Fix most compile time warning messages
[official-gcc.git] / gcc / c-lex.c
blob616ecbe2080491e97c9e2b59d31cd529d2e2698c
1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 1988, 1989, 1992, 1994, 1995, 1996, 1997
3 1998, 1999, 2000 Free Software Foundation, Inc.
5 This file is part of GNU CC.
7 GNU CC is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2, or (at your option)
10 any later version.
12 GNU CC is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with GNU CC; see the file COPYING. If not, write to
19 the Free Software Foundation, 59 Temple Place - Suite 330,
20 Boston, MA 02111-1307, USA. */
22 #include "config.h"
23 #include "system.h"
25 #include "rtl.h"
26 #include "tree.h"
27 #include "input.h"
28 #include "output.h"
29 #include "c-lex.h"
30 #include "c-tree.h"
31 #include "flags.h"
32 #include "c-parse.h"
33 #include "c-pragma.h"
34 #include "toplev.h"
35 #include "intl.h"
36 #include "ggc.h"
37 #include "tm_p.h"
39 /* MULTIBYTE_CHARS support only works for native compilers.
40 ??? Ideally what we want is to model widechar support after
41 the current floating point support. */
42 #ifdef CROSS_COMPILE
43 #undef MULTIBYTE_CHARS
44 #endif
46 #ifdef MULTIBYTE_CHARS
47 #include "mbchar.h"
48 #include <locale.h>
49 #endif /* MULTIBYTE_CHARS */
50 #ifndef GET_ENVIRONMENT
51 #define GET_ENVIRONMENT(ENV_VALUE,ENV_NAME) ((ENV_VALUE) = getenv (ENV_NAME))
52 #endif
54 #if USE_CPPLIB
55 #include "cpplib.h"
56 extern cpp_reader parse_in;
57 extern cpp_options parse_options;
58 #else
59 /* Stream for reading from the input file. */
60 FILE *finput;
61 #endif
63 extern void yyprint PARAMS ((FILE *, int, YYSTYPE));
65 /* Cause the `yydebug' variable to be defined. */
66 #define YYDEBUG 1
68 #if USE_CPPLIB
69 extern unsigned char *yy_cur, *yy_lim;
70 extern enum cpp_token cpp_token;
72 extern int yy_get_token ();
74 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
75 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
77 #else /* ! USE_CPPLIB */
79 #define GETC() getch ()
80 #define UNGETC(c) put_back (c)
82 struct putback_buffer {
83 unsigned char *buffer;
84 int buffer_size;
85 int index;
88 static struct putback_buffer putback = {NULL, 0, -1};
90 static inline int getch PARAMS ((void));
92 static inline int
93 getch ()
95 if (putback.index != -1)
97 int ch = putback.buffer[putback.index];
98 --putback.index;
99 return ch;
101 return getc (finput);
104 static inline void put_back PARAMS ((int));
106 static inline void
107 put_back (ch)
108 int ch;
110 if (ch != EOF)
112 if (putback.index == putback.buffer_size - 1)
114 putback.buffer_size += 16;
115 putback.buffer = xrealloc (putback.buffer, putback.buffer_size);
117 putback.buffer[++putback.index] = ch;
120 #endif /* ! USE_CPPLIB */
122 int linemode;
124 extern int yydebug;
126 /* File used for outputting assembler code. */
127 extern FILE *asm_out_file;
129 #undef WCHAR_TYPE_SIZE
130 #define WCHAR_TYPE_SIZE TYPE_PRECISION (wchar_type_node)
132 /* Number of bytes in a wide character. */
133 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
135 static int maxtoken; /* Current nominal length of token buffer. */
136 char *token_buffer; /* Pointer to token buffer.
137 Actual allocated length is maxtoken + 2.
138 This is not static because objc-parse.y uses it. */
140 static int indent_level; /* Number of { minus number of }. */
142 /* Nonzero tells yylex to ignore \ in string constants. */
143 static int ignore_escape_flag;
145 /* Nonzero if end-of-file has been seen on input. */
146 static int end_of_file;
148 #ifdef HANDLE_GENERIC_PRAGMAS
149 static int handle_generic_pragma PARAMS ((int));
150 #endif /* HANDLE_GENERIC_PRAGMAS */
151 static int whitespace_cr PARAMS ((int));
152 static int skip_white_space PARAMS ((int));
153 static char *extend_token_buffer PARAMS ((const char *));
154 static int readescape PARAMS ((int *));
155 static void parse_float PARAMS ((PTR));
156 static void extend_token_buffer_to PARAMS ((int));
157 static int read_line_number PARAMS ((int *));
159 /* Do not insert generated code into the source, instead, include it.
160 This allows us to build gcc automatically even for targets that
161 need to add or modify the reserved keyword lists. */
162 #include "c-gperf.h"
164 /* Return something to represent absolute declarators containing a *.
165 TARGET is the absolute declarator that the * contains.
166 TYPE_QUALS is a list of modifiers such as const or volatile
167 to apply to the pointer type, represented as identifiers.
169 We return an INDIRECT_REF whose "contents" are TARGET
170 and whose type is the modifier list. */
172 tree
173 make_pointer_declarator (type_quals, target)
174 tree type_quals, target;
176 return build1 (INDIRECT_REF, type_quals, target);
179 void
180 forget_protocol_qualifiers ()
182 int i, n = sizeof wordlist / sizeof (struct resword);
184 for (i = 0; i < n; i++)
185 if ((int) wordlist[i].rid >= (int) RID_IN
186 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
187 wordlist[i].name = "";
190 void
191 remember_protocol_qualifiers ()
193 int i, n = sizeof wordlist / sizeof (struct resword);
195 for (i = 0; i < n; i++)
196 if (wordlist[i].rid == RID_IN)
197 wordlist[i].name = "in";
198 else if (wordlist[i].rid == RID_OUT)
199 wordlist[i].name = "out";
200 else if (wordlist[i].rid == RID_INOUT)
201 wordlist[i].name = "inout";
202 else if (wordlist[i].rid == RID_BYCOPY)
203 wordlist[i].name = "bycopy";
204 else if (wordlist[i].rid == RID_BYREF)
205 wordlist[i].name = "byref";
206 else if (wordlist[i].rid == RID_ONEWAY)
207 wordlist[i].name = "oneway";
210 const char *
211 init_parse (filename)
212 const char *filename;
214 #if !USE_CPPLIB
215 /* Open input file. */
216 if (filename == 0 || !strcmp (filename, "-"))
218 finput = stdin;
219 filename = "stdin";
221 else
222 finput = fopen (filename, "r");
223 if (finput == 0)
224 pfatal_with_name (filename);
226 #ifdef IO_BUFFER_SIZE
227 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
228 #endif
229 #else /* !USE_CPPLIB */
230 parse_in.show_column = 1;
231 if (! cpp_start_read (&parse_in, filename))
232 abort ();
234 if (filename == 0 || !strcmp (filename, "-"))
235 filename = "stdin";
237 /* cpp_start_read always puts at least one line directive into the
238 token buffer. We must arrange to read it out here. */
239 yy_cur = parse_in.token_buffer;
240 yy_lim = CPP_PWRITTEN (&parse_in);
241 cpp_token = CPP_DIRECTIVE;
242 #endif
244 add_c_tree_codes ();
246 init_lex ();
247 init_pragma ();
249 return filename;
252 void
253 finish_parse ()
255 #if USE_CPPLIB
256 cpp_finish (&parse_in);
257 errorcount += parse_in.errors;
258 #else
259 fclose (finput);
260 #endif
263 void
264 init_lex ()
266 /* Make identifier nodes long enough for the language-specific slots. */
267 set_identifier_size (sizeof (struct lang_identifier));
269 /* Start it at 0, because check_newline is called at the very beginning
270 and will increment it to 1. */
271 lineno = 0;
273 #ifdef MULTIBYTE_CHARS
274 /* Change to the native locale for multibyte conversions. */
275 setlocale (LC_CTYPE, "");
276 GET_ENVIRONMENT (literal_codeset, "LANG");
277 #endif
279 maxtoken = 40;
280 token_buffer = (char *) xmalloc (maxtoken + 2);
282 ridpointers = (tree *) xcalloc ((int) RID_MAX, sizeof (tree));
283 ridpointers[(int) RID_INT] = get_identifier ("int");
284 ridpointers[(int) RID_CHAR] = get_identifier ("char");
285 ridpointers[(int) RID_VOID] = get_identifier ("void");
286 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
287 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
288 ridpointers[(int) RID_SHORT] = get_identifier ("short");
289 ridpointers[(int) RID_LONG] = get_identifier ("long");
290 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
291 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
292 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
293 ridpointers[(int) RID_CONST] = get_identifier ("const");
294 ridpointers[(int) RID_RESTRICT] = get_identifier ("restrict");
295 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
296 ridpointers[(int) RID_BOUNDED] = get_identifier ("__bounded");
297 ridpointers[(int) RID_UNBOUNDED] = get_identifier ("__unbounded");
298 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
299 ridpointers[(int) RID_STATIC] = get_identifier ("static");
300 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
301 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
302 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
303 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
304 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
305 ridpointers[(int) RID_ID] = get_identifier ("id");
306 ridpointers[(int) RID_IN] = get_identifier ("in");
307 ridpointers[(int) RID_OUT] = get_identifier ("out");
308 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
309 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
310 ridpointers[(int) RID_BYREF] = get_identifier ("byref");
311 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
312 forget_protocol_qualifiers();
314 /* Some options inhibit certain reserved words.
315 Clear those words out of the hash table so they won't be recognized. */
316 #define UNSET_RESERVED_WORD(STRING) \
317 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
318 if (s) s->name = ""; } while (0)
320 if (! doing_objc_thang)
321 UNSET_RESERVED_WORD ("id");
323 if (flag_traditional)
325 UNSET_RESERVED_WORD ("const");
326 UNSET_RESERVED_WORD ("restrict");
327 UNSET_RESERVED_WORD ("volatile");
328 UNSET_RESERVED_WORD ("typeof");
329 UNSET_RESERVED_WORD ("signed");
330 UNSET_RESERVED_WORD ("inline");
331 UNSET_RESERVED_WORD ("iterator");
332 UNSET_RESERVED_WORD ("complex");
334 else if (!flag_isoc99)
335 UNSET_RESERVED_WORD ("restrict");
337 if (flag_no_asm)
339 UNSET_RESERVED_WORD ("asm");
340 UNSET_RESERVED_WORD ("typeof");
341 if (! flag_isoc99)
342 UNSET_RESERVED_WORD ("inline");
343 UNSET_RESERVED_WORD ("iterator");
344 UNSET_RESERVED_WORD ("complex");
348 void
349 reinit_parse_for_function ()
353 /* Function used when yydebug is set, to print a token in more detail. */
355 void
356 yyprint (file, yychar, yylval)
357 FILE *file;
358 int yychar;
359 YYSTYPE yylval;
361 tree t;
362 switch (yychar)
364 case IDENTIFIER:
365 case TYPENAME:
366 case OBJECTNAME:
367 t = yylval.ttype;
368 if (IDENTIFIER_POINTER (t))
369 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
370 break;
372 case CONSTANT:
373 t = yylval.ttype;
374 if (TREE_CODE (t) == INTEGER_CST)
375 fprintf (file,
376 #if HOST_BITS_PER_WIDE_INT == 64
377 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
378 " 0x%x%016x",
379 #else
380 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
381 " 0x%lx%016lx",
382 #else
383 " 0x%llx%016llx",
384 #endif
385 #endif
386 #else
387 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
388 " 0x%lx%08lx",
389 #else
390 " 0x%x%08x",
391 #endif
392 #endif
393 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
394 break;
398 /* Iff C is a carriage return, warn about it - if appropriate -
399 and return nonzero. */
401 static int
402 whitespace_cr (c)
403 int c;
405 static int newline_warning = 0;
407 if (c == '\r')
409 /* ANSI C says the effects of a carriage return in a source file
410 are undefined. */
411 if (pedantic && !newline_warning)
413 warning ("carriage return in source file");
414 warning ("(we only warn about the first carriage return)");
415 newline_warning = 1;
417 return 1;
419 return 0;
422 /* If C is not whitespace, return C.
423 Otherwise skip whitespace and return first nonwhite char read. */
425 static int
426 skip_white_space (c)
427 register int c;
429 for (;;)
431 switch (c)
433 /* We don't recognize comments here, because
434 cpp output can include / and * consecutively as operators.
435 Also, there's no need, since cpp removes all comments. */
437 case '\n':
438 if (linemode)
440 UNGETC (c);
441 return EOF;
443 c = check_newline ();
444 break;
446 case ' ':
447 case '\t':
448 case '\f':
449 case '\v':
450 case '\b':
451 #if USE_CPPLIB
452 /* While processing a # directive we don't get CPP_HSPACE
453 tokens, so we also need to handle whitespace the normal way. */
454 if (cpp_token == CPP_HSPACE)
455 c = yy_get_token ();
456 else
457 #endif
458 c = GETC();
459 break;
461 case '\r':
462 whitespace_cr (c);
463 c = GETC();
464 break;
466 case '\\':
467 c = GETC();
468 if (c == '\n')
469 lineno++;
470 else
471 error ("stray '\\' in program");
472 c = GETC();
473 break;
475 default:
476 return (c);
481 /* Skips all of the white space at the current location in the input file. */
483 void
484 position_after_white_space ()
486 register int c;
488 c = GETC();
490 UNGETC (skip_white_space (c));
493 /* Make the token buffer longer, preserving the data in it.
494 P should point to just beyond the last valid character in the old buffer.
495 The value we return is a pointer to the new buffer
496 at a place corresponding to P. */
498 static void
499 extend_token_buffer_to (size)
500 int size;
503 maxtoken = maxtoken * 2 + 10;
504 while (maxtoken < size);
505 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
508 static char *
509 extend_token_buffer (p)
510 const char *p;
512 int offset = p - token_buffer;
513 extend_token_buffer_to (offset);
514 return token_buffer + offset;
517 #if defined HANDLE_PRAGMA
518 /* Local versions of these macros, that can be passed as function pointers. */
519 static int
520 pragma_getc ()
522 return GETC ();
525 static void
526 pragma_ungetc (arg)
527 int arg;
529 UNGETC (arg);
531 #endif
533 static int
534 read_line_number (num)
535 int *num;
537 register int token = yylex ();
539 if (token == CONSTANT
540 && TREE_CODE (yylval.ttype) == INTEGER_CST)
542 *num = TREE_INT_CST_LOW (yylval.ttype);
543 return 1;
545 else
547 if (token != END_OF_LINE)
548 error ("invalid #-line");
549 return 0;
553 /* At the beginning of a line, increment the line number
554 and process any #-directive on this line.
555 If the line is a #-directive, read the entire line and return a newline.
556 Otherwise, return the line's first non-whitespace character.
558 Note that in the case of USE_CPPLIB, we get the whole line as one
559 CPP_DIRECTIVE token. */
562 check_newline ()
564 register int c;
565 register int token;
566 int saw_line;
567 enum { act_none, act_push, act_pop } action;
568 int old_lineno, action_number, l;
570 restart:
571 /* Read first nonwhite char on the line. */
573 #ifdef USE_CPPLIB
574 c = GETC ();
575 /* In some cases where we're leaving an include file, we can get multiple
576 CPP_HSPACE tokens in a row, so we need to loop. */
577 while (cpp_token == CPP_HSPACE)
578 c = yy_get_token ();
579 #else
581 c = GETC ();
582 while (c == ' ' || c == '\t');
583 #endif
585 lineno++;
587 if (c != '#')
589 /* Sequences of multiple newlines are very common; optimize them. */
590 if (c == '\n')
591 goto restart;
593 /* If not #, return it so caller will use it. */
594 return c;
597 /* Don't read beyond this line. */
598 saw_line = 0;
599 linemode = 1;
601 #if USE_CPPLIB
602 if (cpp_token == CPP_VSPACE)
604 /* Format is "<space> <line number> <filename> <newline>".
605 Only the line number is interesting, and even that
606 we can get more efficiently than scanning the line. */
607 yy_cur = yy_lim - 1;
608 lineno = parse_in.lineno - 1;
609 goto skipline;
611 #endif
613 token = yylex ();
615 if (token == IDENTIFIER)
617 /* If a letter follows, then if the word here is `line', skip
618 it and ignore it; otherwise, ignore the line, with an error
619 if the word isn't `pragma'. */
621 const char *name = IDENTIFIER_POINTER (yylval.ttype);
623 if (!strcmp (name, "pragma"))
625 token = yylex ();
626 if (token != IDENTIFIER
627 || TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
628 goto skipline;
630 #ifdef HANDLE_PRAGMA
631 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS
632 (if both are defined), in order to give the back
633 end a chance to override the interpretation of
634 SYSV style pragmas. */
635 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
636 IDENTIFIER_POINTER (yylval.ttype)))
637 goto skipline;
638 #endif /* HANDLE_PRAGMA */
640 #ifdef HANDLE_GENERIC_PRAGMAS
641 if (handle_generic_pragma (token))
642 goto skipline;
643 #endif /* HANDLE_GENERIC_PRAGMAS */
645 /* Issue a warning message if we have been asked to do so.
646 Ignoring unknown pragmas in system header file unless
647 an explcit -Wunknown-pragmas has been given. */
648 if (warn_unknown_pragmas > 1
649 || (warn_unknown_pragmas && ! in_system_header))
650 warning ("ignoring pragma: %s", token_buffer);
652 goto skipline;
654 else if (!strcmp (name, "define"))
656 debug_define (lineno, GET_DIRECTIVE_LINE ());
657 goto skipline;
659 else if (!strcmp (name, "undef"))
661 debug_undef (lineno, GET_DIRECTIVE_LINE ());
662 goto skipline;
664 else if (!strcmp (name, "line"))
666 saw_line = 1;
667 token = yylex ();
668 goto linenum;
670 else if (!strcmp (name, "ident"))
672 /* #ident. The pedantic warning is now in cpp. */
674 /* Here we have just seen `#ident '.
675 A string constant should follow. */
677 token = yylex ();
678 if (token == END_OF_LINE)
679 goto skipline;
680 if (token != STRING
681 || TREE_CODE (yylval.ttype) != STRING_CST)
683 error ("invalid #ident");
684 goto skipline;
687 if (! flag_no_ident)
689 #ifdef ASM_OUTPUT_IDENT
690 ASM_OUTPUT_IDENT (asm_out_file,
691 TREE_STRING_POINTER (yylval.ttype));
692 #endif
695 /* Skip the rest of this line. */
696 goto skipline;
699 error ("undefined or invalid # directive `%s'", name);
700 goto skipline;
703 /* If the # is the only nonwhite char on the line,
704 just ignore it. Check the new newline. */
705 if (token == END_OF_LINE)
706 goto skipline;
708 linenum:
709 /* Here we have either `#line' or `# <nonletter>'.
710 In either case, it should be a line number; a digit should follow. */
712 if (token != CONSTANT
713 || TREE_CODE (yylval.ttype) != INTEGER_CST)
715 error ("invalid #-line");
716 goto skipline;
719 /* subtract one, because it is the following line that
720 gets the specified number */
722 l = TREE_INT_CST_LOW (yylval.ttype) - 1;
724 /* More follows: it must be a string constant (filename).
725 It would be neat to use cpplib to quickly process the string, but
726 (1) we don't have a handy tokenization of the string, and
727 (2) I don't know how well that would work in the presense
728 of filenames that contain wide characters. */
730 if (saw_line)
732 /* Don't treat \ as special if we are processing #line 1 "...".
733 If you want it to be treated specially, use # 1 "...". */
734 ignore_escape_flag = 1;
737 /* Read the string constant. */
738 token = yylex ();
740 ignore_escape_flag = 0;
742 if (token == END_OF_LINE)
744 /* No more: store the line number and check following line. */
745 lineno = l;
746 goto skipline;
749 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
751 error ("invalid #line");
752 goto skipline;
755 input_filename = TREE_STRING_POINTER (yylval.ttype);
757 if (main_input_filename == 0)
758 main_input_filename = input_filename;
760 old_lineno = lineno;
761 action = act_none;
762 action_number = 0;
763 lineno = l;
765 /* Each change of file name
766 reinitializes whether we are now in a system header. */
767 in_system_header = 0;
769 if (!read_line_number (&action_number))
771 /* Update the name in the top element of input_file_stack. */
772 if (input_file_stack)
773 input_file_stack->name = input_filename;
776 /* `1' after file name means entering new file.
777 `2' after file name means just left a file. */
779 if (action_number == 1)
781 action = act_push;
782 read_line_number (&action_number);
784 else if (action_number == 2)
786 action = act_pop;
787 read_line_number (&action_number);
789 if (action_number == 3)
791 /* `3' after file name means this is a system header file. */
792 in_system_header = 1;
793 read_line_number (&action_number);
796 /* Do the actions implied by the preceding numbers. */
798 if (action == act_push)
800 /* Pushing to a new file. */
801 struct file_stack *p
802 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
803 input_file_stack->line = old_lineno;
804 p->next = input_file_stack;
805 p->name = input_filename;
806 p->indent_level = indent_level;
807 input_file_stack = p;
808 input_file_stack_tick++;
809 debug_start_source_file (input_filename);
811 else if (action == act_pop)
813 /* Popping out of a file. */
814 if (input_file_stack->next)
816 struct file_stack *p = input_file_stack;
817 if (indent_level != p->indent_level)
819 warning_with_file_and_line
820 (p->name, old_lineno,
821 "This file contains more `%c's than `%c's.",
822 indent_level > p->indent_level ? '{' : '}',
823 indent_level > p->indent_level ? '}' : '{');
825 input_file_stack = p->next;
826 free (p);
827 input_file_stack_tick++;
828 debug_end_source_file (input_file_stack->line);
830 else
831 error ("#-lines for entering and leaving files don't match");
834 /* Now that we've pushed or popped the input stack,
835 update the name in the top element. */
836 if (input_file_stack)
837 input_file_stack->name = input_filename;
839 /* skip the rest of this line. */
840 skipline:
841 linemode = 0;
842 end_of_file = 0;
845 c = GETC();
846 while (c != '\n' && c != EOF);
847 return c;
850 #ifdef HANDLE_GENERIC_PRAGMAS
852 /* Handle a #pragma directive.
853 TOKEN is the token we read after `#pragma'. Processes the entire input
854 line and return non-zero iff the pragma has been successfully parsed. */
856 /* This function has to be in this file, in order to get at
857 the token types. */
859 static int
860 handle_generic_pragma (token)
861 register int token;
863 for (;;)
865 switch (token)
867 case IDENTIFIER:
868 case TYPENAME:
869 case STRING:
870 case CONSTANT:
871 handle_pragma_token (token_buffer, yylval.ttype);
872 break;
874 case END_OF_LINE:
875 return handle_pragma_token (NULL_PTR, NULL_TREE);
877 default:
878 handle_pragma_token (token_buffer, NULL);
881 token = yylex ();
885 #endif /* HANDLE_GENERIC_PRAGMAS */
887 #define ENDFILE -1 /* token that represents end-of-file */
889 /* Read an escape sequence, returning its equivalent as a character,
890 or store 1 in *ignore_ptr if it is backslash-newline. */
892 static int
893 readescape (ignore_ptr)
894 int *ignore_ptr;
896 register int c = GETC();
897 register int code;
898 register unsigned count;
899 unsigned firstdig = 0;
900 int nonnull;
902 switch (c)
904 case 'x':
905 if (warn_traditional)
906 warning ("the meaning of `\\x' varies with -traditional");
908 if (flag_traditional)
909 return c;
911 code = 0;
912 count = 0;
913 nonnull = 0;
914 while (1)
916 c = GETC();
917 if (! ISXDIGIT (c))
919 UNGETC (c);
920 break;
922 code *= 16;
923 if (c >= 'a' && c <= 'f')
924 code += c - 'a' + 10;
925 if (c >= 'A' && c <= 'F')
926 code += c - 'A' + 10;
927 if (c >= '0' && c <= '9')
928 code += c - '0';
929 if (code != 0 || count != 0)
931 if (count == 0)
932 firstdig = code;
933 count++;
935 nonnull = 1;
937 if (! nonnull)
939 warning ("\\x used with no following hex digits");
940 return 'x';
942 else if (count == 0)
943 /* Digits are all 0's. Ok. */
945 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
946 || (count > 1
947 && (((unsigned)1
948 << (TYPE_PRECISION (integer_type_node)
949 - (count - 1) * 4))
950 <= firstdig)))
951 pedwarn ("hex escape out of range");
952 return code;
954 case '0': case '1': case '2': case '3': case '4':
955 case '5': case '6': case '7':
956 code = 0;
957 count = 0;
958 while ((c <= '7') && (c >= '0') && (count++ < 3))
960 code = (code * 8) + (c - '0');
961 c = GETC();
963 UNGETC (c);
964 return code;
966 case '\\': case '\'': case '"':
967 return c;
969 case '\n':
970 lineno++;
971 *ignore_ptr = 1;
972 return 0;
974 case 'n':
975 return TARGET_NEWLINE;
977 case 't':
978 return TARGET_TAB;
980 case 'r':
981 return TARGET_CR;
983 case 'f':
984 return TARGET_FF;
986 case 'b':
987 return TARGET_BS;
989 case 'a':
990 if (warn_traditional)
991 warning ("the meaning of `\\a' varies with -traditional");
993 if (flag_traditional)
994 return c;
995 return TARGET_BELL;
997 case 'v':
998 #if 0 /* Vertical tab is present in common usage compilers. */
999 if (flag_traditional)
1000 return c;
1001 #endif
1002 return TARGET_VT;
1004 case 'e':
1005 case 'E':
1006 if (pedantic)
1007 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1008 return TARGET_ESC;
1010 case '?':
1011 return c;
1013 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1014 case '(':
1015 case '{':
1016 case '[':
1017 /* `\%' is used to prevent SCCS from getting confused. */
1018 case '%':
1019 if (pedantic)
1020 pedwarn ("unknown escape sequence `\\%c'", c);
1021 return c;
1023 if (ISGRAPH (c))
1024 pedwarn ("unknown escape sequence `\\%c'", c);
1025 else
1026 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1027 return c;
1030 void
1031 yyerror (msgid)
1032 const char *msgid;
1034 const char *string = _(msgid);
1036 /* We can't print string and character constants well
1037 because the token_buffer contains the result of processing escapes. */
1038 if (end_of_file)
1039 error ("%s at end of input", string);
1040 else if (token_buffer[0] == 0)
1041 error ("%s at null character", string);
1042 else if (token_buffer[0] == '"')
1043 error ("%s before string constant", string);
1044 else if (token_buffer[0] == '\'')
1045 error ("%s before character constant", string);
1046 else if (!ISGRAPH(token_buffer[0]))
1047 error ("%s before character 0%o", string, (unsigned char) token_buffer[0]);
1048 else
1049 error ("%s before `%s'", string, token_buffer);
1052 #if 0
1054 struct try_type
1056 tree *node_var;
1057 char unsigned_flag;
1058 char long_flag;
1059 char long_long_flag;
1062 struct try_type type_sequence[] =
1064 { &integer_type_node, 0, 0, 0},
1065 { &unsigned_type_node, 1, 0, 0},
1066 { &long_integer_type_node, 0, 1, 0},
1067 { &long_unsigned_type_node, 1, 1, 0},
1068 { &long_long_integer_type_node, 0, 1, 1},
1069 { &long_long_unsigned_type_node, 1, 1, 1}
1071 #endif /* 0 */
1073 struct pf_args
1075 /* Input */
1076 int base;
1077 char * p;
1078 /* I/O */
1079 int c;
1080 /* Output */
1081 int imag;
1082 tree type;
1083 int conversion_errno;
1084 REAL_VALUE_TYPE value;
1087 static void
1088 parse_float (data)
1089 PTR data;
1091 struct pf_args * args = (struct pf_args *) data;
1092 int fflag = 0, lflag = 0;
1093 /* Copy token_buffer now, while it has just the number
1094 and not the suffixes; once we add `f' or `i',
1095 REAL_VALUE_ATOF may not work any more. */
1096 char *copy = (char *) alloca (args->p - token_buffer + 1);
1097 bcopy (token_buffer, copy, args->p - token_buffer + 1);
1098 args->imag = 0;
1099 args->conversion_errno = 0;
1100 args->type = double_type_node;
1102 while (1)
1104 int lose = 0;
1106 /* Read the suffixes to choose a data type. */
1107 switch (args->c)
1109 case 'f': case 'F':
1110 if (fflag)
1111 error ("more than one `f' in numeric constant");
1112 fflag = 1;
1113 break;
1115 case 'l': case 'L':
1116 if (lflag)
1117 error ("more than one `l' in numeric constant");
1118 lflag = 1;
1119 break;
1121 case 'i': case 'I':
1122 if (args->imag)
1123 error ("more than one `i' or `j' in numeric constant");
1124 else if (pedantic)
1125 pedwarn ("ANSI C forbids imaginary numeric constants");
1126 args->imag = 1;
1127 break;
1129 default:
1130 lose = 1;
1133 if (lose)
1134 break;
1136 if (args->p >= token_buffer + maxtoken - 3)
1137 args->p = extend_token_buffer (args->p);
1138 *(args->p++) = args->c;
1139 *(args->p) = 0;
1140 args->c = GETC();
1143 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1144 tells the desired precision of the binary result
1145 of decimal-to-binary conversion. */
1147 if (fflag)
1149 if (lflag)
1150 error ("both `f' and `l' in floating constant");
1152 args->type = float_type_node;
1153 errno = 0;
1154 if (args->base == 16)
1155 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1156 else
1157 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1158 args->conversion_errno = errno;
1159 /* A diagnostic is required here by some ANSI C testsuites.
1160 This is not pedwarn, because some people don't want
1161 an error for this. */
1162 if (REAL_VALUE_ISINF (args->value) && pedantic)
1163 warning ("floating point number exceeds range of `float'");
1165 else if (lflag)
1167 args->type = long_double_type_node;
1168 errno = 0;
1169 if (args->base == 16)
1170 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1171 else
1172 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1173 args->conversion_errno = errno;
1174 if (REAL_VALUE_ISINF (args->value) && pedantic)
1175 warning ("floating point number exceeds range of `long double'");
1177 else
1179 errno = 0;
1180 if (flag_single_precision_constant)
1181 args->type = float_type_node;
1182 if (args->base == 16)
1183 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1184 else
1185 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1186 args->conversion_errno = errno;
1187 if (REAL_VALUE_ISINF (args->value) && pedantic)
1188 warning ("floating point number exceeds range of `double'");
1192 /* Get the next character, staying within the current token if possible.
1193 If we're lexing a token, we don't want to look beyond the end of the
1194 token cpplib has prepared for us; otherwise, we end up reading in the
1195 next token, which screws up feed_input. So just return a null
1196 character. */
1198 static inline int token_getch PARAMS ((void));
1200 static inline int
1201 token_getch ()
1203 #if USE_CPPLIB
1204 if (yy_cur == yy_lim)
1205 return '\0';
1206 #endif
1207 return GETC ();
1210 static inline void token_put_back PARAMS ((int));
1212 static inline void
1213 token_put_back (ch)
1214 int ch;
1216 #if USE_CPPLIB
1217 if (ch == '\0')
1218 return;
1219 #endif
1220 UNGETC (ch);
1223 /* Read a single token from the input stream, and assign it lexical
1224 semantics. */
1227 yylex ()
1229 register int c;
1230 register char *p;
1231 register int value;
1232 int wide_flag = 0;
1233 int objc_flag = 0;
1235 c = GETC();
1237 /* Effectively do c = skip_white_space (c)
1238 but do it faster in the usual cases. */
1239 while (1)
1240 switch (c)
1242 case ' ':
1243 case '\t':
1244 case '\f':
1245 case '\v':
1246 case '\b':
1247 #if USE_CPPLIB
1248 if (cpp_token == CPP_HSPACE)
1249 c = yy_get_token ();
1250 else
1251 #endif
1252 c = GETC();
1253 break;
1255 case '\r':
1256 /* Call skip_white_space so we can warn if appropriate. */
1258 case '\n':
1259 case '/':
1260 case '\\':
1261 c = skip_white_space (c);
1262 default:
1263 goto found_nonwhite;
1265 found_nonwhite:
1267 token_buffer[0] = c;
1268 token_buffer[1] = 0;
1270 /* yylloc.first_line = lineno; */
1272 switch (c)
1274 case EOF:
1275 end_of_file = 1;
1276 token_buffer[0] = 0;
1277 if (linemode)
1278 value = END_OF_LINE;
1279 else
1280 value = ENDFILE;
1281 break;
1283 case 'L':
1284 #if USE_CPPLIB
1285 if (cpp_token == CPP_NAME)
1286 goto letter;
1287 #endif
1288 /* Capital L may start a wide-string or wide-character constant. */
1290 register int c = token_getch();
1291 if (c == '\'')
1293 wide_flag = 1;
1294 goto char_constant;
1296 if (c == '"')
1298 wide_flag = 1;
1299 goto string_constant;
1301 token_put_back (c);
1303 goto letter;
1305 case '@':
1306 if (!doing_objc_thang)
1308 value = c;
1309 break;
1311 else
1313 /* '@' may start a constant string object. */
1314 register int c = token_getch ();
1315 if (c == '"')
1317 objc_flag = 1;
1318 goto string_constant;
1320 token_put_back (c);
1321 /* Fall through to treat '@' as the start of an identifier. */
1324 case 'A': case 'B': case 'C': case 'D': case 'E':
1325 case 'F': case 'G': case 'H': case 'I': case 'J':
1326 case 'K': case 'M': case 'N': case 'O':
1327 case 'P': case 'Q': case 'R': case 'S': case 'T':
1328 case 'U': case 'V': case 'W': case 'X': case 'Y':
1329 case 'Z':
1330 case 'a': case 'b': case 'c': case 'd': case 'e':
1331 case 'f': case 'g': case 'h': case 'i': case 'j':
1332 case 'k': case 'l': case 'm': case 'n': case 'o':
1333 case 'p': case 'q': case 'r': case 's': case 't':
1334 case 'u': case 'v': case 'w': case 'x': case 'y':
1335 case 'z':
1336 case '_':
1337 case '$':
1338 letter:
1339 #if USE_CPPLIB
1340 if (cpp_token == CPP_NAME)
1342 /* Note that one character has already been read from
1343 yy_cur into token_buffer. Also, cpplib complains about
1344 $ in identifiers, so we don't have to. */
1346 int len = yy_lim - yy_cur + 1;
1347 if (len >= maxtoken)
1348 extend_token_buffer_to (len + 1);
1349 memcpy (token_buffer + 1, yy_cur, len);
1350 p = token_buffer + len;
1351 yy_cur = yy_lim;
1353 else
1354 #endif
1356 p = token_buffer;
1357 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1359 /* Make sure this char really belongs in an identifier. */
1360 if (c == '$')
1362 if (! dollars_in_ident)
1363 error ("`$' in identifier");
1364 else if (pedantic)
1365 pedwarn ("`$' in identifier");
1368 if (p >= token_buffer + maxtoken)
1369 p = extend_token_buffer (p);
1371 *p++ = c;
1372 c = token_getch();
1375 *p = 0;
1376 token_put_back (c);
1379 value = IDENTIFIER;
1380 yylval.itype = 0;
1382 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1385 register struct resword *ptr;
1387 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1389 if (ptr->rid)
1390 yylval.ttype = ridpointers[(int) ptr->rid];
1391 value = (int) ptr->token;
1393 /* Only return OBJECTNAME if it is a typedef. */
1394 if (doing_objc_thang && value == OBJECTNAME)
1396 tree decl = lookup_name(yylval.ttype);
1398 if (decl == NULL_TREE
1399 || TREE_CODE (decl) != TYPE_DECL)
1400 value = IDENTIFIER;
1403 /* Even if we decided to recognize asm, still perhaps warn. */
1404 if (pedantic
1405 && (value == ASM_KEYWORD || value == TYPEOF
1406 || (ptr->rid == RID_INLINE && ! flag_isoc99))
1407 && token_buffer[0] != '_')
1408 pedwarn ("ANSI does not permit the keyword `%s'",
1409 token_buffer);
1413 /* If we did not find a keyword, look for an identifier
1414 (or a typename). */
1416 if (value == IDENTIFIER)
1418 tree decl;
1420 if (token_buffer[0] == '@')
1421 error("invalid identifier `%s'", token_buffer);
1423 yylval.ttype = get_identifier (token_buffer);
1424 decl = lookup_name (yylval.ttype);
1426 if (decl != 0 && TREE_CODE (decl) == TYPE_DECL)
1427 value = TYPENAME;
1428 /* A user-invisible read-only initialized variable
1429 should be replaced by its value.
1430 We handle only strings since that's the only case used in C. */
1431 else if (decl != 0 && TREE_CODE (decl) == VAR_DECL
1432 && DECL_IGNORED_P (decl)
1433 && TREE_READONLY (decl)
1434 && DECL_INITIAL (decl) != 0
1435 && TREE_CODE (DECL_INITIAL (decl)) == STRING_CST)
1437 tree stringval = DECL_INITIAL (decl);
1439 /* Copy the string value so that we won't clobber anything
1440 if we put something in the TREE_CHAIN of this one. */
1441 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1442 TREE_STRING_POINTER (stringval));
1443 value = STRING;
1445 else if (doing_objc_thang)
1447 tree objc_interface_decl = is_class_name (yylval.ttype);
1449 if (objc_interface_decl)
1451 value = CLASSNAME;
1452 yylval.ttype = objc_interface_decl;
1457 break;
1459 case '.':
1460 #if USE_CPPLIB
1461 if (yy_cur < yy_lim)
1462 #endif
1464 /* It's hard to preserve tokenization on '.' because
1465 it could be a symbol by itself, or it could be the
1466 start of a floating point number and cpp won't tell us. */
1467 register int c1 = token_getch ();
1468 token_buffer[1] = c1;
1469 if (c1 == '.')
1471 c1 = token_getch ();
1472 if (c1 == '.')
1474 token_buffer[2] = c1;
1475 token_buffer[3] = 0;
1476 value = ELLIPSIS;
1477 goto done;
1479 error ("parse error at `..'");
1481 if (ISDIGIT (c1))
1483 token_put_back (c1);
1484 goto number;
1486 token_put_back (c1);
1488 value = '.';
1489 token_buffer[1] = 0;
1490 break;
1492 case '0': case '1':
1493 /* Optimize for most frequent case. */
1495 register int cond;
1497 #if USE_CPPLIB
1498 cond = (yy_cur == yy_lim);
1499 #else
1500 register int c1 = token_getch ();
1501 token_put_back (c1);
1502 cond = (! ISALNUM (c1) && c1 != '.');
1503 #endif
1504 if (cond)
1506 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1507 value = CONSTANT;
1508 break;
1510 /*FALLTHRU*/
1512 case '2': case '3': case '4':
1513 case '5': case '6': case '7': case '8': case '9':
1514 number:
1516 int base = 10;
1517 int count = 0;
1518 int largest_digit = 0;
1519 int numdigits = 0;
1520 int overflow = 0;
1522 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
1523 The code below which fills the parts array assumes that a host
1524 int is at least twice as wide as a host char, and that
1525 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
1526 Two HOST_WIDE_INTs is the largest int literal we can store.
1527 In order to detect overflow below, the number of parts (TOTAL_PARTS)
1528 must be exactly the number of parts needed to hold the bits
1529 of two HOST_WIDE_INTs. */
1530 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
1531 unsigned int parts[TOTAL_PARTS];
1533 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS, AFTER_EXPON}
1534 floatflag = NOT_FLOAT;
1536 for (count = 0; count < TOTAL_PARTS; count++)
1537 parts[count] = 0;
1539 p = token_buffer;
1540 *p++ = c;
1542 if (c == '0')
1544 *p++ = (c = token_getch());
1545 if ((c == 'x') || (c == 'X'))
1547 base = 16;
1548 *p++ = (c = token_getch());
1550 /* Leading 0 forces octal unless the 0 is the only digit. */
1551 else if (c >= '0' && c <= '9')
1553 base = 8;
1554 numdigits++;
1556 else
1557 numdigits++;
1560 /* Read all the digits-and-decimal-points. */
1562 while (c == '.'
1563 || (ISALNUM (c) && c != 'l' && c != 'L'
1564 && c != 'u' && c != 'U'
1565 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1566 && (floatflag == NOT_FLOAT
1567 || ((base != 16) && (c != 'f') && (c != 'F'))
1568 || base == 16)))
1570 if (c == '.')
1572 if (base == 16 && pedantic && !flag_isoc99)
1573 pedwarn ("floating constant may not be in radix 16");
1574 if (floatflag == TOO_MANY_POINTS)
1575 /* We have already emitted an error. Don't need another. */
1577 else if (floatflag == AFTER_POINT || floatflag == AFTER_EXPON)
1579 error ("malformed floating constant");
1580 floatflag = TOO_MANY_POINTS;
1581 /* Avoid another error from atof by forcing all characters
1582 from here on to be ignored. */
1583 p[-1] = '\0';
1585 else
1586 floatflag = AFTER_POINT;
1588 if (base == 8)
1589 base = 10;
1590 *p++ = c = token_getch();
1591 /* Accept '.' as the start of a floating-point number
1592 only when it is followed by a digit. */
1593 if (p == token_buffer + 2 && !ISDIGIT (c))
1594 abort ();
1596 else
1598 /* It is not a decimal point.
1599 It should be a digit (perhaps a hex digit). */
1601 if (ISDIGIT (c))
1603 c = c - '0';
1605 else if (base <= 10)
1607 if (c == 'e' || c == 'E')
1609 base = 10;
1610 floatflag = AFTER_EXPON;
1611 break; /* start of exponent */
1613 error ("nondigits in number and not hexadecimal");
1614 c = 0;
1616 else if (base == 16 && (c == 'p' || c == 'P'))
1618 floatflag = AFTER_EXPON;
1619 break; /* start of exponent */
1621 else if (c >= 'a' && c <= 'f')
1623 c = c - 'a' + 10;
1625 else
1627 c = c - 'A' + 10;
1629 if (c >= largest_digit)
1630 largest_digit = c;
1631 numdigits++;
1633 for (count = 0; count < TOTAL_PARTS; count++)
1635 parts[count] *= base;
1636 if (count)
1638 parts[count]
1639 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1640 parts[count-1]
1641 &= (1 << HOST_BITS_PER_CHAR) - 1;
1643 else
1644 parts[0] += c;
1647 /* If the highest-order part overflows (gets larger than
1648 a host char will hold) then the whole number has
1649 overflowed. Record this and truncate the highest-order
1650 part. */
1651 if (parts[TOTAL_PARTS - 1] >> HOST_BITS_PER_CHAR)
1653 overflow = 1;
1654 parts[TOTAL_PARTS - 1] &= (1 << HOST_BITS_PER_CHAR) - 1;
1657 if (p >= token_buffer + maxtoken - 3)
1658 p = extend_token_buffer (p);
1659 *p++ = (c = token_getch());
1663 /* This can happen on input like `int i = 0x;' */
1664 if (numdigits == 0)
1665 error ("numeric constant with no digits");
1667 if (largest_digit >= base)
1668 error ("numeric constant contains digits beyond the radix");
1670 /* Remove terminating char from the token buffer and delimit the
1671 string. */
1672 *--p = 0;
1674 if (floatflag != NOT_FLOAT)
1676 tree type;
1677 int imag, conversion_errno;
1678 REAL_VALUE_TYPE value;
1679 struct pf_args args;
1681 /* Read explicit exponent if any, and put it in tokenbuf. */
1683 if ((base == 10 && ((c == 'e') || (c == 'E')))
1684 || (base == 16 && (c == 'p' || c == 'P')))
1686 if (p >= token_buffer + maxtoken - 3)
1687 p = extend_token_buffer (p);
1688 *p++ = c;
1689 c = token_getch();
1690 if ((c == '+') || (c == '-'))
1692 *p++ = c;
1693 c = token_getch();
1695 /* Exponent is decimal, even if string is a hex float. */
1696 if (! ISDIGIT (c))
1697 error ("floating constant exponent has no digits");
1698 while (ISDIGIT (c))
1700 if (p >= token_buffer + maxtoken - 3)
1701 p = extend_token_buffer (p);
1702 *p++ = c;
1703 c = token_getch ();
1706 if (base == 16 && floatflag != AFTER_EXPON)
1707 error ("hexadecimal floating constant has no exponent");
1709 *p = 0;
1711 /* Setup input for parse_float() */
1712 args.base = base;
1713 args.p = p;
1714 args.c = c;
1716 /* Convert string to a double, checking for overflow. */
1717 if (do_float_handler (parse_float, (PTR) &args))
1719 /* Receive output from parse_float() */
1720 value = args.value;
1722 else
1724 /* We got an exception from parse_float() */
1725 error ("floating constant out of range");
1726 value = dconst0;
1729 /* Receive output from parse_float() */
1730 c = args.c;
1731 imag = args.imag;
1732 type = args.type;
1733 conversion_errno = args.conversion_errno;
1735 #ifdef ERANGE
1736 /* ERANGE is also reported for underflow,
1737 so test the value to distinguish overflow from that. */
1738 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1739 && (REAL_VALUES_LESS (dconst1, value)
1740 || REAL_VALUES_LESS (value, dconstm1)))
1741 warning ("floating point number exceeds range of `double'");
1742 #endif
1744 /* If the result is not a number, assume it must have been
1745 due to some error message above, so silently convert
1746 it to a zero. */
1747 if (REAL_VALUE_ISNAN (value))
1748 value = dconst0;
1750 /* Create a node with determined type and value. */
1751 if (imag)
1752 yylval.ttype = build_complex (NULL_TREE,
1753 convert (type, integer_zero_node),
1754 build_real (type, value));
1755 else
1756 yylval.ttype = build_real (type, value);
1758 else
1760 tree traditional_type, ansi_type, type;
1761 HOST_WIDE_INT high, low;
1762 int spec_unsigned = 0;
1763 int spec_long = 0;
1764 int spec_long_long = 0;
1765 int spec_imag = 0;
1766 int warn = 0, i;
1768 traditional_type = ansi_type = type = NULL_TREE;
1769 while (1)
1771 if (c == 'u' || c == 'U')
1773 if (spec_unsigned)
1774 error ("two `u's in integer constant");
1775 spec_unsigned = 1;
1777 else if (c == 'l' || c == 'L')
1779 if (spec_long)
1781 if (spec_long_long)
1782 error ("three `l's in integer constant");
1783 else if (pedantic && ! flag_isoc99
1784 && ! in_system_header && warn_long_long)
1785 pedwarn ("ANSI C forbids long long integer constants");
1786 spec_long_long = 1;
1788 spec_long = 1;
1790 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1792 if (spec_imag)
1793 error ("more than one `i' or `j' in numeric constant");
1794 else if (pedantic)
1795 pedwarn ("ANSI C forbids imaginary numeric constants");
1796 spec_imag = 1;
1798 else
1799 break;
1800 if (p >= token_buffer + maxtoken - 3)
1801 p = extend_token_buffer (p);
1802 *p++ = c;
1803 c = token_getch();
1806 /* If the literal overflowed, pedwarn about it now. */
1807 if (overflow)
1809 warn = 1;
1810 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT * 2);
1813 /* This is simplified by the fact that our constant
1814 is always positive. */
1816 high = low = 0;
1818 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1820 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1821 / HOST_BITS_PER_CHAR)]
1822 << (i * HOST_BITS_PER_CHAR));
1823 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1826 yylval.ttype = build_int_2 (low, high);
1827 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1829 /* If warn_traditional, calculate both the ANSI type and the
1830 traditional type, then see if they disagree.
1831 Otherwise, calculate only the type for the dialect in use. */
1832 if (warn_traditional || flag_traditional)
1834 /* Calculate the traditional type. */
1835 /* Traditionally, any constant is signed;
1836 but if unsigned is specified explicitly, obey that.
1837 Use the smallest size with the right number of bits,
1838 except for one special case with decimal constants. */
1839 if (! spec_long && base != 10
1840 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1841 traditional_type = (spec_unsigned ? unsigned_type_node
1842 : integer_type_node);
1843 /* A decimal constant must be long
1844 if it does not fit in type int.
1845 I think this is independent of whether
1846 the constant is signed. */
1847 else if (! spec_long && base == 10
1848 && int_fits_type_p (yylval.ttype, integer_type_node))
1849 traditional_type = (spec_unsigned ? unsigned_type_node
1850 : integer_type_node);
1851 else if (! spec_long_long)
1852 traditional_type = (spec_unsigned ? long_unsigned_type_node
1853 : long_integer_type_node);
1854 else if (int_fits_type_p (yylval.ttype,
1855 spec_unsigned
1856 ? long_long_unsigned_type_node
1857 : long_long_integer_type_node))
1858 traditional_type = (spec_unsigned
1859 ? long_long_unsigned_type_node
1860 : long_long_integer_type_node);
1861 else
1862 traditional_type = (spec_unsigned
1863 ? widest_unsigned_literal_type_node
1864 : widest_integer_literal_type_node);
1866 if (warn_traditional || ! flag_traditional)
1868 /* Calculate the ANSI type. */
1869 if (! spec_long && ! spec_unsigned
1870 && int_fits_type_p (yylval.ttype, integer_type_node))
1871 ansi_type = integer_type_node;
1872 else if (! spec_long && (base != 10 || spec_unsigned)
1873 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1874 ansi_type = unsigned_type_node;
1875 else if (! spec_unsigned && !spec_long_long
1876 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1877 ansi_type = long_integer_type_node;
1878 else if (! spec_long_long
1879 && int_fits_type_p (yylval.ttype,
1880 long_unsigned_type_node))
1881 ansi_type = long_unsigned_type_node;
1882 else if (! spec_unsigned
1883 && int_fits_type_p (yylval.ttype,
1884 long_long_integer_type_node))
1885 ansi_type = long_long_integer_type_node;
1886 else if (int_fits_type_p (yylval.ttype,
1887 long_long_unsigned_type_node))
1888 ansi_type = long_long_unsigned_type_node;
1889 else if (! spec_unsigned
1890 && int_fits_type_p (yylval.ttype,
1891 widest_integer_literal_type_node))
1892 ansi_type = widest_integer_literal_type_node;
1893 else
1894 ansi_type = widest_unsigned_literal_type_node;
1897 type = flag_traditional ? traditional_type : ansi_type;
1899 /* We assume that constants specified in a non-decimal
1900 base are bit patterns, and that the programmer really
1901 meant what they wrote. */
1902 if (warn_traditional && base == 10
1903 && traditional_type != ansi_type)
1905 if (TYPE_PRECISION (traditional_type)
1906 != TYPE_PRECISION (ansi_type))
1907 warning ("width of integer constant changes with -traditional");
1908 else if (TREE_UNSIGNED (traditional_type)
1909 != TREE_UNSIGNED (ansi_type))
1910 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1911 else
1912 warning ("width of integer constant may change on other systems with -traditional");
1915 if (pedantic && !flag_traditional && !spec_long_long && !warn
1916 && (TYPE_PRECISION (long_integer_type_node)
1917 < TYPE_PRECISION (type)))
1919 warn = 1;
1920 pedwarn ("integer constant larger than the maximum value of an unsigned long int");
1923 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1924 warning ("decimal constant is so large that it is unsigned");
1926 if (spec_imag)
1928 if (TYPE_PRECISION (type)
1929 <= TYPE_PRECISION (integer_type_node))
1930 yylval.ttype
1931 = build_complex (NULL_TREE, integer_zero_node,
1932 convert (integer_type_node,
1933 yylval.ttype));
1934 else
1935 error ("complex integer constant is too wide for `complex int'");
1937 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1938 /* The traditional constant 0x80000000 is signed
1939 but doesn't fit in the range of int.
1940 This will change it to -0x80000000, which does fit. */
1942 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1943 yylval.ttype = convert (type, yylval.ttype);
1944 TREE_OVERFLOW (yylval.ttype)
1945 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1947 else
1948 TREE_TYPE (yylval.ttype) = type;
1951 /* If it's still an integer (not a complex), and it doesn't
1952 fit in the type we choose for it, then pedwarn. */
1954 if (! warn
1955 && TREE_CODE (TREE_TYPE (yylval.ttype)) == INTEGER_TYPE
1956 && ! int_fits_type_p (yylval.ttype, TREE_TYPE (yylval.ttype)))
1957 pedwarn ("integer constant is larger than the maximum value for its type");
1960 token_put_back (c);
1961 *p = 0;
1963 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1964 || (!flag_traditional && (c == '-' || c == '+')
1965 && (p[-1] == 'e' || p[-1] == 'E')))
1966 error ("missing white space after number `%s'", token_buffer);
1968 value = CONSTANT; break;
1971 case '\'':
1972 char_constant:
1974 register int result = 0;
1975 register int num_chars = 0;
1976 int chars_seen = 0;
1977 unsigned width = TYPE_PRECISION (char_type_node);
1978 int max_chars;
1979 #ifdef MULTIBYTE_CHARS
1980 int longest_char = local_mb_cur_max ();
1981 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1982 #endif
1984 max_chars = TYPE_PRECISION (integer_type_node) / width;
1985 if (wide_flag)
1986 width = WCHAR_TYPE_SIZE;
1988 while (1)
1990 tryagain:
1991 c = token_getch();
1993 if (c == '\'' || c == EOF)
1994 break;
1996 ++chars_seen;
1997 if (c == '\\')
1999 int ignore = 0;
2000 c = readescape (&ignore);
2001 if (ignore)
2002 goto tryagain;
2003 if (width < HOST_BITS_PER_INT
2004 && (unsigned) c >= ((unsigned)1 << width))
2005 pedwarn ("escape sequence out of range for character");
2006 #ifdef MAP_CHARACTER
2007 if (ISPRINT (c))
2008 c = MAP_CHARACTER (c);
2009 #endif
2011 else if (c == '\n')
2013 if (pedantic)
2014 pedwarn ("ANSI C forbids newline in character constant");
2015 lineno++;
2017 else
2019 #ifdef MULTIBYTE_CHARS
2020 wchar_t wc;
2021 int i;
2022 int char_len = -1;
2023 for (i = 1; i <= longest_char; ++i)
2025 if (i > maxtoken - 4)
2026 extend_token_buffer (token_buffer);
2028 token_buffer[i] = c;
2029 char_len = local_mbtowc (& wc,
2030 token_buffer + 1,
2032 if (char_len != -1)
2033 break;
2034 c = token_getch ();
2036 if (char_len > 1)
2038 /* mbtowc sometimes needs an extra char before accepting */
2039 if (char_len < i)
2040 token_put_back (c);
2041 if (! wide_flag)
2043 /* Merge character into result; ignore excess chars. */
2044 for (i = 1; i <= char_len; ++i)
2046 if (i > max_chars)
2047 break;
2048 if (width < HOST_BITS_PER_INT)
2049 result = (result << width)
2050 | (token_buffer[i]
2051 & ((1 << width) - 1));
2052 else
2053 result = token_buffer[i];
2055 num_chars += char_len;
2056 goto tryagain;
2058 c = wc;
2060 else
2062 if (char_len == -1)
2064 warning ("Ignoring invalid multibyte character");
2065 /* Replace all but the first byte. */
2066 for (--i; i > 1; --i)
2067 token_put_back (token_buffer[i]);
2068 wc = token_buffer[1];
2070 #ifdef MAP_CHARACTER
2071 c = MAP_CHARACTER (wc);
2072 #else
2073 c = wc;
2074 #endif
2076 #else /* ! MULTIBYTE_CHARS */
2077 #ifdef MAP_CHARACTER
2078 c = MAP_CHARACTER (c);
2079 #endif
2080 #endif /* ! MULTIBYTE_CHARS */
2083 if (wide_flag)
2085 if (chars_seen == 1) /* only keep the first one */
2086 result = c;
2087 goto tryagain;
2090 /* Merge character into result; ignore excess chars. */
2091 num_chars += (width / TYPE_PRECISION (char_type_node));
2092 if (num_chars < max_chars + 1)
2094 if (width < HOST_BITS_PER_INT)
2095 result = (result << width) | (c & ((1 << width) - 1));
2096 else
2097 result = c;
2101 if (c != '\'')
2102 error ("malformed character constant");
2103 else if (chars_seen == 0)
2104 error ("empty character constant");
2105 else if (num_chars > max_chars)
2107 num_chars = max_chars;
2108 error ("character constant too long");
2110 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
2111 warning ("multi-character character constant");
2113 /* If char type is signed, sign-extend the constant. */
2114 if (! wide_flag)
2116 int num_bits = num_chars * width;
2117 if (num_bits == 0)
2118 /* We already got an error; avoid invalid shift. */
2119 yylval.ttype = build_int_2 (0, 0);
2120 else if (TREE_UNSIGNED (char_type_node)
2121 || ((result >> (num_bits - 1)) & 1) == 0)
2122 yylval.ttype
2123 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
2124 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2126 else
2127 yylval.ttype
2128 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
2129 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2130 -1);
2131 TREE_TYPE (yylval.ttype) = integer_type_node;
2133 else
2135 yylval.ttype = build_int_2 (result, 0);
2136 TREE_TYPE (yylval.ttype) = wchar_type_node;
2139 value = CONSTANT;
2140 break;
2143 case '"':
2144 string_constant:
2146 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2147 : TYPE_PRECISION (char_type_node);
2148 #ifdef MULTIBYTE_CHARS
2149 int longest_char = local_mb_cur_max ();
2150 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2151 #endif
2152 c = token_getch ();
2153 p = token_buffer + 1;
2155 while (c != '"' && c != EOF)
2157 /* ignore_escape_flag is set for reading the filename in #line. */
2158 if (!ignore_escape_flag && c == '\\')
2160 int ignore = 0;
2161 c = readescape (&ignore);
2162 if (ignore)
2163 goto skipnewline;
2164 if (width < HOST_BITS_PER_INT
2165 && (unsigned) c >= ((unsigned)1 << width))
2166 pedwarn ("escape sequence out of range for character");
2168 else if (c == '\n')
2170 if (pedantic)
2171 pedwarn ("ANSI C forbids newline in string constant");
2172 lineno++;
2174 else
2176 #ifdef MULTIBYTE_CHARS
2177 wchar_t wc;
2178 int i;
2179 int char_len = -1;
2180 for (i = 0; i < longest_char; ++i)
2182 if (p + i >= token_buffer + maxtoken)
2183 p = extend_token_buffer (p);
2184 p[i] = c;
2186 char_len = local_mbtowc (& wc, p, i + 1);
2187 if (char_len != -1)
2188 break;
2189 c = token_getch ();
2191 if (char_len == -1)
2193 warning ("Ignoring invalid multibyte character");
2194 /* Replace all except the first byte. */
2195 token_put_back (c);
2196 for (--i; i > 0; --i)
2197 token_put_back (p[i]);
2198 char_len = 1;
2200 /* mbtowc sometimes needs an extra char before accepting */
2201 if (char_len <= i)
2202 token_put_back (c);
2203 if (! wide_flag)
2205 p += (i + 1);
2206 c = token_getch ();
2207 continue;
2209 c = wc;
2210 #endif /* MULTIBYTE_CHARS */
2213 /* Add this single character into the buffer either as a wchar_t
2214 or as a single byte. */
2215 if (wide_flag)
2217 unsigned width = TYPE_PRECISION (char_type_node);
2218 unsigned bytemask = (1 << width) - 1;
2219 int byte;
2221 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2222 p = extend_token_buffer (p);
2224 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2226 int value;
2227 if (byte >= (int) sizeof (c))
2228 value = 0;
2229 else
2230 value = (c >> (byte * width)) & bytemask;
2231 if (BYTES_BIG_ENDIAN)
2232 p[WCHAR_BYTES - byte - 1] = value;
2233 else
2234 p[byte] = value;
2236 p += WCHAR_BYTES;
2238 else
2240 if (p >= token_buffer + maxtoken)
2241 p = extend_token_buffer (p);
2242 *p++ = c;
2245 skipnewline:
2246 c = token_getch ();
2249 /* Terminate the string value, either with a single byte zero
2250 or with a wide zero. */
2251 if (wide_flag)
2253 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2254 p = extend_token_buffer (p);
2255 bzero (p, WCHAR_BYTES);
2256 p += WCHAR_BYTES;
2258 else
2260 if (p >= token_buffer + maxtoken)
2261 p = extend_token_buffer (p);
2262 *p++ = 0;
2265 if (c == EOF)
2266 error ("Unterminated string constant");
2268 /* We have read the entire constant.
2269 Construct a STRING_CST for the result. */
2271 yylval.ttype = build_string (p - (token_buffer + 1), token_buffer + 1);
2272 if (wide_flag)
2274 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2275 value = STRING;
2277 else if (objc_flag)
2279 TREE_TYPE (yylval.ttype) = char_array_type_node;
2280 value = OBJC_STRING;
2282 else
2284 TREE_TYPE (yylval.ttype) = char_array_type_node;
2285 value = STRING;
2288 break;
2291 case '+':
2292 case '-':
2293 case '&':
2294 case '|':
2295 case ':':
2296 case '<':
2297 case '>':
2298 case '*':
2299 case '/':
2300 case '%':
2301 case '^':
2302 case '!':
2303 case '=':
2305 register int c1;
2307 combine:
2309 switch (c)
2311 case '+':
2312 yylval.code = PLUS_EXPR; break;
2313 case '-':
2314 yylval.code = MINUS_EXPR; break;
2315 case '&':
2316 yylval.code = BIT_AND_EXPR; break;
2317 case '|':
2318 yylval.code = BIT_IOR_EXPR; break;
2319 case '*':
2320 yylval.code = MULT_EXPR; break;
2321 case '/':
2322 yylval.code = TRUNC_DIV_EXPR; break;
2323 case '%':
2324 yylval.code = TRUNC_MOD_EXPR; break;
2325 case '^':
2326 yylval.code = BIT_XOR_EXPR; break;
2327 case LSHIFT:
2328 yylval.code = LSHIFT_EXPR; break;
2329 case RSHIFT:
2330 yylval.code = RSHIFT_EXPR; break;
2331 case '<':
2332 yylval.code = LT_EXPR; break;
2333 case '>':
2334 yylval.code = GT_EXPR; break;
2337 token_buffer[1] = c1 = token_getch();
2338 token_buffer[2] = 0;
2340 if (c1 == '=')
2342 switch (c)
2344 case '<':
2345 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2346 case '>':
2347 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2348 case '!':
2349 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2350 case '=':
2351 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2353 value = ASSIGN; goto done;
2355 else if (c == c1)
2356 switch (c)
2358 case '+':
2359 value = PLUSPLUS; goto done;
2360 case '-':
2361 value = MINUSMINUS; goto done;
2362 case '&':
2363 value = ANDAND; goto done;
2364 case '|':
2365 value = OROR; goto done;
2366 case '<':
2367 c = LSHIFT;
2368 goto combine;
2369 case '>':
2370 c = RSHIFT;
2371 goto combine;
2373 else
2374 switch (c)
2376 case '-':
2377 if (c1 == '>')
2378 { value = POINTSAT; goto done; }
2379 break;
2381 /* digraphs */
2382 case ':':
2383 if (c1 == '>' && flag_digraphs)
2384 { value = ']'; goto done; }
2385 break;
2386 case '<':
2387 if (flag_digraphs)
2389 if (c1 == '%')
2390 { value = '{'; indent_level++; goto done; }
2391 if (c1 == ':')
2392 { value = '['; goto done; }
2394 break;
2395 case '%':
2396 if (c1 == '>' && flag_digraphs)
2397 { value = '}'; indent_level--; goto done; }
2398 break;
2401 token_put_back (c1);
2402 token_buffer[1] = 0;
2404 if ((c == '<') || (c == '>'))
2405 value = ARITHCOMPARE;
2406 else value = c;
2407 break;
2410 case 0:
2411 /* Don't make yyparse think this is eof. */
2412 value = 1;
2413 break;
2415 case '{':
2416 indent_level++;
2417 value = c;
2418 break;
2420 case '}':
2421 indent_level--;
2422 value = c;
2423 break;
2425 default:
2426 value = c;
2429 done:
2430 /* yylloc.last_line = lineno; */
2432 return value;
2435 /* Sets the value of the 'yydebug' variable to VALUE.
2436 This is a function so we don't have to have YYDEBUG defined
2437 in order to build the compiler. */
2439 void
2440 set_yydebug (value)
2441 int value;
2443 #if YYDEBUG != 0
2444 yydebug = value;
2445 #else
2446 warning ("YYDEBUG not defined.");
2447 #endif