]>
Commit | Line | Data |
---|---|---|
d45cf215 | 1 | /* Lexical analyzer for C and Objective C. |
1c7b145e | 2 | Copyright (C) 1987, 1988, 1989, 1992, 1994 Free Software Foundation, Inc. |
e8bbfc4e RK |
3 | |
4 | This file is part of GNU CC. | |
5 | ||
6 | GNU CC is free software; you can redistribute it and/or modify | |
7 | it under the terms of the GNU General Public License as published by | |
8 | the Free Software Foundation; either version 2, or (at your option) | |
9 | any later version. | |
10 | ||
11 | GNU CC 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 | |
14 | GNU General Public License for more details. | |
15 | ||
16 | You should have received a copy of the GNU General Public License | |
17 | along with GNU CC; see the file COPYING. If not, write to | |
18 | the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */ | |
19 | ||
20 | ||
21 | #include <stdio.h> | |
22 | #include <errno.h> | |
23 | #include <setjmp.h> | |
24 | ||
25 | #include "config.h" | |
26 | #include "rtl.h" | |
27 | #include "tree.h" | |
28 | #include "input.h" | |
29 | #include "c-lex.h" | |
30 | #include "c-tree.h" | |
31 | #include "flags.h" | |
32 | #include "c-parse.h" | |
33 | ||
8ffd69e4 RS |
34 | #include <ctype.h> |
35 | ||
e8bbfc4e RK |
36 | #ifdef MULTIBYTE_CHARS |
37 | #include <stdlib.h> | |
38 | #include <locale.h> | |
39 | #endif | |
40 | ||
41 | #ifndef errno | |
42 | extern int errno; | |
43 | #endif | |
44 | ||
45 | /* The elements of `ridpointers' are identifier nodes | |
46 | for the reserved type names and storage classes. | |
47 | It is indexed by a RID_... value. */ | |
48 | tree ridpointers[(int) RID_MAX]; | |
49 | ||
50 | /* Cause the `yydebug' variable to be defined. */ | |
51 | #define YYDEBUG 1 | |
52 | ||
53 | /* the declaration found for the last IDENTIFIER token read in. | |
54 | yylex must look this up to detect typedefs, which get token type TYPENAME, | |
55 | so it is left around in case the identifier is not a typedef but is | |
56 | used in a context which makes it a reference to a variable. */ | |
57 | tree lastiddecl; | |
58 | ||
59 | /* Nonzero enables objc features. */ | |
60 | ||
61 | int doing_objc_thang; | |
62 | ||
e31c7eec | 63 | extern tree is_class_name (); |
e8bbfc4e RK |
64 | |
65 | extern int yydebug; | |
66 | ||
67 | /* File used for outputting assembler code. */ | |
68 | extern FILE *asm_out_file; | |
69 | ||
70 | #ifndef WCHAR_TYPE_SIZE | |
71 | #ifdef INT_TYPE_SIZE | |
72 | #define WCHAR_TYPE_SIZE INT_TYPE_SIZE | |
73 | #else | |
74 | #define WCHAR_TYPE_SIZE BITS_PER_WORD | |
75 | #endif | |
76 | #endif | |
77 | ||
78 | /* Number of bytes in a wide character. */ | |
79 | #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT) | |
80 | ||
81 | static int maxtoken; /* Current nominal length of token buffer. */ | |
82 | char *token_buffer; /* Pointer to token buffer. | |
83 | Actual allocated length is maxtoken + 2. | |
84 | This is not static because objc-parse.y uses it. */ | |
85 | ||
86 | /* Nonzero if end-of-file has been seen on input. */ | |
87 | static int end_of_file; | |
88 | ||
89 | /* Buffered-back input character; faster than using ungetc. */ | |
90 | static int nextchar = -1; | |
91 | ||
92 | int check_newline (); | |
e8bbfc4e | 93 | \f |
877eef6d RK |
94 | /* Do not insert generated code into the source, instead, include it. |
95 | This allows us to build gcc automatically even for targets that | |
96 | need to add or modify the reserved keyword lists. */ | |
97 | #include "c-gperf.h" | |
e8bbfc4e RK |
98 | \f |
99 | /* Return something to represent absolute declarators containing a *. | |
100 | TARGET is the absolute declarator that the * contains. | |
101 | TYPE_QUALS is a list of modifiers such as const or volatile | |
102 | to apply to the pointer type, represented as identifiers. | |
103 | ||
104 | We return an INDIRECT_REF whose "contents" are TARGET | |
105 | and whose type is the modifier list. */ | |
106 | ||
107 | tree | |
108 | make_pointer_declarator (type_quals, target) | |
109 | tree type_quals, target; | |
110 | { | |
111 | return build1 (INDIRECT_REF, type_quals, target); | |
112 | } | |
113 | \f | |
114 | void | |
e31c7eec TW |
115 | forget_protocol_qualifiers () |
116 | { | |
117 | int i, n = sizeof wordlist / sizeof (struct resword); | |
118 | ||
119 | for (i = 0; i < n; i++) | |
ea64849e RK |
120 | if ((int) wordlist[i].rid >= (int) RID_IN |
121 | && (int) wordlist[i].rid <= (int) RID_ONEWAY) | |
e31c7eec TW |
122 | wordlist[i].name = ""; |
123 | } | |
124 | ||
125 | void | |
126 | remember_protocol_qualifiers () | |
127 | { | |
128 | int i, n = sizeof wordlist / sizeof (struct resword); | |
129 | ||
130 | for (i = 0; i < n; i++) | |
131 | if (wordlist[i].rid == RID_IN) | |
132 | wordlist[i].name = "in"; | |
133 | else if (wordlist[i].rid == RID_OUT) | |
134 | wordlist[i].name = "out"; | |
135 | else if (wordlist[i].rid == RID_INOUT) | |
136 | wordlist[i].name = "inout"; | |
137 | else if (wordlist[i].rid == RID_BYCOPY) | |
138 | wordlist[i].name = "bycopy"; | |
139 | else if (wordlist[i].rid == RID_ONEWAY) | |
140 | wordlist[i].name = "oneway"; | |
141 | } | |
142 | \f | |
143 | void | |
e8bbfc4e RK |
144 | init_lex () |
145 | { | |
146 | /* Make identifier nodes long enough for the language-specific slots. */ | |
147 | set_identifier_size (sizeof (struct lang_identifier)); | |
148 | ||
149 | /* Start it at 0, because check_newline is called at the very beginning | |
150 | and will increment it to 1. */ | |
151 | lineno = 0; | |
152 | ||
153 | #ifdef MULTIBYTE_CHARS | |
154 | /* Change to the native locale for multibyte conversions. */ | |
155 | setlocale (LC_CTYPE, ""); | |
156 | #endif | |
157 | ||
158 | maxtoken = 40; | |
159 | token_buffer = (char *) xmalloc (maxtoken + 2); | |
160 | ||
161 | ridpointers[(int) RID_INT] = get_identifier ("int"); | |
162 | ridpointers[(int) RID_CHAR] = get_identifier ("char"); | |
163 | ridpointers[(int) RID_VOID] = get_identifier ("void"); | |
164 | ridpointers[(int) RID_FLOAT] = get_identifier ("float"); | |
165 | ridpointers[(int) RID_DOUBLE] = get_identifier ("double"); | |
166 | ridpointers[(int) RID_SHORT] = get_identifier ("short"); | |
167 | ridpointers[(int) RID_LONG] = get_identifier ("long"); | |
168 | ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned"); | |
169 | ridpointers[(int) RID_SIGNED] = get_identifier ("signed"); | |
170 | ridpointers[(int) RID_INLINE] = get_identifier ("inline"); | |
171 | ridpointers[(int) RID_CONST] = get_identifier ("const"); | |
172 | ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile"); | |
173 | ridpointers[(int) RID_AUTO] = get_identifier ("auto"); | |
174 | ridpointers[(int) RID_STATIC] = get_identifier ("static"); | |
175 | ridpointers[(int) RID_EXTERN] = get_identifier ("extern"); | |
176 | ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef"); | |
177 | ridpointers[(int) RID_REGISTER] = get_identifier ("register"); | |
6506ecd7 | 178 | ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator"); |
122026c0 | 179 | ridpointers[(int) RID_COMPLEX] = get_identifier ("complex"); |
e31c7eec TW |
180 | ridpointers[(int) RID_ID] = get_identifier ("id"); |
181 | ridpointers[(int) RID_IN] = get_identifier ("in"); | |
182 | ridpointers[(int) RID_OUT] = get_identifier ("out"); | |
183 | ridpointers[(int) RID_INOUT] = get_identifier ("inout"); | |
184 | ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy"); | |
185 | ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway"); | |
186 | forget_protocol_qualifiers(); | |
e8bbfc4e RK |
187 | |
188 | /* Some options inhibit certain reserved words. | |
189 | Clear those words out of the hash table so they won't be recognized. */ | |
190 | #define UNSET_RESERVED_WORD(STRING) \ | |
191 | do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \ | |
192 | if (s) s->name = ""; } while (0) | |
193 | ||
e31c7eec TW |
194 | if (! doing_objc_thang) |
195 | UNSET_RESERVED_WORD ("id"); | |
196 | ||
e8bbfc4e RK |
197 | if (flag_traditional) |
198 | { | |
199 | UNSET_RESERVED_WORD ("const"); | |
200 | UNSET_RESERVED_WORD ("volatile"); | |
201 | UNSET_RESERVED_WORD ("typeof"); | |
202 | UNSET_RESERVED_WORD ("signed"); | |
203 | UNSET_RESERVED_WORD ("inline"); | |
6f2f09cd | 204 | UNSET_RESERVED_WORD ("iterator"); |
122026c0 | 205 | UNSET_RESERVED_WORD ("complex"); |
e8bbfc4e RK |
206 | } |
207 | if (flag_no_asm) | |
208 | { | |
209 | UNSET_RESERVED_WORD ("asm"); | |
210 | UNSET_RESERVED_WORD ("typeof"); | |
211 | UNSET_RESERVED_WORD ("inline"); | |
6f2f09cd | 212 | UNSET_RESERVED_WORD ("iterator"); |
122026c0 | 213 | UNSET_RESERVED_WORD ("complex"); |
e8bbfc4e RK |
214 | } |
215 | } | |
216 | ||
217 | void | |
218 | reinit_parse_for_function () | |
219 | { | |
220 | } | |
221 | \f | |
222 | /* Function used when yydebug is set, to print a token in more detail. */ | |
223 | ||
224 | void | |
225 | yyprint (file, yychar, yylval) | |
226 | FILE *file; | |
227 | int yychar; | |
228 | YYSTYPE yylval; | |
229 | { | |
230 | tree t; | |
231 | switch (yychar) | |
232 | { | |
233 | case IDENTIFIER: | |
234 | case TYPENAME: | |
e31c7eec | 235 | case OBJECTNAME: |
e8bbfc4e RK |
236 | t = yylval.ttype; |
237 | if (IDENTIFIER_POINTER (t)) | |
238 | fprintf (file, " `%s'", IDENTIFIER_POINTER (t)); | |
239 | break; | |
240 | ||
241 | case CONSTANT: | |
242 | t = yylval.ttype; | |
243 | if (TREE_CODE (t) == INTEGER_CST) | |
8d9bfdc5 RK |
244 | fprintf (file, |
245 | #if HOST_BITS_PER_WIDE_INT == 64 | |
246 | #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT | |
247 | " 0x%lx%016lx", | |
248 | #else | |
249 | " 0x%x%016x", | |
250 | #endif | |
251 | #else | |
252 | #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT | |
253 | " 0x%lx%08lx", | |
254 | #else | |
255 | " 0x%x%08x", | |
256 | #endif | |
257 | #endif | |
258 | TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t)); | |
e8bbfc4e RK |
259 | break; |
260 | } | |
261 | } | |
262 | ||
263 | \f | |
264 | /* If C is not whitespace, return C. | |
265 | Otherwise skip whitespace and return first nonwhite char read. */ | |
266 | ||
267 | static int | |
268 | skip_white_space (c) | |
269 | register int c; | |
270 | { | |
0dcd8cee | 271 | static int newline_warning = 0; |
e8bbfc4e RK |
272 | |
273 | for (;;) | |
274 | { | |
275 | switch (c) | |
276 | { | |
0dcd8cee RS |
277 | /* We don't recognize comments here, because |
278 | cpp output can include / and * consecutively as operators. | |
279 | Also, there's no need, since cpp removes all comments. */ | |
e8bbfc4e RK |
280 | |
281 | case '\n': | |
282 | c = check_newline (); | |
283 | break; | |
284 | ||
285 | case ' ': | |
286 | case '\t': | |
287 | case '\f': | |
e8bbfc4e RK |
288 | case '\v': |
289 | case '\b': | |
290 | c = getc (finput); | |
291 | break; | |
292 | ||
0dcd8cee RS |
293 | case '\r': |
294 | /* ANSI C says the effects of a carriage return in a source file | |
295 | are undefined. */ | |
296 | if (pedantic && !newline_warning) | |
297 | { | |
298 | warning ("carriage return in source file"); | |
299 | warning ("(we only warn about the first carriage return)"); | |
300 | newline_warning = 1; | |
301 | } | |
302 | c = getc (finput); | |
303 | break; | |
304 | ||
e8bbfc4e RK |
305 | case '\\': |
306 | c = getc (finput); | |
307 | if (c == '\n') | |
308 | lineno++; | |
309 | else | |
310 | error ("stray '\\' in program"); | |
311 | c = getc (finput); | |
312 | break; | |
313 | ||
314 | default: | |
315 | return (c); | |
316 | } | |
317 | } | |
318 | } | |
319 | ||
320 | /* Skips all of the white space at the current location in the input file. | |
321 | Must use and reset nextchar if it has the next character. */ | |
322 | ||
323 | void | |
324 | position_after_white_space () | |
325 | { | |
326 | register int c; | |
327 | ||
328 | if (nextchar != -1) | |
329 | c = nextchar, nextchar = -1; | |
330 | else | |
331 | c = getc (finput); | |
332 | ||
333 | ungetc (skip_white_space (c), finput); | |
334 | } | |
335 | ||
336 | /* Make the token buffer longer, preserving the data in it. | |
337 | P should point to just beyond the last valid character in the old buffer. | |
338 | The value we return is a pointer to the new buffer | |
339 | at a place corresponding to P. */ | |
340 | ||
341 | static char * | |
342 | extend_token_buffer (p) | |
343 | char *p; | |
344 | { | |
345 | int offset = p - token_buffer; | |
346 | ||
347 | maxtoken = maxtoken * 2 + 10; | |
348 | token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2); | |
349 | ||
350 | return token_buffer + offset; | |
351 | } | |
352 | \f | |
353 | /* At the beginning of a line, increment the line number | |
354 | and process any #-directive on this line. | |
355 | If the line is a #-directive, read the entire line and return a newline. | |
356 | Otherwise, return the line's first non-whitespace character. */ | |
357 | ||
358 | int | |
359 | check_newline () | |
360 | { | |
361 | register int c; | |
362 | register int token; | |
363 | ||
364 | lineno++; | |
365 | ||
366 | /* Read first nonwhite char on the line. */ | |
367 | ||
368 | c = getc (finput); | |
369 | while (c == ' ' || c == '\t') | |
370 | c = getc (finput); | |
371 | ||
372 | if (c != '#') | |
373 | { | |
374 | /* If not #, return it so caller will use it. */ | |
375 | return c; | |
376 | } | |
377 | ||
378 | /* Read first nonwhite char after the `#'. */ | |
379 | ||
380 | c = getc (finput); | |
381 | while (c == ' ' || c == '\t') | |
382 | c = getc (finput); | |
383 | ||
384 | /* If a letter follows, then if the word here is `line', skip | |
385 | it and ignore it; otherwise, ignore the line, with an error | |
386 | if the word isn't `pragma', `ident', `define', or `undef'. */ | |
387 | ||
388 | if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')) | |
389 | { | |
390 | if (c == 'p') | |
391 | { | |
392 | if (getc (finput) == 'r' | |
393 | && getc (finput) == 'a' | |
394 | && getc (finput) == 'g' | |
395 | && getc (finput) == 'm' | |
396 | && getc (finput) == 'a' | |
397 | && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n')) | |
398 | { | |
ca5b800a | 399 | #ifdef HANDLE_SYSV_PRAGMA |
a3100298 | 400 | return handle_sysv_pragma (finput, c); |
90835097 | 401 | #else /* !HANDLE_SYSV_PRAGMA */ |
e8bbfc4e RK |
402 | #ifdef HANDLE_PRAGMA |
403 | HANDLE_PRAGMA (finput); | |
404 | #endif /* HANDLE_PRAGMA */ | |
405 | goto skipline; | |
90835097 | 406 | #endif /* !HANDLE_SYSV_PRAGMA */ |
e8bbfc4e RK |
407 | } |
408 | } | |
409 | ||
410 | else if (c == 'd') | |
411 | { | |
412 | if (getc (finput) == 'e' | |
413 | && getc (finput) == 'f' | |
414 | && getc (finput) == 'i' | |
415 | && getc (finput) == 'n' | |
416 | && getc (finput) == 'e' | |
417 | && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n')) | |
418 | { | |
419 | #ifdef DWARF_DEBUGGING_INFO | |
420 | if ((debug_info_level == DINFO_LEVEL_VERBOSE) | |
421 | && (write_symbols == DWARF_DEBUG)) | |
422 | dwarfout_define (lineno, get_directive_line (finput)); | |
423 | #endif /* DWARF_DEBUGGING_INFO */ | |
424 | goto skipline; | |
425 | } | |
426 | } | |
427 | else if (c == 'u') | |
428 | { | |
429 | if (getc (finput) == 'n' | |
430 | && getc (finput) == 'd' | |
431 | && getc (finput) == 'e' | |
432 | && getc (finput) == 'f' | |
433 | && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n')) | |
434 | { | |
435 | #ifdef DWARF_DEBUGGING_INFO | |
436 | if ((debug_info_level == DINFO_LEVEL_VERBOSE) | |
437 | && (write_symbols == DWARF_DEBUG)) | |
438 | dwarfout_undef (lineno, get_directive_line (finput)); | |
439 | #endif /* DWARF_DEBUGGING_INFO */ | |
440 | goto skipline; | |
441 | } | |
442 | } | |
443 | else if (c == 'l') | |
444 | { | |
445 | if (getc (finput) == 'i' | |
446 | && getc (finput) == 'n' | |
447 | && getc (finput) == 'e' | |
448 | && ((c = getc (finput)) == ' ' || c == '\t')) | |
449 | goto linenum; | |
450 | } | |
451 | else if (c == 'i') | |
452 | { | |
453 | if (getc (finput) == 'd' | |
454 | && getc (finput) == 'e' | |
455 | && getc (finput) == 'n' | |
456 | && getc (finput) == 't' | |
457 | && ((c = getc (finput)) == ' ' || c == '\t')) | |
458 | { | |
459 | /* #ident. The pedantic warning is now in cccp.c. */ | |
460 | ||
461 | /* Here we have just seen `#ident '. | |
462 | A string constant should follow. */ | |
463 | ||
464 | while (c == ' ' || c == '\t') | |
465 | c = getc (finput); | |
466 | ||
467 | /* If no argument, ignore the line. */ | |
468 | if (c == '\n') | |
469 | return c; | |
470 | ||
471 | ungetc (c, finput); | |
472 | token = yylex (); | |
473 | if (token != STRING | |
474 | || TREE_CODE (yylval.ttype) != STRING_CST) | |
475 | { | |
476 | error ("invalid #ident"); | |
477 | goto skipline; | |
478 | } | |
479 | ||
480 | if (!flag_no_ident) | |
481 | { | |
482 | #ifdef ASM_OUTPUT_IDENT | |
483 | ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype)); | |
484 | #endif | |
485 | } | |
486 | ||
487 | /* Skip the rest of this line. */ | |
488 | goto skipline; | |
489 | } | |
490 | } | |
491 | ||
492 | error ("undefined or invalid # directive"); | |
493 | goto skipline; | |
494 | } | |
495 | ||
496 | linenum: | |
497 | /* Here we have either `#line' or `# <nonletter>'. | |
498 | In either case, it should be a line number; a digit should follow. */ | |
499 | ||
500 | while (c == ' ' || c == '\t') | |
501 | c = getc (finput); | |
502 | ||
503 | /* If the # is the only nonwhite char on the line, | |
504 | just ignore it. Check the new newline. */ | |
505 | if (c == '\n') | |
506 | return c; | |
507 | ||
508 | /* Something follows the #; read a token. */ | |
509 | ||
510 | ungetc (c, finput); | |
511 | token = yylex (); | |
512 | ||
513 | if (token == CONSTANT | |
514 | && TREE_CODE (yylval.ttype) == INTEGER_CST) | |
515 | { | |
516 | int old_lineno = lineno; | |
517 | int used_up = 0; | |
518 | /* subtract one, because it is the following line that | |
519 | gets the specified number */ | |
520 | ||
521 | int l = TREE_INT_CST_LOW (yylval.ttype) - 1; | |
522 | ||
523 | /* Is this the last nonwhite stuff on the line? */ | |
524 | c = getc (finput); | |
525 | while (c == ' ' || c == '\t') | |
526 | c = getc (finput); | |
527 | if (c == '\n') | |
528 | { | |
529 | /* No more: store the line number and check following line. */ | |
530 | lineno = l; | |
531 | return c; | |
532 | } | |
533 | ungetc (c, finput); | |
534 | ||
535 | /* More follows: it must be a string constant (filename). */ | |
536 | ||
cf1d841d | 537 | /* Read the string constant. */ |
e8bbfc4e | 538 | token = yylex (); |
e8bbfc4e RK |
539 | |
540 | if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST) | |
541 | { | |
542 | error ("invalid #line"); | |
543 | goto skipline; | |
544 | } | |
545 | ||
546 | input_filename | |
547 | = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1); | |
548 | strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype)); | |
549 | lineno = l; | |
550 | ||
551 | /* Each change of file name | |
552 | reinitializes whether we are now in a system header. */ | |
553 | in_system_header = 0; | |
554 | ||
555 | if (main_input_filename == 0) | |
556 | main_input_filename = input_filename; | |
557 | ||
558 | /* Is this the last nonwhite stuff on the line? */ | |
559 | c = getc (finput); | |
560 | while (c == ' ' || c == '\t') | |
561 | c = getc (finput); | |
562 | if (c == '\n') | |
5b450ae5 RS |
563 | { |
564 | /* Update the name in the top element of input_file_stack. */ | |
565 | if (input_file_stack) | |
566 | input_file_stack->name = input_filename; | |
567 | ||
568 | return c; | |
569 | } | |
e8bbfc4e RK |
570 | ungetc (c, finput); |
571 | ||
572 | token = yylex (); | |
573 | used_up = 0; | |
574 | ||
575 | /* `1' after file name means entering new file. | |
576 | `2' after file name means just left a file. */ | |
577 | ||
578 | if (token == CONSTANT | |
579 | && TREE_CODE (yylval.ttype) == INTEGER_CST) | |
580 | { | |
581 | if (TREE_INT_CST_LOW (yylval.ttype) == 1) | |
582 | { | |
583 | /* Pushing to a new file. */ | |
584 | struct file_stack *p | |
585 | = (struct file_stack *) xmalloc (sizeof (struct file_stack)); | |
586 | input_file_stack->line = old_lineno; | |
587 | p->next = input_file_stack; | |
588 | p->name = input_filename; | |
589 | input_file_stack = p; | |
590 | input_file_stack_tick++; | |
591 | #ifdef DWARF_DEBUGGING_INFO | |
592 | if (debug_info_level == DINFO_LEVEL_VERBOSE | |
593 | && write_symbols == DWARF_DEBUG) | |
594 | dwarfout_start_new_source_file (input_filename); | |
595 | #endif /* DWARF_DEBUGGING_INFO */ | |
596 | ||
597 | used_up = 1; | |
598 | } | |
599 | else if (TREE_INT_CST_LOW (yylval.ttype) == 2) | |
600 | { | |
601 | /* Popping out of a file. */ | |
602 | if (input_file_stack->next) | |
603 | { | |
604 | struct file_stack *p = input_file_stack; | |
605 | input_file_stack = p->next; | |
606 | free (p); | |
607 | input_file_stack_tick++; | |
608 | #ifdef DWARF_DEBUGGING_INFO | |
609 | if (debug_info_level == DINFO_LEVEL_VERBOSE | |
610 | && write_symbols == DWARF_DEBUG) | |
611 | dwarfout_resume_previous_source_file (input_file_stack->line); | |
612 | #endif /* DWARF_DEBUGGING_INFO */ | |
613 | } | |
614 | else | |
615 | error ("#-lines for entering and leaving files don't match"); | |
616 | ||
617 | used_up = 1; | |
618 | } | |
619 | } | |
620 | ||
5b450ae5 RS |
621 | /* Now that we've pushed or popped the input stack, |
622 | update the name in the top element. */ | |
623 | if (input_file_stack) | |
624 | input_file_stack->name = input_filename; | |
625 | ||
e8bbfc4e RK |
626 | /* If we have handled a `1' or a `2', |
627 | see if there is another number to read. */ | |
628 | if (used_up) | |
629 | { | |
630 | /* Is this the last nonwhite stuff on the line? */ | |
631 | c = getc (finput); | |
632 | while (c == ' ' || c == '\t') | |
633 | c = getc (finput); | |
634 | if (c == '\n') | |
635 | return c; | |
636 | ungetc (c, finput); | |
637 | ||
638 | token = yylex (); | |
639 | used_up = 0; | |
640 | } | |
641 | ||
642 | /* `3' after file name means this is a system header file. */ | |
643 | ||
644 | if (token == CONSTANT | |
645 | && TREE_CODE (yylval.ttype) == INTEGER_CST | |
646 | && TREE_INT_CST_LOW (yylval.ttype) == 3) | |
647 | in_system_header = 1; | |
648 | } | |
649 | else | |
650 | error ("invalid #-line"); | |
651 | ||
652 | /* skip the rest of this line. */ | |
653 | skipline: | |
654 | if (c == '\n') | |
655 | return c; | |
656 | while ((c = getc (finput)) != EOF && c != '\n'); | |
657 | return c; | |
658 | } | |
659 | \f | |
ca5b800a RS |
660 | #ifdef HANDLE_SYSV_PRAGMA |
661 | ||
662 | /* Handle a #pragma directive. INPUT is the current input stream, | |
a3100298 TW |
663 | and C is a character to reread. Processes the entire input line |
664 | and returns a character for the caller to reread: either \n or EOF. */ | |
ca5b800a RS |
665 | |
666 | /* This function has to be in this file, in order to get at | |
667 | the token types. */ | |
668 | ||
669 | int | |
670 | handle_sysv_pragma (input, c) | |
671 | FILE *input; | |
672 | int c; | |
673 | { | |
a3100298 | 674 | for (;;) |
ca5b800a | 675 | { |
a3100298 TW |
676 | while (c == ' ' || c == '\t') |
677 | c = getc (input); | |
678 | if (c == '\n' || c == EOF) | |
679 | { | |
680 | handle_pragma_token (0, 0); | |
681 | return c; | |
682 | } | |
683 | ungetc (c, input); | |
684 | switch (yylex ()) | |
685 | { | |
686 | case IDENTIFIER: | |
687 | case TYPENAME: | |
688 | case STRING: | |
689 | case CONSTANT: | |
690 | handle_pragma_token (token_buffer, yylval.ttype); | |
691 | break; | |
692 | default: | |
693 | handle_pragma_token (token_buffer, 0); | |
694 | } | |
695 | if (nextchar >= 0) | |
696 | c = nextchar, nextchar = -1; | |
697 | else | |
698 | c = getc (input); | |
ca5b800a | 699 | } |
ca5b800a RS |
700 | } |
701 | ||
702 | #endif /* HANDLE_SYSV_PRAGMA */ | |
703 | \f | |
e8bbfc4e RK |
704 | #define ENDFILE -1 /* token that represents end-of-file */ |
705 | ||
706 | /* Read an escape sequence, returning its equivalent as a character, | |
b70af677 | 707 | or store 1 in *ignore_ptr if it is backslash-newline. */ |
e8bbfc4e RK |
708 | |
709 | static int | |
b70af677 RS |
710 | readescape (ignore_ptr) |
711 | int *ignore_ptr; | |
e8bbfc4e RK |
712 | { |
713 | register int c = getc (finput); | |
714 | register int code; | |
715 | register unsigned count; | |
1c7b145e | 716 | unsigned firstdig = 0; |
8696da34 | 717 | int nonnull; |
e8bbfc4e RK |
718 | |
719 | switch (c) | |
720 | { | |
721 | case 'x': | |
722 | if (warn_traditional) | |
723 | warning ("the meaning of `\\x' varies with -traditional"); | |
724 | ||
725 | if (flag_traditional) | |
726 | return c; | |
727 | ||
728 | code = 0; | |
729 | count = 0; | |
8696da34 | 730 | nonnull = 0; |
e8bbfc4e RK |
731 | while (1) |
732 | { | |
733 | c = getc (finput); | |
734 | if (!(c >= 'a' && c <= 'f') | |
735 | && !(c >= 'A' && c <= 'F') | |
736 | && !(c >= '0' && c <= '9')) | |
737 | { | |
738 | ungetc (c, finput); | |
739 | break; | |
740 | } | |
741 | code *= 16; | |
742 | if (c >= 'a' && c <= 'f') | |
743 | code += c - 'a' + 10; | |
744 | if (c >= 'A' && c <= 'F') | |
745 | code += c - 'A' + 10; | |
746 | if (c >= '0' && c <= '9') | |
747 | code += c - '0'; | |
8696da34 RS |
748 | if (code != 0 || count != 0) |
749 | { | |
750 | if (count == 0) | |
751 | firstdig = code; | |
752 | count++; | |
753 | } | |
754 | nonnull = 1; | |
e8bbfc4e | 755 | } |
8696da34 | 756 | if (! nonnull) |
e8bbfc4e | 757 | error ("\\x used with no following hex digits"); |
be63d912 RS |
758 | else if (count == 0) |
759 | /* Digits are all 0's. Ok. */ | |
760 | ; | |
e8bbfc4e RK |
761 | else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node) |
762 | || (count > 1 | |
763 | && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4)) | |
764 | <= firstdig))) | |
765 | pedwarn ("hex escape out of range"); | |
766 | return code; | |
767 | ||
768 | case '0': case '1': case '2': case '3': case '4': | |
769 | case '5': case '6': case '7': | |
770 | code = 0; | |
771 | count = 0; | |
772 | while ((c <= '7') && (c >= '0') && (count++ < 3)) | |
773 | { | |
774 | code = (code * 8) + (c - '0'); | |
775 | c = getc (finput); | |
776 | } | |
777 | ungetc (c, finput); | |
778 | return code; | |
779 | ||
780 | case '\\': case '\'': case '"': | |
781 | return c; | |
782 | ||
783 | case '\n': | |
784 | lineno++; | |
b70af677 RS |
785 | *ignore_ptr = 1; |
786 | return 0; | |
e8bbfc4e RK |
787 | |
788 | case 'n': | |
789 | return TARGET_NEWLINE; | |
790 | ||
791 | case 't': | |
792 | return TARGET_TAB; | |
793 | ||
794 | case 'r': | |
795 | return TARGET_CR; | |
796 | ||
797 | case 'f': | |
798 | return TARGET_FF; | |
799 | ||
800 | case 'b': | |
801 | return TARGET_BS; | |
802 | ||
803 | case 'a': | |
804 | if (warn_traditional) | |
805 | warning ("the meaning of `\\a' varies with -traditional"); | |
806 | ||
807 | if (flag_traditional) | |
808 | return c; | |
809 | return TARGET_BELL; | |
810 | ||
811 | case 'v': | |
812 | #if 0 /* Vertical tab is present in common usage compilers. */ | |
813 | if (flag_traditional) | |
814 | return c; | |
815 | #endif | |
816 | return TARGET_VT; | |
817 | ||
dad112ca | 818 | case 'e': |
e8bbfc4e | 819 | case 'E': |
dad112ca RS |
820 | if (pedantic) |
821 | pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c); | |
e8bbfc4e RK |
822 | return 033; |
823 | ||
824 | case '?': | |
825 | return c; | |
826 | ||
827 | /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */ | |
828 | case '(': | |
829 | case '{': | |
830 | case '[': | |
3285fe92 RS |
831 | /* `\%' is used to prevent SCCS from getting confused. */ |
832 | case '%': | |
e8bbfc4e RK |
833 | if (pedantic) |
834 | pedwarn ("non-ANSI escape sequence `\\%c'", c); | |
835 | return c; | |
836 | } | |
7e6b19dc | 837 | if (c >= 040 && c < 0177) |
e8bbfc4e RK |
838 | pedwarn ("unknown escape sequence `\\%c'", c); |
839 | else | |
840 | pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c); | |
841 | return c; | |
842 | } | |
843 | \f | |
844 | void | |
845 | yyerror (string) | |
846 | char *string; | |
847 | { | |
848 | char buf[200]; | |
849 | ||
850 | strcpy (buf, string); | |
851 | ||
852 | /* We can't print string and character constants well | |
853 | because the token_buffer contains the result of processing escapes. */ | |
854 | if (end_of_file) | |
855 | strcat (buf, " at end of input"); | |
856 | else if (token_buffer[0] == 0) | |
857 | strcat (buf, " at null character"); | |
858 | else if (token_buffer[0] == '"') | |
859 | strcat (buf, " before string constant"); | |
860 | else if (token_buffer[0] == '\'') | |
861 | strcat (buf, " before character constant"); | |
862 | else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177) | |
863 | sprintf (buf + strlen (buf), " before character 0%o", | |
864 | (unsigned char) token_buffer[0]); | |
865 | else | |
866 | strcat (buf, " before `%s'"); | |
867 | ||
868 | error (buf, token_buffer); | |
869 | } | |
870 | ||
871 | #if 0 | |
872 | ||
873 | struct try_type | |
874 | { | |
875 | tree *node_var; | |
876 | char unsigned_flag; | |
877 | char long_flag; | |
878 | char long_long_flag; | |
879 | }; | |
880 | ||
881 | struct try_type type_sequence[] = | |
882 | { | |
883 | { &integer_type_node, 0, 0, 0}, | |
884 | { &unsigned_type_node, 1, 0, 0}, | |
885 | { &long_integer_type_node, 0, 1, 0}, | |
886 | { &long_unsigned_type_node, 1, 1, 0}, | |
887 | { &long_long_integer_type_node, 0, 1, 1}, | |
888 | { &long_long_unsigned_type_node, 1, 1, 1} | |
889 | }; | |
890 | #endif /* 0 */ | |
891 | \f | |
892 | int | |
893 | yylex () | |
894 | { | |
895 | register int c; | |
896 | register char *p; | |
897 | register int value; | |
898 | int wide_flag = 0; | |
e31c7eec | 899 | int objc_flag = 0; |
e8bbfc4e RK |
900 | |
901 | if (nextchar >= 0) | |
902 | c = nextchar, nextchar = -1; | |
903 | else | |
904 | c = getc (finput); | |
905 | ||
906 | /* Effectively do c = skip_white_space (c) | |
907 | but do it faster in the usual cases. */ | |
908 | while (1) | |
909 | switch (c) | |
910 | { | |
e8bbfc4e RK |
911 | case ' ': |
912 | case '\t': | |
913 | case '\f': | |
914 | case '\v': | |
915 | case '\b': | |
916 | c = getc (finput); | |
917 | break; | |
918 | ||
80f5b9d8 RS |
919 | case '\r': |
920 | /* Call skip_white_space so we can warn if appropriate. */ | |
921 | ||
e8bbfc4e RK |
922 | case '\n': |
923 | case '/': | |
924 | case '\\': | |
925 | c = skip_white_space (c); | |
926 | default: | |
927 | goto found_nonwhite; | |
928 | } | |
929 | found_nonwhite: | |
930 | ||
931 | token_buffer[0] = c; | |
932 | token_buffer[1] = 0; | |
933 | ||
934 | /* yylloc.first_line = lineno; */ | |
935 | ||
936 | switch (c) | |
937 | { | |
938 | case EOF: | |
939 | end_of_file = 1; | |
940 | token_buffer[0] = 0; | |
941 | value = ENDFILE; | |
942 | break; | |
943 | ||
944 | case '$': | |
945 | if (dollars_in_ident) | |
946 | goto letter; | |
947 | return '$'; | |
948 | ||
949 | case 'L': | |
950 | /* Capital L may start a wide-string or wide-character constant. */ | |
951 | { | |
952 | register int c = getc (finput); | |
953 | if (c == '\'') | |
954 | { | |
955 | wide_flag = 1; | |
956 | goto char_constant; | |
957 | } | |
958 | if (c == '"') | |
959 | { | |
960 | wide_flag = 1; | |
961 | goto string_constant; | |
962 | } | |
963 | ungetc (c, finput); | |
964 | } | |
965 | goto letter; | |
966 | ||
967 | case '@': | |
968 | if (!doing_objc_thang) | |
969 | { | |
970 | value = c; | |
971 | break; | |
972 | } | |
e31c7eec | 973 | else |
e8bbfc4e | 974 | { |
e31c7eec TW |
975 | /* '@' may start a constant string object. */ |
976 | register int c = getc(finput); | |
977 | if (c == '"') | |
978 | { | |
979 | objc_flag = 1; | |
980 | goto string_constant; | |
981 | } | |
982 | ungetc(c, finput); | |
983 | /* Fall through to treat '@' as the start of an indentifier. */ | |
e8bbfc4e RK |
984 | } |
985 | ||
e8bbfc4e RK |
986 | case 'A': case 'B': case 'C': case 'D': case 'E': |
987 | case 'F': case 'G': case 'H': case 'I': case 'J': | |
988 | case 'K': case 'M': case 'N': case 'O': | |
989 | case 'P': case 'Q': case 'R': case 'S': case 'T': | |
990 | case 'U': case 'V': case 'W': case 'X': case 'Y': | |
991 | case 'Z': | |
992 | case 'a': case 'b': case 'c': case 'd': case 'e': | |
993 | case 'f': case 'g': case 'h': case 'i': case 'j': | |
994 | case 'k': case 'l': case 'm': case 'n': case 'o': | |
995 | case 'p': case 'q': case 'r': case 's': case 't': | |
996 | case 'u': case 'v': case 'w': case 'x': case 'y': | |
997 | case 'z': | |
998 | case '_': | |
999 | letter: | |
1000 | p = token_buffer; | |
1001 | while (isalnum (c) || c == '_' || c == '$' || c == '@') | |
1002 | { | |
8355fc27 RS |
1003 | /* Make sure this char really belongs in an identifier. */ |
1004 | if (c == '@' && ! doing_objc_thang) | |
1005 | break; | |
e8bbfc4e RK |
1006 | if (c == '$' && ! dollars_in_ident) |
1007 | break; | |
1008 | ||
8355fc27 RS |
1009 | if (p >= token_buffer + maxtoken) |
1010 | p = extend_token_buffer (p); | |
1011 | ||
e8bbfc4e RK |
1012 | *p++ = c; |
1013 | c = getc (finput); | |
1014 | } | |
1015 | ||
1016 | *p = 0; | |
1017 | nextchar = c; | |
1018 | ||
1019 | value = IDENTIFIER; | |
1020 | yylval.itype = 0; | |
1021 | ||
1022 | /* Try to recognize a keyword. Uses minimum-perfect hash function */ | |
1023 | ||
1024 | { | |
1025 | register struct resword *ptr; | |
1026 | ||
1027 | if (ptr = is_reserved_word (token_buffer, p - token_buffer)) | |
1028 | { | |
1029 | if (ptr->rid) | |
1030 | yylval.ttype = ridpointers[(int) ptr->rid]; | |
1031 | value = (int) ptr->token; | |
1032 | ||
e31c7eec TW |
1033 | /* Only return OBJECTNAME if it is a typedef. */ |
1034 | if (doing_objc_thang && value == OBJECTNAME) | |
1035 | { | |
1036 | lastiddecl = lookup_name(yylval.ttype); | |
1037 | ||
1038 | if (lastiddecl == NULL_TREE | |
1039 | || TREE_CODE (lastiddecl) != TYPE_DECL) | |
1040 | value = IDENTIFIER; | |
1041 | } | |
1042 | ||
e8bbfc4e RK |
1043 | /* Even if we decided to recognize asm, still perhaps warn. */ |
1044 | if (pedantic | |
d45cf215 | 1045 | && (value == ASM_KEYWORD || value == TYPEOF |
e8bbfc4e RK |
1046 | || ptr->rid == RID_INLINE) |
1047 | && token_buffer[0] != '_') | |
1048 | pedwarn ("ANSI does not permit the keyword `%s'", | |
1049 | token_buffer); | |
1050 | } | |
1051 | } | |
1052 | ||
1053 | /* If we did not find a keyword, look for an identifier | |
1054 | (or a typename). */ | |
1055 | ||
1056 | if (value == IDENTIFIER) | |
1057 | { | |
e31c7eec TW |
1058 | if (token_buffer[0] == '@') |
1059 | error("invalid identifier `%s'", token_buffer); | |
1060 | ||
e8bbfc4e RK |
1061 | yylval.ttype = get_identifier (token_buffer); |
1062 | lastiddecl = lookup_name (yylval.ttype); | |
1063 | ||
1064 | if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL) | |
1065 | value = TYPENAME; | |
6dbf678a RS |
1066 | /* A user-invisible read-only initialized variable |
1067 | should be replaced by its value. | |
1068 | We handle only strings since that's the only case used in C. */ | |
1069 | else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL | |
1070 | && DECL_IGNORED_P (lastiddecl) | |
1071 | && TREE_READONLY (lastiddecl) | |
1072 | && DECL_INITIAL (lastiddecl) != 0 | |
1073 | && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST) | |
1074 | { | |
8f7ac220 RS |
1075 | tree stringval = DECL_INITIAL (lastiddecl); |
1076 | ||
1077 | /* Copy the string value so that we won't clobber anything | |
1078 | if we put something in the TREE_CHAIN of this one. */ | |
1079 | yylval.ttype = build_string (TREE_STRING_LENGTH (stringval), | |
1080 | TREE_STRING_POINTER (stringval)); | |
6dbf678a RS |
1081 | value = STRING; |
1082 | } | |
e8bbfc4e RK |
1083 | else if (doing_objc_thang) |
1084 | { | |
e31c7eec | 1085 | tree objc_interface_decl = is_class_name (yylval.ttype); |
e8bbfc4e RK |
1086 | |
1087 | if (objc_interface_decl) | |
1088 | { | |
1089 | value = CLASSNAME; | |
1090 | yylval.ttype = objc_interface_decl; | |
1091 | } | |
1092 | } | |
1093 | } | |
1094 | ||
1095 | break; | |
1096 | ||
1097 | case '0': case '1': case '2': case '3': case '4': | |
1098 | case '5': case '6': case '7': case '8': case '9': | |
1099 | case '.': | |
1100 | { | |
1101 | int base = 10; | |
1102 | int count = 0; | |
1103 | int largest_digit = 0; | |
1104 | int numdigits = 0; | |
1105 | /* for multi-precision arithmetic, | |
4ca827d0 RS |
1106 | we actually store only HOST_BITS_PER_CHAR bits in each part. |
1107 | The number of parts is chosen so as to be sufficient to hold | |
9b0995aa RK |
1108 | the enough bits to fit into the two HOST_WIDE_INTs that contain |
1109 | the integer value (this is always at least as many bits as are | |
1110 | in a target `long long' value, but may be wider). */ | |
1111 | #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2) | |
4ca827d0 | 1112 | int parts[TOTAL_PARTS]; |
e8bbfc4e RK |
1113 | int overflow = 0; |
1114 | ||
1115 | enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag | |
1116 | = NOT_FLOAT; | |
1117 | ||
4ca827d0 RS |
1118 | for (count = 0; count < TOTAL_PARTS; count++) |
1119 | parts[count] = 0; | |
e8bbfc4e RK |
1120 | |
1121 | p = token_buffer; | |
1122 | *p++ = c; | |
1123 | ||
1124 | if (c == '0') | |
1125 | { | |
1126 | *p++ = (c = getc (finput)); | |
1127 | if ((c == 'x') || (c == 'X')) | |
1128 | { | |
1129 | base = 16; | |
1130 | *p++ = (c = getc (finput)); | |
1131 | } | |
1132 | /* Leading 0 forces octal unless the 0 is the only digit. */ | |
1133 | else if (c >= '0' && c <= '9') | |
1134 | { | |
1135 | base = 8; | |
1136 | numdigits++; | |
1137 | } | |
1138 | else | |
1139 | numdigits++; | |
1140 | } | |
1141 | ||
1142 | /* Read all the digits-and-decimal-points. */ | |
1143 | ||
1144 | while (c == '.' | |
122026c0 RS |
1145 | || (isalnum (c) && c != 'l' && c != 'L' |
1146 | && c != 'u' && c != 'U' | |
1147 | && c != 'i' && c != 'I' && c != 'j' && c != 'J' | |
e8bbfc4e RK |
1148 | && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F'))))) |
1149 | { | |
1150 | if (c == '.') | |
1151 | { | |
1152 | if (base == 16) | |
1153 | error ("floating constant may not be in radix 16"); | |
b408c3fd JW |
1154 | if (floatflag == TOO_MANY_POINTS) |
1155 | /* We have already emitted an error. Don't need another. */ | |
1156 | ; | |
1157 | else if (floatflag == AFTER_POINT) | |
e8bbfc4e RK |
1158 | { |
1159 | error ("malformed floating constant"); | |
1160 | floatflag = TOO_MANY_POINTS; | |
b408c3fd JW |
1161 | /* Avoid another error from atof by forcing all characters |
1162 | from here on to be ignored. */ | |
1163 | p[-1] = '\0'; | |
e8bbfc4e RK |
1164 | } |
1165 | else | |
1166 | floatflag = AFTER_POINT; | |
1167 | ||
1168 | base = 10; | |
1169 | *p++ = c = getc (finput); | |
1170 | /* Accept '.' as the start of a floating-point number | |
1171 | only when it is followed by a digit. | |
1172 | Otherwise, unread the following non-digit | |
1173 | and use the '.' as a structural token. */ | |
1174 | if (p == token_buffer + 2 && !isdigit (c)) | |
1175 | { | |
1176 | if (c == '.') | |
1177 | { | |
1178 | c = getc (finput); | |
1179 | if (c == '.') | |
1180 | { | |
1181 | *p++ = c; | |
1182 | *p = 0; | |
1183 | return ELLIPSIS; | |
1184 | } | |
1185 | error ("parse error at `..'"); | |
1186 | } | |
1187 | ungetc (c, finput); | |
1188 | token_buffer[1] = 0; | |
1189 | value = '.'; | |
1190 | goto done; | |
1191 | } | |
1192 | } | |
1193 | else | |
1194 | { | |
1195 | /* It is not a decimal point. | |
1196 | It should be a digit (perhaps a hex digit). */ | |
1197 | ||
1198 | if (isdigit (c)) | |
1199 | { | |
1200 | c = c - '0'; | |
1201 | } | |
1202 | else if (base <= 10) | |
1203 | { | |
f3acae4d | 1204 | if (c == 'e' || c == 'E') |
e8bbfc4e RK |
1205 | { |
1206 | base = 10; | |
1207 | floatflag = AFTER_POINT; | |
1208 | break; /* start of exponent */ | |
1209 | } | |
1210 | error ("nondigits in number and not hexadecimal"); | |
1211 | c = 0; | |
1212 | } | |
1213 | else if (c >= 'a') | |
1214 | { | |
1215 | c = c - 'a' + 10; | |
1216 | } | |
1217 | else | |
1218 | { | |
1219 | c = c - 'A' + 10; | |
1220 | } | |
1221 | if (c >= largest_digit) | |
1222 | largest_digit = c; | |
1223 | numdigits++; | |
1224 | ||
4ca827d0 | 1225 | for (count = 0; count < TOTAL_PARTS; count++) |
e8bbfc4e | 1226 | { |
4ca827d0 | 1227 | parts[count] *= base; |
e8bbfc4e RK |
1228 | if (count) |
1229 | { | |
4ca827d0 RS |
1230 | parts[count] |
1231 | += (parts[count-1] >> HOST_BITS_PER_CHAR); | |
1232 | parts[count-1] | |
1233 | &= (1 << HOST_BITS_PER_CHAR) - 1; | |
e8bbfc4e | 1234 | } |
4ca827d0 RS |
1235 | else |
1236 | parts[0] += c; | |
e8bbfc4e RK |
1237 | } |
1238 | ||
4ca827d0 RS |
1239 | /* If the extra highest-order part ever gets anything in it, |
1240 | the number is certainly too big. */ | |
1241 | if (parts[TOTAL_PARTS - 1] != 0) | |
1242 | overflow = 1; | |
e8bbfc4e RK |
1243 | |
1244 | if (p >= token_buffer + maxtoken - 3) | |
1245 | p = extend_token_buffer (p); | |
1246 | *p++ = (c = getc (finput)); | |
1247 | } | |
1248 | } | |
1249 | ||
1250 | if (numdigits == 0) | |
1251 | error ("numeric constant with no digits"); | |
1252 | ||
1253 | if (largest_digit >= base) | |
1254 | error ("numeric constant contains digits beyond the radix"); | |
1255 | ||
1256 | /* Remove terminating char from the token buffer and delimit the string */ | |
1257 | *--p = 0; | |
1258 | ||
1259 | if (floatflag != NOT_FLOAT) | |
1260 | { | |
1261 | tree type = double_type_node; | |
63ea33ce | 1262 | int garbage_chars = 0, exceeds_double = 0; |
122026c0 | 1263 | int imag = 0; |
e8bbfc4e RK |
1264 | REAL_VALUE_TYPE value; |
1265 | jmp_buf handler; | |
1266 | ||
1267 | /* Read explicit exponent if any, and put it in tokenbuf. */ | |
1268 | ||
1269 | if ((c == 'e') || (c == 'E')) | |
1270 | { | |
1271 | if (p >= token_buffer + maxtoken - 3) | |
1272 | p = extend_token_buffer (p); | |
1273 | *p++ = c; | |
1274 | c = getc (finput); | |
1275 | if ((c == '+') || (c == '-')) | |
1276 | { | |
1277 | *p++ = c; | |
1278 | c = getc (finput); | |
1279 | } | |
1280 | if (! isdigit (c)) | |
1281 | error ("floating constant exponent has no digits"); | |
1282 | while (isdigit (c)) | |
1283 | { | |
1284 | if (p >= token_buffer + maxtoken - 3) | |
1285 | p = extend_token_buffer (p); | |
1286 | *p++ = c; | |
1287 | c = getc (finput); | |
1288 | } | |
1289 | } | |
1290 | ||
1291 | *p = 0; | |
1292 | errno = 0; | |
1293 | ||
1294 | /* Convert string to a double, checking for overflow. */ | |
1295 | if (setjmp (handler)) | |
1296 | { | |
1297 | error ("floating constant out of range"); | |
1298 | value = dconst0; | |
1299 | } | |
1300 | else | |
1301 | { | |
e26ceb28 RS |
1302 | int fflag = 0, lflag = 0; |
1303 | /* Copy token_buffer now, while it has just the number | |
1304 | and not the suffixes; once we add `f' or `i', | |
1305 | REAL_VALUE_ATOF may not work any more. */ | |
1306 | char *copy = (char *) alloca (p - token_buffer + 1); | |
1307 | bcopy (token_buffer, copy, p - token_buffer + 1); | |
1308 | ||
e8bbfc4e | 1309 | set_float_handler (handler); |
1454a5ba | 1310 | |
e26ceb28 RS |
1311 | while (1) |
1312 | { | |
1313 | int lose = 0; | |
e8bbfc4e | 1314 | |
e26ceb28 RS |
1315 | /* Read the suffixes to choose a data type. */ |
1316 | switch (c) | |
1317 | { | |
1318 | case 'f': case 'F': | |
1319 | if (fflag) | |
1320 | error ("more than one `f' in numeric constant"); | |
1321 | fflag = 1; | |
1322 | break; | |
1323 | ||
1324 | case 'l': case 'L': | |
1325 | if (lflag) | |
1326 | error ("more than one `l' in numeric constant"); | |
1327 | lflag = 1; | |
1328 | break; | |
1329 | ||
1330 | case 'i': case 'I': | |
1331 | if (imag) | |
1332 | error ("more than one `i' or `j' in numeric constant"); | |
b8bdbf16 JW |
1333 | else if (pedantic) |
1334 | pedwarn ("ANSI C forbids imaginary numeric constants"); | |
e26ceb28 RS |
1335 | imag = 1; |
1336 | break; | |
1337 | ||
1338 | default: | |
1339 | lose = 1; | |
1340 | } | |
1341 | ||
1342 | if (lose) | |
1343 | break; | |
1344 | ||
1345 | if (p >= token_buffer + maxtoken - 3) | |
1346 | p = extend_token_buffer (p); | |
1347 | *p++ = c; | |
1348 | *p = 0; | |
1349 | c = getc (finput); | |
1350 | } | |
1351 | ||
1352 | /* The second argument, machine_mode, of REAL_VALUE_ATOF | |
1353 | tells the desired precision of the binary result | |
1354 | of decimal-to-binary conversion. */ | |
1355 | ||
1356 | if (fflag) | |
1357 | { | |
1358 | if (lflag) | |
1359 | error ("both `f' and `l' in floating constant"); | |
1360 | ||
1361 | type = float_type_node; | |
1362 | value = REAL_VALUE_ATOF (copy, TYPE_MODE (type)); | |
e0bf6b33 JW |
1363 | /* A diagnostic is required here by some ANSI C testsuites. |
1364 | This is not pedwarn, become some people don't want | |
1365 | an error for this. */ | |
1366 | if (REAL_VALUE_ISINF (value) && pedantic) | |
1367 | warning ("floating point number exceeds range of `float'"); | |
e26ceb28 RS |
1368 | } |
1369 | else if (lflag) | |
1370 | { | |
1371 | type = long_double_type_node; | |
1372 | value = REAL_VALUE_ATOF (copy, TYPE_MODE (type)); | |
e0bf6b33 JW |
1373 | if (REAL_VALUE_ISINF (value) && pedantic) |
1374 | warning ("floating point number exceeds range of `long double'"); | |
e26ceb28 RS |
1375 | } |
1376 | else | |
1377 | { | |
1378 | value = REAL_VALUE_ATOF (copy, TYPE_MODE (type)); | |
e0bf6b33 JW |
1379 | if (REAL_VALUE_ISINF (value) && pedantic) |
1380 | warning ("floating point number exceeds range of `double'"); | |
e26ceb28 RS |
1381 | } |
1382 | ||
1383 | set_float_handler (NULL_PTR); | |
1454a5ba RS |
1384 | } |
1385 | #ifdef ERANGE | |
1386 | if (errno == ERANGE && !flag_traditional && pedantic) | |
1387 | { | |
1388 | /* ERANGE is also reported for underflow, | |
1389 | so test the value to distinguish overflow from that. */ | |
e0bf6b33 JW |
1390 | if (REAL_VALUES_LESS (dconst1, value) |
1391 | || REAL_VALUES_LESS (value, dconstm1)) | |
1454a5ba | 1392 | { |
e0bf6b33 | 1393 | warning ("floating point number exceeds range of `double'"); |
1454a5ba RS |
1394 | exceeds_double = 1; |
1395 | } | |
1396 | } | |
1397 | #endif | |
e26ceb28 | 1398 | garbage_chars = 0; |
f3acae4d RS |
1399 | while (isalnum (c) || c == '.' || c == '_' |
1400 | || (!flag_traditional && (c == '+' || c == '-') | |
1401 | && (p[-1] == 'e' || p[-1] == 'E'))) | |
e8bbfc4e | 1402 | { |
e8bbfc4e RK |
1403 | if (p >= token_buffer + maxtoken - 3) |
1404 | p = extend_token_buffer (p); | |
1405 | *p++ = c; | |
1406 | c = getc (finput); | |
63ea33ce | 1407 | garbage_chars++; |
e8bbfc4e | 1408 | } |
63ea33ce RS |
1409 | if (garbage_chars > 0) |
1410 | error ("garbage at end of number"); | |
e8bbfc4e | 1411 | |
a47a0ed5 RK |
1412 | /* If the result is not a number, assume it must have been |
1413 | due to some error message above, so silently convert | |
1414 | it to a zero. */ | |
1415 | if (REAL_VALUE_ISNAN (value)) | |
1416 | value = dconst0; | |
1417 | ||
e8bbfc4e | 1418 | /* Create a node with determined type and value. */ |
122026c0 RS |
1419 | if (imag) |
1420 | yylval.ttype = build_complex (convert (type, integer_zero_node), | |
1421 | build_real (type, value)); | |
1422 | else | |
1423 | yylval.ttype = build_real (type, value); | |
e8bbfc4e RK |
1424 | |
1425 | ungetc (c, finput); | |
1426 | *p = 0; | |
1427 | } | |
1428 | else | |
1429 | { | |
1430 | tree traditional_type, ansi_type, type; | |
8d9bfdc5 | 1431 | HOST_WIDE_INT high, low; |
e8bbfc4e RK |
1432 | int spec_unsigned = 0; |
1433 | int spec_long = 0; | |
1434 | int spec_long_long = 0; | |
122026c0 | 1435 | int spec_imag = 0; |
e8bbfc4e RK |
1436 | int bytes, warn, i; |
1437 | ||
1438 | while (1) | |
1439 | { | |
1440 | if (c == 'u' || c == 'U') | |
1441 | { | |
1442 | if (spec_unsigned) | |
1443 | error ("two `u's in integer constant"); | |
1444 | spec_unsigned = 1; | |
1445 | } | |
1446 | else if (c == 'l' || c == 'L') | |
1447 | { | |
1448 | if (spec_long) | |
1449 | { | |
1450 | if (spec_long_long) | |
1451 | error ("three `l's in integer constant"); | |
1452 | else if (pedantic) | |
1453 | pedwarn ("ANSI C forbids long long integer constants"); | |
1454 | spec_long_long = 1; | |
1455 | } | |
1456 | spec_long = 1; | |
1457 | } | |
122026c0 RS |
1458 | else if (c == 'i' || c == 'j' || c == 'I' || c == 'J') |
1459 | { | |
1460 | if (spec_imag) | |
1461 | error ("more than one `i' or `j' in numeric constant"); | |
b8bdbf16 JW |
1462 | else if (pedantic) |
1463 | pedwarn ("ANSI C forbids imaginary numeric constants"); | |
122026c0 RS |
1464 | spec_imag = 1; |
1465 | } | |
e8bbfc4e RK |
1466 | else |
1467 | { | |
f3acae4d RS |
1468 | if (isalnum (c) || c == '.' || c == '_' |
1469 | || (!flag_traditional && (c == '+' || c == '-') | |
1470 | && (p[-1] == 'e' || p[-1] == 'E'))) | |
e8bbfc4e RK |
1471 | { |
1472 | error ("garbage at end of number"); | |
f3acae4d RS |
1473 | while (isalnum (c) || c == '.' || c == '_' |
1474 | || (!flag_traditional && (c == '+' || c == '-') | |
1475 | && (p[-1] == 'e' || p[-1] == 'E'))) | |
e8bbfc4e RK |
1476 | { |
1477 | if (p >= token_buffer + maxtoken - 3) | |
1478 | p = extend_token_buffer (p); | |
1479 | *p++ = c; | |
1480 | c = getc (finput); | |
1481 | } | |
1482 | } | |
1483 | break; | |
1484 | } | |
1485 | if (p >= token_buffer + maxtoken - 3) | |
1486 | p = extend_token_buffer (p); | |
1487 | *p++ = c; | |
1488 | c = getc (finput); | |
1489 | } | |
1490 | ||
1491 | ungetc (c, finput); | |
1492 | ||
1493 | /* If the constant is not long long and it won't fit in an | |
1494 | unsigned long, or if the constant is long long and won't fit | |
1495 | in an unsigned long long, then warn that the constant is out | |
1496 | of range. */ | |
1497 | ||
1498 | /* ??? This assumes that long long and long integer types are | |
1499 | a multiple of 8 bits. This better than the original code | |
1500 | though which assumed that long was exactly 32 bits and long | |
1501 | long was exactly 64 bits. */ | |
1502 | ||
1503 | if (spec_long_long) | |
1504 | bytes = TYPE_PRECISION (long_long_integer_type_node) / 8; | |
1505 | else | |
1506 | bytes = TYPE_PRECISION (long_integer_type_node) / 8; | |
1507 | ||
4ca827d0 RS |
1508 | warn = overflow; |
1509 | for (i = bytes; i < TOTAL_PARTS; i++) | |
1510 | if (parts[i]) | |
1511 | warn = 1; | |
1512 | if (warn) | |
1513 | pedwarn ("integer constant out of range"); | |
e8bbfc4e RK |
1514 | |
1515 | /* This is simplified by the fact that our constant | |
1516 | is always positive. */ | |
8d9bfdc5 RK |
1517 | |
1518 | high = low = 0; | |
1519 | ||
1520 | for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++) | |
1521 | { | |
1522 | high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT | |
1523 | / HOST_BITS_PER_CHAR)] | |
1524 | << (i * HOST_BITS_PER_CHAR)); | |
1525 | low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR); | |
1526 | } | |
1527 | ||
1528 | yylval.ttype = build_int_2 (low, high); | |
5a9085c7 | 1529 | TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node; |
e8bbfc4e | 1530 | |
e8bbfc4e RK |
1531 | /* If warn_traditional, calculate both the ANSI type and the |
1532 | traditional type, then see if they disagree. | |
1533 | Otherwise, calculate only the type for the dialect in use. */ | |
1534 | if (warn_traditional || flag_traditional) | |
1535 | { | |
1536 | /* Calculate the traditional type. */ | |
1537 | /* Traditionally, any constant is signed; | |
1538 | but if unsigned is specified explicitly, obey that. | |
1539 | Use the smallest size with the right number of bits, | |
1540 | except for one special case with decimal constants. */ | |
1541 | if (! spec_long && base != 10 | |
1542 | && int_fits_type_p (yylval.ttype, unsigned_type_node)) | |
1543 | traditional_type = (spec_unsigned ? unsigned_type_node | |
1544 | : integer_type_node); | |
1545 | /* A decimal constant must be long | |
1546 | if it does not fit in type int. | |
1547 | I think this is independent of whether | |
1548 | the constant is signed. */ | |
1549 | else if (! spec_long && base == 10 | |
1550 | && int_fits_type_p (yylval.ttype, integer_type_node)) | |
1551 | traditional_type = (spec_unsigned ? unsigned_type_node | |
1552 | : integer_type_node); | |
4ca827d0 | 1553 | else if (! spec_long_long) |
e8bbfc4e RK |
1554 | traditional_type = (spec_unsigned ? long_unsigned_type_node |
1555 | : long_integer_type_node); | |
1556 | else | |
1557 | traditional_type = (spec_unsigned | |
1558 | ? long_long_unsigned_type_node | |
1559 | : long_long_integer_type_node); | |
1560 | } | |
1561 | if (warn_traditional || ! flag_traditional) | |
1562 | { | |
1563 | /* Calculate the ANSI type. */ | |
1564 | if (! spec_long && ! spec_unsigned | |
1565 | && int_fits_type_p (yylval.ttype, integer_type_node)) | |
1566 | ansi_type = integer_type_node; | |
1567 | else if (! spec_long && (base != 10 || spec_unsigned) | |
1568 | && int_fits_type_p (yylval.ttype, unsigned_type_node)) | |
1569 | ansi_type = unsigned_type_node; | |
1570 | else if (! spec_unsigned && !spec_long_long | |
1571 | && int_fits_type_p (yylval.ttype, long_integer_type_node)) | |
1572 | ansi_type = long_integer_type_node; | |
4ca827d0 | 1573 | else if (! spec_long_long) |
e8bbfc4e RK |
1574 | ansi_type = long_unsigned_type_node; |
1575 | else if (! spec_unsigned | |
a120b788 RS |
1576 | /* Verify value does not overflow into sign bit. */ |
1577 | && TREE_INT_CST_HIGH (yylval.ttype) >= 0 | |
e8bbfc4e RK |
1578 | && int_fits_type_p (yylval.ttype, |
1579 | long_long_integer_type_node)) | |
1580 | ansi_type = long_long_integer_type_node; | |
1581 | else | |
1582 | ansi_type = long_long_unsigned_type_node; | |
1583 | } | |
1584 | ||
1585 | type = flag_traditional ? traditional_type : ansi_type; | |
1586 | ||
1587 | if (warn_traditional && traditional_type != ansi_type) | |
1588 | { | |
1589 | if (TYPE_PRECISION (traditional_type) | |
1590 | != TYPE_PRECISION (ansi_type)) | |
1591 | warning ("width of integer constant changes with -traditional"); | |
1592 | else if (TREE_UNSIGNED (traditional_type) | |
1593 | != TREE_UNSIGNED (ansi_type)) | |
1594 | warning ("integer constant is unsigned in ANSI C, signed with -traditional"); | |
a3ee5899 JW |
1595 | else |
1596 | warning ("width of integer constant may change on other systems with -traditional"); | |
e8bbfc4e | 1597 | } |
e8bbfc4e | 1598 | |
4ca827d0 RS |
1599 | if (!flag_traditional && !int_fits_type_p (yylval.ttype, type) |
1600 | && !warn) | |
88d92ca5 RS |
1601 | pedwarn ("integer constant out of range"); |
1602 | ||
c832a30e | 1603 | if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type)) |
ba99333a | 1604 | warning ("decimal constant is so large that it is unsigned"); |
c832a30e | 1605 | |
122026c0 RS |
1606 | if (spec_imag) |
1607 | { | |
1608 | if (TYPE_PRECISION (type) | |
1609 | <= TYPE_PRECISION (integer_type_node)) | |
1610 | yylval.ttype | |
1611 | = build_complex (integer_zero_node, | |
1612 | convert (integer_type_node, yylval.ttype)); | |
1613 | else | |
1614 | error ("complex integer constant is too wide for `complex int'"); | |
1615 | } | |
1616 | else if (flag_traditional && !int_fits_type_p (yylval.ttype, type)) | |
88dd95c1 RS |
1617 | /* The traditional constant 0x80000000 is signed |
1618 | but doesn't fit in the range of int. | |
1619 | This will change it to -0x80000000, which does fit. */ | |
1620 | { | |
1621 | TREE_TYPE (yylval.ttype) = unsigned_type (type); | |
1622 | yylval.ttype = convert (type, yylval.ttype); | |
ba85b2e4 RK |
1623 | TREE_OVERFLOW (yylval.ttype) |
1624 | = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0; | |
88dd95c1 RS |
1625 | } |
1626 | else | |
1627 | TREE_TYPE (yylval.ttype) = type; | |
1628 | ||
e8bbfc4e RK |
1629 | *p = 0; |
1630 | } | |
1631 | ||
1632 | value = CONSTANT; break; | |
1633 | } | |
1634 | ||
1635 | case '\'': | |
1636 | char_constant: | |
1637 | { | |
1638 | register int result = 0; | |
a2543283 | 1639 | register int num_chars = 0; |
e8bbfc4e RK |
1640 | unsigned width = TYPE_PRECISION (char_type_node); |
1641 | int max_chars; | |
1642 | ||
1643 | if (wide_flag) | |
1644 | { | |
1645 | width = WCHAR_TYPE_SIZE; | |
1646 | #ifdef MULTIBYTE_CHARS | |
1647 | max_chars = MB_CUR_MAX; | |
1648 | #else | |
1649 | max_chars = 1; | |
1650 | #endif | |
1651 | } | |
1652 | else | |
1653 | max_chars = TYPE_PRECISION (integer_type_node) / width; | |
1654 | ||
1655 | while (1) | |
1656 | { | |
1657 | tryagain: | |
1658 | ||
1659 | c = getc (finput); | |
1660 | ||
1661 | if (c == '\'' || c == EOF) | |
1662 | break; | |
1663 | ||
1664 | if (c == '\\') | |
1665 | { | |
b70af677 RS |
1666 | int ignore = 0; |
1667 | c = readescape (&ignore); | |
1668 | if (ignore) | |
e8bbfc4e RK |
1669 | goto tryagain; |
1670 | if (width < HOST_BITS_PER_INT | |
1671 | && (unsigned) c >= (1 << width)) | |
1672 | pedwarn ("escape sequence out of range for character"); | |
f6cb3487 RS |
1673 | #ifdef MAP_CHARACTER |
1674 | if (isprint (c)) | |
1675 | c = MAP_CHARACTER (c); | |
1676 | #endif | |
e8bbfc4e RK |
1677 | } |
1678 | else if (c == '\n') | |
1679 | { | |
1680 | if (pedantic) | |
1681 | pedwarn ("ANSI C forbids newline in character constant"); | |
1682 | lineno++; | |
1683 | } | |
b539a651 RS |
1684 | #ifdef MAP_CHARACTER |
1685 | else | |
1686 | c = MAP_CHARACTER (c); | |
1687 | #endif | |
e8bbfc4e RK |
1688 | |
1689 | num_chars++; | |
1690 | if (num_chars > maxtoken - 4) | |
1691 | extend_token_buffer (token_buffer); | |
1692 | ||
1693 | token_buffer[num_chars] = c; | |
1694 | ||
1695 | /* Merge character into result; ignore excess chars. */ | |
1696 | if (num_chars < max_chars + 1) | |
1697 | { | |
1698 | if (width < HOST_BITS_PER_INT) | |
1699 | result = (result << width) | (c & ((1 << width) - 1)); | |
1700 | else | |
1701 | result = c; | |
1702 | } | |
1703 | } | |
1704 | ||
1705 | token_buffer[num_chars + 1] = '\''; | |
1706 | token_buffer[num_chars + 2] = 0; | |
1707 | ||
1708 | if (c != '\'') | |
1709 | error ("malformatted character constant"); | |
1710 | else if (num_chars == 0) | |
1711 | error ("empty character constant"); | |
1712 | else if (num_chars > max_chars) | |
1713 | { | |
1714 | num_chars = max_chars; | |
1715 | error ("character constant too long"); | |
1716 | } | |
1717 | else if (num_chars != 1 && ! flag_traditional) | |
1718 | warning ("multi-character character constant"); | |
1719 | ||
1720 | /* If char type is signed, sign-extend the constant. */ | |
1721 | if (! wide_flag) | |
1722 | { | |
1723 | int num_bits = num_chars * width; | |
5f0abdc3 RS |
1724 | if (num_bits == 0) |
1725 | /* We already got an error; avoid invalid shift. */ | |
1726 | yylval.ttype = build_int_2 (0, 0); | |
1727 | else if (TREE_UNSIGNED (char_type_node) | |
1728 | || ((result >> (num_bits - 1)) & 1) == 0) | |
e8bbfc4e | 1729 | yylval.ttype |
8d9bfdc5 RK |
1730 | = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0 |
1731 | >> (HOST_BITS_PER_WIDE_INT - num_bits)), | |
e8bbfc4e RK |
1732 | 0); |
1733 | else | |
1734 | yylval.ttype | |
8d9bfdc5 RK |
1735 | = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0 |
1736 | >> (HOST_BITS_PER_WIDE_INT - num_bits)), | |
e8bbfc4e | 1737 | -1); |
f3acae4d | 1738 | TREE_TYPE (yylval.ttype) = integer_type_node; |
e8bbfc4e RK |
1739 | } |
1740 | else | |
1741 | { | |
1742 | #ifdef MULTIBYTE_CHARS | |
1743 | /* Set the initial shift state and convert the next sequence. */ | |
1744 | result = 0; | |
1745 | /* In all locales L'\0' is zero and mbtowc will return zero, | |
1746 | so don't use it. */ | |
1747 | if (num_chars > 1 | |
1748 | || (num_chars == 1 && token_buffer[1] != '\0')) | |
1749 | { | |
1750 | wchar_t wc; | |
8d9bfdc5 | 1751 | (void) mbtowc (NULL_PTR, NULL_PTR, 0); |
e8bbfc4e RK |
1752 | if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars) |
1753 | result = wc; | |
1754 | else | |
1755 | warning ("Ignoring invalid multibyte character"); | |
1756 | } | |
1757 | #endif | |
1758 | yylval.ttype = build_int_2 (result, 0); | |
f3acae4d | 1759 | TREE_TYPE (yylval.ttype) = wchar_type_node; |
e8bbfc4e RK |
1760 | } |
1761 | ||
e8bbfc4e RK |
1762 | value = CONSTANT; |
1763 | break; | |
1764 | } | |
1765 | ||
1766 | case '"': | |
1767 | string_constant: | |
1768 | { | |
1769 | c = getc (finput); | |
1770 | p = token_buffer + 1; | |
1771 | ||
1772 | while (c != '"' && c >= 0) | |
1773 | { | |
cf1d841d | 1774 | if (c == '\\') |
e8bbfc4e | 1775 | { |
fad04099 | 1776 | int ignore = 0; |
b70af677 RS |
1777 | c = readescape (&ignore); |
1778 | if (ignore) | |
e8bbfc4e RK |
1779 | goto skipnewline; |
1780 | if (!wide_flag | |
1781 | && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT | |
1782 | && c >= (1 << TYPE_PRECISION (char_type_node))) | |
1783 | pedwarn ("escape sequence out of range for character"); | |
1784 | } | |
1785 | else if (c == '\n') | |
1786 | { | |
1787 | if (pedantic) | |
1788 | pedwarn ("ANSI C forbids newline in string constant"); | |
1789 | lineno++; | |
1790 | } | |
1791 | ||
1792 | if (p == token_buffer + maxtoken) | |
1793 | p = extend_token_buffer (p); | |
1794 | *p++ = c; | |
1795 | ||
1796 | skipnewline: | |
1797 | c = getc (finput); | |
1798 | } | |
1799 | *p = 0; | |
1800 | ||
1801 | /* We have read the entire constant. | |
1802 | Construct a STRING_CST for the result. */ | |
1803 | ||
1804 | if (wide_flag) | |
1805 | { | |
1806 | /* If this is a L"..." wide-string, convert the multibyte string | |
1807 | to a wide character string. */ | |
1808 | char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES); | |
1809 | int len; | |
1810 | ||
1811 | #ifdef MULTIBYTE_CHARS | |
1812 | len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer); | |
e36a3b5d | 1813 | if (len < 0 || len >= (p - token_buffer)) |
e8bbfc4e RK |
1814 | { |
1815 | warning ("Ignoring invalid multibyte string"); | |
1816 | len = 0; | |
1817 | } | |
1818 | bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES); | |
1819 | #else | |
1820 | { | |
1821 | union { long l; char c[sizeof (long)]; } u; | |
1822 | int big_endian; | |
1823 | char *wp, *cp; | |
1824 | ||
1825 | /* Determine whether host is little or big endian. */ | |
1826 | u.l = 1; | |
1827 | big_endian = u.c[sizeof (long) - 1]; | |
1828 | wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0); | |
1829 | ||
1830 | bzero (widep, (p - token_buffer) * WCHAR_BYTES); | |
1831 | for (cp = token_buffer + 1; cp < p; cp++) | |
1832 | *wp = *cp, wp += WCHAR_BYTES; | |
1833 | len = p - token_buffer - 1; | |
1834 | } | |
1835 | #endif | |
1836 | yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep); | |
1837 | TREE_TYPE (yylval.ttype) = wchar_array_type_node; | |
e31c7eec TW |
1838 | value = STRING; |
1839 | } | |
1840 | else if (objc_flag) | |
1841 | { | |
1842 | extern tree build_objc_string(); | |
1843 | /* Return an Objective-C @"..." constant string object. */ | |
1844 | yylval.ttype = build_objc_string (p - token_buffer, | |
1845 | token_buffer + 1); | |
1846 | TREE_TYPE (yylval.ttype) = char_array_type_node; | |
1847 | value = OBJC_STRING; | |
e8bbfc4e RK |
1848 | } |
1849 | else | |
1850 | { | |
1851 | yylval.ttype = build_string (p - token_buffer, token_buffer + 1); | |
1852 | TREE_TYPE (yylval.ttype) = char_array_type_node; | |
e31c7eec | 1853 | value = STRING; |
e8bbfc4e RK |
1854 | } |
1855 | ||
1856 | *p++ = '"'; | |
1857 | *p = 0; | |
1858 | ||
e31c7eec | 1859 | break; |
e8bbfc4e RK |
1860 | } |
1861 | ||
1862 | case '+': | |
1863 | case '-': | |
1864 | case '&': | |
1865 | case '|': | |
1866 | case '<': | |
1867 | case '>': | |
1868 | case '*': | |
1869 | case '/': | |
1870 | case '%': | |
1871 | case '^': | |
1872 | case '!': | |
1873 | case '=': | |
1874 | { | |
1875 | register int c1; | |
1876 | ||
1877 | combine: | |
1878 | ||
1879 | switch (c) | |
1880 | { | |
1881 | case '+': | |
1882 | yylval.code = PLUS_EXPR; break; | |
1883 | case '-': | |
1884 | yylval.code = MINUS_EXPR; break; | |
1885 | case '&': | |
1886 | yylval.code = BIT_AND_EXPR; break; | |
1887 | case '|': | |
1888 | yylval.code = BIT_IOR_EXPR; break; | |
1889 | case '*': | |
1890 | yylval.code = MULT_EXPR; break; | |
1891 | case '/': | |
1892 | yylval.code = TRUNC_DIV_EXPR; break; | |
1893 | case '%': | |
1894 | yylval.code = TRUNC_MOD_EXPR; break; | |
1895 | case '^': | |
1896 | yylval.code = BIT_XOR_EXPR; break; | |
1897 | case LSHIFT: | |
1898 | yylval.code = LSHIFT_EXPR; break; | |
1899 | case RSHIFT: | |
1900 | yylval.code = RSHIFT_EXPR; break; | |
1901 | case '<': | |
1902 | yylval.code = LT_EXPR; break; | |
1903 | case '>': | |
1904 | yylval.code = GT_EXPR; break; | |
1905 | } | |
1906 | ||
1907 | token_buffer[1] = c1 = getc (finput); | |
1908 | token_buffer[2] = 0; | |
1909 | ||
1910 | if (c1 == '=') | |
1911 | { | |
1912 | switch (c) | |
1913 | { | |
1914 | case '<': | |
1915 | value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done; | |
1916 | case '>': | |
1917 | value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done; | |
1918 | case '!': | |
1919 | value = EQCOMPARE; yylval.code = NE_EXPR; goto done; | |
1920 | case '=': | |
1921 | value = EQCOMPARE; yylval.code = EQ_EXPR; goto done; | |
1922 | } | |
1923 | value = ASSIGN; goto done; | |
1924 | } | |
1925 | else if (c == c1) | |
1926 | switch (c) | |
1927 | { | |
1928 | case '+': | |
1929 | value = PLUSPLUS; goto done; | |
1930 | case '-': | |
1931 | value = MINUSMINUS; goto done; | |
1932 | case '&': | |
1933 | value = ANDAND; goto done; | |
1934 | case '|': | |
1935 | value = OROR; goto done; | |
1936 | case '<': | |
1937 | c = LSHIFT; | |
1938 | goto combine; | |
1939 | case '>': | |
1940 | c = RSHIFT; | |
1941 | goto combine; | |
1942 | } | |
1943 | else if ((c == '-') && (c1 == '>')) | |
1944 | { value = POINTSAT; goto done; } | |
1945 | ungetc (c1, finput); | |
1946 | token_buffer[1] = 0; | |
1947 | ||
1948 | if ((c == '<') || (c == '>')) | |
1949 | value = ARITHCOMPARE; | |
1950 | else value = c; | |
1951 | goto done; | |
1952 | } | |
1953 | ||
1954 | case 0: | |
1955 | /* Don't make yyparse think this is eof. */ | |
1956 | value = 1; | |
1957 | break; | |
1958 | ||
1959 | default: | |
1960 | value = c; | |
1961 | } | |
1962 | ||
1963 | done: | |
1964 | /* yylloc.last_line = lineno; */ | |
1965 | ||
1966 | return value; | |
1967 | } | |
1968 | ||
d45cf215 | 1969 | /* Sets the value of the 'yydebug' variable to VALUE. |
e8bbfc4e RK |
1970 | This is a function so we don't have to have YYDEBUG defined |
1971 | in order to build the compiler. */ | |
1972 | ||
1973 | void | |
1974 | set_yydebug (value) | |
1975 | int value; | |
1976 | { | |
1977 | #if YYDEBUG != 0 | |
1978 | yydebug = value; | |
1979 | #else | |
1980 | warning ("YYDEBUG not defined."); | |
1981 | #endif | |
1982 | } |