![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to c-lex.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Apple XNU] / GNUtools / cc|
Return to c-lex.c CVS log | Up to [Apple XNU] / GNUtools / cc |
1.1 root 1: /* Lexical analyzer for C and Objective C.
2: Copyright (C) 1987, 1988, 1989, 1992 Free Software Foundation, Inc.
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:
34: #include <ctype.h>
35:
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:
63: extern tree is_class_name ();
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 ();
93:
94: /* C code produced by gperf version 2.5 (GNU C++ version) */
95: /* Command-line: gperf -p -j1 -i 1 -g -o -t -G -N is_reserved_word -k1,3,$ c-parse.gperf */
96: struct resword { char *name; short token; enum rid rid; };
97:
98: #define TOTAL_KEYWORDS 79
99: #define MIN_WORD_LENGTH 2
100: #define MAX_WORD_LENGTH 20
101: #define MIN_HASH_VALUE 10
102: #define MAX_HASH_VALUE 144
103: /* maximum key range = 135, duplicates = 0 */
104:
105: #ifdef __GNUC__
106: __inline
107: #endif
108: static unsigned int
109: hash (str, len)
110: register char *str;
111: register int unsigned len;
112: {
113: static unsigned char asso_values[] =
114: {
115: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
116: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
117: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
118: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
119: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
120: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
121: 145, 145, 145, 145, 25, 145, 145, 145, 145, 145,
122: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
123: 145, 145, 145, 145, 145, 145, 145, 145, 145, 145,
124: 145, 145, 145, 145, 145, 1, 145, 46, 8, 15,
125: 61, 6, 36, 48, 3, 5, 145, 18, 63, 25,
126: 29, 76, 1, 145, 13, 2, 1, 51, 37, 9,
127: 9, 1, 3, 145, 145, 145, 145, 145,
128: };
129: register int hval = len;
130:
131: switch (hval)
132: {
133: default:
134: case 3:
135: hval += asso_values[str[2]];
136: case 2:
137: case 1:
138: hval += asso_values[str[0]];
139: }
140: return hval + asso_values[str[len - 1]];
141: }
142:
143: static struct resword wordlist[] =
144: {
145: {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",},
146: {"",},
147: {"int", TYPESPEC, RID_INT},
148: {"",}, {"",},
149: {"__typeof__", TYPEOF, NORID},
150: {"__signed__", TYPESPEC, RID_SIGNED},
151: {"__imag__", IMAGPART, NORID},
152: {"switch", SWITCH, NORID},
153: {"__inline__", SCSPEC, RID_INLINE},
154: {"else", ELSE, NORID},
155: {"__iterator__", SCSPEC, RID_ITERATOR},
156: {"__inline", SCSPEC, RID_INLINE},
157: {"__extension__", EXTENSION, NORID},
158: {"struct", STRUCT, NORID},
159: {"__real__", REALPART, NORID},
160: {"__const", TYPE_QUAL, RID_CONST},
161: {"while", WHILE, NORID},
162: {"__const__", TYPE_QUAL, RID_CONST},
163: {"case", CASE, NORID},
164: {"__complex__", TYPESPEC, RID_COMPLEX},
165: {"__iterator", SCSPEC, RID_ITERATOR},
166: {"bycopy", TYPE_QUAL, RID_BYCOPY},
167: {"",}, {"",}, {"",},
168: {"__complex", TYPESPEC, RID_COMPLEX},
169: {"",},
170: {"in", TYPE_QUAL, RID_IN},
171: {"break", BREAK, NORID},
172: {"@defs", DEFS, NORID},
173: {"",}, {"",}, {"",},
174: {"extern", SCSPEC, RID_EXTERN},
175: {"if", IF, NORID},
176: {"typeof", TYPEOF, NORID},
177: {"typedef", SCSPEC, RID_TYPEDEF},
178: {"__typeof", TYPEOF, NORID},
179: {"sizeof", SIZEOF, NORID},
180: {"",},
181: {"return", RETURN, NORID},
182: {"const", TYPE_QUAL, RID_CONST},
183: {"__volatile__", TYPE_QUAL, RID_VOLATILE},
184: {"@private", PRIVATE, NORID},
185: {"@selector", SELECTOR, NORID},
186: {"__volatile", TYPE_QUAL, RID_VOLATILE},
187: {"__asm__", ASM_KEYWORD, NORID},
188: {"",}, {"",},
189: {"continue", CONTINUE, NORID},
190: {"__alignof__", ALIGNOF, NORID},
191: {"__imag", IMAGPART, NORID},
192: {"__attribute__", ATTRIBUTE, NORID},
193: {"",}, {"",},
194: {"__attribute", ATTRIBUTE, NORID},
195: {"for", FOR, NORID},
196: {"",},
197: {"@encode", ENCODE, NORID},
198: {"id", OBJECTNAME, RID_ID},
199: {"static", SCSPEC, RID_STATIC},
200: {"@interface", INTERFACE, NORID},
201: {"",},
202: {"__signed", TYPESPEC, RID_SIGNED},
203: {"",},
204: {"__label__", LABEL, NORID},
205: {"",}, {"",},
206: {"__asm", ASM_KEYWORD, NORID},
207: {"char", TYPESPEC, RID_CHAR},
208: {"",},
209: {"inline", SCSPEC, RID_INLINE},
210: {"out", TYPE_QUAL, RID_OUT},
211: {"register", SCSPEC, RID_REGISTER},
212: {"__real", REALPART, NORID},
213: {"short", TYPESPEC, RID_SHORT},
214: {"",},
215: {"enum", ENUM, NORID},
216: {"inout", TYPE_QUAL, RID_INOUT},
217: {"",},
218: {"oneway", TYPE_QUAL, RID_ONEWAY},
219: {"union", UNION, NORID},
220: {"",},
221: {"__alignof", ALIGNOF, NORID},
222: {"",},
223: {"@implementation", IMPLEMENTATION, NORID},
224: {"",},
225: {"@class", CLASS, NORID},
226: {"",},
227: {"@public", PUBLIC, NORID},
228: {"asm", ASM_KEYWORD, NORID},
229: {"",}, {"",}, {"",}, {"",}, {"",},
230: {"default", DEFAULT, NORID},
231: {"",},
232: {"void", TYPESPEC, RID_VOID},
233: {"",},
234: {"@protected", PROTECTED, NORID},
235: {"@protocol", PROTOCOL, NORID},
236: {"",}, {"",}, {"",},
237: {"volatile", TYPE_QUAL, RID_VOLATILE},
238: {"",}, {"",},
239: {"signed", TYPESPEC, RID_SIGNED},
240: {"float", TYPESPEC, RID_FLOAT},
241: {"@end", END, NORID},
242: {"",}, {"",},
243: {"unsigned", TYPESPEC, RID_UNSIGNED},
244: {"@compatibility_alias", ALIAS, NORID},
245: {"double", TYPESPEC, RID_DOUBLE},
246: {"",}, {"",},
247: {"auto", SCSPEC, RID_AUTO},
248: {"",},
249: {"goto", GOTO, NORID},
250: {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",}, {"",},
251: {"do", DO, NORID},
252: {"",}, {"",}, {"",}, {"",},
253: {"long", TYPESPEC, RID_LONG},
254: };
255:
256: #ifdef __GNUC__
257: __inline
258: #endif
259: struct resword *
260: is_reserved_word (str, len)
261: register char *str;
262: register unsigned int len;
263: {
264: if (len <= MAX_WORD_LENGTH && len >= MIN_WORD_LENGTH)
265: {
266: register int key = hash (str, len);
267:
268: if (key <= MAX_HASH_VALUE && key >= 0)
269: {
270: register char *s = wordlist[key].name;
271:
272: if (*s == *str && !strcmp (str + 1, s + 1))
273: return &wordlist[key];
274: }
275: }
276: return 0;
277: }
278:
279: /* Return something to represent absolute declarators containing a *.
280: TARGET is the absolute declarator that the * contains.
281: TYPE_QUALS is a list of modifiers such as const or volatile
282: to apply to the pointer type, represented as identifiers.
283:
284: We return an INDIRECT_REF whose "contents" are TARGET
285: and whose type is the modifier list. */
286:
287: tree
288: make_pointer_declarator (type_quals, target)
289: tree type_quals, target;
290: {
291: return build1 (INDIRECT_REF, type_quals, target);
292: }
293:
294: void
295: forget_protocol_qualifiers ()
296: {
297: int i, n = sizeof wordlist / sizeof (struct resword);
298:
299: for (i = 0; i < n; i++)
300: if ((int) wordlist[i].rid >= (int) RID_IN
301: && (int) wordlist[i].rid <= (int) RID_ONEWAY)
302: wordlist[i].name = "";
303: }
304:
305: void
306: remember_protocol_qualifiers ()
307: {
308: int i, n = sizeof wordlist / sizeof (struct resword);
309:
310: for (i = 0; i < n; i++)
311: if (wordlist[i].rid == RID_IN)
312: wordlist[i].name = "in";
313: else if (wordlist[i].rid == RID_OUT)
314: wordlist[i].name = "out";
315: else if (wordlist[i].rid == RID_INOUT)
316: wordlist[i].name = "inout";
317: else if (wordlist[i].rid == RID_BYCOPY)
318: wordlist[i].name = "bycopy";
319: else if (wordlist[i].rid == RID_ONEWAY)
320: wordlist[i].name = "oneway";
321: }
322:
323: void
324: init_lex ()
325: {
326: /* Make identifier nodes long enough for the language-specific slots. */
327: set_identifier_size (sizeof (struct lang_identifier));
328:
329: /* Start it at 0, because check_newline is called at the very beginning
330: and will increment it to 1. */
331: lineno = 0;
332:
333: #ifdef MULTIBYTE_CHARS
334: /* Change to the native locale for multibyte conversions. */
335: setlocale (LC_CTYPE, "");
336: #endif
337:
338: maxtoken = 40;
339: token_buffer = (char *) xmalloc (maxtoken + 2);
340:
341: ridpointers[(int) RID_INT] = get_identifier ("int");
342: ridpointers[(int) RID_CHAR] = get_identifier ("char");
343: ridpointers[(int) RID_VOID] = get_identifier ("void");
344: ridpointers[(int) RID_FLOAT] = get_identifier ("float");
345: ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
346: ridpointers[(int) RID_SHORT] = get_identifier ("short");
347: ridpointers[(int) RID_LONG] = get_identifier ("long");
348: ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
349: ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
350: ridpointers[(int) RID_INLINE] = get_identifier ("inline");
351: ridpointers[(int) RID_CONST] = get_identifier ("const");
352: ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
353: ridpointers[(int) RID_AUTO] = get_identifier ("auto");
354: ridpointers[(int) RID_STATIC] = get_identifier ("static");
355: ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
356: ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
357: ridpointers[(int) RID_REGISTER] = get_identifier ("register");
358: ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
359: ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
360: ridpointers[(int) RID_ID] = get_identifier ("id");
361: ridpointers[(int) RID_IN] = get_identifier ("in");
362: ridpointers[(int) RID_OUT] = get_identifier ("out");
363: ridpointers[(int) RID_INOUT] = get_identifier ("inout");
364: ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
365: ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
366: forget_protocol_qualifiers();
367:
368: /* Some options inhibit certain reserved words.
369: Clear those words out of the hash table so they won't be recognized. */
370: #define UNSET_RESERVED_WORD(STRING) \
371: do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
372: if (s) s->name = ""; } while (0)
373:
374: if (! doing_objc_thang)
375: UNSET_RESERVED_WORD ("id");
376:
377: if (flag_traditional)
378: {
379: UNSET_RESERVED_WORD ("const");
380: UNSET_RESERVED_WORD ("volatile");
381: UNSET_RESERVED_WORD ("typeof");
382: UNSET_RESERVED_WORD ("signed");
383: UNSET_RESERVED_WORD ("inline");
384: UNSET_RESERVED_WORD ("iterator");
385: UNSET_RESERVED_WORD ("complex");
386: }
387: if (flag_no_asm)
388: {
389: UNSET_RESERVED_WORD ("asm");
390: UNSET_RESERVED_WORD ("typeof");
391: UNSET_RESERVED_WORD ("inline");
392: UNSET_RESERVED_WORD ("iterator");
393: UNSET_RESERVED_WORD ("complex");
394: }
395: }
396:
397: void
398: reinit_parse_for_function ()
399: {
400: }
401:
402: /* Function used when yydebug is set, to print a token in more detail. */
403:
404: void
405: yyprint (file, yychar, yylval)
406: FILE *file;
407: int yychar;
408: YYSTYPE yylval;
409: {
410: tree t;
411: switch (yychar)
412: {
413: case IDENTIFIER:
414: case TYPENAME:
415: case OBJECTNAME:
416: t = yylval.ttype;
417: if (IDENTIFIER_POINTER (t))
418: fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
419: break;
420:
421: case CONSTANT:
422: t = yylval.ttype;
423: if (TREE_CODE (t) == INTEGER_CST)
424: fprintf (file,
425: #if HOST_BITS_PER_WIDE_INT == 64
426: #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
427: " 0x%lx%016lx",
428: #else
429: " 0x%x%016x",
430: #endif
431: #else
432: #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
433: " 0x%lx%08lx",
434: #else
435: " 0x%x%08x",
436: #endif
437: #endif
438: TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
439: break;
440: }
441: }
442:
443:
444: /* If C is not whitespace, return C.
445: Otherwise skip whitespace and return first nonwhite char read. */
446:
447: static int
448: skip_white_space (c)
449: register int c;
450: {
451: static int newline_warning = 0;
452:
453: for (;;)
454: {
455: switch (c)
456: {
457: /* We don't recognize comments here, because
458: cpp output can include / and * consecutively as operators.
459: Also, there's no need, since cpp removes all comments. */
460:
461: case '\n':
462: c = check_newline ();
463: break;
464:
465: case ' ':
466: case '\t':
467: case '\f':
468: case '\v':
469: case '\b':
470: c = getc (finput);
471: break;
472:
473: case '\r':
474: /* ANSI C says the effects of a carriage return in a source file
475: are undefined. */
476: if (pedantic && !newline_warning)
477: {
478: warning ("carriage return in source file");
479: warning ("(we only warn about the first carriage return)");
480: newline_warning = 1;
481: }
482: c = getc (finput);
483: break;
484:
485: case '\\':
486: c = getc (finput);
487: if (c == '\n')
488: lineno++;
489: else
490: error ("stray '\\' in program");
491: c = getc (finput);
492: break;
493:
494: default:
495: return (c);
496: }
497: }
498: }
499:
500: /* Skips all of the white space at the current location in the input file.
501: Must use and reset nextchar if it has the next character. */
502:
503: void
504: position_after_white_space ()
505: {
506: register int c;
507:
508: if (nextchar != -1)
509: c = nextchar, nextchar = -1;
510: else
511: c = getc (finput);
512:
513: ungetc (skip_white_space (c), finput);
514: }
515:
516: /* Make the token buffer longer, preserving the data in it.
517: P should point to just beyond the last valid character in the old buffer.
518: The value we return is a pointer to the new buffer
519: at a place corresponding to P. */
520:
521: static char *
522: extend_token_buffer (p)
523: char *p;
524: {
525: int offset = p - token_buffer;
526:
527: maxtoken = maxtoken * 2 + 10;
528: token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
529:
530: return token_buffer + offset;
531: }
532:
533: /* At the beginning of a line, increment the line number
534: and process any #-directive on this line.
535: If the line is a #-directive, read the entire line and return a newline.
536: Otherwise, return the line's first non-whitespace character. */
537:
538: int
539: check_newline ()
540: {
541: register int c;
542: register int token;
543:
544: lineno++;
545:
546: /* Read first nonwhite char on the line. */
547:
548: c = getc (finput);
549: while (c == ' ' || c == '\t')
550: c = getc (finput);
551:
552: if (c != '#')
553: {
554: /* If not #, return it so caller will use it. */
555: return c;
556: }
557:
558: /* Read first nonwhite char after the `#'. */
559:
560: c = getc (finput);
561: while (c == ' ' || c == '\t')
562: c = getc (finput);
563:
564: /* If a letter follows, then if the word here is `line', skip
565: it and ignore it; otherwise, ignore the line, with an error
566: if the word isn't `pragma', `ident', `define', or `undef'. */
567:
568: if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
569: {
570: if (c == 'p')
571: {
572: if (getc (finput) == 'r'
573: && getc (finput) == 'a'
574: && getc (finput) == 'g'
575: && getc (finput) == 'm'
576: && getc (finput) == 'a'
577: && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
578: {
579: #ifdef HANDLE_SYSV_PRAGMA
580: return handle_sysv_pragma (finput, c);
581: #endif /* HANDLE_SYSV_PRAGMA */
582: #ifdef HANDLE_PRAGMA
583: HANDLE_PRAGMA (finput);
584: #endif /* HANDLE_PRAGMA */
585: goto skipline;
586: }
587: }
588:
589: else if (c == 'd')
590: {
591: if (getc (finput) == 'e'
592: && getc (finput) == 'f'
593: && getc (finput) == 'i'
594: && getc (finput) == 'n'
595: && getc (finput) == 'e'
596: && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
597: {
598: #ifdef DWARF_DEBUGGING_INFO
599: if ((debug_info_level == DINFO_LEVEL_VERBOSE)
600: && (write_symbols == DWARF_DEBUG))
601: dwarfout_define (lineno, get_directive_line (finput));
602: #endif /* DWARF_DEBUGGING_INFO */
603: goto skipline;
604: }
605: }
606: else if (c == 'u')
607: {
608: if (getc (finput) == 'n'
609: && getc (finput) == 'd'
610: && getc (finput) == 'e'
611: && getc (finput) == 'f'
612: && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
613: {
614: #ifdef DWARF_DEBUGGING_INFO
615: if ((debug_info_level == DINFO_LEVEL_VERBOSE)
616: && (write_symbols == DWARF_DEBUG))
617: dwarfout_undef (lineno, get_directive_line (finput));
618: #endif /* DWARF_DEBUGGING_INFO */
619: goto skipline;
620: }
621: }
622: else if (c == 'l')
623: {
624: if (getc (finput) == 'i'
625: && getc (finput) == 'n'
626: && getc (finput) == 'e'
627: && ((c = getc (finput)) == ' ' || c == '\t'))
628: goto linenum;
629: }
630: else if (c == 'i')
631: {
632: if (getc (finput) == 'd'
633: && getc (finput) == 'e'
634: && getc (finput) == 'n'
635: && getc (finput) == 't'
636: && ((c = getc (finput)) == ' ' || c == '\t'))
637: {
638: /* #ident. The pedantic warning is now in cccp.c. */
639:
640: /* Here we have just seen `#ident '.
641: A string constant should follow. */
642:
643: while (c == ' ' || c == '\t')
644: c = getc (finput);
645:
646: /* If no argument, ignore the line. */
647: if (c == '\n')
648: return c;
649:
650: ungetc (c, finput);
651: token = yylex ();
652: if (token != STRING
653: || TREE_CODE (yylval.ttype) != STRING_CST)
654: {
655: error ("invalid #ident");
656: goto skipline;
657: }
658:
659: if (!flag_no_ident)
660: {
661: #ifdef ASM_OUTPUT_IDENT
662: ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
663: #endif
664: }
665:
666: /* Skip the rest of this line. */
667: goto skipline;
668: }
669: }
670:
671: error ("undefined or invalid # directive");
672: goto skipline;
673: }
674:
675: linenum:
676: /* Here we have either `#line' or `# <nonletter>'.
677: In either case, it should be a line number; a digit should follow. */
678:
679: while (c == ' ' || c == '\t')
680: c = getc (finput);
681:
682: /* If the # is the only nonwhite char on the line,
683: just ignore it. Check the new newline. */
684: if (c == '\n')
685: return c;
686:
687: /* Something follows the #; read a token. */
688:
689: ungetc (c, finput);
690: token = yylex ();
691:
692: if (token == CONSTANT
693: && TREE_CODE (yylval.ttype) == INTEGER_CST)
694: {
695: int old_lineno = lineno;
696: int used_up = 0;
697: /* subtract one, because it is the following line that
698: gets the specified number */
699:
700: int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
701:
702: /* Is this the last nonwhite stuff on the line? */
703: c = getc (finput);
704: while (c == ' ' || c == '\t')
705: c = getc (finput);
706: if (c == '\n')
707: {
708: /* No more: store the line number and check following line. */
709: lineno = l;
710: return c;
711: }
712: ungetc (c, finput);
713:
714: /* More follows: it must be a string constant (filename). */
715:
716: /* Read the string constant. */
717: token = yylex ();
718:
719: if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
720: {
721: error ("invalid #line");
722: goto skipline;
723: }
724:
725: input_filename
726: = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
727: strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
728: lineno = l;
729:
730: /* Each change of file name
731: reinitializes whether we are now in a system header. */
732: in_system_header = 0;
733:
734: if (main_input_filename == 0)
735: main_input_filename = input_filename;
736:
737: /* Is this the last nonwhite stuff on the line? */
738: c = getc (finput);
739: while (c == ' ' || c == '\t')
740: c = getc (finput);
741: if (c == '\n')
742: {
743: /* Update the name in the top element of input_file_stack. */
744: if (input_file_stack)
745: input_file_stack->name = input_filename;
746:
747: return c;
748: }
749: ungetc (c, finput);
750:
751: token = yylex ();
752: used_up = 0;
753:
754: /* `1' after file name means entering new file.
755: `2' after file name means just left a file. */
756:
757: if (token == CONSTANT
758: && TREE_CODE (yylval.ttype) == INTEGER_CST)
759: {
760: if (TREE_INT_CST_LOW (yylval.ttype) == 1)
761: {
762: /* Pushing to a new file. */
763: struct file_stack *p
764: = (struct file_stack *) xmalloc (sizeof (struct file_stack));
765: input_file_stack->line = old_lineno;
766: p->next = input_file_stack;
767: p->name = input_filename;
768: input_file_stack = p;
769: input_file_stack_tick++;
770: #ifdef DWARF_DEBUGGING_INFO
771: if (debug_info_level == DINFO_LEVEL_VERBOSE
772: && write_symbols == DWARF_DEBUG)
773: dwarfout_start_new_source_file (input_filename);
774: #endif /* DWARF_DEBUGGING_INFO */
775:
776: used_up = 1;
777: }
778: else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
779: {
780: /* Popping out of a file. */
781: if (input_file_stack->next)
782: {
783: struct file_stack *p = input_file_stack;
784: input_file_stack = p->next;
785: free (p);
786: input_file_stack_tick++;
787: #ifdef DWARF_DEBUGGING_INFO
788: if (debug_info_level == DINFO_LEVEL_VERBOSE
789: && write_symbols == DWARF_DEBUG)
790: dwarfout_resume_previous_source_file (input_file_stack->line);
791: #endif /* DWARF_DEBUGGING_INFO */
792: }
793: else
794: error ("#-lines for entering and leaving files don't match");
795:
796: used_up = 1;
797: }
798: }
799:
800: /* Now that we've pushed or popped the input stack,
801: update the name in the top element. */
802: if (input_file_stack)
803: input_file_stack->name = input_filename;
804:
805: /* If we have handled a `1' or a `2',
806: see if there is another number to read. */
807: if (used_up)
808: {
809: /* Is this the last nonwhite stuff on the line? */
810: c = getc (finput);
811: while (c == ' ' || c == '\t')
812: c = getc (finput);
813: if (c == '\n')
814: return c;
815: ungetc (c, finput);
816:
817: token = yylex ();
818: used_up = 0;
819: }
820:
821: /* `3' after file name means this is a system header file. */
822:
823: if (token == CONSTANT
824: && TREE_CODE (yylval.ttype) == INTEGER_CST
825: && TREE_INT_CST_LOW (yylval.ttype) == 3)
826: in_system_header = 1;
827: }
828: else
829: error ("invalid #-line");
830:
831: /* skip the rest of this line. */
832: skipline:
833: if (c == '\n')
834: return c;
835: while ((c = getc (finput)) != EOF && c != '\n');
836: return c;
837: }
838:
839: #ifdef HANDLE_SYSV_PRAGMA
840:
841: /* Handle a #pragma directive. INPUT is the current input stream,
842: and C is a character to reread. Processes the entire input line
843: and returns a character for the caller to reread: either \n or EOF. */
844:
845: /* This function has to be in this file, in order to get at
846: the token types. */
847:
848: int
849: handle_sysv_pragma (input, c)
850: FILE *input;
851: int c;
852: {
853: for (;;)
854: {
855: while (c == ' ' || c == '\t')
856: c = getc (input);
857: if (c == '\n' || c == EOF)
858: {
859: handle_pragma_token (0, 0);
860: return c;
861: }
862: ungetc (c, input);
863: switch (yylex ())
864: {
865: case IDENTIFIER:
866: case TYPENAME:
867: case STRING:
868: case CONSTANT:
869: handle_pragma_token (token_buffer, yylval.ttype);
870: break;
871: default:
872: handle_pragma_token (token_buffer, 0);
873: }
874: if (nextchar >= 0)
875: c = nextchar, nextchar = -1;
876: else
877: c = getc (input);
878: }
879: }
880:
881: #endif /* HANDLE_SYSV_PRAGMA */
882:
883: #define ENDFILE -1 /* token that represents end-of-file */
884:
885: /* Read an escape sequence, returning its equivalent as a character,
886: or store 1 in *ignore_ptr if it is backslash-newline. */
887:
888: static int
889: readescape (ignore_ptr)
890: int *ignore_ptr;
891: {
892: register int c = getc (finput);
893: register int code;
894: register unsigned count;
895: unsigned firstdig;
896: int nonnull;
897:
898: switch (c)
899: {
900: case 'x':
901: if (warn_traditional)
902: warning ("the meaning of `\\x' varies with -traditional");
903:
904: if (flag_traditional)
905: return c;
906:
907: code = 0;
908: count = 0;
909: nonnull = 0;
910: while (1)
911: {
912: c = getc (finput);
913: if (!(c >= 'a' && c <= 'f')
914: && !(c >= 'A' && c <= 'F')
915: && !(c >= '0' && c <= '9'))
916: {
917: ungetc (c, finput);
918: break;
919: }
920: code *= 16;
921: if (c >= 'a' && c <= 'f')
922: code += c - 'a' + 10;
923: if (c >= 'A' && c <= 'F')
924: code += c - 'A' + 10;
925: if (c >= '0' && c <= '9')
926: code += c - '0';
927: if (code != 0 || count != 0)
928: {
929: if (count == 0)
930: firstdig = code;
931: count++;
932: }
933: nonnull = 1;
934: }
935: if (! nonnull)
936: error ("\\x used with no following hex digits");
937: else if (count == 0)
938: /* Digits are all 0's. Ok. */
939: ;
940: else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
941: || (count > 1
942: && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
943: <= firstdig)))
944: pedwarn ("hex escape out of range");
945: return code;
946:
947: case '0': case '1': case '2': case '3': case '4':
948: case '5': case '6': case '7':
949: code = 0;
950: count = 0;
951: while ((c <= '7') && (c >= '0') && (count++ < 3))
952: {
953: code = (code * 8) + (c - '0');
954: c = getc (finput);
955: }
956: ungetc (c, finput);
957: return code;
958:
959: case '\\': case '\'': case '"':
960: return c;
961:
962: case '\n':
963: lineno++;
964: *ignore_ptr = 1;
965: return 0;
966:
967: case 'n':
968: return TARGET_NEWLINE;
969:
970: case 't':
971: return TARGET_TAB;
972:
973: case 'r':
974: return TARGET_CR;
975:
976: case 'f':
977: return TARGET_FF;
978:
979: case 'b':
980: return TARGET_BS;
981:
982: case 'a':
983: if (warn_traditional)
984: warning ("the meaning of `\\a' varies with -traditional");
985:
986: if (flag_traditional)
987: return c;
988: return TARGET_BELL;
989:
990: case 'v':
991: #if 0 /* Vertical tab is present in common usage compilers. */
992: if (flag_traditional)
993: return c;
994: #endif
995: return TARGET_VT;
996:
997: case 'e':
998: case 'E':
999: if (pedantic)
1000: pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1001: return 033;
1002:
1003: case '?':
1004: return c;
1005:
1006: /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1007: case '(':
1008: case '{':
1009: case '[':
1010: /* `\%' is used to prevent SCCS from getting confused. */
1011: case '%':
1012: if (pedantic)
1013: pedwarn ("non-ANSI escape sequence `\\%c'", c);
1014: return c;
1015: }
1016: if (c >= 040 && c < 0177)
1017: pedwarn ("unknown escape sequence `\\%c'", c);
1018: else
1019: pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1020: return c;
1021: }
1022:
1023: void
1024: yyerror (string)
1025: char *string;
1026: {
1027: char buf[200];
1028:
1029: strcpy (buf, string);
1030:
1031: /* We can't print string and character constants well
1032: because the token_buffer contains the result of processing escapes. */
1033: if (end_of_file)
1034: strcat (buf, " at end of input");
1035: else if (token_buffer[0] == 0)
1036: strcat (buf, " at null character");
1037: else if (token_buffer[0] == '"')
1038: strcat (buf, " before string constant");
1039: else if (token_buffer[0] == '\'')
1040: strcat (buf, " before character constant");
1041: else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
1042: sprintf (buf + strlen (buf), " before character 0%o",
1043: (unsigned char) token_buffer[0]);
1044: else
1045: strcat (buf, " before `%s'");
1046:
1047: error (buf, token_buffer);
1048: }
1049:
1050: #if 0
1051:
1052: struct try_type
1053: {
1054: tree *node_var;
1055: char unsigned_flag;
1056: char long_flag;
1057: char long_long_flag;
1058: };
1059:
1060: struct try_type type_sequence[] =
1061: {
1062: { &integer_type_node, 0, 0, 0},
1063: { &unsigned_type_node, 1, 0, 0},
1064: { &long_integer_type_node, 0, 1, 0},
1065: { &long_unsigned_type_node, 1, 1, 0},
1066: { &long_long_integer_type_node, 0, 1, 1},
1067: { &long_long_unsigned_type_node, 1, 1, 1}
1068: };
1069: #endif /* 0 */
1070:
1071: int
1072: yylex ()
1073: {
1074: register int c;
1075: register char *p;
1076: register int value;
1077: int wide_flag = 0;
1078: int objc_flag = 0;
1079:
1080: if (nextchar >= 0)
1081: c = nextchar, nextchar = -1;
1082: else
1083: c = getc (finput);
1084:
1085: /* Effectively do c = skip_white_space (c)
1086: but do it faster in the usual cases. */
1087: while (1)
1088: switch (c)
1089: {
1090: case ' ':
1091: case '\t':
1092: case '\f':
1093: case '\v':
1094: case '\b':
1095: c = getc (finput);
1096: break;
1097:
1098: case '\r':
1099: /* Call skip_white_space so we can warn if appropriate. */
1100:
1101: case '\n':
1102: case '/':
1103: case '\\':
1104: c = skip_white_space (c);
1105: default:
1106: goto found_nonwhite;
1107: }
1108: found_nonwhite:
1109:
1110: token_buffer[0] = c;
1111: token_buffer[1] = 0;
1112:
1113: /* yylloc.first_line = lineno; */
1114:
1115: switch (c)
1116: {
1117: case EOF:
1118: end_of_file = 1;
1119: token_buffer[0] = 0;
1120: value = ENDFILE;
1121: break;
1122:
1123: case '$':
1124: if (dollars_in_ident)
1125: goto letter;
1126: return '$';
1127:
1128: case 'L':
1129: /* Capital L may start a wide-string or wide-character constant. */
1130: {
1131: register int c = getc (finput);
1132: if (c == '\'')
1133: {
1134: wide_flag = 1;
1135: goto char_constant;
1136: }
1137: if (c == '"')
1138: {
1139: wide_flag = 1;
1140: goto string_constant;
1141: }
1142: ungetc (c, finput);
1143: }
1144: goto letter;
1145:
1146: case '@':
1147: if (!doing_objc_thang)
1148: {
1149: warning ("possible Objective-C token in C input. Use -ObjC");
1150: value = c;
1151: break;
1152: }
1153: else
1154: {
1155: /* '@' may start a constant string object. */
1156: register int c = getc(finput);
1157: if (c == '"')
1158: {
1159: objc_flag = 1;
1160: goto string_constant;
1161: }
1162: ungetc(c, finput);
1163: /* Fall through to treat '@' as the start of an indentifier. */
1164: }
1165:
1166: case 'A': case 'B': case 'C': case 'D': case 'E':
1167: case 'F': case 'G': case 'H': case 'I': case 'J':
1168: case 'K': case 'M': case 'N': case 'O':
1169: case 'P': case 'Q': case 'R': case 'S': case 'T':
1170: case 'U': case 'V': case 'W': case 'X': case 'Y':
1171: case 'Z':
1172: case 'a': case 'b': case 'c': case 'd': case 'e':
1173: case 'f': case 'g': case 'h': case 'i': case 'j':
1174: case 'k': case 'l': case 'm': case 'n': case 'o':
1175: case 'p': case 'q': case 'r': case 's': case 't':
1176: case 'u': case 'v': case 'w': case 'x': case 'y':
1177: case 'z':
1178: case '_':
1179: letter:
1180: p = token_buffer;
1181: while (isalnum (c) || c == '_' || c == '$' || c == '@')
1182: {
1183: /* Make sure this char really belongs in an identifier. */
1184: if (c == '@' && ! doing_objc_thang)
1185: break;
1186: if (c == '$' && ! dollars_in_ident)
1187: break;
1188:
1189: if (p >= token_buffer + maxtoken)
1190: p = extend_token_buffer (p);
1191:
1192: *p++ = c;
1193: c = getc (finput);
1194: }
1195:
1196: *p = 0;
1197: nextchar = c;
1198:
1199: value = IDENTIFIER;
1200: yylval.itype = 0;
1201:
1202: /* Try to recognize a keyword. Uses minimum-perfect hash function */
1203:
1204: {
1205: register struct resword *ptr;
1206:
1207: if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1208: {
1209: if (ptr->rid)
1210: yylval.ttype = ridpointers[(int) ptr->rid];
1211: value = (int) ptr->token;
1212:
1213: /* Only return OBJECTNAME if it is a typedef. */
1214: if (doing_objc_thang && value == OBJECTNAME)
1215: {
1216: lastiddecl = lookup_name(yylval.ttype);
1217:
1218: if (lastiddecl == NULL_TREE
1219: || TREE_CODE (lastiddecl) != TYPE_DECL)
1220: value = IDENTIFIER;
1221: }
1222:
1223: /* Even if we decided to recognize asm, still perhaps warn. */
1224: if (pedantic
1225: && (value == ASM_KEYWORD || value == TYPEOF
1226: || ptr->rid == RID_INLINE)
1227: && token_buffer[0] != '_')
1228: pedwarn ("ANSI does not permit the keyword `%s'",
1229: token_buffer);
1230: }
1231: }
1232:
1233: /* If we did not find a keyword, look for an identifier
1234: (or a typename). */
1235:
1236: if (value == IDENTIFIER)
1237: {
1238: if (token_buffer[0] == '@')
1239: error("invalid identifier `%s'", token_buffer);
1240:
1241: yylval.ttype = get_identifier (token_buffer);
1242: lastiddecl = lookup_name (yylval.ttype);
1243:
1244: if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1245: value = TYPENAME;
1246: /* A user-invisible read-only initialized variable
1247: should be replaced by its value.
1248: We handle only strings since that's the only case used in C. */
1249: else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1250: && DECL_IGNORED_P (lastiddecl)
1251: && TREE_READONLY (lastiddecl)
1252: && DECL_INITIAL (lastiddecl) != 0
1253: && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1254: {
1255: tree stringval = DECL_INITIAL (lastiddecl);
1256:
1257: /* Copy the string value so that we won't clobber anything
1258: if we put something in the TREE_CHAIN of this one. */
1259: yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1260: TREE_STRING_POINTER (stringval));
1261: value = STRING;
1262: }
1263: else if (doing_objc_thang)
1264: {
1265: tree objc_interface_decl = is_class_name (yylval.ttype);
1266:
1267: if (objc_interface_decl)
1268: {
1269: value = CLASSNAME;
1270: yylval.ttype = objc_interface_decl;
1271: }
1272: }
1273: }
1274:
1275: break;
1276:
1277: case '0': case '1': case '2': case '3': case '4':
1278: case '5': case '6': case '7': case '8': case '9':
1279: case '.':
1280: {
1281: int base = 10;
1282: int count = 0;
1283: int largest_digit = 0;
1284: int numdigits = 0;
1285: /* for multi-precision arithmetic,
1286: we actually store only HOST_BITS_PER_CHAR bits in each part.
1287: The number of parts is chosen so as to be sufficient to hold
1288: the enough bits to fit into the two HOST_WIDE_INTs that contain
1289: the integer value (this is always at least as many bits as are
1290: in a target `long long' value, but may be wider). */
1291: #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1292: int parts[TOTAL_PARTS];
1293: int overflow = 0;
1294:
1295: enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1296: = NOT_FLOAT;
1297:
1298: for (count = 0; count < TOTAL_PARTS; count++)
1299: parts[count] = 0;
1300:
1301: p = token_buffer;
1302: *p++ = c;
1303:
1304: if (c == '0')
1305: {
1306: *p++ = (c = getc (finput));
1307: if ((c == 'x') || (c == 'X'))
1308: {
1309: base = 16;
1310: *p++ = (c = getc (finput));
1311: }
1312: /* Leading 0 forces octal unless the 0 is the only digit. */
1313: else if (c >= '0' && c <= '9')
1314: {
1315: base = 8;
1316: numdigits++;
1317: }
1318: else
1319: numdigits++;
1320: }
1321:
1322: /* Read all the digits-and-decimal-points. */
1323:
1324: while (c == '.'
1325: || (isalnum (c) && c != 'l' && c != 'L'
1326: && c != 'u' && c != 'U'
1327: && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1328: && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1329: {
1330: if (c == '.')
1331: {
1332: if (base == 16)
1333: error ("floating constant may not be in radix 16");
1334: if (floatflag == TOO_MANY_POINTS)
1335: /* We have already emitted an error. Don't need another. */
1336: ;
1337: else if (floatflag == AFTER_POINT)
1338: {
1339: error ("malformed floating constant");
1340: floatflag = TOO_MANY_POINTS;
1341: /* Avoid another error from atof by forcing all characters
1342: from here on to be ignored. */
1343: p[-1] = '\0';
1344: }
1345: else
1346: floatflag = AFTER_POINT;
1347:
1348: base = 10;
1349: *p++ = c = getc (finput);
1350: /* Accept '.' as the start of a floating-point number
1351: only when it is followed by a digit.
1352: Otherwise, unread the following non-digit
1353: and use the '.' as a structural token. */
1354: if (p == token_buffer + 2 && !isdigit (c))
1355: {
1356: if (c == '.')
1357: {
1358: c = getc (finput);
1359: if (c == '.')
1360: {
1361: *p++ = c;
1362: *p = 0;
1363: return ELLIPSIS;
1364: }
1365: error ("parse error at `..'");
1366: }
1367: ungetc (c, finput);
1368: token_buffer[1] = 0;
1369: value = '.';
1370: goto done;
1371: }
1372: }
1373: else
1374: {
1375: /* It is not a decimal point.
1376: It should be a digit (perhaps a hex digit). */
1377:
1378: if (isdigit (c))
1379: {
1380: c = c - '0';
1381: }
1382: else if (base <= 10)
1383: {
1384: if (c == 'e' || c == 'E')
1385: {
1386: base = 10;
1387: floatflag = AFTER_POINT;
1388: break; /* start of exponent */
1389: }
1390: error ("nondigits in number and not hexadecimal");
1391: c = 0;
1392: }
1393: else if (c >= 'a')
1394: {
1395: c = c - 'a' + 10;
1396: }
1397: else
1398: {
1399: c = c - 'A' + 10;
1400: }
1401: if (c >= largest_digit)
1402: largest_digit = c;
1403: numdigits++;
1404:
1405: for (count = 0; count < TOTAL_PARTS; count++)
1406: {
1407: parts[count] *= base;
1408: if (count)
1409: {
1410: parts[count]
1411: += (parts[count-1] >> HOST_BITS_PER_CHAR);
1412: parts[count-1]
1413: &= (1 << HOST_BITS_PER_CHAR) - 1;
1414: }
1415: else
1416: parts[0] += c;
1417: }
1418:
1419: /* If the extra highest-order part ever gets anything in it,
1420: the number is certainly too big. */
1421: if (parts[TOTAL_PARTS - 1] != 0)
1422: overflow = 1;
1423:
1424: if (p >= token_buffer + maxtoken - 3)
1425: p = extend_token_buffer (p);
1426: *p++ = (c = getc (finput));
1427: }
1428: }
1429:
1430: if (numdigits == 0)
1431: error ("numeric constant with no digits");
1432:
1433: if (largest_digit >= base)
1434: error ("numeric constant contains digits beyond the radix");
1435:
1436: /* Remove terminating char from the token buffer and delimit the string */
1437: *--p = 0;
1438:
1439: if (floatflag != NOT_FLOAT)
1440: {
1441: tree type = double_type_node;
1442: int garbage_chars = 0, exceeds_double = 0;
1443: int imag = 0;
1444: REAL_VALUE_TYPE value;
1445: jmp_buf handler;
1446:
1447: /* Read explicit exponent if any, and put it in tokenbuf. */
1448:
1449: if ((c == 'e') || (c == 'E'))
1450: {
1451: if (p >= token_buffer + maxtoken - 3)
1452: p = extend_token_buffer (p);
1453: *p++ = c;
1454: c = getc (finput);
1455: if ((c == '+') || (c == '-'))
1456: {
1457: *p++ = c;
1458: c = getc (finput);
1459: }
1460: if (! isdigit (c))
1461: error ("floating constant exponent has no digits");
1462: while (isdigit (c))
1463: {
1464: if (p >= token_buffer + maxtoken - 3)
1465: p = extend_token_buffer (p);
1466: *p++ = c;
1467: c = getc (finput);
1468: }
1469: }
1470:
1471: *p = 0;
1472: errno = 0;
1473:
1474: /* Convert string to a double, checking for overflow. */
1475: if (setjmp (handler))
1476: {
1477: error ("floating constant out of range");
1478: value = dconst0;
1479: }
1480: else
1481: {
1482: int fflag = 0, lflag = 0;
1483: /* Copy token_buffer now, while it has just the number
1484: and not the suffixes; once we add `f' or `i',
1485: REAL_VALUE_ATOF may not work any more. */
1486: char *copy = (char *) alloca (p - token_buffer + 1);
1487: bcopy (token_buffer, copy, p - token_buffer + 1);
1488:
1489: set_float_handler (handler);
1490:
1491: while (1)
1492: {
1493: int lose = 0;
1494:
1495: /* Read the suffixes to choose a data type. */
1496: switch (c)
1497: {
1498: case 'f': case 'F':
1499: if (fflag)
1500: error ("more than one `f' in numeric constant");
1501: fflag = 1;
1502: break;
1503:
1504: case 'l': case 'L':
1505: if (lflag)
1506: error ("more than one `l' in numeric constant");
1507: lflag = 1;
1508: break;
1509:
1510: case 'i': case 'I':
1511: if (imag)
1512: error ("more than one `i' or `j' in numeric constant");
1513: imag = 1;
1514: break;
1515:
1516: default:
1517: lose = 1;
1518: }
1519:
1520: if (lose)
1521: break;
1522:
1523: if (p >= token_buffer + maxtoken - 3)
1524: p = extend_token_buffer (p);
1525: *p++ = c;
1526: *p = 0;
1527: c = getc (finput);
1528: }
1529:
1530: /* The second argument, machine_mode, of REAL_VALUE_ATOF
1531: tells the desired precision of the binary result
1532: of decimal-to-binary conversion. */
1533:
1534: if (fflag)
1535: {
1536: if (lflag)
1537: error ("both `f' and `l' in floating constant");
1538:
1539: type = float_type_node;
1540: value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1541: if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1542: && REAL_VALUE_ISINF (value) && pedantic)
1543: pedwarn ("floating point number exceeds range of `float'");
1544: }
1545: else if (lflag)
1546: {
1547: type = long_double_type_node;
1548: value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1549: if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1550: && REAL_VALUE_ISINF (value) && pedantic)
1551: pedwarn ("floating point number exceeds range of `long double'");
1552: }
1553: else
1554: {
1555: value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1556: if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1557: && REAL_VALUE_ISINF (value) && pedantic)
1558: pedwarn ("floating point number exceeds range of `double'");
1559: }
1560:
1561: set_float_handler (NULL_PTR);
1562: }
1563: #ifdef ERANGE
1564: if (errno == ERANGE && !flag_traditional && pedantic)
1565: {
1566: /* ERANGE is also reported for underflow,
1567: so test the value to distinguish overflow from that. */
1568: if (TARGET_FLOAT_FORMAT != IEEE_FLOAT_FORMAT
1569: && (REAL_VALUES_LESS (dconst1, value)
1570: || REAL_VALUES_LESS (value, dconstm1)))
1571: {
1572: pedwarn ("floating point number exceeds range of `double'");
1573: exceeds_double = 1;
1574: }
1575: }
1576: #endif
1577: garbage_chars = 0;
1578: while (isalnum (c) || c == '.' || c == '_'
1579: || (!flag_traditional && (c == '+' || c == '-')
1580: && (p[-1] == 'e' || p[-1] == 'E')))
1581: {
1582: if (p >= token_buffer + maxtoken - 3)
1583: p = extend_token_buffer (p);
1584: *p++ = c;
1585: c = getc (finput);
1586: garbage_chars++;
1587: }
1588: if (garbage_chars > 0)
1589: error ("garbage at end of number");
1590:
1591: /* Create a node with determined type and value. */
1592: if (imag)
1593: yylval.ttype = build_complex (convert (type, integer_zero_node),
1594: build_real (type, value));
1595: else
1596: yylval.ttype = build_real (type, value);
1597:
1598: ungetc (c, finput);
1599: *p = 0;
1600: }
1601: else
1602: {
1603: tree traditional_type, ansi_type, type;
1604: HOST_WIDE_INT high, low;
1605: int spec_unsigned = 0;
1606: int spec_long = 0;
1607: int spec_long_long = 0;
1608: int spec_imag = 0;
1609: int bytes, warn, i;
1610:
1611: while (1)
1612: {
1613: if (c == 'u' || c == 'U')
1614: {
1615: if (spec_unsigned)
1616: error ("two `u's in integer constant");
1617: spec_unsigned = 1;
1618: }
1619: else if (c == 'l' || c == 'L')
1620: {
1621: if (spec_long)
1622: {
1623: if (spec_long_long)
1624: error ("three `l's in integer constant");
1625: else if (pedantic)
1626: pedwarn ("ANSI C forbids long long integer constants");
1627: spec_long_long = 1;
1628: }
1629: spec_long = 1;
1630: }
1631: else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1632: {
1633: if (spec_imag)
1634: error ("more than one `i' or `j' in numeric constant");
1635: spec_imag = 1;
1636: }
1637: else
1638: {
1639: if (isalnum (c) || c == '.' || c == '_'
1640: || (!flag_traditional && (c == '+' || c == '-')
1641: && (p[-1] == 'e' || p[-1] == 'E')))
1642: {
1643: error ("garbage at end of number");
1644: while (isalnum (c) || c == '.' || c == '_'
1645: || (!flag_traditional && (c == '+' || c == '-')
1646: && (p[-1] == 'e' || p[-1] == 'E')))
1647: {
1648: if (p >= token_buffer + maxtoken - 3)
1649: p = extend_token_buffer (p);
1650: *p++ = c;
1651: c = getc (finput);
1652: }
1653: }
1654: break;
1655: }
1656: if (p >= token_buffer + maxtoken - 3)
1657: p = extend_token_buffer (p);
1658: *p++ = c;
1659: c = getc (finput);
1660: }
1661:
1662: ungetc (c, finput);
1663:
1664: /* If the constant is not long long and it won't fit in an
1665: unsigned long, or if the constant is long long and won't fit
1666: in an unsigned long long, then warn that the constant is out
1667: of range. */
1668:
1669: /* ??? This assumes that long long and long integer types are
1670: a multiple of 8 bits. This better than the original code
1671: though which assumed that long was exactly 32 bits and long
1672: long was exactly 64 bits. */
1673:
1674: if (spec_long_long)
1675: bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1676: else
1677: bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1678:
1679: warn = overflow;
1680: for (i = bytes; i < TOTAL_PARTS; i++)
1681: if (parts[i])
1682: warn = 1;
1683: if (warn)
1684: pedwarn ("integer constant out of range");
1685:
1686: /* This is simplified by the fact that our constant
1687: is always positive. */
1688:
1689: high = low = 0;
1690:
1691: for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1692: {
1693: high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1694: / HOST_BITS_PER_CHAR)]
1695: << (i * HOST_BITS_PER_CHAR));
1696: low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1697: }
1698:
1699: yylval.ttype = build_int_2 (low, high);
1700: TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1701:
1702: /* If warn_traditional, calculate both the ANSI type and the
1703: traditional type, then see if they disagree.
1704: Otherwise, calculate only the type for the dialect in use. */
1705: if (warn_traditional || flag_traditional)
1706: {
1707: /* Calculate the traditional type. */
1708: /* Traditionally, any constant is signed;
1709: but if unsigned is specified explicitly, obey that.
1710: Use the smallest size with the right number of bits,
1711: except for one special case with decimal constants. */
1712: if (! spec_long && base != 10
1713: && int_fits_type_p (yylval.ttype, unsigned_type_node))
1714: traditional_type = (spec_unsigned ? unsigned_type_node
1715: : integer_type_node);
1716: /* A decimal constant must be long
1717: if it does not fit in type int.
1718: I think this is independent of whether
1719: the constant is signed. */
1720: else if (! spec_long && base == 10
1721: && int_fits_type_p (yylval.ttype, integer_type_node))
1722: traditional_type = (spec_unsigned ? unsigned_type_node
1723: : integer_type_node);
1724: else if (! spec_long_long)
1725: traditional_type = (spec_unsigned ? long_unsigned_type_node
1726: : long_integer_type_node);
1727: else
1728: traditional_type = (spec_unsigned
1729: ? long_long_unsigned_type_node
1730: : long_long_integer_type_node);
1731: }
1732: if (warn_traditional || ! flag_traditional)
1733: {
1734: /* Calculate the ANSI type. */
1735: if (! spec_long && ! spec_unsigned
1736: && int_fits_type_p (yylval.ttype, integer_type_node))
1737: ansi_type = integer_type_node;
1738: else if (! spec_long && (base != 10 || spec_unsigned)
1739: && int_fits_type_p (yylval.ttype, unsigned_type_node))
1740: ansi_type = unsigned_type_node;
1741: else if (! spec_unsigned && !spec_long_long
1742: && int_fits_type_p (yylval.ttype, long_integer_type_node))
1743: ansi_type = long_integer_type_node;
1744: else if (! spec_long_long)
1745: ansi_type = long_unsigned_type_node;
1746: else if (! spec_unsigned
1747: /* Verify value does not overflow into sign bit. */
1748: && TREE_INT_CST_HIGH (yylval.ttype) >= 0
1749: && int_fits_type_p (yylval.ttype,
1750: long_long_integer_type_node))
1751: ansi_type = long_long_integer_type_node;
1752: else
1753: ansi_type = long_long_unsigned_type_node;
1754: }
1755:
1756: type = flag_traditional ? traditional_type : ansi_type;
1757:
1758: if (warn_traditional && traditional_type != ansi_type)
1759: {
1760: if (TYPE_PRECISION (traditional_type)
1761: != TYPE_PRECISION (ansi_type))
1762: warning ("width of integer constant changes with -traditional");
1763: else if (TREE_UNSIGNED (traditional_type)
1764: != TREE_UNSIGNED (ansi_type))
1765: warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1766: else
1767: warning ("width of integer constant may change on other systems with -traditional");
1768: }
1769:
1770: if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1771: && !warn)
1772: pedwarn ("integer constant out of range");
1773:
1774: if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1775: warning ("decimal constant is so large that it is unsigned");
1776:
1777: if (spec_imag)
1778: {
1779: if (TYPE_PRECISION (type)
1780: <= TYPE_PRECISION (integer_type_node))
1781: yylval.ttype
1782: = build_complex (integer_zero_node,
1783: convert (integer_type_node, yylval.ttype));
1784: else
1785: error ("complex integer constant is too wide for `complex int'");
1786: }
1787: else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1788: /* The traditional constant 0x80000000 is signed
1789: but doesn't fit in the range of int.
1790: This will change it to -0x80000000, which does fit. */
1791: {
1792: TREE_TYPE (yylval.ttype) = unsigned_type (type);
1793: yylval.ttype = convert (type, yylval.ttype);
1794: }
1795: else
1796: TREE_TYPE (yylval.ttype) = type;
1797:
1798: *p = 0;
1799: }
1800:
1801: value = CONSTANT; break;
1802: }
1803:
1804: case '\'':
1805: char_constant:
1806: {
1807: register int result = 0;
1808: register int num_chars = 0;
1809: unsigned width = TYPE_PRECISION (char_type_node);
1810: int max_chars;
1811:
1812: if (wide_flag)
1813: {
1814: width = WCHAR_TYPE_SIZE;
1815: #ifdef MULTIBYTE_CHARS
1816: max_chars = MB_CUR_MAX;
1817: #else
1818: max_chars = 1;
1819: #endif
1820: }
1821: else
1822: max_chars = TYPE_PRECISION (integer_type_node) / width;
1823:
1824: while (1)
1825: {
1826: tryagain:
1827:
1828: c = getc (finput);
1829:
1830: if (c == '\'' || c == EOF)
1831: break;
1832:
1833: if (c == '\\')
1834: {
1835: int ignore = 0;
1836: c = readescape (&ignore);
1837: if (ignore)
1838: goto tryagain;
1839: if (width < HOST_BITS_PER_INT
1840: && (unsigned) c >= (1 << width))
1841: pedwarn ("escape sequence out of range for character");
1842: #ifdef MAP_CHARACTER
1843: if (isprint (c))
1844: c = MAP_CHARACTER (c);
1845: #endif
1846: }
1847: else if (c == '\n')
1848: {
1849: if (pedantic)
1850: pedwarn ("ANSI C forbids newline in character constant");
1851: lineno++;
1852: }
1853: #ifdef MAP_CHARACTER
1854: else
1855: c = MAP_CHARACTER (c);
1856: #endif
1857:
1858: num_chars++;
1859: if (num_chars > maxtoken - 4)
1860: extend_token_buffer (token_buffer);
1861:
1862: token_buffer[num_chars] = c;
1863:
1864: /* Merge character into result; ignore excess chars. */
1865: if (num_chars < max_chars + 1)
1866: {
1867: if (width < HOST_BITS_PER_INT)
1868: result = (result << width) | (c & ((1 << width) - 1));
1869: else
1870: result = c;
1871: }
1872: }
1873:
1874: token_buffer[num_chars + 1] = '\'';
1875: token_buffer[num_chars + 2] = 0;
1876:
1877: if (c != '\'')
1878: error ("malformatted character constant");
1879: else if (num_chars == 0)
1880: error ("empty character constant");
1881: else if (num_chars > max_chars)
1882: {
1883: num_chars = max_chars;
1884: error ("character constant too long");
1885: }
1886: else if (num_chars != 1 && ! flag_traditional)
1887: warning ("multi-character character constant");
1888:
1889: /* If char type is signed, sign-extend the constant. */
1890: if (! wide_flag)
1891: {
1892: int num_bits = num_chars * width;
1893: if (num_bits == 0)
1894: /* We already got an error; avoid invalid shift. */
1895: yylval.ttype = build_int_2 (0, 0);
1896: else if (TREE_UNSIGNED (char_type_node)
1897: || ((result >> (num_bits - 1)) & 1) == 0)
1898: yylval.ttype
1899: = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1900: >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1901: 0);
1902: else
1903: yylval.ttype
1904: = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1905: >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1906: -1);
1907: TREE_TYPE (yylval.ttype) = integer_type_node;
1908: }
1909: else
1910: {
1911: #ifdef MULTIBYTE_CHARS
1912: /* Set the initial shift state and convert the next sequence. */
1913: result = 0;
1914: /* In all locales L'\0' is zero and mbtowc will return zero,
1915: so don't use it. */
1916: if (num_chars > 1
1917: || (num_chars == 1 && token_buffer[1] != '\0'))
1918: {
1919: wchar_t wc;
1920: (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1921: if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1922: result = wc;
1923: else
1924: warning ("Ignoring invalid multibyte character");
1925: }
1926: #endif
1927: yylval.ttype = build_int_2 (result, 0);
1928: TREE_TYPE (yylval.ttype) = wchar_type_node;
1929: }
1930:
1931: value = CONSTANT;
1932: break;
1933: }
1934:
1935: case '"':
1936: string_constant:
1937: {
1938: c = getc (finput);
1939: p = token_buffer + 1;
1940:
1941: while (c != '"' && c >= 0)
1942: {
1943: if (c == '\\')
1944: {
1945: int ignore = 0;
1946: c = readescape (&ignore);
1947: if (ignore)
1948: goto skipnewline;
1949: if (!wide_flag
1950: && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1951: && c >= (1 << TYPE_PRECISION (char_type_node)))
1952: pedwarn ("escape sequence out of range for character");
1953: }
1954: else if (c == '\n')
1955: {
1956: if (pedantic)
1957: pedwarn ("ANSI C forbids newline in string constant");
1958: lineno++;
1959: }
1960:
1961: if (p == token_buffer + maxtoken)
1962: p = extend_token_buffer (p);
1963: *p++ = c;
1964:
1965: skipnewline:
1966: c = getc (finput);
1967: }
1968: *p = 0;
1969:
1970: /* We have read the entire constant.
1971: Construct a STRING_CST for the result. */
1972:
1973: if (wide_flag)
1974: {
1975: /* If this is a L"..." wide-string, convert the multibyte string
1976: to a wide character string. */
1977: char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1978: int len;
1979:
1980: #ifdef MULTIBYTE_CHARS
1981: len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1982: if (len < 0 || len >= (p - token_buffer))
1983: {
1984: warning ("Ignoring invalid multibyte string");
1985: len = 0;
1986: }
1987: bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1988: #else
1989: {
1990: union { long l; char c[sizeof (long)]; } u;
1991: int big_endian;
1992: char *wp, *cp;
1993:
1994: /* Determine whether host is little or big endian. */
1995: u.l = 1;
1996: big_endian = u.c[sizeof (long) - 1];
1997: wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1998:
1999: bzero (widep, (p - token_buffer) * WCHAR_BYTES);
2000: for (cp = token_buffer + 1; cp < p; cp++)
2001: *wp = *cp, wp += WCHAR_BYTES;
2002: len = p - token_buffer - 1;
2003: }
2004: #endif
2005: yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
2006: TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2007: value = STRING;
2008: }
2009: else if (objc_flag)
2010: {
2011: extern tree build_objc_string();
2012: /* Return an Objective-C @"..." constant string object. */
2013: yylval.ttype = build_objc_string (p - token_buffer,
2014: token_buffer + 1);
2015: TREE_TYPE (yylval.ttype) = char_array_type_node;
2016: value = OBJC_STRING;
2017: }
2018: else
2019: {
2020: yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
2021: TREE_TYPE (yylval.ttype) = char_array_type_node;
2022: value = STRING;
2023: }
2024:
2025: *p++ = '"';
2026: *p = 0;
2027:
2028: break;
2029: }
2030:
2031: case '+':
2032: case '-':
2033: case '&':
2034: case '|':
2035: case '<':
2036: case '>':
2037: case '*':
2038: case '/':
2039: case '%':
2040: case '^':
2041: case '!':
2042: case '=':
2043: {
2044: register int c1;
2045:
2046: combine:
2047:
2048: switch (c)
2049: {
2050: case '+':
2051: yylval.code = PLUS_EXPR; break;
2052: case '-':
2053: yylval.code = MINUS_EXPR; break;
2054: case '&':
2055: yylval.code = BIT_AND_EXPR; break;
2056: case '|':
2057: yylval.code = BIT_IOR_EXPR; break;
2058: case '*':
2059: yylval.code = MULT_EXPR; break;
2060: case '/':
2061: yylval.code = TRUNC_DIV_EXPR; break;
2062: case '%':
2063: yylval.code = TRUNC_MOD_EXPR; break;
2064: case '^':
2065: yylval.code = BIT_XOR_EXPR; break;
2066: case LSHIFT:
2067: yylval.code = LSHIFT_EXPR; break;
2068: case RSHIFT:
2069: yylval.code = RSHIFT_EXPR; break;
2070: case '<':
2071: yylval.code = LT_EXPR; break;
2072: case '>':
2073: yylval.code = GT_EXPR; break;
2074: }
2075:
2076: token_buffer[1] = c1 = getc (finput);
2077: token_buffer[2] = 0;
2078:
2079: if (c1 == '=')
2080: {
2081: switch (c)
2082: {
2083: case '<':
2084: value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2085: case '>':
2086: value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2087: case '!':
2088: value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2089: case '=':
2090: value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2091: }
2092: value = ASSIGN; goto done;
2093: }
2094: else if (c == c1)
2095: switch (c)
2096: {
2097: case '+':
2098: value = PLUSPLUS; goto done;
2099: case '-':
2100: value = MINUSMINUS; goto done;
2101: case '&':
2102: value = ANDAND; goto done;
2103: case '|':
2104: value = OROR; goto done;
2105: case '<':
2106: c = LSHIFT;
2107: goto combine;
2108: case '>':
2109: c = RSHIFT;
2110: goto combine;
2111: }
2112: else if ((c == '-') && (c1 == '>'))
2113: { value = POINTSAT; goto done; }
2114: ungetc (c1, finput);
2115: token_buffer[1] = 0;
2116:
2117: if ((c == '<') || (c == '>'))
2118: value = ARITHCOMPARE;
2119: else value = c;
2120: goto done;
2121: }
2122:
2123: case 0:
2124: /* Don't make yyparse think this is eof. */
2125: value = 1;
2126: break;
2127:
2128: default:
2129: value = c;
2130: }
2131:
2132: done:
2133: /* yylloc.last_line = lineno; */
2134:
2135: return value;
2136: }
2137:
2138: /* Sets the value of the 'yydebug' variable to VALUE.
2139: This is a function so we don't have to have YYDEBUG defined
2140: in order to build the compiler. */
2141:
2142: void
2143: set_yydebug (value)
2144: int value;
2145: {
2146: #if YYDEBUG != 0
2147: yydebug = value;
2148: #else
2149: warning ("YYDEBUG not defined.");
2150: #endif
2151: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.