(*zeroextract[qs]i_compare0_scratch): Use const_int_operand
[official-gcc.git] / gcc / c-lex.c
blob89ff60b0ccef7a2b7f7b7d21c01f68868017c7b9
1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 88, 89, 92, 94, 95, 1996 Free Software Foundation, Inc.
4 This file is part of GNU CC.
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.
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.
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, 59 Temple Place - Suite 330,
19 Boston, MA 02111-1307, USA. */
22 #include <stdio.h>
23 #include <errno.h>
24 #include <setjmp.h>
26 #include "config.h"
27 #include "rtl.h"
28 #include "tree.h"
29 #include "input.h"
30 #include "c-lex.h"
31 #include "c-tree.h"
32 #include "flags.h"
33 #include "c-parse.h"
34 #include "c-pragma.h"
36 #include <ctype.h>
38 #ifdef MULTIBYTE_CHARS
39 #include <stdlib.h>
40 #include <locale.h>
41 #endif
43 #ifndef errno
44 extern int errno;
45 #endif
47 /* The elements of `ridpointers' are identifier nodes
48 for the reserved type names and storage classes.
49 It is indexed by a RID_... value. */
50 tree ridpointers[(int) RID_MAX];
52 /* Cause the `yydebug' variable to be defined. */
53 #define YYDEBUG 1
55 /* the declaration found for the last IDENTIFIER token read in.
56 yylex must look this up to detect typedefs, which get token type TYPENAME,
57 so it is left around in case the identifier is not a typedef but is
58 used in a context which makes it a reference to a variable. */
59 tree lastiddecl;
61 /* Nonzero enables objc features. */
63 int doing_objc_thang;
65 extern tree is_class_name ();
67 extern int yydebug;
69 /* File used for outputting assembler code. */
70 extern FILE *asm_out_file;
72 #ifndef WCHAR_TYPE_SIZE
73 #ifdef INT_TYPE_SIZE
74 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
75 #else
76 #define WCHAR_TYPE_SIZE BITS_PER_WORD
77 #endif
78 #endif
80 /* Number of bytes in a wide character. */
81 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
83 static int maxtoken; /* Current nominal length of token buffer. */
84 char *token_buffer; /* Pointer to token buffer.
85 Actual allocated length is maxtoken + 2.
86 This is not static because objc-parse.y uses it. */
88 /* Nonzero if end-of-file has been seen on input. */
89 static int end_of_file;
91 /* Buffered-back input character; faster than using ungetc. */
92 static int nextchar = -1;
94 int check_newline ();
96 /* Do not insert generated code into the source, instead, include it.
97 This allows us to build gcc automatically even for targets that
98 need to add or modify the reserved keyword lists. */
99 #include "c-gperf.h"
101 /* Return something to represent absolute declarators containing a *.
102 TARGET is the absolute declarator that the * contains.
103 TYPE_QUALS is a list of modifiers such as const or volatile
104 to apply to the pointer type, represented as identifiers.
106 We return an INDIRECT_REF whose "contents" are TARGET
107 and whose type is the modifier list. */
109 tree
110 make_pointer_declarator (type_quals, target)
111 tree type_quals, target;
113 return build1 (INDIRECT_REF, type_quals, target);
116 void
117 forget_protocol_qualifiers ()
119 int i, n = sizeof wordlist / sizeof (struct resword);
121 for (i = 0; i < n; i++)
122 if ((int) wordlist[i].rid >= (int) RID_IN
123 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
124 wordlist[i].name = "";
127 void
128 remember_protocol_qualifiers ()
130 int i, n = sizeof wordlist / sizeof (struct resword);
132 for (i = 0; i < n; i++)
133 if (wordlist[i].rid == RID_IN)
134 wordlist[i].name = "in";
135 else if (wordlist[i].rid == RID_OUT)
136 wordlist[i].name = "out";
137 else if (wordlist[i].rid == RID_INOUT)
138 wordlist[i].name = "inout";
139 else if (wordlist[i].rid == RID_BYCOPY)
140 wordlist[i].name = "bycopy";
141 else if (wordlist[i].rid == RID_ONEWAY)
142 wordlist[i].name = "oneway";
145 void
146 init_lex ()
148 /* Make identifier nodes long enough for the language-specific slots. */
149 set_identifier_size (sizeof (struct lang_identifier));
151 /* Start it at 0, because check_newline is called at the very beginning
152 and will increment it to 1. */
153 lineno = 0;
155 #ifdef MULTIBYTE_CHARS
156 /* Change to the native locale for multibyte conversions. */
157 setlocale (LC_CTYPE, "");
158 #endif
160 maxtoken = 40;
161 token_buffer = (char *) xmalloc (maxtoken + 2);
163 ridpointers[(int) RID_INT] = get_identifier ("int");
164 ridpointers[(int) RID_CHAR] = get_identifier ("char");
165 ridpointers[(int) RID_VOID] = get_identifier ("void");
166 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
167 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
168 ridpointers[(int) RID_SHORT] = get_identifier ("short");
169 ridpointers[(int) RID_LONG] = get_identifier ("long");
170 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
171 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
172 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
173 ridpointers[(int) RID_CONST] = get_identifier ("const");
174 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
175 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
176 ridpointers[(int) RID_STATIC] = get_identifier ("static");
177 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
178 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
179 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
180 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
181 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
182 ridpointers[(int) RID_ID] = get_identifier ("id");
183 ridpointers[(int) RID_IN] = get_identifier ("in");
184 ridpointers[(int) RID_OUT] = get_identifier ("out");
185 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
186 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
187 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
188 forget_protocol_qualifiers();
190 /* Some options inhibit certain reserved words.
191 Clear those words out of the hash table so they won't be recognized. */
192 #define UNSET_RESERVED_WORD(STRING) \
193 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
194 if (s) s->name = ""; } while (0)
196 if (! doing_objc_thang)
197 UNSET_RESERVED_WORD ("id");
199 if (flag_traditional)
201 UNSET_RESERVED_WORD ("const");
202 UNSET_RESERVED_WORD ("volatile");
203 UNSET_RESERVED_WORD ("typeof");
204 UNSET_RESERVED_WORD ("signed");
205 UNSET_RESERVED_WORD ("inline");
206 UNSET_RESERVED_WORD ("iterator");
207 UNSET_RESERVED_WORD ("complex");
209 if (flag_no_asm)
211 UNSET_RESERVED_WORD ("asm");
212 UNSET_RESERVED_WORD ("typeof");
213 UNSET_RESERVED_WORD ("inline");
214 UNSET_RESERVED_WORD ("iterator");
215 UNSET_RESERVED_WORD ("complex");
219 void
220 reinit_parse_for_function ()
224 /* Function used when yydebug is set, to print a token in more detail. */
226 void
227 yyprint (file, yychar, yylval)
228 FILE *file;
229 int yychar;
230 YYSTYPE yylval;
232 tree t;
233 switch (yychar)
235 case IDENTIFIER:
236 case TYPENAME:
237 case OBJECTNAME:
238 t = yylval.ttype;
239 if (IDENTIFIER_POINTER (t))
240 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
241 break;
243 case CONSTANT:
244 t = yylval.ttype;
245 if (TREE_CODE (t) == INTEGER_CST)
246 fprintf (file,
247 #if HOST_BITS_PER_WIDE_INT == 64
248 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
249 " 0x%lx%016lx",
250 #else
251 " 0x%x%016x",
252 #endif
253 #else
254 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
255 " 0x%lx%08lx",
256 #else
257 " 0x%x%08x",
258 #endif
259 #endif
260 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
261 break;
266 /* If C is not whitespace, return C.
267 Otherwise skip whitespace and return first nonwhite char read. */
269 static int
270 skip_white_space (c)
271 register int c;
273 static int newline_warning = 0;
275 for (;;)
277 switch (c)
279 /* We don't recognize comments here, because
280 cpp output can include / and * consecutively as operators.
281 Also, there's no need, since cpp removes all comments. */
283 case '\n':
284 c = check_newline ();
285 break;
287 case ' ':
288 case '\t':
289 case '\f':
290 case '\v':
291 case '\b':
292 c = getc (finput);
293 break;
295 case '\r':
296 /* ANSI C says the effects of a carriage return in a source file
297 are undefined. */
298 if (pedantic && !newline_warning)
300 warning ("carriage return in source file");
301 warning ("(we only warn about the first carriage return)");
302 newline_warning = 1;
304 c = getc (finput);
305 break;
307 case '\\':
308 c = getc (finput);
309 if (c == '\n')
310 lineno++;
311 else
312 error ("stray '\\' in program");
313 c = getc (finput);
314 break;
316 default:
317 return (c);
322 /* Skips all of the white space at the current location in the input file.
323 Must use and reset nextchar if it has the next character. */
325 void
326 position_after_white_space ()
328 register int c;
330 if (nextchar != -1)
331 c = nextchar, nextchar = -1;
332 else
333 c = getc (finput);
335 ungetc (skip_white_space (c), finput);
338 /* Make the token buffer longer, preserving the data in it.
339 P should point to just beyond the last valid character in the old buffer.
340 The value we return is a pointer to the new buffer
341 at a place corresponding to P. */
343 static char *
344 extend_token_buffer (p)
345 char *p;
347 int offset = p - token_buffer;
349 maxtoken = maxtoken * 2 + 10;
350 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
352 return token_buffer + offset;
355 /* At the beginning of a line, increment the line number
356 and process any #-directive on this line.
357 If the line is a #-directive, read the entire line and return a newline.
358 Otherwise, return the line's first non-whitespace character. */
361 check_newline ()
363 register int c;
364 register int token;
366 lineno++;
368 /* Read first nonwhite char on the line. */
370 c = getc (finput);
371 while (c == ' ' || c == '\t')
372 c = getc (finput);
374 if (c != '#')
376 /* If not #, return it so caller will use it. */
377 return c;
380 /* Read first nonwhite char after the `#'. */
382 c = getc (finput);
383 while (c == ' ' || c == '\t')
384 c = getc (finput);
386 /* If a letter follows, then if the word here is `line', skip
387 it and ignore it; otherwise, ignore the line, with an error
388 if the word isn't `pragma', `ident', `define', or `undef'. */
390 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
392 if (c == 'p')
394 if (getc (finput) == 'r'
395 && getc (finput) == 'a'
396 && getc (finput) == 'g'
397 && getc (finput) == 'm'
398 && getc (finput) == 'a'
399 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
401 #ifdef HANDLE_SYSV_PRAGMA
402 return handle_sysv_pragma (finput, c);
403 #else /* !HANDLE_SYSV_PRAGMA */
404 #ifdef HANDLE_PRAGMA
405 return HANDLE_PRAGMA (finput, c);
406 #endif /* HANDLE_PRAGMA */
407 goto skipline;
408 #endif /* !HANDLE_SYSV_PRAGMA */
412 else if (c == 'd')
414 if (getc (finput) == 'e'
415 && getc (finput) == 'f'
416 && getc (finput) == 'i'
417 && getc (finput) == 'n'
418 && getc (finput) == 'e'
419 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
421 #ifdef DWARF_DEBUGGING_INFO
422 if (c != '\n'
423 && (debug_info_level == DINFO_LEVEL_VERBOSE)
424 && (write_symbols == DWARF_DEBUG))
425 dwarfout_define (lineno, get_directive_line (finput));
426 #endif /* DWARF_DEBUGGING_INFO */
427 goto skipline;
430 else if (c == 'u')
432 if (getc (finput) == 'n'
433 && getc (finput) == 'd'
434 && getc (finput) == 'e'
435 && getc (finput) == 'f'
436 && ((c = getc (finput)) == ' ' || c == '\t' || c == '\n'))
438 #ifdef DWARF_DEBUGGING_INFO
439 if (c != '\n'
440 && (debug_info_level == DINFO_LEVEL_VERBOSE)
441 && (write_symbols == DWARF_DEBUG))
442 dwarfout_undef (lineno, get_directive_line (finput));
443 #endif /* DWARF_DEBUGGING_INFO */
444 goto skipline;
447 else if (c == 'l')
449 if (getc (finput) == 'i'
450 && getc (finput) == 'n'
451 && getc (finput) == 'e'
452 && ((c = getc (finput)) == ' ' || c == '\t'))
453 goto linenum;
455 else if (c == 'i')
457 if (getc (finput) == 'd'
458 && getc (finput) == 'e'
459 && getc (finput) == 'n'
460 && getc (finput) == 't'
461 && ((c = getc (finput)) == ' ' || c == '\t'))
463 /* #ident. The pedantic warning is now in cccp.c. */
465 /* Here we have just seen `#ident '.
466 A string constant should follow. */
468 while (c == ' ' || c == '\t')
469 c = getc (finput);
471 /* If no argument, ignore the line. */
472 if (c == '\n')
473 return c;
475 ungetc (c, finput);
476 token = yylex ();
477 if (token != STRING
478 || TREE_CODE (yylval.ttype) != STRING_CST)
480 error ("invalid #ident");
481 goto skipline;
484 if (!flag_no_ident)
486 #ifdef ASM_OUTPUT_IDENT
487 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
488 #endif
491 /* Skip the rest of this line. */
492 goto skipline;
496 error ("undefined or invalid # directive");
497 goto skipline;
500 linenum:
501 /* Here we have either `#line' or `# <nonletter>'.
502 In either case, it should be a line number; a digit should follow. */
504 while (c == ' ' || c == '\t')
505 c = getc (finput);
507 /* If the # is the only nonwhite char on the line,
508 just ignore it. Check the new newline. */
509 if (c == '\n')
510 return c;
512 /* Something follows the #; read a token. */
514 ungetc (c, finput);
515 token = yylex ();
517 if (token == CONSTANT
518 && TREE_CODE (yylval.ttype) == INTEGER_CST)
520 int old_lineno = lineno;
521 int used_up = 0;
522 /* subtract one, because it is the following line that
523 gets the specified number */
525 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
527 /* Is this the last nonwhite stuff on the line? */
528 c = getc (finput);
529 while (c == ' ' || c == '\t')
530 c = getc (finput);
531 if (c == '\n')
533 /* No more: store the line number and check following line. */
534 lineno = l;
535 return c;
537 ungetc (c, finput);
539 /* More follows: it must be a string constant (filename). */
541 /* Read the string constant. */
542 token = yylex ();
544 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
546 error ("invalid #line");
547 goto skipline;
550 input_filename
551 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
552 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
553 lineno = l;
555 /* Each change of file name
556 reinitializes whether we are now in a system header. */
557 in_system_header = 0;
559 if (main_input_filename == 0)
560 main_input_filename = input_filename;
562 /* Is this the last nonwhite stuff on the line? */
563 c = getc (finput);
564 while (c == ' ' || c == '\t')
565 c = getc (finput);
566 if (c == '\n')
568 /* Update the name in the top element of input_file_stack. */
569 if (input_file_stack)
570 input_file_stack->name = input_filename;
572 return c;
574 ungetc (c, finput);
576 token = yylex ();
577 used_up = 0;
579 /* `1' after file name means entering new file.
580 `2' after file name means just left a file. */
582 if (token == CONSTANT
583 && TREE_CODE (yylval.ttype) == INTEGER_CST)
585 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
587 /* Pushing to a new file. */
588 struct file_stack *p
589 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
590 input_file_stack->line = old_lineno;
591 p->next = input_file_stack;
592 p->name = input_filename;
593 input_file_stack = p;
594 input_file_stack_tick++;
595 #ifdef DBX_DEBUGGING_INFO
596 if (write_symbols == DBX_DEBUG)
597 dbxout_start_new_source_file (input_filename);
598 #endif
599 #ifdef DWARF_DEBUGGING_INFO
600 if (debug_info_level == DINFO_LEVEL_VERBOSE
601 && write_symbols == DWARF_DEBUG)
602 dwarfout_start_new_source_file (input_filename);
603 #endif /* DWARF_DEBUGGING_INFO */
605 used_up = 1;
607 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
609 /* Popping out of a file. */
610 if (input_file_stack->next)
612 struct file_stack *p = input_file_stack;
613 input_file_stack = p->next;
614 free (p);
615 input_file_stack_tick++;
616 #ifdef DBX_DEBUGGING_INFO
617 if (write_symbols == DBX_DEBUG)
618 dbxout_resume_previous_source_file ();
619 #endif
620 #ifdef DWARF_DEBUGGING_INFO
621 if (debug_info_level == DINFO_LEVEL_VERBOSE
622 && write_symbols == DWARF_DEBUG)
623 dwarfout_resume_previous_source_file (input_file_stack->line);
624 #endif /* DWARF_DEBUGGING_INFO */
626 else
627 error ("#-lines for entering and leaving files don't match");
629 used_up = 1;
633 /* Now that we've pushed or popped the input stack,
634 update the name in the top element. */
635 if (input_file_stack)
636 input_file_stack->name = input_filename;
638 /* If we have handled a `1' or a `2',
639 see if there is another number to read. */
640 if (used_up)
642 /* Is this the last nonwhite stuff on the line? */
643 c = getc (finput);
644 while (c == ' ' || c == '\t')
645 c = getc (finput);
646 if (c == '\n')
647 return c;
648 ungetc (c, finput);
650 token = yylex ();
651 used_up = 0;
654 /* `3' after file name means this is a system header file. */
656 if (token == CONSTANT
657 && TREE_CODE (yylval.ttype) == INTEGER_CST
658 && TREE_INT_CST_LOW (yylval.ttype) == 3)
659 in_system_header = 1, used_up = 1;
661 if (used_up)
663 /* Is this the last nonwhite stuff on the line? */
664 c = getc (finput);
665 while (c == ' ' || c == '\t')
666 c = getc (finput);
667 if (c == '\n')
668 return c;
669 ungetc (c, finput);
672 warning ("unrecognized text at end of #line");
674 else
675 error ("invalid #-line");
677 /* skip the rest of this line. */
678 skipline:
679 while (c != '\n' && c != EOF)
680 c = getc (finput);
681 return c;
684 #ifdef HANDLE_SYSV_PRAGMA
686 /* Handle a #pragma directive. INPUT is the current input stream,
687 and C is a character to reread. Processes the entire input line
688 and returns a character for the caller to reread: either \n or EOF. */
690 /* This function has to be in this file, in order to get at
691 the token types. */
694 handle_sysv_pragma (input, c)
695 FILE *input;
696 int c;
698 for (;;)
700 while (c == ' ' || c == '\t')
701 c = getc (input);
702 if (c == '\n' || c == EOF)
704 handle_pragma_token (0, 0);
705 return c;
707 ungetc (c, input);
708 switch (yylex ())
710 case IDENTIFIER:
711 case TYPENAME:
712 case STRING:
713 case CONSTANT:
714 handle_pragma_token (token_buffer, yylval.ttype);
715 break;
716 default:
717 handle_pragma_token (token_buffer, 0);
719 if (nextchar >= 0)
720 c = nextchar, nextchar = -1;
721 else
722 c = getc (input);
726 #endif /* HANDLE_SYSV_PRAGMA */
728 #define ENDFILE -1 /* token that represents end-of-file */
730 /* Read an escape sequence, returning its equivalent as a character,
731 or store 1 in *ignore_ptr if it is backslash-newline. */
733 static int
734 readescape (ignore_ptr)
735 int *ignore_ptr;
737 register int c = getc (finput);
738 register int code;
739 register unsigned count;
740 unsigned firstdig = 0;
741 int nonnull;
743 switch (c)
745 case 'x':
746 if (warn_traditional)
747 warning ("the meaning of `\\x' varies with -traditional");
749 if (flag_traditional)
750 return c;
752 code = 0;
753 count = 0;
754 nonnull = 0;
755 while (1)
757 c = getc (finput);
758 if (!(c >= 'a' && c <= 'f')
759 && !(c >= 'A' && c <= 'F')
760 && !(c >= '0' && c <= '9'))
762 ungetc (c, finput);
763 break;
765 code *= 16;
766 if (c >= 'a' && c <= 'f')
767 code += c - 'a' + 10;
768 if (c >= 'A' && c <= 'F')
769 code += c - 'A' + 10;
770 if (c >= '0' && c <= '9')
771 code += c - '0';
772 if (code != 0 || count != 0)
774 if (count == 0)
775 firstdig = code;
776 count++;
778 nonnull = 1;
780 if (! nonnull)
781 error ("\\x used with no following hex digits");
782 else if (count == 0)
783 /* Digits are all 0's. Ok. */
785 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
786 || (count > 1
787 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
788 <= firstdig)))
789 pedwarn ("hex escape out of range");
790 return code;
792 case '0': case '1': case '2': case '3': case '4':
793 case '5': case '6': case '7':
794 code = 0;
795 count = 0;
796 while ((c <= '7') && (c >= '0') && (count++ < 3))
798 code = (code * 8) + (c - '0');
799 c = getc (finput);
801 ungetc (c, finput);
802 return code;
804 case '\\': case '\'': case '"':
805 return c;
807 case '\n':
808 lineno++;
809 *ignore_ptr = 1;
810 return 0;
812 case 'n':
813 return TARGET_NEWLINE;
815 case 't':
816 return TARGET_TAB;
818 case 'r':
819 return TARGET_CR;
821 case 'f':
822 return TARGET_FF;
824 case 'b':
825 return TARGET_BS;
827 case 'a':
828 if (warn_traditional)
829 warning ("the meaning of `\\a' varies with -traditional");
831 if (flag_traditional)
832 return c;
833 return TARGET_BELL;
835 case 'v':
836 #if 0 /* Vertical tab is present in common usage compilers. */
837 if (flag_traditional)
838 return c;
839 #endif
840 return TARGET_VT;
842 case 'e':
843 case 'E':
844 if (pedantic)
845 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
846 return 033;
848 case '?':
849 return c;
851 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
852 case '(':
853 case '{':
854 case '[':
855 /* `\%' is used to prevent SCCS from getting confused. */
856 case '%':
857 if (pedantic)
858 pedwarn ("non-ANSI escape sequence `\\%c'", c);
859 return c;
861 if (c >= 040 && c < 0177)
862 pedwarn ("unknown escape sequence `\\%c'", c);
863 else
864 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
865 return c;
868 void
869 yyerror (string)
870 char *string;
872 char buf[200];
874 strcpy (buf, string);
876 /* We can't print string and character constants well
877 because the token_buffer contains the result of processing escapes. */
878 if (end_of_file)
879 strcat (buf, " at end of input");
880 else if (token_buffer[0] == 0)
881 strcat (buf, " at null character");
882 else if (token_buffer[0] == '"')
883 strcat (buf, " before string constant");
884 else if (token_buffer[0] == '\'')
885 strcat (buf, " before character constant");
886 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
887 sprintf (buf + strlen (buf), " before character 0%o",
888 (unsigned char) token_buffer[0]);
889 else
890 strcat (buf, " before `%s'");
892 error (buf, token_buffer);
895 #if 0
897 struct try_type
899 tree *node_var;
900 char unsigned_flag;
901 char long_flag;
902 char long_long_flag;
905 struct try_type type_sequence[] =
907 { &integer_type_node, 0, 0, 0},
908 { &unsigned_type_node, 1, 0, 0},
909 { &long_integer_type_node, 0, 1, 0},
910 { &long_unsigned_type_node, 1, 1, 0},
911 { &long_long_integer_type_node, 0, 1, 1},
912 { &long_long_unsigned_type_node, 1, 1, 1}
914 #endif /* 0 */
917 yylex ()
919 register int c;
920 register char *p;
921 register int value;
922 int wide_flag = 0;
923 int objc_flag = 0;
925 if (nextchar >= 0)
926 c = nextchar, nextchar = -1;
927 else
928 c = getc (finput);
930 /* Effectively do c = skip_white_space (c)
931 but do it faster in the usual cases. */
932 while (1)
933 switch (c)
935 case ' ':
936 case '\t':
937 case '\f':
938 case '\v':
939 case '\b':
940 c = getc (finput);
941 break;
943 case '\r':
944 /* Call skip_white_space so we can warn if appropriate. */
946 case '\n':
947 case '/':
948 case '\\':
949 c = skip_white_space (c);
950 default:
951 goto found_nonwhite;
953 found_nonwhite:
955 token_buffer[0] = c;
956 token_buffer[1] = 0;
958 /* yylloc.first_line = lineno; */
960 switch (c)
962 case EOF:
963 end_of_file = 1;
964 token_buffer[0] = 0;
965 value = ENDFILE;
966 break;
968 case '$':
969 if (dollars_in_ident)
970 goto letter;
971 return '$';
973 case 'L':
974 /* Capital L may start a wide-string or wide-character constant. */
976 register int c = getc (finput);
977 if (c == '\'')
979 wide_flag = 1;
980 goto char_constant;
982 if (c == '"')
984 wide_flag = 1;
985 goto string_constant;
987 ungetc (c, finput);
989 goto letter;
991 case '@':
992 if (!doing_objc_thang)
994 value = c;
995 break;
997 else
999 /* '@' may start a constant string object. */
1000 register int c = getc(finput);
1001 if (c == '"')
1003 objc_flag = 1;
1004 goto string_constant;
1006 ungetc(c, finput);
1007 /* Fall through to treat '@' as the start of an identifier. */
1010 case 'A': case 'B': case 'C': case 'D': case 'E':
1011 case 'F': case 'G': case 'H': case 'I': case 'J':
1012 case 'K': case 'M': case 'N': case 'O':
1013 case 'P': case 'Q': case 'R': case 'S': case 'T':
1014 case 'U': case 'V': case 'W': case 'X': case 'Y':
1015 case 'Z':
1016 case 'a': case 'b': case 'c': case 'd': case 'e':
1017 case 'f': case 'g': case 'h': case 'i': case 'j':
1018 case 'k': case 'l': case 'm': case 'n': case 'o':
1019 case 'p': case 'q': case 'r': case 's': case 't':
1020 case 'u': case 'v': case 'w': case 'x': case 'y':
1021 case 'z':
1022 case '_':
1023 letter:
1024 p = token_buffer;
1025 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1027 /* Make sure this char really belongs in an identifier. */
1028 if (c == '@' && ! doing_objc_thang)
1029 break;
1030 if (c == '$' && ! dollars_in_ident)
1031 break;
1033 if (p >= token_buffer + maxtoken)
1034 p = extend_token_buffer (p);
1036 *p++ = c;
1037 c = getc (finput);
1040 *p = 0;
1041 nextchar = c;
1043 value = IDENTIFIER;
1044 yylval.itype = 0;
1046 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1049 register struct resword *ptr;
1051 if (ptr = is_reserved_word (token_buffer, p - token_buffer))
1053 if (ptr->rid)
1054 yylval.ttype = ridpointers[(int) ptr->rid];
1055 value = (int) ptr->token;
1057 /* Only return OBJECTNAME if it is a typedef. */
1058 if (doing_objc_thang && value == OBJECTNAME)
1060 lastiddecl = lookup_name(yylval.ttype);
1062 if (lastiddecl == NULL_TREE
1063 || TREE_CODE (lastiddecl) != TYPE_DECL)
1064 value = IDENTIFIER;
1067 /* Even if we decided to recognize asm, still perhaps warn. */
1068 if (pedantic
1069 && (value == ASM_KEYWORD || value == TYPEOF
1070 || ptr->rid == RID_INLINE)
1071 && token_buffer[0] != '_')
1072 pedwarn ("ANSI does not permit the keyword `%s'",
1073 token_buffer);
1077 /* If we did not find a keyword, look for an identifier
1078 (or a typename). */
1080 if (value == IDENTIFIER)
1082 if (token_buffer[0] == '@')
1083 error("invalid identifier `%s'", token_buffer);
1085 yylval.ttype = get_identifier (token_buffer);
1086 lastiddecl = lookup_name (yylval.ttype);
1088 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1089 value = TYPENAME;
1090 /* A user-invisible read-only initialized variable
1091 should be replaced by its value.
1092 We handle only strings since that's the only case used in C. */
1093 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1094 && DECL_IGNORED_P (lastiddecl)
1095 && TREE_READONLY (lastiddecl)
1096 && DECL_INITIAL (lastiddecl) != 0
1097 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1099 tree stringval = DECL_INITIAL (lastiddecl);
1101 /* Copy the string value so that we won't clobber anything
1102 if we put something in the TREE_CHAIN of this one. */
1103 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1104 TREE_STRING_POINTER (stringval));
1105 value = STRING;
1107 else if (doing_objc_thang)
1109 tree objc_interface_decl = is_class_name (yylval.ttype);
1111 if (objc_interface_decl)
1113 value = CLASSNAME;
1114 yylval.ttype = objc_interface_decl;
1119 break;
1121 case '0': case '1':
1123 int next_c;
1124 /* Check first for common special case: single-digit 0 or 1. */
1126 next_c = getc (finput);
1127 ungetc (next_c, finput); /* Always undo this lookahead. */
1128 if (!isalnum (next_c) && next_c != '.')
1130 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1131 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1132 value = CONSTANT;
1133 break;
1135 /*FALLTHRU*/
1137 case '2': case '3': case '4':
1138 case '5': case '6': case '7': case '8': case '9':
1139 case '.':
1141 int base = 10;
1142 int count = 0;
1143 int largest_digit = 0;
1144 int numdigits = 0;
1145 /* for multi-precision arithmetic,
1146 we actually store only HOST_BITS_PER_CHAR bits in each part.
1147 The number of parts is chosen so as to be sufficient to hold
1148 the enough bits to fit into the two HOST_WIDE_INTs that contain
1149 the integer value (this is always at least as many bits as are
1150 in a target `long long' value, but may be wider). */
1151 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1152 int parts[TOTAL_PARTS];
1153 int overflow = 0;
1155 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1156 = NOT_FLOAT;
1158 for (count = 0; count < TOTAL_PARTS; count++)
1159 parts[count] = 0;
1161 p = token_buffer;
1162 *p++ = c;
1164 if (c == '0')
1166 *p++ = (c = getc (finput));
1167 if ((c == 'x') || (c == 'X'))
1169 base = 16;
1170 *p++ = (c = getc (finput));
1172 /* Leading 0 forces octal unless the 0 is the only digit. */
1173 else if (c >= '0' && c <= '9')
1175 base = 8;
1176 numdigits++;
1178 else
1179 numdigits++;
1182 /* Read all the digits-and-decimal-points. */
1184 while (c == '.'
1185 || (isalnum (c) && c != 'l' && c != 'L'
1186 && c != 'u' && c != 'U'
1187 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1188 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1190 if (c == '.')
1192 if (base == 16)
1193 error ("floating constant may not be in radix 16");
1194 if (floatflag == TOO_MANY_POINTS)
1195 /* We have already emitted an error. Don't need another. */
1197 else if (floatflag == AFTER_POINT)
1199 error ("malformed floating constant");
1200 floatflag = TOO_MANY_POINTS;
1201 /* Avoid another error from atof by forcing all characters
1202 from here on to be ignored. */
1203 p[-1] = '\0';
1205 else
1206 floatflag = AFTER_POINT;
1208 base = 10;
1209 *p++ = c = getc (finput);
1210 /* Accept '.' as the start of a floating-point number
1211 only when it is followed by a digit.
1212 Otherwise, unread the following non-digit
1213 and use the '.' as a structural token. */
1214 if (p == token_buffer + 2 && !isdigit (c))
1216 if (c == '.')
1218 c = getc (finput);
1219 if (c == '.')
1221 *p++ = c;
1222 *p = 0;
1223 return ELLIPSIS;
1225 error ("parse error at `..'");
1227 ungetc (c, finput);
1228 token_buffer[1] = 0;
1229 value = '.';
1230 goto done;
1233 else
1235 /* It is not a decimal point.
1236 It should be a digit (perhaps a hex digit). */
1238 if (isdigit (c))
1240 c = c - '0';
1242 else if (base <= 10)
1244 if (c == 'e' || c == 'E')
1246 base = 10;
1247 floatflag = AFTER_POINT;
1248 break; /* start of exponent */
1250 error ("nondigits in number and not hexadecimal");
1251 c = 0;
1253 else if (c >= 'a')
1255 c = c - 'a' + 10;
1257 else
1259 c = c - 'A' + 10;
1261 if (c >= largest_digit)
1262 largest_digit = c;
1263 numdigits++;
1265 for (count = 0; count < TOTAL_PARTS; count++)
1267 parts[count] *= base;
1268 if (count)
1270 parts[count]
1271 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1272 parts[count-1]
1273 &= (1 << HOST_BITS_PER_CHAR) - 1;
1275 else
1276 parts[0] += c;
1279 /* If the extra highest-order part ever gets anything in it,
1280 the number is certainly too big. */
1281 if (parts[TOTAL_PARTS - 1] != 0)
1282 overflow = 1;
1284 if (p >= token_buffer + maxtoken - 3)
1285 p = extend_token_buffer (p);
1286 *p++ = (c = getc (finput));
1290 if (numdigits == 0)
1291 error ("numeric constant with no digits");
1293 if (largest_digit >= base)
1294 error ("numeric constant contains digits beyond the radix");
1296 /* Remove terminating char from the token buffer and delimit the string */
1297 *--p = 0;
1299 if (floatflag != NOT_FLOAT)
1301 tree type = double_type_node;
1302 int exceeds_double = 0;
1303 int imag = 0;
1304 REAL_VALUE_TYPE value;
1305 jmp_buf handler;
1307 /* Read explicit exponent if any, and put it in tokenbuf. */
1309 if ((c == 'e') || (c == 'E'))
1311 if (p >= token_buffer + maxtoken - 3)
1312 p = extend_token_buffer (p);
1313 *p++ = c;
1314 c = getc (finput);
1315 if ((c == '+') || (c == '-'))
1317 *p++ = c;
1318 c = getc (finput);
1320 if (! isdigit (c))
1321 error ("floating constant exponent has no digits");
1322 while (isdigit (c))
1324 if (p >= token_buffer + maxtoken - 3)
1325 p = extend_token_buffer (p);
1326 *p++ = c;
1327 c = getc (finput);
1331 *p = 0;
1332 errno = 0;
1334 /* Convert string to a double, checking for overflow. */
1335 if (setjmp (handler))
1337 error ("floating constant out of range");
1338 value = dconst0;
1340 else
1342 int fflag = 0, lflag = 0;
1343 /* Copy token_buffer now, while it has just the number
1344 and not the suffixes; once we add `f' or `i',
1345 REAL_VALUE_ATOF may not work any more. */
1346 char *copy = (char *) alloca (p - token_buffer + 1);
1347 bcopy (token_buffer, copy, p - token_buffer + 1);
1349 set_float_handler (handler);
1351 while (1)
1353 int lose = 0;
1355 /* Read the suffixes to choose a data type. */
1356 switch (c)
1358 case 'f': case 'F':
1359 if (fflag)
1360 error ("more than one `f' in numeric constant");
1361 fflag = 1;
1362 break;
1364 case 'l': case 'L':
1365 if (lflag)
1366 error ("more than one `l' in numeric constant");
1367 lflag = 1;
1368 break;
1370 case 'i': case 'I':
1371 if (imag)
1372 error ("more than one `i' or `j' in numeric constant");
1373 else if (pedantic)
1374 pedwarn ("ANSI C forbids imaginary numeric constants");
1375 imag = 1;
1376 break;
1378 default:
1379 lose = 1;
1382 if (lose)
1383 break;
1385 if (p >= token_buffer + maxtoken - 3)
1386 p = extend_token_buffer (p);
1387 *p++ = c;
1388 *p = 0;
1389 c = getc (finput);
1392 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1393 tells the desired precision of the binary result
1394 of decimal-to-binary conversion. */
1396 if (fflag)
1398 if (lflag)
1399 error ("both `f' and `l' in floating constant");
1401 type = float_type_node;
1402 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1403 /* A diagnostic is required here by some ANSI C testsuites.
1404 This is not pedwarn, become some people don't want
1405 an error for this. */
1406 if (REAL_VALUE_ISINF (value) && pedantic)
1407 warning ("floating point number exceeds range of `float'");
1409 else if (lflag)
1411 type = long_double_type_node;
1412 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1413 if (REAL_VALUE_ISINF (value) && pedantic)
1414 warning ("floating point number exceeds range of `long double'");
1416 else
1418 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1419 if (REAL_VALUE_ISINF (value) && pedantic)
1420 warning ("floating point number exceeds range of `double'");
1423 set_float_handler (NULL_PTR);
1425 #ifdef ERANGE
1426 if (errno == ERANGE && !flag_traditional && pedantic)
1428 /* ERANGE is also reported for underflow,
1429 so test the value to distinguish overflow from that. */
1430 if (REAL_VALUES_LESS (dconst1, value)
1431 || REAL_VALUES_LESS (value, dconstm1))
1433 warning ("floating point number exceeds range of `double'");
1434 exceeds_double = 1;
1437 #endif
1439 /* If the result is not a number, assume it must have been
1440 due to some error message above, so silently convert
1441 it to a zero. */
1442 if (REAL_VALUE_ISNAN (value))
1443 value = dconst0;
1445 /* Create a node with determined type and value. */
1446 if (imag)
1447 yylval.ttype = build_complex (convert (type, integer_zero_node),
1448 build_real (type, value));
1449 else
1450 yylval.ttype = build_real (type, value);
1452 else
1454 tree traditional_type, ansi_type, type;
1455 HOST_WIDE_INT high, low;
1456 int spec_unsigned = 0;
1457 int spec_long = 0;
1458 int spec_long_long = 0;
1459 int spec_imag = 0;
1460 int bytes, warn, i;
1462 while (1)
1464 if (c == 'u' || c == 'U')
1466 if (spec_unsigned)
1467 error ("two `u's in integer constant");
1468 spec_unsigned = 1;
1470 else if (c == 'l' || c == 'L')
1472 if (spec_long)
1474 if (spec_long_long)
1475 error ("three `l's in integer constant");
1476 else if (pedantic)
1477 pedwarn ("ANSI C forbids long long integer constants");
1478 spec_long_long = 1;
1480 spec_long = 1;
1482 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1484 if (spec_imag)
1485 error ("more than one `i' or `j' in numeric constant");
1486 else if (pedantic)
1487 pedwarn ("ANSI C forbids imaginary numeric constants");
1488 spec_imag = 1;
1490 else
1491 break;
1492 if (p >= token_buffer + maxtoken - 3)
1493 p = extend_token_buffer (p);
1494 *p++ = c;
1495 c = getc (finput);
1498 /* If the constant is not long long and it won't fit in an
1499 unsigned long, or if the constant is long long and won't fit
1500 in an unsigned long long, then warn that the constant is out
1501 of range. */
1503 /* ??? This assumes that long long and long integer types are
1504 a multiple of 8 bits. This better than the original code
1505 though which assumed that long was exactly 32 bits and long
1506 long was exactly 64 bits. */
1508 if (spec_long_long)
1509 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1510 else
1511 bytes = TYPE_PRECISION (long_integer_type_node) / 8;
1513 warn = overflow;
1514 for (i = bytes; i < TOTAL_PARTS; i++)
1515 if (parts[i])
1516 warn = 1;
1517 if (warn)
1518 pedwarn ("integer constant out of range");
1520 /* This is simplified by the fact that our constant
1521 is always positive. */
1523 high = low = 0;
1525 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1527 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1528 / HOST_BITS_PER_CHAR)]
1529 << (i * HOST_BITS_PER_CHAR));
1530 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1533 yylval.ttype = build_int_2 (low, high);
1534 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1536 /* If warn_traditional, calculate both the ANSI type and the
1537 traditional type, then see if they disagree.
1538 Otherwise, calculate only the type for the dialect in use. */
1539 if (warn_traditional || flag_traditional)
1541 /* Calculate the traditional type. */
1542 /* Traditionally, any constant is signed;
1543 but if unsigned is specified explicitly, obey that.
1544 Use the smallest size with the right number of bits,
1545 except for one special case with decimal constants. */
1546 if (! spec_long && base != 10
1547 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1548 traditional_type = (spec_unsigned ? unsigned_type_node
1549 : integer_type_node);
1550 /* A decimal constant must be long
1551 if it does not fit in type int.
1552 I think this is independent of whether
1553 the constant is signed. */
1554 else if (! spec_long && base == 10
1555 && int_fits_type_p (yylval.ttype, integer_type_node))
1556 traditional_type = (spec_unsigned ? unsigned_type_node
1557 : integer_type_node);
1558 else if (! spec_long_long)
1559 traditional_type = (spec_unsigned ? long_unsigned_type_node
1560 : long_integer_type_node);
1561 else
1562 traditional_type = (spec_unsigned
1563 ? long_long_unsigned_type_node
1564 : long_long_integer_type_node);
1566 if (warn_traditional || ! flag_traditional)
1568 /* Calculate the ANSI type. */
1569 if (! spec_long && ! spec_unsigned
1570 && int_fits_type_p (yylval.ttype, integer_type_node))
1571 ansi_type = integer_type_node;
1572 else if (! spec_long && (base != 10 || spec_unsigned)
1573 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1574 ansi_type = unsigned_type_node;
1575 else if (! spec_unsigned && !spec_long_long
1576 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1577 ansi_type = long_integer_type_node;
1578 else if (! spec_long_long)
1579 ansi_type = long_unsigned_type_node;
1580 else if (! spec_unsigned
1581 /* Verify value does not overflow into sign bit. */
1582 && TREE_INT_CST_HIGH (yylval.ttype) >= 0
1583 && int_fits_type_p (yylval.ttype,
1584 long_long_integer_type_node))
1585 ansi_type = long_long_integer_type_node;
1586 else
1587 ansi_type = long_long_unsigned_type_node;
1590 type = flag_traditional ? traditional_type : ansi_type;
1592 if (warn_traditional && traditional_type != ansi_type)
1594 if (TYPE_PRECISION (traditional_type)
1595 != TYPE_PRECISION (ansi_type))
1596 warning ("width of integer constant changes with -traditional");
1597 else if (TREE_UNSIGNED (traditional_type)
1598 != TREE_UNSIGNED (ansi_type))
1599 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1600 else
1601 warning ("width of integer constant may change on other systems with -traditional");
1604 if (!flag_traditional && !int_fits_type_p (yylval.ttype, type)
1605 && !warn)
1606 pedwarn ("integer constant out of range");
1608 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1609 warning ("decimal constant is so large that it is unsigned");
1611 if (spec_imag)
1613 if (TYPE_PRECISION (type)
1614 <= TYPE_PRECISION (integer_type_node))
1615 yylval.ttype
1616 = build_complex (integer_zero_node,
1617 convert (integer_type_node, yylval.ttype));
1618 else
1619 error ("complex integer constant is too wide for `complex int'");
1621 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1622 /* The traditional constant 0x80000000 is signed
1623 but doesn't fit in the range of int.
1624 This will change it to -0x80000000, which does fit. */
1626 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1627 yylval.ttype = convert (type, yylval.ttype);
1628 TREE_OVERFLOW (yylval.ttype)
1629 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1631 else
1632 TREE_TYPE (yylval.ttype) = type;
1635 ungetc (c, finput);
1636 *p = 0;
1638 if (isalnum (c) || c == '.' || c == '_'
1639 || (!flag_traditional && (c == '-' || c == '+')
1640 && (p[-1] == 'e' || p[-1] == 'E')))
1641 error ("missing white space after number `%s'", token_buffer);
1643 value = CONSTANT; break;
1646 case '\'':
1647 char_constant:
1649 register int result = 0;
1650 register int num_chars = 0;
1651 unsigned width = TYPE_PRECISION (char_type_node);
1652 int max_chars;
1654 if (wide_flag)
1656 width = WCHAR_TYPE_SIZE;
1657 #ifdef MULTIBYTE_CHARS
1658 max_chars = MB_CUR_MAX;
1659 #else
1660 max_chars = 1;
1661 #endif
1663 else
1664 max_chars = TYPE_PRECISION (integer_type_node) / width;
1666 while (1)
1668 tryagain:
1670 c = getc (finput);
1672 if (c == '\'' || c == EOF)
1673 break;
1675 if (c == '\\')
1677 int ignore = 0;
1678 c = readescape (&ignore);
1679 if (ignore)
1680 goto tryagain;
1681 if (width < HOST_BITS_PER_INT
1682 && (unsigned) c >= (1 << width))
1683 pedwarn ("escape sequence out of range for character");
1684 #ifdef MAP_CHARACTER
1685 if (isprint (c))
1686 c = MAP_CHARACTER (c);
1687 #endif
1689 else if (c == '\n')
1691 if (pedantic)
1692 pedwarn ("ANSI C forbids newline in character constant");
1693 lineno++;
1695 #ifdef MAP_CHARACTER
1696 else
1697 c = MAP_CHARACTER (c);
1698 #endif
1700 num_chars++;
1701 if (num_chars > maxtoken - 4)
1702 extend_token_buffer (token_buffer);
1704 token_buffer[num_chars] = c;
1706 /* Merge character into result; ignore excess chars. */
1707 if (num_chars < max_chars + 1)
1709 if (width < HOST_BITS_PER_INT)
1710 result = (result << width) | (c & ((1 << width) - 1));
1711 else
1712 result = c;
1716 token_buffer[num_chars + 1] = '\'';
1717 token_buffer[num_chars + 2] = 0;
1719 if (c != '\'')
1720 error ("malformatted character constant");
1721 else if (num_chars == 0)
1722 error ("empty character constant");
1723 else if (num_chars > max_chars)
1725 num_chars = max_chars;
1726 error ("character constant too long");
1728 else if (num_chars != 1 && ! flag_traditional)
1729 warning ("multi-character character constant");
1731 /* If char type is signed, sign-extend the constant. */
1732 if (! wide_flag)
1734 int num_bits = num_chars * width;
1735 if (num_bits == 0)
1736 /* We already got an error; avoid invalid shift. */
1737 yylval.ttype = build_int_2 (0, 0);
1738 else if (TREE_UNSIGNED (char_type_node)
1739 || ((result >> (num_bits - 1)) & 1) == 0)
1740 yylval.ttype
1741 = build_int_2 (result & ((unsigned HOST_WIDE_INT) ~0
1742 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1744 else
1745 yylval.ttype
1746 = build_int_2 (result | ~((unsigned HOST_WIDE_INT) ~0
1747 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1748 -1);
1749 TREE_TYPE (yylval.ttype) = integer_type_node;
1751 else
1753 #ifdef MULTIBYTE_CHARS
1754 /* Set the initial shift state and convert the next sequence. */
1755 result = 0;
1756 /* In all locales L'\0' is zero and mbtowc will return zero,
1757 so don't use it. */
1758 if (num_chars > 1
1759 || (num_chars == 1 && token_buffer[1] != '\0'))
1761 wchar_t wc;
1762 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
1763 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
1764 result = wc;
1765 else
1766 warning ("Ignoring invalid multibyte character");
1768 #endif
1769 yylval.ttype = build_int_2 (result, 0);
1770 TREE_TYPE (yylval.ttype) = wchar_type_node;
1773 value = CONSTANT;
1774 break;
1777 case '"':
1778 string_constant:
1780 c = getc (finput);
1781 p = token_buffer + 1;
1783 while (c != '"' && c >= 0)
1785 if (c == '\\')
1787 int ignore = 0;
1788 c = readescape (&ignore);
1789 if (ignore)
1790 goto skipnewline;
1791 if (!wide_flag
1792 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
1793 && c >= (1 << TYPE_PRECISION (char_type_node)))
1794 pedwarn ("escape sequence out of range for character");
1796 else if (c == '\n')
1798 if (pedantic)
1799 pedwarn ("ANSI C forbids newline in string constant");
1800 lineno++;
1803 if (p == token_buffer + maxtoken)
1804 p = extend_token_buffer (p);
1805 *p++ = c;
1807 skipnewline:
1808 c = getc (finput);
1810 *p = 0;
1812 if (c < 0)
1813 error ("Unterminated string constant");
1815 /* We have read the entire constant.
1816 Construct a STRING_CST for the result. */
1818 if (wide_flag)
1820 /* If this is a L"..." wide-string, convert the multibyte string
1821 to a wide character string. */
1822 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
1823 int len;
1825 #ifdef MULTIBYTE_CHARS
1826 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
1827 if (len < 0 || len >= (p - token_buffer))
1829 warning ("Ignoring invalid multibyte string");
1830 len = 0;
1832 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
1833 #else
1835 union { long l; char c[sizeof (long)]; } u;
1836 int big_endian;
1837 char *wp, *cp;
1839 /* Determine whether host is little or big endian. */
1840 u.l = 1;
1841 big_endian = u.c[sizeof (long) - 1];
1842 wp = widep + (big_endian ? WCHAR_BYTES - 1 : 0);
1844 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
1845 for (cp = token_buffer + 1; cp < p; cp++)
1846 *wp = *cp, wp += WCHAR_BYTES;
1847 len = p - token_buffer - 1;
1849 #endif
1850 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
1851 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
1852 value = STRING;
1854 else if (objc_flag)
1856 extern tree build_objc_string();
1857 /* Return an Objective-C @"..." constant string object. */
1858 yylval.ttype = build_objc_string (p - token_buffer,
1859 token_buffer + 1);
1860 TREE_TYPE (yylval.ttype) = char_array_type_node;
1861 value = OBJC_STRING;
1863 else
1865 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
1866 TREE_TYPE (yylval.ttype) = char_array_type_node;
1867 value = STRING;
1870 *p++ = '"';
1871 *p = 0;
1873 break;
1876 case '+':
1877 case '-':
1878 case '&':
1879 case '|':
1880 case ':':
1881 case '<':
1882 case '>':
1883 case '*':
1884 case '/':
1885 case '%':
1886 case '^':
1887 case '!':
1888 case '=':
1890 register int c1;
1892 combine:
1894 switch (c)
1896 case '+':
1897 yylval.code = PLUS_EXPR; break;
1898 case '-':
1899 yylval.code = MINUS_EXPR; break;
1900 case '&':
1901 yylval.code = BIT_AND_EXPR; break;
1902 case '|':
1903 yylval.code = BIT_IOR_EXPR; break;
1904 case '*':
1905 yylval.code = MULT_EXPR; break;
1906 case '/':
1907 yylval.code = TRUNC_DIV_EXPR; break;
1908 case '%':
1909 yylval.code = TRUNC_MOD_EXPR; break;
1910 case '^':
1911 yylval.code = BIT_XOR_EXPR; break;
1912 case LSHIFT:
1913 yylval.code = LSHIFT_EXPR; break;
1914 case RSHIFT:
1915 yylval.code = RSHIFT_EXPR; break;
1916 case '<':
1917 yylval.code = LT_EXPR; break;
1918 case '>':
1919 yylval.code = GT_EXPR; break;
1922 token_buffer[1] = c1 = getc (finput);
1923 token_buffer[2] = 0;
1925 if (c1 == '=')
1927 switch (c)
1929 case '<':
1930 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
1931 case '>':
1932 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
1933 case '!':
1934 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
1935 case '=':
1936 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
1938 value = ASSIGN; goto done;
1940 else if (c == c1)
1941 switch (c)
1943 case '+':
1944 value = PLUSPLUS; goto done;
1945 case '-':
1946 value = MINUSMINUS; goto done;
1947 case '&':
1948 value = ANDAND; goto done;
1949 case '|':
1950 value = OROR; goto done;
1951 case '<':
1952 c = LSHIFT;
1953 goto combine;
1954 case '>':
1955 c = RSHIFT;
1956 goto combine;
1958 else
1959 switch (c)
1961 case '-':
1962 if (c1 == '>')
1963 { value = POINTSAT; goto done; }
1964 break;
1965 case ':':
1966 if (c1 == '>')
1967 { value = ']'; goto done; }
1968 break;
1969 case '<':
1970 if (c1 == '%')
1971 { value = '{'; goto done; }
1972 if (c1 == ':')
1973 { value = '['; goto done; }
1974 break;
1975 case '%':
1976 if (c1 == '>')
1977 { value = '}'; goto done; }
1978 break;
1980 ungetc (c1, finput);
1981 token_buffer[1] = 0;
1983 if ((c == '<') || (c == '>'))
1984 value = ARITHCOMPARE;
1985 else value = c;
1986 goto done;
1989 case 0:
1990 /* Don't make yyparse think this is eof. */
1991 value = 1;
1992 break;
1994 default:
1995 value = c;
1998 done:
1999 /* yylloc.last_line = lineno; */
2001 return value;
2004 /* Sets the value of the 'yydebug' variable to VALUE.
2005 This is a function so we don't have to have YYDEBUG defined
2006 in order to build the compiler. */
2008 void
2009 set_yydebug (value)
2010 int value;
2012 #if YYDEBUG != 0
2013 yydebug = value;
2014 #else
2015 warning ("YYDEBUG not defined.");
2016 #endif