1235 #if LONG_SIZE == 4
1236 case TOK_CULONG:
1237 #endif
1238 case TOK_CUINT:
1239 cv->i = (unsigned)*p++;
1240 break;
1241 case TOK_CFLOAT:
1242 tab[0] = *p++;
1243 break;
1244 case TOK_STR:
1245 case TOK_LSTR:
1246 case TOK_PPNUM:
1247 case TOK_PPSTR:
1248 cv->str.size = *p++;
1249 cv->str.data = p;
1250 p += (cv->str.size + sizeof(int) - 1) / sizeof(int);
1251 break;
1252 case TOK_CDOUBLE:
1253 case TOK_CLLONG:
1254 case TOK_CULLONG:
1255 #if LONG_SIZE == 8
1256 case TOK_CLONG:
1257 case TOK_CULONG:
1258 #endif
1259 n = 2;
1260 goto copy;
1261 case TOK_CLDOUBLE:
1262 #if LDOUBLE_SIZE == 16
1263 n = 4;
1264 #elif LDOUBLE_SIZE == 12
1265 n = 3;
1266 #elif LDOUBLE_SIZE == 8
1267 n = 2;
1268 #else
1269 # error add long double size support
1270 #endif
1271 copy:
1272 do
1273 *tab++ = *p++;
1274 while (--n);
1275 break;
1276 default:
1277 break;
1278 }
1279 *pp = p;
1280 }
1281
1282 static int macro_is_equal(const int *a, const int *b)
1283 {
1284 CValue cv;
1285 int t;
1286
1287 if (!a || !b)
1288 return 1;
1289
1290 while (*a && *b) {
1291 /* first time preallocate macro_equal_buf, next time only reset position to start */
1292 cstr_reset(&macro_equal_buf);
1293 TOK_GET(&t, &a, &cv);
1294 cstr_cat(&macro_equal_buf, get_tok_str(t, &cv), 0);
1295 TOK_GET(&t, &b, &cv);
1296 if (strcmp(macro_equal_buf.data, get_tok_str(t, &cv)))
1297 return 0;
1298 }
1299 return !(*a || *b);
1300 }
1301
1302 /* defines handling */
1303 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1304 {
1305 Sym *s, *o;
1306
1307 o = define_find(v);
1308 s = sym_push2(&define_stack, v, macro_type, 0);
1309 s->d = str;
1310 s->next = first_arg;
1311 table_ident[v - TOK_IDENT]->sym_define = s;
1312
1313 if (o && !macro_is_equal(o->d, s->d))
1314 tcc_warning("%s redefined", get_tok_str(v, NULL));
1315 }
1316
1317 /* undefined a define symbol. Its name is just set to zero */
1318 ST_FUNC void define_undef(Sym *s)
1319 {
1320 int v = s->v;
1321 if (v >= TOK_IDENT && v < tok_ident)
1322 table_ident[v - TOK_IDENT]->sym_define = NULL;
1323 }
1324
1325 ST_INLN Sym *define_find(int v)
1326 {
1327 v -= TOK_IDENT;
1328 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1329 return NULL;
1330 return table_ident[v]->sym_define;
1331 }
1332
1333 /* free define stack until top reaches 'b' */
1334 ST_FUNC void free_defines(Sym *b)
1335 {
1336 while (define_stack != b) {
1337 Sym *top = define_stack;
1338 define_stack = top->prev;
1339 tok_str_free_str(top->d);
1340 define_undef(top);
1341 sym_free(top);
1342 }
1343
1344 /* restore remaining (-D or predefined) symbols if they were
1345 #undef'd in the file */
1346 while (b) {
1347 int v = b->v;
1348 if (v >= TOK_IDENT && v < tok_ident) {
1349 Sym **d = &table_ident[v - TOK_IDENT]->sym_define;
1350 if (!*d)
1351 *d = b;
1352 }
1353 b = b->prev;
1354 }
1355 }
1356
1357 /* label lookup */
1358 ST_FUNC Sym *label_find(int v)
1359 {
1360 v -= TOK_IDENT;
1361 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1362 return NULL;
1363 return table_ident[v]->sym_label;
1364 }
1365
1366 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1367 {
1368 Sym *s, **ps;
1369 s = sym_push2(ptop, v, 0, 0);
1370 s->r = flags;
1371 ps = &table_ident[v - TOK_IDENT]->sym_label;
1372 if (ptop == &global_label_stack) {
1373 /* modify the top most local identifier, so that
1374 sym_identifier will point to 's' when popped */
1375 while (*ps != NULL)
1376 ps = &(*ps)->prev_tok;
1377 }
1378 s->prev_tok = *ps;
1379 *ps = s;
1380 return s;
1381 }
1382
1383 /* pop labels until element last is reached. Look if any labels are
1384 undefined. Define symbols if '&&label' was used. */
1385 ST_FUNC void label_pop(Sym **ptop, Sym *slast, int keep)
1386 {
1387 Sym *s, *s1;
1388 for(s = *ptop; s != slast; s = s1) {
1389 s1 = s->prev;
1390 if (s->r == LABEL_DECLARED) {
1391 tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1392 } else if (s->r == LABEL_FORWARD) {
1393 tcc_error("label '%s' used but not defined",
1394 get_tok_str(s->v, NULL));
1395 } else {
1396 if (s->c) {
1397 /* define corresponding symbol. A size of
1398 1 is put. */
1399 put_extern_sym(s, cur_text_section, s->jnext, 1);
1400 }
1401 }
1402 /* remove label */
1403 table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1404 if (!keep)
1405 sym_free(s);
1406 }
1407 if (!keep)
1408 *ptop = slast;
1409 }
1410
1411 /* fake the nth "#if defined test_..." for tcc -dt -run */
1412 static void maybe_run_test(TCCState *s)
1413 {
1414 const char *p;
1415 if (s->include_stack_ptr != s->include_stack)
1416 return;
1417 p = get_tok_str(tok, NULL);
1418 if (0 != memcmp(p, "test_", 5))
1419 return;
1420 if (0 != --s->run_test)
1421 return;
1422 fprintf(s->ppfp, "\n[%s]\n" + !(s->dflag & 32), p), fflush(s->ppfp);
1423 define_push(tok, MACRO_OBJ, NULL, NULL);
1424 }
1425
1426 /* eval an expression for #if/#elif */
1427 static int expr_preprocess(void)
1428 {
1429 int c, t;
1430 TokenString *str;
1431
1432 str = tok_str_alloc();
1433 pp_expr = 1;
1434 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1435 next(); /* do macro subst */
1436 if (tok == TOK_DEFINED) {
1437 next_nomacro();
1438 t = tok;
1439 if (t == '(')
1440 next_nomacro();
1441 if (tok < TOK_IDENT)
1442 expect("identifier");
1443 if (tcc_state->run_test)
1444 maybe_run_test(tcc_state);
1445 c = define_find(tok) != 0;
1446 if (t == '(') {
1447 next_nomacro();
1448 if (tok != ')')
1449 expect("')'");
1450 }
1451 tok = TOK_CINT;
1452 tokc.i = c;
1453 } else if (tok >= TOK_IDENT) {
1454 /* if undefined macro */
1455 tok = TOK_CINT;
1456 tokc.i = 0;
1457 }
1458 tok_str_add_tok(str);
1459 }
1460 pp_expr = 0;
1461 tok_str_add(str, -1); /* simulate end of file */
1462 tok_str_add(str, 0);
1463 /* now evaluate C constant expression */
1464 begin_macro(str, 1);
1465 next();
1466 c = expr_const();
1467 end_macro();
1468 return c != 0;
1469 }
1470
1471
1472 /* parse after #define */
1473 ST_FUNC void parse_define(void)
1474 {
1475 Sym *s, *first, **ps;
1476 int v, t, varg, is_vaargs, spc;
1477 int saved_parse_flags = parse_flags;
1478
1479 v = tok;
1480 if (v < TOK_IDENT || v == TOK_DEFINED)
1481 tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1482 /* XXX: should check if same macro (ANSI) */
1483 first = NULL;
1484 t = MACRO_OBJ;
1485 /* We have to parse the whole define as if not in asm mode, in particular
1486 no line comment with '#' must be ignored. Also for function
1487 macros the argument list must be parsed without '.' being an ID
1488 character. */
1489 parse_flags = ((parse_flags & ~PARSE_FLAG_ASM_FILE) | PARSE_FLAG_SPACES);
1490 /* '(' must be just after macro definition for MACRO_FUNC */
1491 next_nomacro_spc();
1492 if (tok == '(') {
1493 int dotid = set_idnum('.', 0);
1494 next_nomacro();
1495 ps = &first;
1496 if (tok != ')') for (;;) {
1497 varg = tok;
1498 next_nomacro();
1499 is_vaargs = 0;
1500 if (varg == TOK_DOTS) {
1501 varg = TOK___VA_ARGS__;
1502 is_vaargs = 1;
1503 } else if (tok == TOK_DOTS && gnu_ext) {
1504 is_vaargs = 1;
1505 next_nomacro();
1506 }
1507 if (varg < TOK_IDENT)
1508 bad_list:
1509 tcc_error("bad macro parameter list");
1510 s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1511 *ps = s;
1512 ps = &s->next;
1513 if (tok == ')')
1514 break;
1515 if (tok != ',' || is_vaargs)
1516 goto bad_list;
1517 next_nomacro();
1518 }
1519 next_nomacro_spc();
1520 t = MACRO_FUNC;
1521 set_idnum('.', dotid);
1522 }
1523
1524 tokstr_buf.len = 0;
1525 spc = 2;
1526 parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1527 /* The body of a macro definition should be parsed such that identifiers
1528 are parsed like the file mode determines (i.e. with '.' being an
1529 ID character in asm mode). But '#' should be retained instead of
1530 regarded as line comment leader, so still don't set ASM_FILE
1531 in parse_flags. */
1532 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1533 /* remove spaces around ## and after '#' */
1534 if (TOK_TWOSHARPS == tok) {
1535 if (2 == spc)
1536 goto bad_twosharp;
1537 if (1 == spc)
1538 --tokstr_buf.len;
1539 spc = 3;
1540 tok = TOK_PPJOIN;
1541 } else if ('#' == tok) {
1542 spc = 4;
1543 } else if (check_space(tok, &spc)) {
1544 goto skip;
1545 }
1546 tok_str_add2(&tokstr_buf, tok, &tokc);
1547 skip:
1548 next_nomacro_spc();
1549 }
1550
1551 parse_flags = saved_parse_flags;
1552 if (spc == 1)
1553 --tokstr_buf.len; /* remove trailing space */
1554 tok_str_add(&tokstr_buf, 0);
1555 if (3 == spc)
1556 bad_twosharp:
1557 tcc_error("'##' cannot appear at either end of macro");
1558 define_push(v, t, tok_str_dup(&tokstr_buf), first);
1559 }
1560
1561 static CachedInclude *search_cached_include(TCCState *s1, const char *filename, int add)
1562 {
1563 const unsigned char *s;
1564 unsigned int h;
1565 CachedInclude *e;
1566 int i;
1567
1568 h = TOK_HASH_INIT;
1569 s = (unsigned char *) filename;
1570 while (*s) {
1571 #ifdef _WIN32
1572 h = TOK_HASH_FUNC(h, toup(*s));
1573 #else
1574 h = TOK_HASH_FUNC(h, *s);
1575 #endif
1576 s++;
1577 }
1578 h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1579
1580 i = s1->cached_includes_hash[h];
1581 for(;;) {
1582 if (i == 0)
1583 break;
1584 e = s1->cached_includes[i - 1];
1585 if (0 == PATHCMP(e->filename, filename))
1586 return e;
1587 i = e->hash_next;
1588 }
1589 if (!add)
1590 return NULL;
1591
1592 e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1593 strcpy(e->filename, filename);
1594 e->ifndef_macro = e->once = 0;
1595 dynarray_add(&s1->cached_includes, &s1->nb_cached_includes, e);
1596 /* add in hash table */
1597 e->hash_next = s1->cached_includes_hash[h];
1598 s1->cached_includes_hash[h] = s1->nb_cached_includes;
1599 #ifdef INC_DEBUG
1600 printf("adding cached '%s'\n", filename);
1601 #endif
1602 return e;
1603 }
1604
1605 static void pragma_parse(TCCState *s1)
1606 {
1607 next_nomacro();
1608 if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1609 int t = tok, v;
1610 Sym *s;
1611
1612 if (next(), tok != '(')
1613 goto pragma_err;
1614 if (next(), tok != TOK_STR)
1615 goto pragma_err;
1616 v = tok_alloc(tokc.str.data, tokc.str.size - 1)->tok;
1617 if (next(), tok != ')')
1618 goto pragma_err;
1619 if (t == TOK_push_macro) {
1620 while (NULL == (s = define_find(v)))
1621 define_push(v, 0, NULL, NULL);
1622 s->type.ref = s; /* set push boundary */
1623 } else {
1624 for (s = define_stack; s; s = s->prev)
1625 if (s->v == v && s->type.ref == s) {
1626 s->type.ref = NULL;
1627 break;
1628 }
1629 }
1630 if (s)
1631 table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1632 else
1633 tcc_warning("unbalanced #pragma pop_macro");
1634 pp_debug_tok = t, pp_debug_symv = v;
1635
1636 } else if (tok == TOK_once) {
1637 search_cached_include(s1, file->filename, 1)->once = pp_once;
1638
1639 } else if (s1->output_type == TCC_OUTPUT_PREPROCESS) {
1640 /* tcc -E: keep pragmas below unchanged */
1641 unget_tok(' ');
1642 unget_tok(TOK_PRAGMA);
1643 unget_tok('#');
1644 unget_tok(TOK_LINEFEED);
1645
1646 } else if (tok == TOK_pack) {
1647 /* This may be:
1648 #pragma pack(1) // set
1649 #pragma pack() // reset to default
1650 #pragma pack(push,1) // push & set
1651 #pragma pack(pop) // restore previous */
1652 next();
1653 skip('(');
1654 if (tok == TOK_ASM_pop) {
1655 next();
1656 if (s1->pack_stack_ptr <= s1->pack_stack) {
1657 stk_error:
1658 tcc_error("out of pack stack");
1659 }
1660 s1->pack_stack_ptr--;
1661 } else {
1662 int val = 0;
1663 if (tok != ')') {
1664 if (tok == TOK_ASM_push) {
1665 next();
1666 if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1667 goto stk_error;
1668 s1->pack_stack_ptr++;
1669 skip(',');
1670 }
1671 if (tok != TOK_CINT)
1672 goto pragma_err;
1673 val = tokc.i;
1674 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1675 goto pragma_err;
1676 next();
1677 }
1678 *s1->pack_stack_ptr = val;
1679 }
1680 if (tok != ')')
1681 goto pragma_err;
1682
1683 } else if (tok == TOK_comment) {
1684 char *p; int t;
1685 next();
1686 skip('(');
1687 t = tok;
1688 next();
1689 skip(',');
1690 if (tok != TOK_STR)
1691 goto pragma_err;
1692 p = tcc_strdup((char *)tokc.str.data);
1693 next();
1694 if (tok != ')')
1695 goto pragma_err;
1696 if (t == TOK_lib) {
1697 dynarray_add(&s1->pragma_libs, &s1->nb_pragma_libs, p);
1698 } else {
1699 if (t == TOK_option)
1700 tcc_set_options(s1, p);
1701 tcc_free(p);
1702 }
1703
1704 } else if (s1->warn_unsupported) {
1705 tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1706 }
1707 return;
1708
1709 pragma_err:
1710 tcc_error("malformed #pragma directive");
1711 return;
1712 }
1713
1714 /* is_bof is true if first non space token at beginning of file */
1715 ST_FUNC void preprocess(int is_bof)
1716 {
1717 TCCState *s1 = tcc_state;
1718 int i, c, n, saved_parse_flags;
1719 char buf[1024], *q;
1720 Sym *s;
1721
1722 saved_parse_flags = parse_flags;
1723 parse_flags = PARSE_FLAG_PREPROCESS
1724 | PARSE_FLAG_TOK_NUM
1725 | PARSE_FLAG_TOK_STR
1726 | PARSE_FLAG_LINEFEED
1727 | (parse_flags & PARSE_FLAG_ASM_FILE)
1728 ;
1729
1730 next_nomacro();
1731 redo:
1732 switch(tok) {
1733 case TOK_DEFINE:
1734 pp_debug_tok = tok;
1735 next_nomacro();
1736 pp_debug_symv = tok;
1737 parse_define();
1738 break;
1739 case TOK_UNDEF:
1740 pp_debug_tok = tok;
1741 next_nomacro();
1742 pp_debug_symv = tok;
1743 s = define_find(tok);
1744 /* undefine symbol by putting an invalid name */
1745 if (s)
1746 define_undef(s);
1747 break;
1748 case TOK_INCLUDE:
1749 case TOK_INCLUDE_NEXT:
1750 ch = file->buf_ptr[0];
1751 /* XXX: incorrect if comments : use next_nomacro with a special mode */
1752 skip_spaces();
1753 if (ch == '<') {
1754 c = '>';
1755 goto read_name;
1756 } else if (ch == '\"') {
1757 c = ch;
1758 read_name:
1759 inp();
1760 q = buf;
1761 while (ch != c && ch != '\n' && ch != CH_EOF) {
1762 if ((q - buf) < sizeof(buf) - 1)
1763 *q++ = ch;
1764 if (ch == '\\') {
1765 if (handle_stray_noerror() == 0)
1766 --q;
1767 } else
1768 inp();
1769 }
1770 *q = '\0';
1771 minp();
1772 #if 0
1773 /* eat all spaces and comments after include */
1774 /* XXX: slightly incorrect */
1775 while (ch1 != '\n' && ch1 != CH_EOF)
1776 inp();
1777 #endif
1778 } else {
1779 int len;
1780 /* computed #include : concatenate everything up to linefeed,
1781 the result must be one of the two accepted forms.
1782 Don't convert pp-tokens to tokens here. */
1783 parse_flags = (PARSE_FLAG_PREPROCESS
1784 | PARSE_FLAG_LINEFEED
1785 | (parse_flags & PARSE_FLAG_ASM_FILE));
1786 next();
1787 buf[0] = '\0';
1788 while (tok != TOK_LINEFEED) {
1789 pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1790 next();
1791 }
1792 len = strlen(buf);
1793 /* check syntax and remove '<>|""' */
1794 if ((len < 2 || ((buf[0] != '"' || buf[len-1] != '"') &&
1795 (buf[0] != '<' || buf[len-1] != '>'))))
1796 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1797 c = buf[len-1];
1798 memmove(buf, buf + 1, len - 2);
1799 buf[len - 2] = '\0';
1800 }
1801
1802 if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1803 tcc_error("#include recursion too deep");
1804 /* store current file in stack, but increment stack later below */
1805 *s1->include_stack_ptr = file;
1806 i = tok == TOK_INCLUDE_NEXT ? file->include_next_index : 0;
1807 n = 2 + s1->nb_include_paths + 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 j = i - 2, k = j - s1->nb_include_paths;
1830 path = k < 0 ? s1->include_paths[j] : s1->sysinclude_paths[k];
1831 pstrcpy(buf1, sizeof(buf1), path);
1832 pstrcat(buf1, sizeof(buf1), "/");
1833 }
1834
1835 pstrcat(buf1, sizeof(buf1), buf);
1836 e = search_cached_include(s1, buf1, 0);
1837 if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) {
1838 /* no need to parse the include because the 'ifndef macro'
1839 is defined (or had #pragma once) */
1840 #ifdef INC_DEBUG
1841 printf("%s: skipping cached %s\n", file->filename, buf1);
1842 #endif
1843 goto include_done;
1844 }
1845
1846 if (tcc_open(s1, buf1) < 0)
1847 continue;
1848
1849 file->include_next_index = i + 1;
1850 #ifdef INC_DEBUG
1851 printf("%s: including %s\n", file->prev->filename, file->filename);
1852 #endif
1853 /* update target deps */
1854 dynarray_add(&s1->target_deps, &s1->nb_target_deps,
1855 tcc_strdup(buf1));
1856 /* push current file in stack */
1857 ++s1->include_stack_ptr;
1858 /* add include file debug info */
1859 if (s1->do_debug)
1860 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1861 tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1862 ch = file->buf_ptr[0];
1863 goto the_end;
1864 }
1865 tcc_error("include file '%s' not found", buf);
1866 include_done:
1867 break;
1868 case TOK_IFNDEF:
1869 c = 1;
1870 goto do_ifdef;
1871 case TOK_IF:
1872 c = expr_preprocess();
1873 goto do_if;
1874 case TOK_IFDEF:
1875 c = 0;
1876 do_ifdef:
1877 next_nomacro();
1878 if (tok < TOK_IDENT)
1879 tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1880 if (is_bof) {
1881 if (c) {
1882 #ifdef INC_DEBUG
1883 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1884 #endif
1885 file->ifndef_macro = tok;
1886 }
1887 }
1888 c = (define_find(tok) != 0) ^ c;
1889 do_if:
1890 if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1891 tcc_error("memory full (ifdef)");
1892 *s1->ifdef_stack_ptr++ = c;
1893 goto test_skip;
1894 case TOK_ELSE:
1895 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1896 tcc_error("#else without matching #if");
1897 if (s1->ifdef_stack_ptr[-1] & 2)
1898 tcc_error("#else after #else");
1899 c = (s1->ifdef_stack_ptr[-1] ^= 3);
1900 goto test_else;
1901 case TOK_ELIF:
1902 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1903 tcc_error("#elif without matching #if");
1904 c = s1->ifdef_stack_ptr[-1];
1905 if (c > 1)
1906 tcc_error("#elif after #else");
1907 /* last #if/#elif expression was true: we skip */
1908 if (c == 1) {
1909 c = 0;
1910 } else {
1911 c = expr_preprocess();
1912 s1->ifdef_stack_ptr[-1] = c;
1913 }
1914 test_else:
1915 if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1916 file->ifndef_macro = 0;
1917 test_skip:
1918 if (!(c & 1)) {
1919 preprocess_skip();
1920 is_bof = 0;
1921 goto redo;
1922 }
1923 break;
1924 case TOK_ENDIF:
1925 if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1926 tcc_error("#endif without matching #if");
1927 s1->ifdef_stack_ptr--;
1928 /* '#ifndef macro' was at the start of file. Now we check if
1929 an '#endif' is exactly at the end of file */
1930 if (file->ifndef_macro &&
1931 s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1932 file->ifndef_macro_saved = file->ifndef_macro;
1933 /* need to set to zero to avoid false matches if another
1934 #ifndef at middle of file */
1935 file->ifndef_macro = 0;
1936 while (tok != TOK_LINEFEED)
1937 next_nomacro();
1938 tok_flags |= TOK_FLAG_ENDIF;
1939 goto the_end;
1940 }
1941 break;
1942 case TOK_PPNUM:
1943 n = strtoul((char*)tokc.str.data, &q, 10);
1944 goto _line_num;
1945 case TOK_LINE:
1946 next();
1947 if (tok != TOK_CINT)
1948 _line_err:
1949 tcc_error("wrong #line format");
1950 n = tokc.i;
1951 _line_num:
1952 next();
1953 if (tok != TOK_LINEFEED) {
1954 if (tok == TOK_STR) {
1955 if (file->true_filename == file->filename)
1956 file->true_filename = tcc_strdup(file->filename);
1957 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.str.data);
1958 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
1959 break;
1960 else
1961 goto _line_err;
1962 --n;
1963 }
1964 if (file->fd > 0)
1965 total_lines += file->line_num - n;
1966 file->line_num = n;
1967 if (s1->do_debug)
1968 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1969 break;
1970 case TOK_ERROR:
1971 case TOK_WARNING:
1972 c = tok;
1973 ch = file->buf_ptr[0];
1974 skip_spaces();
1975 q = buf;
1976 while (ch != '\n' && ch != CH_EOF) {
1977 if ((q - buf) < sizeof(buf) - 1)
1978 *q++ = ch;
1979 if (ch == '\\') {
1980 if (handle_stray_noerror() == 0)
1981 --q;
1982 } else
1983 inp();
1984 }
1985 *q = '\0';
1986 if (c == TOK_ERROR)
1987 tcc_error("#error %s", buf);
1988 else
1989 tcc_warning("#warning %s", buf);
1990 break;
1991 case TOK_PRAGMA:
1992 pragma_parse(s1);
1993 break;
1994 case TOK_LINEFEED:
1995 goto the_end;
1996 default:
1997 /* ignore gas line comment in an 'S' file. */
1998 if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
1999 goto ignore;
2000 if (tok == '!' && is_bof)
2001 /* '!' is ignored at beginning to allow C scripts. */
2002 goto ignore;
2003 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
2004 ignore:
2005 file->buf_ptr = parse_line_comment(file->buf_ptr - 1);
2006 goto the_end;
2007 }
2008 /* ignore other preprocess commands or #! for C scripts */
2009 while (tok != TOK_LINEFEED)
2010 next_nomacro();
2011 the_end:
2012 parse_flags = saved_parse_flags;
2013 }
2014
2015 /* evaluate escape codes in a string. */
2016 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
2017 {
2018 int c, n;
2019 const uint8_t *p;
2020
2021 p = buf;
2022 for(;;) {
2023 c = *p;
2024 if (c == '\0')
2025 break;
2026 if (c == '\\') {
2027 p++;
2028 /* escape */
2029 c = *p;
2030 switch(c) {
2031 case '0': case '1': case '2': case '3':
2032 case '4': case '5': case '6': case '7':
2033 /* at most three octal digits */
2034 n = c - '0';
2035 p++;
2036 c = *p;
2037 if (isoct(c)) {
2038 n = n * 8 + c - '0';
2039 p++;
2040 c = *p;
2041 if (isoct(c)) {
2042 n = n * 8 + c - '0';
2043 p++;
2044 }
2045 }
2046 c = n;
2047 goto add_char_nonext;
2048 case 'x':
2049 case 'u':
2050 case 'U':
2051 p++;
2052 n = 0;
2053 for(;;) {
2054 c = *p;
2055 if (c >= 'a' && c <= 'f')
2056 c = c - 'a' + 10;
2057 else if (c >= 'A' && c <= 'F')
2058 c = c - 'A' + 10;
2059 else if (isnum(c))
2060 c = c - '0';
2061 else
2062 break;
2063 n = n * 16 + c;
2064 p++;
2065 }
2066 c = n;
2067 goto add_char_nonext;
2068 case 'a':
2069 c = '\a';
2070 break;
2071 case 'b':
2072 c = '\b';
2073 break;
2074 case 'f':
2075 c = '\f';
2076 break;
2077 case 'n':
2078 c = '\n';
2079 break;
2080 case 'r':
2081 c = '\r';
2082 break;
2083 case 't':
2084 c = '\t';
2085 break;
2086 case 'v':
2087 c = '\v';
2088 break;
2089 case 'e':
2090 if (!gnu_ext)
2091 goto invalid_escape;
2092 c = 27;
2093 break;
2094 case '\'':
2095 case '\"':
2096 case '\\':
2097 case '?':
2098 break;
2099 default:
2100 invalid_escape:
2101 if (c >= '!' && c <= '~')
2102 tcc_warning("unknown escape sequence: \'\\%c\'", c);
2103 else
2104 tcc_warning("unknown escape sequence: \'\\x%x\'", c);
2105 break;
2106 }
2107 } else if (is_long && c >= 0x80) {
2108 /* assume we are processing UTF-8 sequence */
2109 /* reference: The Unicode Standard, Version 10.0, ch3.9 */
2110
2111 int cont; /* count of continuation bytes */
2112 int skip; /* how many bytes should skip when error occurred */
2113 int i;
2114
2115 /* decode leading byte */
2116 if (c < 0xC2) {
2117 skip = 1; goto invalid_utf8_sequence;
2118 } else if (c <= 0xDF) {
2119 cont = 1; n = c & 0x1f;
2120 } else if (c <= 0xEF) {
2121 cont = 2; n = c & 0xf;
2122 } else if (c <= 0xF4) {
2123 cont = 3; n = c & 0x7;
2124 } else {
2125 skip = 1; goto invalid_utf8_sequence;
2126 }
2127
2128 /* decode continuation bytes */
2129 for (i = 1; i <= cont; i++) {
2130 int l = 0x80, h = 0xBF;
2131
2132 /* adjust limit for second byte */
2133 if (i == 1) {
2134 switch (c) {
2135 case 0xE0: l = 0xA0; break;
2136 case 0xED: h = 0x9F; break;
2137 case 0xF0: l = 0x90; break;
2138 case 0xF4: h = 0x8F; break;
2139 }
2140 }
2141
2142 if (p[i] < l || p[i] > h) {
2143 skip = i; goto invalid_utf8_sequence;
2144 }
2145
2146 n = (n << 6) | (p[i] & 0x3f);
2147 }
2148
2149 /* advance pointer */
2150 p += 1 + cont;
2151 c = n;
2152 goto add_char_nonext;
2153
2154 /* error handling */
2155 invalid_utf8_sequence:
2156 tcc_warning("ill-formed UTF-8 subsequence starting with: \'\\x%x\'", c);
2157 c = 0xFFFD;
2158 p += skip;
2159 goto add_char_nonext;
2160
2161 }
2162 p++;
2163 add_char_nonext:
2164 if (!is_long)
2165 cstr_ccat(outstr, c);
2166 else {
2167 #ifdef TCC_TARGET_PE
2168 /* store as UTF-16 */
2169 if (c < 0x10000) {
2170 cstr_wccat(outstr, c);
2171 } else {
2172 c -= 0x10000;
2173 cstr_wccat(outstr, (c >> 10) + 0xD800);
2174 cstr_wccat(outstr, (c & 0x3FF) + 0xDC00);
2175 }
2176 #else
2177 cstr_wccat(outstr, c);
2178 #endif
2179 }
2180 }
2181 /* add a trailing '\0' */
2182 if (!is_long)
2183 cstr_ccat(outstr, '\0');
2184 else
2185 cstr_wccat(outstr, '\0');
2186 }
2187
2188 static void parse_string(const char *s, int len)
2189 {
2190 uint8_t buf[1000], *p = buf;
2191 int is_long, sep;
2192
2193 if ((is_long = *s == 'L'))
2194 ++s, --len;
2195 sep = *s++;
2196 len -= 2;
2197 if (len >= sizeof buf)
2198 p = tcc_malloc(len + 1);
2199 memcpy(p, s, len);
2200 p[len] = 0;
2201
2202 cstr_reset(&tokcstr);
2203 parse_escape_string(&tokcstr, p, is_long);
2204 if (p != buf)
2205 tcc_free(p);
2206
2207 if (sep == '\'') {
2208 int char_size, i, n, c;
2209 /* XXX: make it portable */
2210 if (!is_long)
2211 tok = TOK_CCHAR, char_size = 1;
2212 else
2213 tok = TOK_LCHAR, char_size = sizeof(nwchar_t);
2214 n = tokcstr.size / char_size - 1;
2215 if (n < 1)
2216 tcc_error("empty character constant");
2217 if (n > 1)
2218 tcc_warning("multi-character character constant");
2219 for (c = i = 0; i < n; ++i) {
2220 if (is_long)
2221 c = ((nwchar_t *)tokcstr.data)[i];
2222 else
2223 c = (c << 8) | ((char *)tokcstr.data)[i];
2224 }
2225 tokc.i = c;
2226 } else {
2227 tokc.str.size = tokcstr.size;
2228 tokc.str.data = tokcstr.data;
2229 if (!is_long)
2230 tok = TOK_STR;
2231 else
2232 tok = TOK_LSTR;
2233 }
2234 }
2235
2236 /* we use 64 bit numbers */
2237 #define BN_SIZE 2
2238
2239 /* bn = (bn << shift) | or_val */
2240 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2241 {
2242 int i;
2243 unsigned int v;
2244 for(i=0;i<BN_SIZE;i++) {
2245 v = bn[i];
2246 bn[i] = (v << shift) | or_val;
2247 or_val = v >> (32 - shift);
2248 }
2249 }
2250
2251 static void bn_zero(unsigned int *bn)
2252 {
2253 int i;
2254 for(i=0;i<BN_SIZE;i++) {
2255 bn[i] = 0;
2256 }
2257 }
2258
2259 /* parse number in null terminated string 'p' and return it in the
2260 current token */
2261 static void parse_number(const char *p)
2262 {
2263 int b, t, shift, frac_bits, s, exp_val, ch;
2264 char *q;
2265 unsigned int bn[BN_SIZE];
2266 double d;
2267
2268 /* number */
2269 q = token_buf;
2270 ch = *p++;
2271 t = ch;
2272 ch = *p++;
2273 *q++ = t;
2274 b = 10;
2275 if (t == '.') {
2276 goto float_frac_parse;
2277 } else if (t == '0') {
2278 if (ch == 'x' || ch == 'X') {
2279 q--;
2280 ch = *p++;
2281 b = 16;
2282 } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2283 q--;
2284 ch = *p++;
2285 b = 2;
2286 }
2287 }
2288 /* parse all digits. cannot check octal numbers at this stage
2289 because of floating point constants */
2290 while (1) {
2291 if (ch >= 'a' && ch <= 'f')
2292 t = ch - 'a' + 10;
2293 else if (ch >= 'A' && ch <= 'F')
2294 t = ch - 'A' + 10;
2295 else if (isnum(ch))
2296 t = ch - '0';
2297 else
2298 break;
2299 if (t >= b)
2300 break;
2301 if (q >= token_buf + STRING_MAX_SIZE) {
2302 num_too_long:
2303 tcc_error("number too long");
2304 }
2305 *q++ = ch;
2306 ch = *p++;
2307 }
2308 if (ch == '.' ||
2309 ((ch == 'e' || ch == 'E') && b == 10) ||
2310 ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2311 if (b != 10) {
2312 /* NOTE: strtox should support that for hexa numbers, but
2313 non ISOC99 libcs do not support it, so we prefer to do
2314 it by hand */
2315 /* hexadecimal or binary floats */
2316 /* XXX: handle overflows */
2317 *q = '\0';
2318 if (b == 16)
2319 shift = 4;
2320 else
2321 shift = 1;
2322 bn_zero(bn);
2323 q = token_buf;
2324 while (1) {
2325 t = *q++;
2326 if (t == '\0') {
2327 break;
2328 } else if (t >= 'a') {
2329 t = t - 'a' + 10;
2330 } else if (t >= 'A') {
2331 t = t - 'A' + 10;
2332 } else {
2333 t = t - '0';
2334 }
2335 bn_lshift(bn, shift, t);
2336 }
2337 frac_bits = 0;
2338 if (ch == '.') {
2339 ch = *p++;
2340 while (1) {
2341 t = ch;
2342 if (t >= 'a' && t <= 'f') {
2343 t = t - 'a' + 10;
2344 } else if (t >= 'A' && t <= 'F') {
2345 t = t - 'A' + 10;
2346 } else if (t >= '0' && t <= '9') {
2347 t = t - '0';
2348 } else {
2349 break;
2350 }
2351 if (t >= b)
2352 tcc_error("invalid digit");
2353 bn_lshift(bn, shift, t);
2354 frac_bits += shift;
2355 ch = *p++;
2356 }
2357 }
2358 if (ch != 'p' && ch != 'P')
2359 expect("exponent");
2360 ch = *p++;
2361 s = 1;
2362 exp_val = 0;
2363 if (ch == '+') {
2364 ch = *p++;
2365 } else if (ch == '-') {
2366 s = -1;
2367 ch = *p++;
2368 }
2369 if (ch < '0' || ch > '9')
2370 expect("exponent digits");
2371 while (ch >= '0' && ch <= '9') {
2372 exp_val = exp_val * 10 + ch - '0';
2373 ch = *p++;
2374 }
2375 exp_val = exp_val * s;
2376
2377 /* now we can generate the number */
2378 /* XXX: should patch directly float number */
2379 d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2380 d = ldexp(d, exp_val - frac_bits);
2381 t = toup(ch);
2382 if (t == 'F') {
2383 ch = *p++;
2384 tok = TOK_CFLOAT;
2385 /* float : should handle overflow */
2386 tokc.f = (float)d;
2387 } else if (t == 'L') {
2388 ch = *p++;
2389 #ifdef TCC_TARGET_PE
2390 tok = TOK_CDOUBLE;
2391 tokc.d = d;
2392 #else
2393 tok = TOK_CLDOUBLE;
2394 /* XXX: not large enough */
2395 tokc.ld = (long double)d;
2396 #endif
2397 } else {
2398 tok = TOK_CDOUBLE;
2399 tokc.d = d;
2400 }
2401 } else {
2402 /* decimal floats */
2403 if (ch == '.') {
2404 if (q >= token_buf + STRING_MAX_SIZE)
2405 goto num_too_long;
2406 *q++ = ch;
2407 ch = *p++;
2408 float_frac_parse:
2409 while (ch >= '0' && ch <= '9') {
2410 if (q >= token_buf + STRING_MAX_SIZE)
2411 goto num_too_long;
2412 *q++ = ch;
2413 ch = *p++;
2414 }
2415 }
2416 if (ch == 'e' || ch == 'E') {
2417 if (q >= token_buf + STRING_MAX_SIZE)
2418 goto num_too_long;
2419 *q++ = ch;
2420 ch = *p++;
2421 if (ch == '-' || ch == '+') {
2422 if (q >= token_buf + STRING_MAX_SIZE)
2423 goto num_too_long;
2424 *q++ = ch;
2425 ch = *p++;
2426 }
2427 if (ch < '0' || ch > '9')
2428 expect("exponent digits");
2429 while (ch >= '0' && ch <= '9') {
2430 if (q >= token_buf + STRING_MAX_SIZE)
2431 goto num_too_long;
2432 *q++ = ch;
2433 ch = *p++;
2434 }
2435 }
2436 *q = '\0';
2437 t = toup(ch);
2438 errno = 0;
2439 if (t == 'F') {
2440 ch = *p++;
2441 tok = TOK_CFLOAT;
2442 tokc.f = strtof(token_buf, NULL);
2443 } else if (t == 'L') {
2444 ch = *p++;
2445 #ifdef TCC_TARGET_PE
2446 tok = TOK_CDOUBLE;
2447 tokc.d = strtod(token_buf, NULL);
2448 #else
2449 tok = TOK_CLDOUBLE;
2450 tokc.ld = strtold(token_buf, NULL);
2451 #endif
2452 } else {
2453 tok = TOK_CDOUBLE;
2454 tokc.d = strtod(token_buf, NULL);
2455 }
2456 }
2457 } else {
2458 unsigned long long n, n1;
2459 int lcount, ucount, ov = 0;
2460 const char *p1;
2461
2462 /* integer number */
2463 *q = '\0';
2464 q = token_buf;
2465 if (b == 10 && *q == '0') {
2466 b = 8;
2467 q++;
2468 }
2469 n = 0;
2470 while(1) {
2471 t = *q++;
2472 /* no need for checks except for base 10 / 8 errors */
2473 if (t == '\0')
2474 break;
2475 else if (t >= 'a')
2476 t = t - 'a' + 10;
2477 else if (t >= 'A')
2478 t = t - 'A' + 10;
2479 else
2480 t = t - '0';
2481 if (t >= b)
2482 tcc_error("invalid digit");
2483 n1 = n;
2484 n = n * b + t;
2485 /* detect overflow */
2486 if (n1 >= 0x1000000000000000ULL && n / b != n1)
2487 ov = 1;
2488 }
2489
2490 /* Determine the characteristics (unsigned and/or 64bit) the type of
2491 the constant must have according to the constant suffix(es) */
2492 lcount = ucount = 0;
2493 p1 = p;
2494 for(;;) {
2495 t = toup(ch);
2496 if (t == 'L') {
2497 if (lcount >= 2)
2498 tcc_error("three 'l's in integer constant");
2499 if (lcount && *(p - 1) != ch)
2500 tcc_error("incorrect integer suffix: %s", p1);
2501 lcount++;
2502 ch = *p++;
2503 } else if (t == 'U') {
2504 if (ucount >= 1)
2505 tcc_error("two 'u's in integer constant");
2506 ucount++;
2507 ch = *p++;
2508 } else {
2509 break;
2510 }
2511 }
2512
2513 /* Determine if it needs 64 bits and/or unsigned in order to fit */
2514 if (ucount == 0 && b == 10) {
2515 if (lcount <= (LONG_SIZE == 4)) {
2516 if (n >= 0x80000000U)
2517 lcount = (LONG_SIZE == 4) + 1;
2518 }
2519 if (n >= 0x8000000000000000ULL)
2520 ov = 1, ucount = 1;
2521 } else {
2522 if (lcount <= (LONG_SIZE == 4)) {
2523 if (n >= 0x100000000ULL)
2524 lcount = (LONG_SIZE == 4) + 1;
2525 else if (n >= 0x80000000U)
2526 ucount = 1;
2527 }
2528 if (n >= 0x8000000000000000ULL)
2529 ucount = 1;
2530 }
2531
2532 if (ov)
2533 tcc_warning("integer constant overflow");
2534
2535 tok = TOK_CINT;
2536 if (lcount) {
2537 tok = TOK_CLONG;
2538 if (lcount == 2)
2539 tok = TOK_CLLONG;
2540 }
2541 if (ucount)
2542 ++tok; /* TOK_CU... */
2543 tokc.i = n;
2544 }
2545 if (ch)
2546 tcc_error("invalid number\n");
2547 }
2548
2549
2550 #define PARSE2(c1, tok1, c2, tok2) \
2551 case c1: \
2552 PEEKC(c, p); \
2553 if (c == c2) { \
2554 p++; \
2555 tok = tok2; \
2556 } else { \
2557 tok = tok1; \
2558 } \
2559 break;
2560
2561 /* return next token without macro substitution */
2562 static inline void next_nomacro1(void)
2563 {
2564 int t, c, is_long, len;
2565 TokenSym *ts;
2566 uint8_t *p, *p1;
2567 unsigned int h;
2568
2569 p = file->buf_ptr;
2570 redo_no_start:
2571 c = *p;
2572 switch(c) {
2573 case ' ':
2574 case '\t':
2575 tok = c;
2576 p++;
2577 if (parse_flags & PARSE_FLAG_SPACES)
2578 goto keep_tok_flags;
2579 while (isidnum_table[*p - CH_EOF] & IS_SPC)
2580 ++p;
2581 goto redo_no_start;
2582 case '\f':
2583 case '\v':
2584 case '\r':
2585 p++;
2586 goto redo_no_start;
2587 case '\\':
2588 /* first look if it is in fact an end of buffer */
2589 c = handle_stray1(p);
2590 p = file->buf_ptr;
2591 if (c == '\\')
2592 goto parse_simple;
2593 if (c != CH_EOF)
2594 goto redo_no_start;
2595 {
2596 TCCState *s1 = tcc_state;
2597 if ((parse_flags & PARSE_FLAG_LINEFEED)
2598 && !(tok_flags & TOK_FLAG_EOF)) {
2599 tok_flags |= TOK_FLAG_EOF;
2600 tok = TOK_LINEFEED;
2601 goto keep_tok_flags;
2602 } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2603 tok = TOK_EOF;
2604 } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2605 tcc_error("missing #endif");
2606 } else if (s1->include_stack_ptr == s1->include_stack) {
2607 /* no include left : end of file. */
2608 tok = TOK_EOF;
2609 } else {
2610 tok_flags &= ~TOK_FLAG_EOF;
2611 /* pop include file */
2612
2613 /* test if previous '#endif' was after a #ifdef at
2614 start of file */
2615 if (tok_flags & TOK_FLAG_ENDIF) {
2616 #ifdef INC_DEBUG
2617 printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2618 #endif
2619 search_cached_include(s1, file->filename, 1)
2620 ->ifndef_macro = file->ifndef_macro_saved;
2621 tok_flags &= ~TOK_FLAG_ENDIF;
2622 }
2623
2624 /* add end of include file debug info */
2625 if (tcc_state->do_debug) {
2626 put_stabd(N_EINCL, 0, 0);
2627 }
2628 /* pop include stack */
2629 tcc_close();
2630 s1->include_stack_ptr--;
2631 p = file->buf_ptr;
2632 if (p == file->buffer)
2633 tok_flags = TOK_FLAG_BOF|TOK_FLAG_BOL;
2634 goto redo_no_start;
2635 }
2636 }
2637 break;
2638
2639 case '\n':
2640 file->line_num++;
2641 tok_flags |= TOK_FLAG_BOL;
2642 p++;
2643 maybe_newline:
2644 if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2645 goto redo_no_start;
2646 tok = TOK_LINEFEED;
2647 goto keep_tok_flags;
2648
2649 case '#':
2650 /* XXX: simplify */
2651 PEEKC(c, p);
2652 if ((tok_flags & TOK_FLAG_BOL) &&
2653 (parse_flags & PARSE_FLAG_PREPROCESS)) {
2654 file->buf_ptr = p;
2655 preprocess(tok_flags & TOK_FLAG_BOF);
2656 p = file->buf_ptr;
2657 goto maybe_newline;
2658 } else {
2659 if (c == '#') {
2660 p++;
2661 tok = TOK_TWOSHARPS;
2662 } else {
2663 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2664 p = parse_line_comment(p - 1);
2665 goto redo_no_start;
2666 } else {
2667 tok = '#';
2668 }
2669 }
2670 }
2671 break;
2672
2673 /* dollar is allowed to start identifiers when not parsing asm */
2674 case '$':
2675 if (!(isidnum_table[c - CH_EOF] & IS_ID)
2676 || (parse_flags & PARSE_FLAG_ASM_FILE))
2677 goto parse_simple;
2678
2679 case 'a': case 'b': case 'c': case 'd':
2680 case 'e': case 'f': case 'g': case 'h':
2681 case 'i': case 'j': case 'k': case 'l':
2682 case 'm': case 'n': case 'o': case 'p':
2683 case 'q': case 'r': case 's': case 't':
2684 case 'u': case 'v': case 'w': case 'x':
2685 case 'y': case 'z':
2686 case 'A': case 'B': case 'C': case 'D':
2687 case 'E': case 'F': case 'G': case 'H':
2688 case 'I': case 'J': case 'K':
2689 case 'M': case 'N': case 'O': case 'P':
2690 case 'Q': case 'R': case 'S': case 'T':
2691 case 'U': case 'V': case 'W': case 'X':
2692 case 'Y': case 'Z':
2693 case '_':
2694 parse_ident_fast:
2695 p1 = p;
2696 h = TOK_HASH_INIT;
2697 h = TOK_HASH_FUNC(h, c);
2698 while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2699 h = TOK_HASH_FUNC(h, c);
2700 len = p - p1;
2701 if (c != '\\') {
2702 TokenSym **pts;
2703
2704 /* fast case : no stray found, so we have the full token
2705 and we have already hashed it */
2706 h &= (TOK_HASH_SIZE - 1);
2707 pts = &hash_ident[h];
2708 for(;;) {
2709 ts = *pts;
2710 if (!ts)
2711 break;
2712 if (ts->len == len && !memcmp(ts->str, p1, len))
2713 goto token_found;
2714 pts = &(ts->hash_next);
2715 }
2716 ts = tok_alloc_new(pts, (char *) p1, len);
2717 token_found: ;
2718 } else {
2719 /* slower case */
2720 cstr_reset(&tokcstr);
2721 cstr_cat(&tokcstr, (char *) p1, len);
2722 p--;
2723 PEEKC(c, p);
2724 parse_ident_slow:
2725 while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2726 {
2727 cstr_ccat(&tokcstr, c);
2728 PEEKC(c, p);
2729 }
2730 ts = tok_alloc(tokcstr.data, tokcstr.size);
2731 }
2732 tok = ts->tok;
2733 break;
2734 case 'L':
2735 t = p[1];
2736 if (t != '\\' && t != '\'' && t != '\"') {
2737 /* fast case */
2738 goto parse_ident_fast;
2739 } else {
2740 PEEKC(c, p);
2741 if (c == '\'' || c == '\"') {
2742 is_long = 1;
2743 goto str_const;
2744 } else {
2745 cstr_reset(&tokcstr);
2746 cstr_ccat(&tokcstr, 'L');
2747 goto parse_ident_slow;
2748 }
2749 }
2750 break;
2751
2752 case '0': case '1': case '2': case '3':
2753 case '4': case '5': case '6': case '7':
2754 case '8': case '9':
2755 t = c;
2756 PEEKC(c, p);
2757 /* after the first digit, accept digits, alpha, '.' or sign if
2758 prefixed by 'eEpP' */
2759 parse_num:
2760 cstr_reset(&tokcstr);
2761 for(;;) {
2762 cstr_ccat(&tokcstr, t);
2763 if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2764 || c == '.'
2765 || ((c == '+' || c == '-')
2766 && (((t == 'e' || t == 'E')
2767 && !(parse_flags & PARSE_FLAG_ASM_FILE
2768 /* 0xe+1 is 3 tokens in asm */
2769 && ((char*)tokcstr.data)[0] == '0'
2770 && toup(((char*)tokcstr.data)[1]) == 'X'))
2771 || t == 'p' || t == 'P'))))
2772 break;
2773 t = c;
2774 PEEKC(c, p);
2775 }
2776 /* We add a trailing '\0' to ease parsing */
2777 cstr_ccat(&tokcstr, '\0');
2778 tokc.str.size = tokcstr.size;
2779 tokc.str.data = tokcstr.data;
2780 tok = TOK_PPNUM;
2781 break;
2782
2783 case '.':
2784 /* special dot handling because it can also start a number */
2785 PEEKC(c, p);
2786 if (isnum(c)) {
2787 t = '.';
2788 goto parse_num;
2789 } else if ((isidnum_table['.' - CH_EOF] & IS_ID)
2790 && (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))) {
2791 *--p = c = '.';
2792 goto parse_ident_fast;
2793 } else if (c == '.') {
2794 PEEKC(c, p);
2795 if (c == '.') {
2796 p++;
2797 tok = TOK_DOTS;
2798 } else {
2799 *--p = '.'; /* may underflow into file->unget[] */
2800 tok = '.';
2801 }
2802 } else {
2803 tok = '.';
2804 }
2805 break;
2806 case '\'':
2807 case '\"':
2808 is_long = 0;
2809 str_const:
2810 cstr_reset(&tokcstr);
2811 if (is_long)
2812 cstr_ccat(&tokcstr, 'L');
2813 cstr_ccat(&tokcstr, c);
2814 p = parse_pp_string(p, c, &tokcstr);
2815 cstr_ccat(&tokcstr, c);
2816 cstr_ccat(&tokcstr, '\0');
2817 tokc.str.size = tokcstr.size;
2818 tokc.str.data = tokcstr.data;
2819 tok = TOK_PPSTR;
2820 break;
2821
2822 case '<':
2823 PEEKC(c, p);
2824 if (c == '=') {
2825 p++;
2826 tok = TOK_LE;
2827 } else if (c == '<') {
2828 PEEKC(c, p);
2829 if (c == '=') {
2830 p++;
2831 tok = TOK_A_SHL;
2832 } else {
2833 tok = TOK_SHL;
2834 }
2835 } else {
2836 tok = TOK_LT;
2837 }
2838 break;
2839 case '>':
2840 PEEKC(c, p);
2841 if (c == '=') {
2842 p++;
2843 tok = TOK_GE;
2844 } else if (c == '>') {
2845 PEEKC(c, p);
2846 if (c == '=') {
2847 p++;
2848 tok = TOK_A_SAR;
2849 } else {
2850 tok = TOK_SAR;
2851 }
2852 } else {
2853 tok = TOK_GT;
2854 }
2855 break;
2856
2857 case '&':
2858 PEEKC(c, p);
2859 if (c == '&') {
2860 p++;
2861 tok = TOK_LAND;
2862 } else if (c == '=') {
2863 p++;
2864 tok = TOK_A_AND;
2865 } else {
2866 tok = '&';
2867 }
2868 break;
2869
2870 case '|':
2871 PEEKC(c, p);
2872 if (c == '|') {
2873 p++;
2874 tok = TOK_LOR;
2875 } else if (c == '=') {
2876 p++;
2877 tok = TOK_A_OR;
2878 } else {
2879 tok = '|';
2880 }
2881 break;
2882
2883 case '+':
2884 PEEKC(c, p);
2885 if (c == '+') {
2886 p++;
2887 tok = TOK_INC;
2888 } else if (c == '=') {
2889 p++;
2890 tok = TOK_A_ADD;
2891 } else {
2892 tok = '+';
2893 }
2894 break;
2895
2896 case '-':
2897 PEEKC(c, p);
2898 if (c == '-') {
2899 p++;
2900 tok = TOK_DEC;
2901 } else if (c == '=') {
2902 p++;
2903 tok = TOK_A_SUB;
2904 } else if (c == '>') {
2905 p++;
2906 tok = TOK_ARROW;
2907 } else {
2908 tok = '-';
2909 }
2910 break;
2911
2912 PARSE2('!', '!', '=', TOK_NE)
2913 PARSE2('=', '=', '=', TOK_EQ)
2914 PARSE2('*', '*', '=', TOK_A_MUL)
2915 PARSE2('%', '%', '=', TOK_A_MOD)
2916 PARSE2('^', '^', '=', TOK_A_XOR)
2917
2918 /* comments or operator */
2919 case '/':
2920 PEEKC(c, p);
2921 if (c == '*') {
2922 p = parse_comment(p);
2923 /* comments replaced by a blank */
2924 tok = ' ';
2925 goto keep_tok_flags;
2926 } else if (c == '/') {
2927 p = parse_line_comment(p);
2928 tok = ' ';
2929 goto keep_tok_flags;
2930 } else if (c == '=') {
2931 p++;
2932 tok = TOK_A_DIV;
2933 } else {
2934 tok = '/';
2935 }
2936 break;
2937
2938 /* simple tokens */
2939 case '(':
2940 case ')':
2941 case '[':
2942 case ']':
2943 case '{':
2944 case '}':
2945 case ',':
2946 case ';':
2947 case ':':
2948 case '?':
2949 case '~':
2950 case '@': /* only used in assembler */
2951 parse_simple:
2952 tok = c;
2953 p++;
2954 break;
2955 default:
2956 if (c >= 0x80 && c <= 0xFF) /* utf8 identifiers */
2957 goto parse_ident_fast;
2958 if (parse_flags & PARSE_FLAG_ASM_FILE)
2959 goto parse_simple;
2960 tcc_error("unrecognized character \\x%02x", c);
2961 break;
2962 }
2963 tok_flags = 0;
2964 keep_tok_flags:
2965 file->buf_ptr = p;
2966 #if defined(PARSE_DEBUG)
2967 printf("token = %d %s\n", tok, get_tok_str(tok, &tokc));
2968 #endif
2969 }
2970
2971 /* return next token without macro substitution. Can read input from
2972 macro_ptr buffer */
2973 static void next_nomacro_spc(void)
2974 {
2975 if (macro_ptr) {
2976 redo:
2977 tok = *macro_ptr;
2978 if (tok) {
2979 TOK_GET(&tok, &macro_ptr, &tokc);
2980 if (tok == TOK_LINENUM) {
2981 file->line_num = tokc.i;
2982 goto redo;
2983 }
2984 }
2985 } else {
2986 next_nomacro1();
2987 }
2988 //printf("token = %s\n", get_tok_str(tok, &tokc));
2989 }
2990
2991 ST_FUNC void next_nomacro(void)
2992 {
2993 do {
2994 next_nomacro_spc();
2995 } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
2996 }
2997
2998
2999 static void macro_subst(
3000 TokenString *tok_str,
3001 Sym **nested_list,
3002 const int *macro_str
3003 );
3004
3005 /* substitute arguments in replacement lists in macro_str by the values in
3006 args (field d) and return allocated string */
3007 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
3008 {
3009 int t, t0, t1, spc;
3010 const int *st;
3011 Sym *s;
3012 CValue cval;
3013 TokenString str;
3014 CString cstr;
3015
3016 tok_str_new(&str);
3017 t0 = t1 = 0;
3018 while(1) {
3019 TOK_GET(&t, &macro_str, &cval);
3020 if (!t)
3021 break;
3022 if (t == '#') {
3023 /* stringize */
3024 TOK_GET(&t, &macro_str, &cval);
3025 if (!t)
3026 goto bad_stringy;
3027 s = sym_find2(args, t);
3028 if (s) {
3029 cstr_new(&cstr);
3030 cstr_ccat(&cstr, '\"');
3031 st = s->d;
3032 spc = 0;
3033 while (*st >= 0) {
3034 TOK_GET(&t, &st, &cval);
3035 if (t != TOK_PLCHLDR
3036 && t != TOK_NOSUBST
3037 && 0 == check_space(t, &spc)) {
3038 const char *s = get_tok_str(t, &cval);
3039 while (*s) {
3040 if (t == TOK_PPSTR && *s != '\'')
3041 add_char(&cstr, *s);
3042 else
3043 cstr_ccat(&cstr, *s);
3044 ++s;
3045 }
3046 }
3047 }
3048 cstr.size -= spc;
3049 cstr_ccat(&cstr, '\"');
3050 cstr_ccat(&cstr, '\0');
3051 #ifdef PP_DEBUG
3052 printf("\nstringize: <%s>\n", (char *)cstr.data);
3053 #endif
3054 /* add string */
3055 cval.str.size = cstr.size;
3056 cval.str.data = cstr.data;
3057 tok_str_add2(&str, TOK_PPSTR, &cval);
3058 cstr_free(&cstr);
3059 } else {
3060 bad_stringy:
3061 expect("macro parameter after '#'");
3062 }
3063 } else if (t >= TOK_IDENT) {
3064 s = sym_find2(args, t);
3065 if (s) {
3066 int l0 = str.len;
3067 st = s->d;
3068 /* if '##' is present before or after, no arg substitution */
3069 if (*macro_str == TOK_PPJOIN || t1 == TOK_PPJOIN) {
3070 /* special case for var arg macros : ## eats the ','
3071 if empty VA_ARGS variable. */
3072 if (t1 == TOK_PPJOIN && t0 == ',' && gnu_ext && s->type.t) {
3073 if (*st <= 0) {
3074 /* suppress ',' '##' */
3075 str.len -= 2;
3076 } else {
3077 /* suppress '##' and add variable */
3078 str.len--;
3079 goto add_var;
3080 }
3081 }
3082 } else {
3083 add_var:
3084 if (!s->next) {
3085 /* Expand arguments tokens and store them. In most
3086 cases we could also re-expand each argument if
3087 used multiple times, but not if the argument
3088 contains the __COUNTER__ macro. */
3089 TokenString str2;
3090 sym_push2(&s->next, s->v, s->type.t, 0);
3091 tok_str_new(&str2);
3092 macro_subst(&str2, nested_list, st);
3093 tok_str_add(&str2, 0);
3094 s->next->d = str2.str;
3095 }
3096 st = s->next->d;
3097 }
3098 for(;;) {
3099 int t2;
3100 TOK_GET(&t2, &st, &cval);
3101 if (t2 <= 0)
3102 break;
3103 tok_str_add2(&str, t2, &cval);
3104 }
3105 if (str.len == l0) /* expanded to empty string */
3106 tok_str_add(&str, TOK_PLCHLDR);
3107 } else {
3108 tok_str_add(&str, t);
3109 }
3110 } else {
3111 tok_str_add2(&str, t, &cval);
3112 }
3113 t0 = t1, t1 = t;
3114 }
3115 tok_str_add(&str, 0);
3116 return str.str;
3117 }
3118
3119 static char const ab_month_name[12][4] =
3120 {
3121 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3122 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3123 };
3124
3125 static int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3126 {
3127 CString cstr;
3128 int n, ret = 1;
3129
3130 cstr_new(&cstr);
3131 if (t1 != TOK_PLCHLDR)
3132 cstr_cat(&cstr, get_tok_str(t1, v1), -1);
3133 n = cstr.size;
3134 if (t2 != TOK_PLCHLDR)
3135 cstr_cat(&cstr, get_tok_str(t2, v2), -1);
3136 cstr_ccat(&cstr, '\0');
3137
3138 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3139 memcpy(file->buffer, cstr.data, cstr.size);
3140 tok_flags = 0;
3141 for (;;) {
3142 next_nomacro1();
3143 if (0 == *file->buf_ptr)
3144 break;
3145 if (is_space(tok))
3146 continue;
3147 tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid"
3148 " preprocessing token", n, cstr.data, (char*)cstr.data + n);
3149 ret = 0;
3150 break;
3151 }
3152 tcc_close();
3153 //printf("paste <%s>\n", (char*)cstr.data);
3154 cstr_free(&cstr);
3155 return ret;
3156 }
3157
3158 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3159 return the resulting string (which must be freed). */
3160 static inline int *macro_twosharps(const int *ptr0)
3161 {
3162 int t;
3163 CValue cval;
3164 TokenString macro_str1;
3165 int start_of_nosubsts = -1;
3166 const int *ptr;
3167
3168 /* we search the first '##' */
3169 for (ptr = ptr0;;) {
3170 TOK_GET(&t, &ptr, &cval);
3171 if (t == TOK_PPJOIN)
3172 break;
3173 if (t == 0)
3174 return NULL;
3175 }
3176
3177 tok_str_new(&macro_str1);
3178
3179 //tok_print(" $$$", ptr0);
3180 for (ptr = ptr0;;) {
3181 TOK_GET(&t, &ptr, &cval);
3182 if (t == 0)
3183 break;
3184 if (t == TOK_PPJOIN)
3185 continue;
3186 while (*ptr == TOK_PPJOIN) {
3187 int t1; CValue cv1;
3188 /* given 'a##b', remove nosubsts preceding 'a' */
3189 if (start_of_nosubsts >= 0)
3190 macro_str1.len = start_of_nosubsts;
3191 /* given 'a##b', remove nosubsts preceding 'b' */
3192 while ((t1 = *++ptr) == TOK_NOSUBST)
3193 ;
3194 if (t1 && t1 != TOK_PPJOIN) {
3195 TOK_GET(&t1, &ptr, &cv1);
3196 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3197 if (paste_tokens(t, &cval, t1, &cv1)) {
3198 t = tok, cval = tokc;
3199 } else {
3200 tok_str_add2(&macro_str1, t, &cval);
3201 t = t1, cval = cv1;
3202 }
3203 }
3204 }
3205 }
3206 if (t == TOK_NOSUBST) {
3207 if (start_of_nosubsts < 0)
3208 start_of_nosubsts = macro_str1.len;
3209 } else {
3210 start_of_nosubsts = -1;
3211 }
3212 tok_str_add2(&macro_str1, t, &cval);
3213 }
3214 tok_str_add(&macro_str1, 0);
3215 //tok_print(" ###", macro_str1.str);
3216 return macro_str1.str;
3217 }
3218
3219 /* peek or read [ws_str == NULL] next token from function macro call,
3220 walking up macro levels up to the file if necessary */
3221 static int next_argstream(Sym **nested_list, TokenString *ws_str)
3222 {
3223 int t;
3224 const int *p;
3225 Sym *sa;
3226
3227 for (;;) {
3228 if (macro_ptr) {
3229 p = macro_ptr, t = *p;
3230 if (ws_str) {
3231 while (is_space(t) || TOK_LINEFEED == t || TOK_PLCHLDR == t)
3232 tok_str_add(ws_str, t), t = *++p;
3233 }
3234 if (t == 0) {
3235 end_macro();
3236 /* also, end of scope for nested defined symbol */
3237 sa = *nested_list;
3238 while (sa && sa->v == 0)
3239 sa = sa->prev;
3240 if (sa)
3241 sa->v = 0;
3242 continue;
3243 }
3244 } else {
3245 ch = handle_eob();
3246 if (ws_str) {
3247 while (is_space(ch) || ch == '\n' || ch == '/') {
3248 if (ch == '/') {
3249 int c;
3250 uint8_t *p = file->buf_ptr;
3251 PEEKC(c, p);
3252 if (c == '*') {
3253 p = parse_comment(p);
3254 file->buf_ptr = p - 1;
3255 } else if (c == '/') {
3256 p = parse_line_comment(p);
3257 file->buf_ptr = p - 1;
3258 } else
3259 break;
3260 ch = ' ';
3261 }
3262 if (ch == '\n')
3263 file->line_num++;
3264 if (!(ch == '\f' || ch == '\v' || ch == '\r'))
3265 tok_str_add(ws_str, ch);
3266 cinp();
3267 }
3268 }
3269 t = ch;
3270 }
3271
3272 if (ws_str)
3273 return t;
3274 next_nomacro_spc();
3275 return tok;
3276 }
3277 }
3278
3279 /* do macro substitution of current token with macro 's' and add
3280 result to (tok_str,tok_len). 'nested_list' is the list of all
3281 macros we got inside to avoid recursing. Return non zero if no
3282 substitution needs to be done */
3283 static int macro_subst_tok(
3284 TokenString *tok_str,
3285 Sym **nested_list,
3286 Sym *s)
3287 {
3288 Sym *args, *sa, *sa1;
3289 int parlevel, t, t1, spc;
3290 TokenString str;
3291 char *cstrval;
3292 CValue cval;
3293 CString cstr;
3294 char buf[32];
3295
3296 /* if symbol is a macro, prepare substitution */
3297 /* special macros */
3298 if (tok == TOK___LINE__ || tok == TOK___COUNTER__) {
3299 t = tok == TOK___LINE__ ? file->line_num : pp_counter++;
3300 snprintf(buf, sizeof(buf), "%d", t);
3301 cstrval = buf;
3302 t1 = TOK_PPNUM;
3303 goto add_cstr1;
3304 } else if (tok == TOK___FILE__) {
3305 cstrval = file->filename;
3306 goto add_cstr;
3307 } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
3308 time_t ti;
3309 struct tm *tm;
3310
3311 time(&ti);
3312 tm = localtime(&ti);
3313 if (tok == TOK___DATE__) {
3314 snprintf(buf, sizeof(buf), "%s %2d %d",
3315 ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
3316 } else {
3317 snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
3318 tm->tm_hour, tm->tm_min, tm->tm_sec);
3319 }
3320 cstrval = buf;
3321 add_cstr:
3322 t1 = TOK_STR;
3323 add_cstr1:
3324 cstr_new(&cstr);
3325 cstr_cat(&cstr, cstrval, 0);
3326 cval.str.size = cstr.size;
3327 cval.str.data = cstr.data;
3328 tok_str_add2(tok_str, t1, &cval);
3329 cstr_free(&cstr);
3330 } else if (s->d) {
3331 int saved_parse_flags = parse_flags;
3332 int *joined_str = NULL;
3333 int *mstr = s->d;
3334
3335 if (s->type.t == MACRO_FUNC) {
3336 /* whitespace between macro name and argument list */
3337 TokenString ws_str;
3338 tok_str_new(&ws_str);
3339
3340 spc = 0;
3341 parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
3342 | PARSE_FLAG_ACCEPT_STRAYS;
3343
3344 /* get next token from argument stream */
3345 t = next_argstream(nested_list, &ws_str);
3346 if (t != '(') {
3347 /* not a macro substitution after all, restore the
3348 * macro token plus all whitespace we've read.
3349 * whitespace is intentionally not merged to preserve
3350 * newlines. */
3351 parse_flags = saved_parse_flags;
3352 tok_str_add(tok_str, tok);
3353 if (parse_flags & PARSE_FLAG_SPACES) {
3354 int i;
3355 for (i = 0; i < ws_str.len; i++)
3356 tok_str_add(tok_str, ws_str.str[i]);
3357 }
3358 tok_str_free_str(ws_str.str);
3359 return 0;
3360 } else {
3361 tok_str_free_str(ws_str.str);
3362 }
3363 do {
3364 next_nomacro(); /* eat '(' */
3365 } while (tok == TOK_PLCHLDR);
3366
3367 /* argument macro */
3368 args = NULL;
3369 sa = s->next;
3370 /* NOTE: empty args are allowed, except if no args */
3371 for(;;) {
3372 do {
3373 next_argstream(nested_list, NULL);
3374 } while (is_space(tok) || TOK_LINEFEED == tok);
3375 empty_arg:
3376 /* handle '()' case */
3377 if (!args && !sa && tok == ')')
3378 break;
3379 if (!sa)
3380 tcc_error("macro '%s' used with too many args",
3381 get_tok_str(s->v, 0));
3382 tok_str_new(&str);
3383 parlevel = spc = 0;
3384 /* NOTE: non zero sa->t indicates VA_ARGS */
3385 while ((parlevel > 0 ||
3386 (tok != ')' &&
3387 (tok != ',' || sa->type.t)))) {
3388 if (tok == TOK_EOF || tok == 0)
3389 break;
3390 if (tok == '(')
3391 parlevel++;
3392 else if (tok == ')')
3393 parlevel--;
3394 if (tok == TOK_LINEFEED)
3395 tok = ' ';
3396 if (!check_space(tok, &spc))
3397 tok_str_add2(&str, tok, &tokc);
3398 next_argstream(nested_list, NULL);
3399 }
3400 if (parlevel)
3401 expect(")");
3402 str.len -= spc;
3403 tok_str_add(&str, -1);
3404 tok_str_add(&str, 0);
3405 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3406 sa1->d = str.str;
3407 sa = sa->next;
3408 if (tok == ')') {
3409 /* special case for gcc var args: add an empty
3410 var arg argument if it is omitted */
3411 if (sa && sa->type.t && gnu_ext)
3412 goto empty_arg;
3413 break;
3414 }
3415 if (tok != ',')
3416 expect(",");
3417 }
3418 if (sa) {
3419 tcc_error("macro '%s' used with too few args",
3420 get_tok_str(s->v, 0));
3421 }
3422
3423 parse_flags = saved_parse_flags;
3424
3425 /* now subst each arg */
3426 mstr = macro_arg_subst(nested_list, mstr, args);
3427 /* free memory */
3428 sa = args;
3429 while (sa) {
3430 sa1 = sa->prev;
3431 tok_str_free_str(sa->d);
3432 if (sa->next) {
3433 tok_str_free_str(sa->next->d);
3434 sym_free(sa->next);
3435 }
3436 sym_free(sa);
3437 sa = sa1;
3438 }
3439 }
3440
3441 sym_push2(nested_list, s->v, 0, 0);
3442 parse_flags = saved_parse_flags;
3443 joined_str = macro_twosharps(mstr);
3444 macro_subst(tok_str, nested_list, joined_str ? joined_str : mstr);
3445
3446 /* pop nested defined symbol */
3447 sa1 = *nested_list;
3448 *nested_list = sa1->prev;
3449 sym_free(sa1);
3450 if (joined_str)
3451 tok_str_free_str(joined_str);
3452 if (mstr != s->d)
3453 tok_str_free_str(mstr);
3454 }
3455 return 0;
3456 }
3457
3458 /* do macro substitution of macro_str and add result to
3459 (tok_str,tok_len). 'nested_list' is the list of all macros we got
3460 inside to avoid recursing. */
3461 static void macro_subst(
3462 TokenString *tok_str,
3463 Sym **nested_list,
3464 const int *macro_str
3465 )
3466 {
3467 Sym *s;
3468 int t, spc, nosubst;
3469 CValue cval;
3470
3471 spc = nosubst = 0;
3472
3473 while (1) {
3474 TOK_GET(&t, &macro_str, &cval);
3475 if (t <= 0)
3476 break;
3477
3478 if (t >= TOK_IDENT && 0 == nosubst) {
3479 s = define_find(t);
3480 if (s == NULL)
3481 goto no_subst;
3482
3483 /* if nested substitution, do nothing */
3484 if (sym_find2(*nested_list, t)) {
3485 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3486 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3487 goto no_subst;
3488 }
3489
3490 {
3491 TokenString str;
3492 str.str = (int*)macro_str;
3493 begin_macro(&str, 2);
3494
3495 tok = t;
3496 macro_subst_tok(tok_str, nested_list, s);
3497
3498 if (str.alloc == 3) {
3499 /* already finished by reading function macro arguments */
3500 break;
3501 }
3502
3503 macro_str = macro_ptr;
3504 end_macro ();
3505 }
3506 if (tok_str->len)
3507 spc = is_space(t = tok_str->str[tok_str->lastlen]);
3508 } else {
3509 if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3510 tcc_error("stray '\\' in program");
3511 no_subst:
3512 if (!check_space(t, &spc))
3513 tok_str_add2(tok_str, t, &cval);
3514
3515 if (nosubst) {
3516 if (nosubst > 1 && (spc || (++nosubst == 3 && t == '(')))
3517 continue;
3518 nosubst = 0;
3519 }
3520 if (t == TOK_NOSUBST)
3521 nosubst = 1;
3522 }
3523 /* GCC supports 'defined' as result of a macro substitution */
3524 if (t == TOK_DEFINED && pp_expr)
3525 nosubst = 2;
3526 }
3527 }
3528
3529 /* return next token with macro substitution */
3530 ST_FUNC void next(void)
3531 {
3532 redo:
3533 if (parse_flags & PARSE_FLAG_SPACES)
3534 next_nomacro_spc();
3535 else
3536 next_nomacro();
3537
3538 if (macro_ptr) {
3539 if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3540 /* discard preprocessor markers */
3541 goto redo;
3542 } else if (tok == 0) {
3543 /* end of macro or unget token string */
3544 end_macro();
3545 goto redo;
3546 }
3547 } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3548 Sym *s;
3549 /* if reading from file, try to substitute macros */
3550 s = define_find(tok);
3551 if (s) {
3552 Sym *nested_list = NULL;
3553 tokstr_buf.len = 0;
3554 macro_subst_tok(&tokstr_buf, &nested_list, s);
3555 tok_str_add(&tokstr_buf, 0);
3556 begin_macro(&tokstr_buf, 2);
3557 goto redo;
3558 }
3559 }
3560 /* convert preprocessor tokens into C tokens */
3561 if (tok == TOK_PPNUM) {
3562 if (parse_flags & PARSE_FLAG_TOK_NUM)
3563 parse_number((char *)tokc.str.data);
3564 } else if (tok == TOK_PPSTR) {
3565 if (parse_flags & PARSE_FLAG_TOK_STR)
3566 parse_string((char *)tokc.str.data, tokc.str.size - 1);
3567 }
3568 }
3569
3570 /* push back current token and set current token to 'last_tok'. Only
3571 identifier case handled for labels. */
3572 ST_INLN void unget_tok(int last_tok)
3573 {
3574
3575 TokenString *str = tok_str_alloc();
3576 tok_str_add2(str, tok, &tokc);
3577 tok_str_add(str, 0);
3578 begin_macro(str, 1);
3579 tok = last_tok;
3580 }
3581
3582 ST_FUNC void preprocess_start(TCCState *s1, int is_asm)
3583 {
3584 CString cstr;
3585 int i;
3586
3587 s1->include_stack_ptr = s1->include_stack;
3588 s1->ifdef_stack_ptr = s1->ifdef_stack;
3589 file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3590 pp_expr = 0;
3591 pp_counter = 0;
3592 pp_debug_tok = pp_debug_symv = 0;
3593 pp_once++;
3594 pvtop = vtop = vstack - 1;
3595 s1->pack_stack[0] = 0;
3596 s1->pack_stack_ptr = s1->pack_stack;
3597
3598 set_idnum('$', s1->dollars_in_identifiers ? IS_ID : 0);
3599 set_idnum('.', is_asm ? IS_ID : 0);
3600
3601 cstr_new(&cstr);
3602 cstr_cat(&cstr, "\"", -1);
3603 cstr_cat(&cstr, file->filename, -1);
3604 cstr_cat(&cstr, "\"", 0);
3605 tcc_define_symbol(s1, "__BASE_FILE__", cstr.data);
3606
3607 cstr_reset(&cstr);
3608 for (i = 0; i < s1->nb_cmd_include_files; i++) {
3609 cstr_cat(&cstr, "#include \"", -1);
3610 cstr_cat(&cstr, s1->cmd_include_files[i], -1);
3611 cstr_cat(&cstr, "\"\n", -1);
3612 }
3613 if (cstr.size) {
3614 *s1->include_stack_ptr++ = file;
3615 tcc_open_bf(s1, "<command line>", cstr.size);
3616 memcpy(file->buffer, cstr.data, cstr.size);
3617 }
3618 cstr_free(&cstr);
3619
3620 if (is_asm)
3621 tcc_define_symbol(s1, "__ASSEMBLER__", NULL);
3622
3623 parse_flags = is_asm ? PARSE_FLAG_ASM_FILE : 0;
3624 tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3625 }
3626
3627 /* cleanup from error/setjmp */
3628 ST_FUNC void preprocess_end(TCCState *s1)
3629 {
3630 while (macro_stack)
3631 end_macro();
3632 macro_ptr = NULL;
3633 }
3634
3635 ST_FUNC void tccpp_new(TCCState *s)
3636 {
3637 int i, c;
3638 const char *p, *r;
3639
3640 /* might be used in error() before preprocess_start() */
3641 s->include_stack_ptr = s->include_stack;
3642 s->ppfp = stdout;
3643
3644 /* init isid table */
3645 for(i = CH_EOF; i<128; i++)
3646 set_idnum(i,
3647 is_space(i) ? IS_SPC
3648 : isid(i) ? IS_ID
3649 : isnum(i) ? IS_NUM
3650 : 0);
3651
3652 for(i = 128; i<256; i++)
3653 set_idnum(i, IS_ID);
3654
3655 /* init allocators */
3656 tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE);
3657 tal_new(&tokstr_alloc, TOKSTR_TAL_LIMIT, TOKSTR_TAL_SIZE);
3658 tal_new(&cstr_alloc, CSTR_TAL_LIMIT, CSTR_TAL_SIZE);
3659
3660 memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3661 cstr_new(&cstr_buf);
3662 cstr_realloc(&cstr_buf, STRING_MAX_SIZE);
3663 tok_str_new(&tokstr_buf);
3664 tok_str_realloc(&tokstr_buf, TOKSTR_MAX_SIZE);
3665
3666 tok_ident = TOK_IDENT;
3667 p = tcc_keywords;
3668 while (*p) {
3669 r = p;
3670 for(;;) {
3671 c = *r++;
3672 if (c == '\0')
3673 break;
3674 }
3675 tok_alloc(p, r - p - 1);
3676 p = r;
3677 }
3678 }
3679
3680 ST_FUNC void tccpp_delete(TCCState *s)
3681 {
3682 int i, n;
3683
3684 /* free -D and compiler defines */
3685 free_defines(NULL);
3686
3687 /* free tokens */
3688 n = tok_ident - TOK_IDENT;
3689 for(i = 0; i < n; i++)
3690 tal_free(toksym_alloc, table_ident[i]);
3691 tcc_free(table_ident);
3692 table_ident = NULL;
3693
3694 /* free static buffers */
3695 cstr_free(&tokcstr);
3696 cstr_free(&cstr_buf);
3697 cstr_free(&macro_equal_buf);
3698 tok_str_free_str(tokstr_buf.str);
3699
3700 /* free allocators */
3701 tal_delete(toksym_alloc);
3702 toksym_alloc = NULL;
3703 tal_delete(tokstr_alloc);
3704 tokstr_alloc = NULL;
3705 tal_delete(cstr_alloc);
3706 cstr_alloc = NULL;
3707 }
3708
3709 /* ------------------------------------------------------------------------- */
3710 /* tcc -E [-P[1]] [-dD} support */
3711
3712 static void tok_print(const char *msg, const int *str)
3713 {
3714 FILE *fp;
3715 int t, s = 0;
3716 CValue cval;
3717
3718 fp = tcc_state->ppfp;
3719 fprintf(fp, "%s", msg);
3720 while (str) {
3721 TOK_GET(&t, &str, &cval);
3722 if (!t)
3723 break;
3724 fprintf(fp, " %s" + s, get_tok_str(t, &cval)), s = 1;
3725 }
3726 fprintf(fp, "\n");
3727 }
3728
3729 static void pp_line(TCCState *s1, BufferedFile *f, int level)
3730 {
3731 int d = f->line_num - f->line_ref;
3732
3733 if (s1->dflag & 4)
3734 return;
3735
3736 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE) {
3737 ;
3738 } else if (level == 0 && f->line_ref && d < 8) {
3739 while (d > 0)
3740 fputs("\n", s1->ppfp), --d;
3741 } else if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
3742 fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
3743 } else {
3744 fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
3745 level > 0 ? " 1" : level < 0 ? " 2" : "");
3746 }
3747 f->line_ref = f->line_num;
3748 }
3749
3750 static void define_print(TCCState *s1, int v)
3751 {
3752 FILE *fp;
3753 Sym *s;
3754
3755 s = define_find(v);
3756 if (NULL == s || NULL == s->d)
3757 return;
3758
3759 fp = s1->ppfp;
3760 fprintf(fp, "#define %s", get_tok_str(v, NULL));
3761 if (s->type.t == MACRO_FUNC) {
3762 Sym *a = s->next;
3763 fprintf(fp,"(");
3764 if (a)
3765 for (;;) {
3766 fprintf(fp,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
3767 if (!(a = a->next))
3768 break;
3769 fprintf(fp,",");
3770 }
3771 fprintf(fp,")");
3772 }
3773 tok_print("", s->d);
3774 }
3775
3776 static void pp_debug_defines(TCCState *s1)
3777 {
3778 int v, t;
3779 const char *vs;
3780 FILE *fp;
3781
3782 t = pp_debug_tok;
3783 if (t == 0)
3784 return;
3785
3786 file->line_num--;
3787 pp_line(s1, file, 0);
3788 file->line_ref = ++file->line_num;
3789
3790 fp = s1->ppfp;
3791 v = pp_debug_symv;
3792 vs = get_tok_str(v, NULL);
3793 if (t == TOK_DEFINE) {
3794 define_print(s1, v);
3795 } else if (t == TOK_UNDEF) {
3796 fprintf(fp, "#undef %s\n", vs);
3797 } else if (t == TOK_push_macro) {
3798 fprintf(fp, "#pragma push_macro(\"%s\")\n", vs);
3799 } else if (t == TOK_pop_macro) {
3800 fprintf(fp, "#pragma pop_macro(\"%s\")\n", vs);
3801 }
3802 pp_debug_tok = 0;
3803 }
3804
3805 static void pp_debug_builtins(TCCState *s1)
3806 {
3807 int v;
3808 for (v = TOK_IDENT; v < tok_ident; ++v)
3809 define_print(s1, v);
3810 }
3811
3812 /* Add a space between tokens a and b to avoid unwanted textual pasting */
3813 static int pp_need_space(int a, int b)
3814 {
3815 return 'E' == a ? '+' == b || '-' == b
3816 : '+' == a ? TOK_INC == b || '+' == b
3817 : '-' == a ? TOK_DEC == b || '-' == b
3818 : a >= TOK_IDENT ? b >= TOK_IDENT
3819 : a == TOK_PPNUM ? b >= TOK_IDENT
3820 : 0;
3821 }
3822
3823 /* maybe hex like 0x1e */
3824 static int pp_check_he0xE(int t, const char *p)
3825 {
3826 if (t == TOK_PPNUM && toup(strchr(p, 0)[-1]) == 'E')
3827 return 'E';
3828 return t;
3829 }
3830
3831 /* Preprocess the current file */
3832 ST_FUNC int tcc_preprocess(TCCState *s1)
3833 {
3834 BufferedFile **iptr;
3835 int token_seen, spcs, level;
3836 const char *p;
3837 char white[400];
3838
3839 parse_flags = PARSE_FLAG_PREPROCESS
3840 | (parse_flags & PARSE_FLAG_ASM_FILE)
3841 | PARSE_FLAG_LINEFEED
3842 | PARSE_FLAG_SPACES
3843 | PARSE_FLAG_ACCEPT_STRAYS
3844 ;
3845 /* Credits to Fabrice Bellard's initial revision to demonstrate its
3846 capability to compile and run itself, provided all numbers are
3847 given as decimals. tcc -E -P10 will do. */
3848 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_P10)
3849 parse_flags |= PARSE_FLAG_TOK_NUM, s1->Pflag = 1;
3850
3851 #ifdef PP_BENCH
3852 /* for PP benchmarks */
3853 do next(); while (tok != TOK_EOF);
3854 return 0;
3855 #endif
3856
3857 if (s1->dflag & 1) {
3858 pp_debug_builtins(s1);
3859 s1->dflag &= ~1;
3860 }
3861
3862 token_seen = TOK_LINEFEED, spcs = 0;
3863 pp_line(s1, file, 0);
3864 for (;;) {
3865 iptr = s1->include_stack_ptr;
3866 next();
3867 if (tok == TOK_EOF)
3868 break;
3869
3870 level = s1->include_stack_ptr - iptr;
3871 if (level) {
3872 if (level > 0)
3873 pp_line(s1, *iptr, 0);
3874 pp_line(s1, file, level);
3875 }
3876 if (s1->dflag & 7) {
3877 pp_debug_defines(s1);
3878 if (s1->dflag & 4)
3879 continue;
3880 }
3881
3882 if (is_space(tok)) {
3883 if (spcs < sizeof white - 1)
3884 white[spcs++] = tok;
3885 continue;
3886 } else if (tok == TOK_LINEFEED) {
3887 spcs = 0;
3888 if (token_seen == TOK_LINEFEED)
3889 continue;
3890 ++file->line_ref;
3891 } else if (token_seen == TOK_LINEFEED) {
3892 pp_line(s1, file, 0);
3893 } else if (spcs == 0 && pp_need_space(token_seen, tok)) {
3894 white[spcs++] = ' ';
3895 }
3896
3897 white[spcs] = 0, fputs(white, s1->ppfp), spcs = 0;
3898 fputs(p = get_tok_str(tok, &tokc), s1->ppfp);
3899 token_seen = pp_check_he0xE(tok, p);
3900 }
3901 return 0;
3902 }
3903
3904 /* ------------------------------------------------------------------------- */
Tiny C Compiler main repository