remove extraneous code checked in with previous delta
[official-gcc.git] / gcc / c-lex.c
blob2f8a8403d69d22c3497243bf24b3e392fe66e10a
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 /* The elements of `ridpointers' are identifier nodes
66 for the reserved type names and storage classes.
67 It is indexed by a RID_... value. */
68 tree ridpointers[(int) RID_MAX];
70 /* Cause the `yydebug' variable to be defined. */
71 #define YYDEBUG 1
73 #if USE_CPPLIB
74 extern unsigned char *yy_cur, *yy_lim;
75 extern enum cpp_token cpp_token;
77 extern int yy_get_token ();
79 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
80 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
82 #else /* ! USE_CPPLIB */
84 #define GETC() getch ()
85 #define UNGETC(c) put_back (c)
87 struct putback_buffer {
88 unsigned char *buffer;
89 int buffer_size;
90 int index;
93 static struct putback_buffer putback = {NULL, 0, -1};
95 static inline int getch PARAMS ((void));
97 static inline int
98 getch ()
100 if (putback.index != -1)
102 int ch = putback.buffer[putback.index];
103 --putback.index;
104 return ch;
106 return getc (finput);
109 static inline void put_back PARAMS ((int));
111 static inline void
112 put_back (ch)
113 int ch;
115 if (ch != EOF)
117 if (putback.index == putback.buffer_size - 1)
119 putback.buffer_size += 16;
120 putback.buffer = xrealloc (putback.buffer, putback.buffer_size);
122 putback.buffer[++putback.index] = ch;
125 #endif /* ! USE_CPPLIB */
127 int linemode;
129 /* the declaration found for the last IDENTIFIER token read in.
130 yylex must look this up to detect typedefs, which get token type TYPENAME,
131 so it is left around in case the identifier is not a typedef but is
132 used in a context which makes it a reference to a variable. */
133 tree lastiddecl;
135 /* Nonzero enables objc features. */
137 int doing_objc_thang;
139 extern int yydebug;
141 /* File used for outputting assembler code. */
142 extern FILE *asm_out_file;
144 #undef WCHAR_TYPE_SIZE
145 #define WCHAR_TYPE_SIZE TYPE_PRECISION (wchar_type_node)
147 /* Number of bytes in a wide character. */
148 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
150 static int maxtoken; /* Current nominal length of token buffer. */
151 char *token_buffer; /* Pointer to token buffer.
152 Actual allocated length is maxtoken + 2.
153 This is not static because objc-parse.y uses it. */
155 static int indent_level; /* Number of { minus number of }. */
157 /* Nonzero tells yylex to ignore \ in string constants. */
158 static int ignore_escape_flag;
160 /* Nonzero if end-of-file has been seen on input. */
161 static int end_of_file;
163 #ifdef HANDLE_GENERIC_PRAGMAS
164 static int handle_generic_pragma PARAMS ((int));
165 #endif /* HANDLE_GENERIC_PRAGMAS */
166 static int whitespace_cr PARAMS ((int));
167 static int skip_white_space PARAMS ((int));
168 static char *extend_token_buffer PARAMS ((const char *));
169 static int readescape PARAMS ((int *));
170 static void parse_float PARAMS ((PTR));
171 static void extend_token_buffer_to PARAMS ((int));
172 static int read_line_number PARAMS ((int *));
174 /* Do not insert generated code into the source, instead, include it.
175 This allows us to build gcc automatically even for targets that
176 need to add or modify the reserved keyword lists. */
177 #include "c-gperf.h"
179 /* Return something to represent absolute declarators containing a *.
180 TARGET is the absolute declarator that the * contains.
181 TYPE_QUALS is a list of modifiers such as const or volatile
182 to apply to the pointer type, represented as identifiers.
184 We return an INDIRECT_REF whose "contents" are TARGET
185 and whose type is the modifier list. */
187 tree
188 make_pointer_declarator (type_quals, target)
189 tree type_quals, target;
191 return build1 (INDIRECT_REF, type_quals, target);
194 void
195 forget_protocol_qualifiers ()
197 int i, n = sizeof wordlist / sizeof (struct resword);
199 for (i = 0; i < n; i++)
200 if ((int) wordlist[i].rid >= (int) RID_IN
201 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
202 wordlist[i].name = "";
205 void
206 remember_protocol_qualifiers ()
208 int i, n = sizeof wordlist / sizeof (struct resword);
210 for (i = 0; i < n; i++)
211 if (wordlist[i].rid == RID_IN)
212 wordlist[i].name = "in";
213 else if (wordlist[i].rid == RID_OUT)
214 wordlist[i].name = "out";
215 else if (wordlist[i].rid == RID_INOUT)
216 wordlist[i].name = "inout";
217 else if (wordlist[i].rid == RID_BYCOPY)
218 wordlist[i].name = "bycopy";
219 else if (wordlist[i].rid == RID_BYREF)
220 wordlist[i].name = "byref";
221 else if (wordlist[i].rid == RID_ONEWAY)
222 wordlist[i].name = "oneway";
225 char *
226 init_parse (filename)
227 char *filename;
229 #if !USE_CPPLIB
230 /* Open input file. */
231 if (filename == 0 || !strcmp (filename, "-"))
233 finput = stdin;
234 filename = "stdin";
236 else
237 finput = fopen (filename, "r");
238 if (finput == 0)
239 pfatal_with_name (filename);
241 #ifdef IO_BUFFER_SIZE
242 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
243 #endif
244 #else /* !USE_CPPLIB */
245 parse_in.show_column = 1;
246 if (! cpp_start_read (&parse_in, filename))
247 abort ();
249 if (filename == 0 || !strcmp (filename, "-"))
250 filename = "stdin";
252 /* cpp_start_read always puts at least one line directive into the
253 token buffer. We must arrange to read it out here. */
254 yy_cur = parse_in.token_buffer;
255 yy_lim = CPP_PWRITTEN (&parse_in);
256 cpp_token = CPP_DIRECTIVE;
257 #endif
259 init_lex ();
260 init_pragma ();
262 return filename;
265 void
266 finish_parse ()
268 #if USE_CPPLIB
269 cpp_finish (&parse_in);
270 errorcount += parse_in.errors;
271 #else
272 fclose (finput);
273 #endif
276 void
277 init_lex ()
279 /* Make identifier nodes long enough for the language-specific slots. */
280 set_identifier_size (sizeof (struct lang_identifier));
282 /* Start it at 0, because check_newline is called at the very beginning
283 and will increment it to 1. */
284 lineno = 0;
286 #ifdef MULTIBYTE_CHARS
287 /* Change to the native locale for multibyte conversions. */
288 setlocale (LC_CTYPE, "");
289 GET_ENVIRONMENT (literal_codeset, "LANG");
290 #endif
292 maxtoken = 40;
293 token_buffer = (char *) xmalloc (maxtoken + 2);
295 ridpointers[(int) RID_INT] = get_identifier ("int");
296 ridpointers[(int) RID_CHAR] = get_identifier ("char");
297 ridpointers[(int) RID_VOID] = get_identifier ("void");
298 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
299 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
300 ridpointers[(int) RID_SHORT] = get_identifier ("short");
301 ridpointers[(int) RID_LONG] = get_identifier ("long");
302 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
303 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
304 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
305 ridpointers[(int) RID_CONST] = get_identifier ("const");
306 ridpointers[(int) RID_RESTRICT] = get_identifier ("restrict");
307 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
308 ridpointers[(int) RID_BOUNDED] = get_identifier ("__bounded");
309 ridpointers[(int) RID_UNBOUNDED] = get_identifier ("__unbounded");
310 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
311 ridpointers[(int) RID_STATIC] = get_identifier ("static");
312 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
313 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
314 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
315 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
316 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
317 ridpointers[(int) RID_ID] = get_identifier ("id");
318 ridpointers[(int) RID_IN] = get_identifier ("in");
319 ridpointers[(int) RID_OUT] = get_identifier ("out");
320 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
321 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
322 ridpointers[(int) RID_BYREF] = get_identifier ("byref");
323 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
324 forget_protocol_qualifiers();
326 /* Some options inhibit certain reserved words.
327 Clear those words out of the hash table so they won't be recognized. */
328 #define UNSET_RESERVED_WORD(STRING) \
329 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
330 if (s) s->name = ""; } while (0)
332 if (! doing_objc_thang)
333 UNSET_RESERVED_WORD ("id");
335 if (flag_traditional)
337 UNSET_RESERVED_WORD ("const");
338 UNSET_RESERVED_WORD ("restrict");
339 UNSET_RESERVED_WORD ("volatile");
340 UNSET_RESERVED_WORD ("typeof");
341 UNSET_RESERVED_WORD ("signed");
342 UNSET_RESERVED_WORD ("inline");
343 UNSET_RESERVED_WORD ("iterator");
344 UNSET_RESERVED_WORD ("complex");
346 else if (!flag_isoc99)
347 UNSET_RESERVED_WORD ("restrict");
349 if (flag_no_asm)
351 UNSET_RESERVED_WORD ("asm");
352 UNSET_RESERVED_WORD ("typeof");
353 UNSET_RESERVED_WORD ("inline");
354 UNSET_RESERVED_WORD ("iterator");
355 UNSET_RESERVED_WORD ("complex");
359 void
360 reinit_parse_for_function ()
364 /* Function used when yydebug is set, to print a token in more detail. */
366 void
367 yyprint (file, yychar, yylval)
368 FILE *file;
369 int yychar;
370 YYSTYPE yylval;
372 tree t;
373 switch (yychar)
375 case IDENTIFIER:
376 case TYPENAME:
377 case OBJECTNAME:
378 t = yylval.ttype;
379 if (IDENTIFIER_POINTER (t))
380 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
381 break;
383 case CONSTANT:
384 t = yylval.ttype;
385 if (TREE_CODE (t) == INTEGER_CST)
386 fprintf (file,
387 #if HOST_BITS_PER_WIDE_INT == 64
388 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
389 " 0x%x%016x",
390 #else
391 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
392 " 0x%lx%016lx",
393 #else
394 " 0x%llx%016llx",
395 #endif
396 #endif
397 #else
398 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
399 " 0x%lx%08lx",
400 #else
401 " 0x%x%08x",
402 #endif
403 #endif
404 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
405 break;
409 /* Iff C is a carriage return, warn about it - if appropriate -
410 and return nonzero. */
412 static int
413 whitespace_cr (c)
414 int c;
416 static int newline_warning = 0;
418 if (c == '\r')
420 /* ANSI C says the effects of a carriage return in a source file
421 are undefined. */
422 if (pedantic && !newline_warning)
424 warning ("carriage return in source file");
425 warning ("(we only warn about the first carriage return)");
426 newline_warning = 1;
428 return 1;
430 return 0;
433 /* If C is not whitespace, return C.
434 Otherwise skip whitespace and return first nonwhite char read. */
436 static int
437 skip_white_space (c)
438 register int c;
440 for (;;)
442 switch (c)
444 /* We don't recognize comments here, because
445 cpp output can include / and * consecutively as operators.
446 Also, there's no need, since cpp removes all comments. */
448 case '\n':
449 if (linemode)
451 UNGETC (c);
452 return EOF;
454 c = check_newline ();
455 break;
457 case ' ':
458 case '\t':
459 case '\f':
460 case '\v':
461 case '\b':
462 #if USE_CPPLIB
463 /* While processing a # directive we don't get CPP_HSPACE
464 tokens, so we also need to handle whitespace the normal way. */
465 if (cpp_token == CPP_HSPACE)
466 c = yy_get_token ();
467 else
468 #endif
469 c = GETC();
470 break;
472 case '\r':
473 whitespace_cr (c);
474 c = GETC();
475 break;
477 case '\\':
478 c = GETC();
479 if (c == '\n')
480 lineno++;
481 else
482 error ("stray '\\' in program");
483 c = GETC();
484 break;
486 default:
487 return (c);
492 /* Skips all of the white space at the current location in the input file. */
494 void
495 position_after_white_space ()
497 register int c;
499 c = GETC();
501 UNGETC (skip_white_space (c));
504 /* Make the token buffer longer, preserving the data in it.
505 P should point to just beyond the last valid character in the old buffer.
506 The value we return is a pointer to the new buffer
507 at a place corresponding to P. */
509 static void
510 extend_token_buffer_to (size)
511 int size;
514 maxtoken = maxtoken * 2 + 10;
515 while (maxtoken < size);
516 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
519 static char *
520 extend_token_buffer (p)
521 const char *p;
523 int offset = p - token_buffer;
524 extend_token_buffer_to (offset);
525 return token_buffer + offset;
528 #if defined HANDLE_PRAGMA
529 /* Local versions of these macros, that can be passed as function pointers. */
530 static int
531 pragma_getc ()
533 return GETC ();
536 static void
537 pragma_ungetc (arg)
538 int arg;
540 UNGETC (arg);
542 #endif
544 static int
545 read_line_number (num)
546 int *num;
548 register int token = yylex ();
550 if (token == CONSTANT
551 && TREE_CODE (yylval.ttype) == INTEGER_CST)
553 *num = TREE_INT_CST_LOW (yylval.ttype);
554 return 1;
556 else
558 if (token != END_OF_LINE)
559 error ("invalid #-line");
560 return 0;
564 /* At the beginning of a line, increment the line number
565 and process any #-directive on this line.
566 If the line is a #-directive, read the entire line and return a newline.
567 Otherwise, return the line's first non-whitespace character.
569 Note that in the case of USE_CPPLIB, we get the whole line as one
570 CPP_DIRECTIVE token. */
573 check_newline ()
575 register int c;
576 register int token;
577 int saw_line;
578 enum { act_none, act_push, act_pop } action;
579 int old_lineno, action_number, l;
581 restart:
582 /* Read first nonwhite char on the line. */
584 #ifdef USE_CPPLIB
585 c = GETC ();
586 /* In some cases where we're leaving an include file, we can get multiple
587 CPP_HSPACE tokens in a row, so we need to loop. */
588 while (cpp_token == CPP_HSPACE)
589 c = yy_get_token ();
590 #else
592 c = GETC ();
593 while (c == ' ' || c == '\t');
594 #endif
596 lineno++;
598 if (c != '#')
600 /* Sequences of multiple newlines are very common; optimize them. */
601 if (c == '\n')
602 goto restart;
604 /* If not #, return it so caller will use it. */
605 return c;
608 /* Don't read beyond this line. */
609 saw_line = 0;
610 linemode = 1;
612 #if USE_CPPLIB
613 if (cpp_token == CPP_VSPACE)
615 /* Format is "<space> <line number> <filename> <newline>".
616 Only the line number is interesting, and even that
617 we can get more efficiently than scanning the line. */
618 yy_cur = yy_lim - 1;
619 lineno = parse_in.lineno - 1;
620 goto skipline;
622 #endif
624 token = yylex ();
626 if (token == IDENTIFIER)
628 /* If a letter follows, then if the word here is `line', skip
629 it and ignore it; otherwise, ignore the line, with an error
630 if the word isn't `pragma'. */
632 const char *name = IDENTIFIER_POINTER (yylval.ttype);
634 if (!strcmp (name, "pragma"))
636 token = yylex ();
637 if (token != IDENTIFIER
638 || TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
639 goto skipline;
641 #ifdef HANDLE_PRAGMA
642 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS
643 (if both are defined), in order to give the back
644 end a chance to override the interpretation of
645 SYSV style pragmas. */
646 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
647 IDENTIFIER_POINTER (yylval.ttype)))
648 goto skipline;
649 #endif /* HANDLE_PRAGMA */
651 #ifdef HANDLE_GENERIC_PRAGMAS
652 if (handle_generic_pragma (token))
653 goto skipline;
654 #endif /* HANDLE_GENERIC_PRAGMAS */
656 /* Issue a warning message if we have been asked to do so.
657 Ignoring unknown pragmas in system header file unless
658 an explcit -Wunknown-pragmas has been given. */
659 if (warn_unknown_pragmas > 1
660 || (warn_unknown_pragmas && ! in_system_header))
661 warning ("ignoring pragma: %s", token_buffer);
663 goto skipline;
665 else if (!strcmp (name, "define"))
667 debug_define (lineno, GET_DIRECTIVE_LINE ());
668 goto skipline;
670 else if (!strcmp (name, "undef"))
672 debug_undef (lineno, GET_DIRECTIVE_LINE ());
673 goto skipline;
675 else if (!strcmp (name, "line"))
677 saw_line = 1;
678 token = yylex ();
679 goto linenum;
681 else if (!strcmp (name, "ident"))
683 /* #ident. The pedantic warning is now in cccp.c. */
685 /* Here we have just seen `#ident '.
686 A string constant should follow. */
688 token = yylex ();
689 if (token == END_OF_LINE)
690 goto skipline;
691 if (token != STRING
692 || TREE_CODE (yylval.ttype) != STRING_CST)
694 error ("invalid #ident");
695 goto skipline;
698 if (! flag_no_ident)
700 #ifdef ASM_OUTPUT_IDENT
701 ASM_OUTPUT_IDENT (asm_out_file,
702 TREE_STRING_POINTER (yylval.ttype));
703 #endif
706 /* Skip the rest of this line. */
707 goto skipline;
710 error ("undefined or invalid # directive `%s'", name);
711 goto skipline;
714 /* If the # is the only nonwhite char on the line,
715 just ignore it. Check the new newline. */
716 if (token == END_OF_LINE)
717 goto skipline;
719 linenum:
720 /* Here we have either `#line' or `# <nonletter>'.
721 In either case, it should be a line number; a digit should follow. */
723 if (token != CONSTANT
724 || TREE_CODE (yylval.ttype) != INTEGER_CST)
726 error ("invalid #-line");
727 goto skipline;
730 /* subtract one, because it is the following line that
731 gets the specified number */
733 l = TREE_INT_CST_LOW (yylval.ttype) - 1;
735 /* More follows: it must be a string constant (filename).
736 It would be neat to use cpplib to quickly process the string, but
737 (1) we don't have a handy tokenization of the string, and
738 (2) I don't know how well that would work in the presense
739 of filenames that contain wide characters. */
741 if (saw_line)
743 /* Don't treat \ as special if we are processing #line 1 "...".
744 If you want it to be treated specially, use # 1 "...". */
745 ignore_escape_flag = 1;
748 /* Read the string constant. */
749 token = yylex ();
751 ignore_escape_flag = 0;
753 if (token == END_OF_LINE)
755 /* No more: store the line number and check following line. */
756 lineno = l;
757 goto skipline;
760 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
762 error ("invalid #line");
763 goto skipline;
766 input_filename = TREE_STRING_POINTER (yylval.ttype);
768 if (main_input_filename == 0)
769 main_input_filename = input_filename;
771 old_lineno = lineno;
772 action = act_none;
773 action_number = 0;
774 lineno = l;
776 /* Each change of file name
777 reinitializes whether we are now in a system header. */
778 in_system_header = 0;
780 if (!read_line_number (&action_number))
782 /* Update the name in the top element of input_file_stack. */
783 if (input_file_stack)
784 input_file_stack->name = input_filename;
787 /* `1' after file name means entering new file.
788 `2' after file name means just left a file. */
790 if (action_number == 1)
792 action = act_push;
793 read_line_number (&action_number);
795 else if (action_number == 2)
797 action = act_pop;
798 read_line_number (&action_number);
800 if (action_number == 3)
802 /* `3' after file name means this is a system header file. */
803 in_system_header = 1;
804 read_line_number (&action_number);
807 /* Do the actions implied by the preceding numbers. */
809 if (action == act_push)
811 /* Pushing to a new file. */
812 struct file_stack *p
813 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
814 input_file_stack->line = old_lineno;
815 p->next = input_file_stack;
816 p->name = input_filename;
817 p->indent_level = indent_level;
818 input_file_stack = p;
819 input_file_stack_tick++;
820 debug_start_source_file (input_filename);
822 else if (action == act_pop)
824 /* Popping out of a file. */
825 if (input_file_stack->next)
827 struct file_stack *p = input_file_stack;
828 if (indent_level != p->indent_level)
830 warning_with_file_and_line
831 (p->name, old_lineno,
832 "This file contains more `%c's than `%c's.",
833 indent_level > p->indent_level ? '{' : '}',
834 indent_level > p->indent_level ? '}' : '{');
836 input_file_stack = p->next;
837 free (p);
838 input_file_stack_tick++;
839 debug_end_source_file (input_file_stack->line);
841 else
842 error ("#-lines for entering and leaving files don't match");
845 /* Now that we've pushed or popped the input stack,
846 update the name in the top element. */
847 if (input_file_stack)
848 input_file_stack->name = input_filename;
850 /* skip the rest of this line. */
851 skipline:
852 linemode = 0;
853 end_of_file = 0;
856 c = GETC();
857 while (c != '\n' && c != EOF);
858 return c;
861 #ifdef HANDLE_GENERIC_PRAGMAS
863 /* Handle a #pragma directive.
864 TOKEN is the token we read after `#pragma'. Processes the entire input
865 line and return non-zero iff the pragma has been successfully parsed. */
867 /* This function has to be in this file, in order to get at
868 the token types. */
870 static int
871 handle_generic_pragma (token)
872 register int token;
874 for (;;)
876 switch (token)
878 case IDENTIFIER:
879 case TYPENAME:
880 case STRING:
881 case CONSTANT:
882 handle_pragma_token (token_buffer, yylval.ttype);
883 break;
885 case END_OF_LINE:
886 return handle_pragma_token (NULL_PTR, NULL_TREE);
888 default:
889 handle_pragma_token (token_buffer, NULL);
892 token = yylex ();
896 #endif /* HANDLE_GENERIC_PRAGMAS */
898 #define ENDFILE -1 /* token that represents end-of-file */
900 /* Read an escape sequence, returning its equivalent as a character,
901 or store 1 in *ignore_ptr if it is backslash-newline. */
903 static int
904 readescape (ignore_ptr)
905 int *ignore_ptr;
907 register int c = GETC();
908 register int code;
909 register unsigned count;
910 unsigned firstdig = 0;
911 int nonnull;
913 switch (c)
915 case 'x':
916 if (warn_traditional)
917 warning ("the meaning of `\\x' varies with -traditional");
919 if (flag_traditional)
920 return c;
922 code = 0;
923 count = 0;
924 nonnull = 0;
925 while (1)
927 c = GETC();
928 if (! ISXDIGIT (c))
930 UNGETC (c);
931 break;
933 code *= 16;
934 if (c >= 'a' && c <= 'f')
935 code += c - 'a' + 10;
936 if (c >= 'A' && c <= 'F')
937 code += c - 'A' + 10;
938 if (c >= '0' && c <= '9')
939 code += c - '0';
940 if (code != 0 || count != 0)
942 if (count == 0)
943 firstdig = code;
944 count++;
946 nonnull = 1;
948 if (! nonnull)
950 warning ("\\x used with no following hex digits");
951 return 'x';
953 else if (count == 0)
954 /* Digits are all 0's. Ok. */
956 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
957 || (count > 1
958 && (((unsigned)1
959 << (TYPE_PRECISION (integer_type_node)
960 - (count - 1) * 4))
961 <= firstdig)))
962 pedwarn ("hex escape out of range");
963 return code;
965 case '0': case '1': case '2': case '3': case '4':
966 case '5': case '6': case '7':
967 code = 0;
968 count = 0;
969 while ((c <= '7') && (c >= '0') && (count++ < 3))
971 code = (code * 8) + (c - '0');
972 c = GETC();
974 UNGETC (c);
975 return code;
977 case '\\': case '\'': case '"':
978 return c;
980 case '\n':
981 lineno++;
982 *ignore_ptr = 1;
983 return 0;
985 case 'n':
986 return TARGET_NEWLINE;
988 case 't':
989 return TARGET_TAB;
991 case 'r':
992 return TARGET_CR;
994 case 'f':
995 return TARGET_FF;
997 case 'b':
998 return TARGET_BS;
1000 case 'a':
1001 if (warn_traditional)
1002 warning ("the meaning of `\\a' varies with -traditional");
1004 if (flag_traditional)
1005 return c;
1006 return TARGET_BELL;
1008 case 'v':
1009 #if 0 /* Vertical tab is present in common usage compilers. */
1010 if (flag_traditional)
1011 return c;
1012 #endif
1013 return TARGET_VT;
1015 case 'e':
1016 case 'E':
1017 if (pedantic)
1018 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1019 return TARGET_ESC;
1021 case '?':
1022 return c;
1024 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1025 case '(':
1026 case '{':
1027 case '[':
1028 /* `\%' is used to prevent SCCS from getting confused. */
1029 case '%':
1030 if (pedantic)
1031 pedwarn ("unknown escape sequence `\\%c'", c);
1032 return c;
1034 if (ISGRAPH (c))
1035 pedwarn ("unknown escape sequence `\\%c'", c);
1036 else
1037 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1038 return c;
1041 void
1042 yyerror (msgid)
1043 const char *msgid;
1045 const char *string = _(msgid);
1047 /* We can't print string and character constants well
1048 because the token_buffer contains the result of processing escapes. */
1049 if (end_of_file)
1050 error ("%s at end of input", string);
1051 else if (token_buffer[0] == 0)
1052 error ("%s at null character", string);
1053 else if (token_buffer[0] == '"')
1054 error ("%s before string constant", string);
1055 else if (token_buffer[0] == '\'')
1056 error ("%s before character constant", string);
1057 else if (!ISGRAPH(token_buffer[0]))
1058 error ("%s before character 0%o", string, (unsigned char) token_buffer[0]);
1059 else
1060 error ("%s before `%s'", string, token_buffer);
1063 #if 0
1065 struct try_type
1067 tree *node_var;
1068 char unsigned_flag;
1069 char long_flag;
1070 char long_long_flag;
1073 struct try_type type_sequence[] =
1075 { &integer_type_node, 0, 0, 0},
1076 { &unsigned_type_node, 1, 0, 0},
1077 { &long_integer_type_node, 0, 1, 0},
1078 { &long_unsigned_type_node, 1, 1, 0},
1079 { &long_long_integer_type_node, 0, 1, 1},
1080 { &long_long_unsigned_type_node, 1, 1, 1}
1082 #endif /* 0 */
1084 struct pf_args
1086 /* Input */
1087 int base;
1088 char * p;
1089 /* I/O */
1090 int c;
1091 /* Output */
1092 int imag;
1093 tree type;
1094 int conversion_errno;
1095 REAL_VALUE_TYPE value;
1098 static void
1099 parse_float (data)
1100 PTR data;
1102 struct pf_args * args = (struct pf_args *) data;
1103 int fflag = 0, lflag = 0;
1104 /* Copy token_buffer now, while it has just the number
1105 and not the suffixes; once we add `f' or `i',
1106 REAL_VALUE_ATOF may not work any more. */
1107 char *copy = (char *) alloca (args->p - token_buffer + 1);
1108 bcopy (token_buffer, copy, args->p - token_buffer + 1);
1109 args->imag = 0;
1110 args->conversion_errno = 0;
1111 args->type = double_type_node;
1113 while (1)
1115 int lose = 0;
1117 /* Read the suffixes to choose a data type. */
1118 switch (args->c)
1120 case 'f': case 'F':
1121 if (fflag)
1122 error ("more than one `f' in numeric constant");
1123 fflag = 1;
1124 break;
1126 case 'l': case 'L':
1127 if (lflag)
1128 error ("more than one `l' in numeric constant");
1129 lflag = 1;
1130 break;
1132 case 'i': case 'I':
1133 if (args->imag)
1134 error ("more than one `i' or `j' in numeric constant");
1135 else if (pedantic)
1136 pedwarn ("ANSI C forbids imaginary numeric constants");
1137 args->imag = 1;
1138 break;
1140 default:
1141 lose = 1;
1144 if (lose)
1145 break;
1147 if (args->p >= token_buffer + maxtoken - 3)
1148 args->p = extend_token_buffer (args->p);
1149 *(args->p++) = args->c;
1150 *(args->p) = 0;
1151 args->c = GETC();
1154 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1155 tells the desired precision of the binary result
1156 of decimal-to-binary conversion. */
1158 if (fflag)
1160 if (lflag)
1161 error ("both `f' and `l' in floating constant");
1163 args->type = float_type_node;
1164 errno = 0;
1165 if (args->base == 16)
1166 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1167 else
1168 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1169 args->conversion_errno = errno;
1170 /* A diagnostic is required here by some ANSI C testsuites.
1171 This is not pedwarn, because some people don't want
1172 an error for this. */
1173 if (REAL_VALUE_ISINF (args->value) && pedantic)
1174 warning ("floating point number exceeds range of `float'");
1176 else if (lflag)
1178 args->type = long_double_type_node;
1179 errno = 0;
1180 if (args->base == 16)
1181 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1182 else
1183 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1184 args->conversion_errno = errno;
1185 if (REAL_VALUE_ISINF (args->value) && pedantic)
1186 warning ("floating point number exceeds range of `long double'");
1188 else
1190 errno = 0;
1191 if (args->base == 16)
1192 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1193 else
1194 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1195 args->conversion_errno = errno;
1196 if (REAL_VALUE_ISINF (args->value) && pedantic)
1197 warning ("floating point number exceeds range of `double'");
1201 /* Get the next character, staying within the current token if possible.
1202 If we're lexing a token, we don't want to look beyond the end of the
1203 token cpplib has prepared for us; otherwise, we end up reading in the
1204 next token, which screws up feed_input. So just return a null
1205 character. */
1207 static inline int token_getch PARAMS ((void));
1209 static inline int
1210 token_getch ()
1212 #if USE_CPPLIB
1213 if (yy_cur == yy_lim)
1214 return '\0';
1215 #endif
1216 return GETC ();
1219 static inline void token_put_back PARAMS ((int));
1221 static inline void
1222 token_put_back (ch)
1223 int ch;
1225 #if USE_CPPLIB
1226 if (ch == '\0')
1227 return;
1228 #endif
1229 UNGETC (ch);
1232 /* Read a single token from the input stream, and assign it lexical
1233 semantics. */
1236 yylex ()
1238 register int c;
1239 register char *p;
1240 register int value;
1241 int wide_flag = 0;
1242 int objc_flag = 0;
1244 c = GETC();
1246 /* Effectively do c = skip_white_space (c)
1247 but do it faster in the usual cases. */
1248 while (1)
1249 switch (c)
1251 case ' ':
1252 case '\t':
1253 case '\f':
1254 case '\v':
1255 case '\b':
1256 #if USE_CPPLIB
1257 if (cpp_token == CPP_HSPACE)
1258 c = yy_get_token ();
1259 else
1260 #endif
1261 c = GETC();
1262 break;
1264 case '\r':
1265 /* Call skip_white_space so we can warn if appropriate. */
1267 case '\n':
1268 case '/':
1269 case '\\':
1270 c = skip_white_space (c);
1271 default:
1272 goto found_nonwhite;
1274 found_nonwhite:
1276 token_buffer[0] = c;
1277 token_buffer[1] = 0;
1279 /* yylloc.first_line = lineno; */
1281 switch (c)
1283 case EOF:
1284 end_of_file = 1;
1285 token_buffer[0] = 0;
1286 if (linemode)
1287 value = END_OF_LINE;
1288 else
1289 value = ENDFILE;
1290 break;
1292 case 'L':
1293 #if USE_CPPLIB
1294 if (cpp_token == CPP_NAME)
1295 goto letter;
1296 #endif
1297 /* Capital L may start a wide-string or wide-character constant. */
1299 register int c = token_getch();
1300 if (c == '\'')
1302 wide_flag = 1;
1303 goto char_constant;
1305 if (c == '"')
1307 wide_flag = 1;
1308 goto string_constant;
1310 token_put_back (c);
1312 goto letter;
1314 case '@':
1315 if (!doing_objc_thang)
1317 value = c;
1318 break;
1320 else
1322 /* '@' may start a constant string object. */
1323 register int c = token_getch ();
1324 if (c == '"')
1326 objc_flag = 1;
1327 goto string_constant;
1329 token_put_back (c);
1330 /* Fall through to treat '@' as the start of an identifier. */
1333 case 'A': case 'B': case 'C': case 'D': case 'E':
1334 case 'F': case 'G': case 'H': case 'I': case 'J':
1335 case 'K': case 'M': case 'N': case 'O':
1336 case 'P': case 'Q': case 'R': case 'S': case 'T':
1337 case 'U': case 'V': case 'W': case 'X': case 'Y':
1338 case 'Z':
1339 case 'a': case 'b': case 'c': case 'd': case 'e':
1340 case 'f': case 'g': case 'h': case 'i': case 'j':
1341 case 'k': case 'l': case 'm': case 'n': case 'o':
1342 case 'p': case 'q': case 'r': case 's': case 't':
1343 case 'u': case 'v': case 'w': case 'x': case 'y':
1344 case 'z':
1345 case '_':
1346 case '$':
1347 letter:
1348 #if USE_CPPLIB
1349 if (cpp_token == CPP_NAME)
1351 /* Note that one character has already been read from
1352 yy_cur into token_buffer. Also, cpplib complains about
1353 $ in identifiers, so we don't have to. */
1355 int len = yy_lim - yy_cur + 1;
1356 if (len >= maxtoken)
1357 extend_token_buffer_to (len + 1);
1358 memcpy (token_buffer + 1, yy_cur, len);
1359 p = token_buffer + len;
1360 yy_cur = yy_lim;
1362 else
1363 #endif
1365 p = token_buffer;
1366 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1368 /* Make sure this char really belongs in an identifier. */
1369 if (c == '$')
1371 if (! dollars_in_ident)
1372 error ("`$' in identifier");
1373 else if (pedantic)
1374 pedwarn ("`$' in identifier");
1377 if (p >= token_buffer + maxtoken)
1378 p = extend_token_buffer (p);
1380 *p++ = c;
1381 c = token_getch();
1384 *p = 0;
1385 token_put_back (c);
1388 value = IDENTIFIER;
1389 yylval.itype = 0;
1391 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1394 register struct resword *ptr;
1396 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1398 if (ptr->rid)
1399 yylval.ttype = ridpointers[(int) ptr->rid];
1400 value = (int) ptr->token;
1402 /* Only return OBJECTNAME if it is a typedef. */
1403 if (doing_objc_thang && value == OBJECTNAME)
1405 lastiddecl = lookup_name(yylval.ttype);
1407 if (lastiddecl == NULL_TREE
1408 || TREE_CODE (lastiddecl) != TYPE_DECL)
1409 value = IDENTIFIER;
1412 /* Even if we decided to recognize asm, still perhaps warn. */
1413 if (pedantic
1414 && (value == ASM_KEYWORD || value == TYPEOF
1415 || ptr->rid == RID_INLINE)
1416 && token_buffer[0] != '_')
1417 pedwarn ("ANSI does not permit the keyword `%s'",
1418 token_buffer);
1422 /* If we did not find a keyword, look for an identifier
1423 (or a typename). */
1425 if (value == IDENTIFIER)
1427 if (token_buffer[0] == '@')
1428 error("invalid identifier `%s'", token_buffer);
1430 yylval.ttype = get_identifier (token_buffer);
1431 lastiddecl = lookup_name (yylval.ttype);
1433 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1434 value = TYPENAME;
1435 /* A user-invisible read-only initialized variable
1436 should be replaced by its value.
1437 We handle only strings since that's the only case used in C. */
1438 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1439 && DECL_IGNORED_P (lastiddecl)
1440 && TREE_READONLY (lastiddecl)
1441 && DECL_INITIAL (lastiddecl) != 0
1442 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1444 tree stringval = DECL_INITIAL (lastiddecl);
1446 /* Copy the string value so that we won't clobber anything
1447 if we put something in the TREE_CHAIN of this one. */
1448 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1449 TREE_STRING_POINTER (stringval));
1450 value = STRING;
1452 else if (doing_objc_thang)
1454 tree objc_interface_decl = is_class_name (yylval.ttype);
1456 if (objc_interface_decl)
1458 value = CLASSNAME;
1459 yylval.ttype = objc_interface_decl;
1464 break;
1466 case '.':
1467 #if USE_CPPLIB
1468 if (yy_cur < yy_lim)
1469 #endif
1471 /* It's hard to preserve tokenization on '.' because
1472 it could be a symbol by itself, or it could be the
1473 start of a floating point number and cpp won't tell us. */
1474 register int c1 = token_getch ();
1475 token_buffer[1] = c1;
1476 if (c1 == '.')
1478 c1 = token_getch ();
1479 if (c1 == '.')
1481 token_buffer[2] = c1;
1482 token_buffer[3] = 0;
1483 value = ELLIPSIS;
1484 goto done;
1486 error ("parse error at `..'");
1488 if (ISDIGIT (c1))
1490 token_put_back (c1);
1491 goto number;
1493 token_put_back (c1);
1495 value = '.';
1496 token_buffer[1] = 0;
1497 break;
1499 case '0': case '1':
1500 /* Optimize for most frequent case. */
1502 register int cond;
1504 #if USE_CPPLIB
1505 cond = (yy_cur == yy_lim);
1506 #else
1507 register int c1 = token_getch ();
1508 token_put_back (c1);
1509 cond = (! ISALNUM (c1) && c1 != '.');
1510 #endif
1511 if (cond)
1513 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1514 value = CONSTANT;
1515 break;
1517 /*FALLTHRU*/
1519 case '2': case '3': case '4':
1520 case '5': case '6': case '7': case '8': case '9':
1521 number:
1523 int base = 10;
1524 int count = 0;
1525 int largest_digit = 0;
1526 int numdigits = 0;
1527 int overflow = 0;
1529 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
1530 The code below which fills the parts array assumes that a host
1531 int is at least twice as wide as a host char, and that
1532 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
1533 Two HOST_WIDE_INTs is the largest int literal we can store.
1534 In order to detect overflow below, the number of parts (TOTAL_PARTS)
1535 must be exactly the number of parts needed to hold the bits
1536 of two HOST_WIDE_INTs. */
1537 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
1538 unsigned int parts[TOTAL_PARTS];
1540 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS, AFTER_EXPON}
1541 floatflag = NOT_FLOAT;
1543 for (count = 0; count < TOTAL_PARTS; count++)
1544 parts[count] = 0;
1546 p = token_buffer;
1547 *p++ = c;
1549 if (c == '0')
1551 *p++ = (c = token_getch());
1552 if ((c == 'x') || (c == 'X'))
1554 base = 16;
1555 *p++ = (c = token_getch());
1557 /* Leading 0 forces octal unless the 0 is the only digit. */
1558 else if (c >= '0' && c <= '9')
1560 base = 8;
1561 numdigits++;
1563 else
1564 numdigits++;
1567 /* Read all the digits-and-decimal-points. */
1569 while (c == '.'
1570 || (ISALNUM (c) && c != 'l' && c != 'L'
1571 && c != 'u' && c != 'U'
1572 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1573 && (floatflag == NOT_FLOAT
1574 || ((base != 16) && (c != 'f') && (c != 'F'))
1575 || base == 16)))
1577 if (c == '.')
1579 if (base == 16 && pedantic)
1580 pedwarn ("floating constant may not be in radix 16");
1581 if (floatflag == TOO_MANY_POINTS)
1582 /* We have already emitted an error. Don't need another. */
1584 else if (floatflag == AFTER_POINT || floatflag == AFTER_EXPON)
1586 error ("malformed floating constant");
1587 floatflag = TOO_MANY_POINTS;
1588 /* Avoid another error from atof by forcing all characters
1589 from here on to be ignored. */
1590 p[-1] = '\0';
1592 else
1593 floatflag = AFTER_POINT;
1595 if (base == 8)
1596 base = 10;
1597 *p++ = c = token_getch();
1598 /* Accept '.' as the start of a floating-point number
1599 only when it is followed by a digit. */
1600 if (p == token_buffer + 2 && !ISDIGIT (c))
1601 abort ();
1603 else
1605 /* It is not a decimal point.
1606 It should be a digit (perhaps a hex digit). */
1608 if (ISDIGIT (c))
1610 c = c - '0';
1612 else if (base <= 10)
1614 if (c == 'e' || c == 'E')
1616 base = 10;
1617 floatflag = AFTER_EXPON;
1618 break; /* start of exponent */
1620 error ("nondigits in number and not hexadecimal");
1621 c = 0;
1623 else if (base == 16 && (c == 'p' || c == 'P'))
1625 floatflag = AFTER_EXPON;
1626 break; /* start of exponent */
1628 else if (c >= 'a')
1630 c = c - 'a' + 10;
1632 else
1634 c = c - 'A' + 10;
1636 if (c >= largest_digit)
1637 largest_digit = c;
1638 numdigits++;
1640 for (count = 0; count < TOTAL_PARTS; count++)
1642 parts[count] *= base;
1643 if (count)
1645 parts[count]
1646 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1647 parts[count-1]
1648 &= (1 << HOST_BITS_PER_CHAR) - 1;
1650 else
1651 parts[0] += c;
1654 /* If the highest-order part overflows (gets larger than
1655 a host char will hold) then the whole number has
1656 overflowed. Record this and truncate the highest-order
1657 part. */
1658 if (parts[TOTAL_PARTS - 1] >> HOST_BITS_PER_CHAR)
1660 overflow = 1;
1661 parts[TOTAL_PARTS - 1] &= (1 << HOST_BITS_PER_CHAR) - 1;
1664 if (p >= token_buffer + maxtoken - 3)
1665 p = extend_token_buffer (p);
1666 *p++ = (c = token_getch());
1670 /* This can happen on input like `int i = 0x;' */
1671 if (numdigits == 0)
1672 error ("numeric constant with no digits");
1674 if (largest_digit >= base)
1675 error ("numeric constant contains digits beyond the radix");
1677 /* Remove terminating char from the token buffer and delimit the
1678 string. */
1679 *--p = 0;
1681 if (floatflag != NOT_FLOAT)
1683 tree type;
1684 int imag, conversion_errno;
1685 REAL_VALUE_TYPE value;
1686 struct pf_args args;
1688 /* Read explicit exponent if any, and put it in tokenbuf. */
1690 if ((base == 10 && ((c == 'e') || (c == 'E')))
1691 || (base == 16 && (c == 'p' || c == 'P')))
1693 if (p >= token_buffer + maxtoken - 3)
1694 p = extend_token_buffer (p);
1695 *p++ = c;
1696 c = token_getch();
1697 if ((c == '+') || (c == '-'))
1699 *p++ = c;
1700 c = token_getch();
1702 /* Exponent is decimal, even if string is a hex float. */
1703 if (! ISDIGIT (c))
1704 error ("floating constant exponent has no digits");
1705 while (ISDIGIT (c))
1707 if (p >= token_buffer + maxtoken - 3)
1708 p = extend_token_buffer (p);
1709 *p++ = c;
1710 c = token_getch ();
1713 if (base == 16 && floatflag != AFTER_EXPON)
1714 error ("hexadecimal floating constant has no exponent");
1716 *p = 0;
1718 /* Setup input for parse_float() */
1719 args.base = base;
1720 args.p = p;
1721 args.c = c;
1723 /* Convert string to a double, checking for overflow. */
1724 if (do_float_handler (parse_float, (PTR) &args))
1726 /* Receive output from parse_float() */
1727 value = args.value;
1729 else
1731 /* We got an exception from parse_float() */
1732 error ("floating constant out of range");
1733 value = dconst0;
1736 /* Receive output from parse_float() */
1737 c = args.c;
1738 imag = args.imag;
1739 type = args.type;
1740 conversion_errno = args.conversion_errno;
1742 #ifdef ERANGE
1743 /* ERANGE is also reported for underflow,
1744 so test the value to distinguish overflow from that. */
1745 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1746 && (REAL_VALUES_LESS (dconst1, value)
1747 || REAL_VALUES_LESS (value, dconstm1)))
1748 warning ("floating point number exceeds range of `double'");
1749 #endif
1751 /* If the result is not a number, assume it must have been
1752 due to some error message above, so silently convert
1753 it to a zero. */
1754 if (REAL_VALUE_ISNAN (value))
1755 value = dconst0;
1757 /* Create a node with determined type and value. */
1758 if (imag)
1759 yylval.ttype = build_complex (NULL_TREE,
1760 convert (type, integer_zero_node),
1761 build_real (type, value));
1762 else
1763 yylval.ttype = build_real (type, value);
1765 else
1767 tree traditional_type, ansi_type, type;
1768 HOST_WIDE_INT high, low;
1769 int spec_unsigned = 0;
1770 int spec_long = 0;
1771 int spec_long_long = 0;
1772 int spec_imag = 0;
1773 int warn = 0, i;
1775 traditional_type = ansi_type = type = NULL_TREE;
1776 while (1)
1778 if (c == 'u' || c == 'U')
1780 if (spec_unsigned)
1781 error ("two `u's in integer constant");
1782 spec_unsigned = 1;
1784 else if (c == 'l' || c == 'L')
1786 if (spec_long)
1788 if (spec_long_long)
1789 error ("three `l's in integer constant");
1790 else if (pedantic && ! in_system_header && warn_long_long)
1791 pedwarn ("ANSI C forbids long long integer constants");
1792 spec_long_long = 1;
1794 spec_long = 1;
1796 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1798 if (spec_imag)
1799 error ("more than one `i' or `j' in numeric constant");
1800 else if (pedantic)
1801 pedwarn ("ANSI C forbids imaginary numeric constants");
1802 spec_imag = 1;
1804 else
1805 break;
1806 if (p >= token_buffer + maxtoken - 3)
1807 p = extend_token_buffer (p);
1808 *p++ = c;
1809 c = token_getch();
1812 /* If the literal overflowed, pedwarn about it now. */
1813 if (overflow)
1815 warn = 1;
1816 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT * 2);
1819 /* This is simplified by the fact that our constant
1820 is always positive. */
1822 high = low = 0;
1824 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1826 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1827 / HOST_BITS_PER_CHAR)]
1828 << (i * HOST_BITS_PER_CHAR));
1829 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1832 yylval.ttype = build_int_2 (low, high);
1833 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1835 /* If warn_traditional, calculate both the ANSI type and the
1836 traditional type, then see if they disagree.
1837 Otherwise, calculate only the type for the dialect in use. */
1838 if (warn_traditional || flag_traditional)
1840 /* Calculate the traditional type. */
1841 /* Traditionally, any constant is signed;
1842 but if unsigned is specified explicitly, obey that.
1843 Use the smallest size with the right number of bits,
1844 except for one special case with decimal constants. */
1845 if (! spec_long && base != 10
1846 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1847 traditional_type = (spec_unsigned ? unsigned_type_node
1848 : integer_type_node);
1849 /* A decimal constant must be long
1850 if it does not fit in type int.
1851 I think this is independent of whether
1852 the constant is signed. */
1853 else if (! spec_long && base == 10
1854 && int_fits_type_p (yylval.ttype, integer_type_node))
1855 traditional_type = (spec_unsigned ? unsigned_type_node
1856 : integer_type_node);
1857 else if (! spec_long_long)
1858 traditional_type = (spec_unsigned ? long_unsigned_type_node
1859 : long_integer_type_node);
1860 else if (int_fits_type_p (yylval.ttype,
1861 spec_unsigned
1862 ? long_long_unsigned_type_node
1863 : long_long_integer_type_node))
1864 traditional_type = (spec_unsigned
1865 ? long_long_unsigned_type_node
1866 : long_long_integer_type_node);
1867 else
1868 traditional_type = (spec_unsigned
1869 ? widest_unsigned_literal_type_node
1870 : widest_integer_literal_type_node);
1872 if (warn_traditional || ! flag_traditional)
1874 /* Calculate the ANSI type. */
1875 if (! spec_long && ! spec_unsigned
1876 && int_fits_type_p (yylval.ttype, integer_type_node))
1877 ansi_type = integer_type_node;
1878 else if (! spec_long && (base != 10 || spec_unsigned)
1879 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1880 ansi_type = unsigned_type_node;
1881 else if (! spec_unsigned && !spec_long_long
1882 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1883 ansi_type = long_integer_type_node;
1884 else if (! spec_long_long
1885 && int_fits_type_p (yylval.ttype,
1886 long_unsigned_type_node))
1887 ansi_type = long_unsigned_type_node;
1888 else if (! spec_unsigned
1889 && int_fits_type_p (yylval.ttype,
1890 long_long_integer_type_node))
1891 ansi_type = long_long_integer_type_node;
1892 else if (int_fits_type_p (yylval.ttype,
1893 long_long_unsigned_type_node))
1894 ansi_type = long_long_unsigned_type_node;
1895 else if (! spec_unsigned
1896 && int_fits_type_p (yylval.ttype,
1897 widest_integer_literal_type_node))
1898 ansi_type = widest_integer_literal_type_node;
1899 else
1900 ansi_type = widest_unsigned_literal_type_node;
1903 type = flag_traditional ? traditional_type : ansi_type;
1905 /* We assume that constants specified in a non-decimal
1906 base are bit patterns, and that the programmer really
1907 meant what they wrote. */
1908 if (warn_traditional && base == 10
1909 && traditional_type != ansi_type)
1911 if (TYPE_PRECISION (traditional_type)
1912 != TYPE_PRECISION (ansi_type))
1913 warning ("width of integer constant changes with -traditional");
1914 else if (TREE_UNSIGNED (traditional_type)
1915 != TREE_UNSIGNED (ansi_type))
1916 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1917 else
1918 warning ("width of integer constant may change on other systems with -traditional");
1921 if (pedantic && !flag_traditional && !spec_long_long && !warn
1922 && (TYPE_PRECISION (long_integer_type_node)
1923 < TYPE_PRECISION (type)))
1925 warn = 1;
1926 pedwarn ("integer constant larger than the maximum value of an unsigned long int");
1929 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1930 warning ("decimal constant is so large that it is unsigned");
1932 if (spec_imag)
1934 if (TYPE_PRECISION (type)
1935 <= TYPE_PRECISION (integer_type_node))
1936 yylval.ttype
1937 = build_complex (NULL_TREE, integer_zero_node,
1938 convert (integer_type_node,
1939 yylval.ttype));
1940 else
1941 error ("complex integer constant is too wide for `complex int'");
1943 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1944 /* The traditional constant 0x80000000 is signed
1945 but doesn't fit in the range of int.
1946 This will change it to -0x80000000, which does fit. */
1948 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1949 yylval.ttype = convert (type, yylval.ttype);
1950 TREE_OVERFLOW (yylval.ttype)
1951 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1953 else
1954 TREE_TYPE (yylval.ttype) = type;
1957 /* If it's still an integer (not a complex), and it doesn't
1958 fit in the type we choose for it, then pedwarn. */
1960 if (! warn
1961 && TREE_CODE (TREE_TYPE (yylval.ttype)) == INTEGER_TYPE
1962 && ! int_fits_type_p (yylval.ttype, TREE_TYPE (yylval.ttype)))
1963 pedwarn ("integer constant is larger than the maximum value for its type");
1966 token_put_back (c);
1967 *p = 0;
1969 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1970 || (!flag_traditional && (c == '-' || c == '+')
1971 && (p[-1] == 'e' || p[-1] == 'E')))
1972 error ("missing white space after number `%s'", token_buffer);
1974 value = CONSTANT; break;
1977 case '\'':
1978 char_constant:
1980 register int result = 0;
1981 register int num_chars = 0;
1982 int chars_seen = 0;
1983 unsigned width = TYPE_PRECISION (char_type_node);
1984 int max_chars;
1985 #ifdef MULTIBYTE_CHARS
1986 int longest_char = local_mb_cur_max ();
1987 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1988 #endif
1990 max_chars = TYPE_PRECISION (integer_type_node) / width;
1991 if (wide_flag)
1992 width = WCHAR_TYPE_SIZE;
1994 while (1)
1996 tryagain:
1997 c = token_getch();
1999 if (c == '\'' || c == EOF)
2000 break;
2002 ++chars_seen;
2003 if (c == '\\')
2005 int ignore = 0;
2006 c = readescape (&ignore);
2007 if (ignore)
2008 goto tryagain;
2009 if (width < HOST_BITS_PER_INT
2010 && (unsigned) c >= ((unsigned)1 << width))
2011 pedwarn ("escape sequence out of range for character");
2012 #ifdef MAP_CHARACTER
2013 if (ISPRINT (c))
2014 c = MAP_CHARACTER (c);
2015 #endif
2017 else if (c == '\n')
2019 if (pedantic)
2020 pedwarn ("ANSI C forbids newline in character constant");
2021 lineno++;
2023 else
2025 #ifdef MULTIBYTE_CHARS
2026 wchar_t wc;
2027 int i;
2028 int char_len = -1;
2029 for (i = 1; i <= longest_char; ++i)
2031 if (i > maxtoken - 4)
2032 extend_token_buffer (token_buffer);
2034 token_buffer[i] = c;
2035 char_len = local_mbtowc (& wc,
2036 token_buffer + 1,
2038 if (char_len != -1)
2039 break;
2040 c = token_getch ();
2042 if (char_len > 1)
2044 /* mbtowc sometimes needs an extra char before accepting */
2045 if (char_len < i)
2046 token_put_back (c);
2047 if (! wide_flag)
2049 /* Merge character into result; ignore excess chars. */
2050 for (i = 1; i <= char_len; ++i)
2052 if (i > max_chars)
2053 break;
2054 if (width < HOST_BITS_PER_INT)
2055 result = (result << width)
2056 | (token_buffer[i]
2057 & ((1 << width) - 1));
2058 else
2059 result = token_buffer[i];
2061 num_chars += char_len;
2062 goto tryagain;
2064 c = wc;
2066 else
2068 if (char_len == -1)
2070 warning ("Ignoring invalid multibyte character");
2071 /* Replace all but the first byte. */
2072 for (--i; i > 1; --i)
2073 token_put_back (token_buffer[i]);
2074 wc = token_buffer[1];
2076 #ifdef MAP_CHARACTER
2077 c = MAP_CHARACTER (wc);
2078 #else
2079 c = wc;
2080 #endif
2082 #else /* ! MULTIBYTE_CHARS */
2083 #ifdef MAP_CHARACTER
2084 c = MAP_CHARACTER (c);
2085 #endif
2086 #endif /* ! MULTIBYTE_CHARS */
2089 if (wide_flag)
2091 if (chars_seen == 1) /* only keep the first one */
2092 result = c;
2093 goto tryagain;
2096 /* Merge character into result; ignore excess chars. */
2097 num_chars += (width / TYPE_PRECISION (char_type_node));
2098 if (num_chars < max_chars + 1)
2100 if (width < HOST_BITS_PER_INT)
2101 result = (result << width) | (c & ((1 << width) - 1));
2102 else
2103 result = c;
2107 if (c != '\'')
2108 error ("malformed character constant");
2109 else if (chars_seen == 0)
2110 error ("empty character constant");
2111 else if (num_chars > max_chars)
2113 num_chars = max_chars;
2114 error ("character constant too long");
2116 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
2117 warning ("multi-character character constant");
2119 /* If char type is signed, sign-extend the constant. */
2120 if (! wide_flag)
2122 int num_bits = num_chars * width;
2123 if (num_bits == 0)
2124 /* We already got an error; avoid invalid shift. */
2125 yylval.ttype = build_int_2 (0, 0);
2126 else if (TREE_UNSIGNED (char_type_node)
2127 || ((result >> (num_bits - 1)) & 1) == 0)
2128 yylval.ttype
2129 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
2130 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2132 else
2133 yylval.ttype
2134 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
2135 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2136 -1);
2137 TREE_TYPE (yylval.ttype) = integer_type_node;
2139 else
2141 yylval.ttype = build_int_2 (result, 0);
2142 TREE_TYPE (yylval.ttype) = wchar_type_node;
2145 value = CONSTANT;
2146 break;
2149 case '"':
2150 string_constant:
2152 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2153 : TYPE_PRECISION (char_type_node);
2154 #ifdef MULTIBYTE_CHARS
2155 int longest_char = local_mb_cur_max ();
2156 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2157 #endif
2158 c = token_getch ();
2159 p = token_buffer + 1;
2161 while (c != '"' && c != EOF)
2163 /* ignore_escape_flag is set for reading the filename in #line. */
2164 if (!ignore_escape_flag && c == '\\')
2166 int ignore = 0;
2167 c = readescape (&ignore);
2168 if (ignore)
2169 goto skipnewline;
2170 if (width < HOST_BITS_PER_INT
2171 && (unsigned) c >= ((unsigned)1 << width))
2172 pedwarn ("escape sequence out of range for character");
2174 else if (c == '\n')
2176 if (pedantic)
2177 pedwarn ("ANSI C forbids newline in string constant");
2178 lineno++;
2180 else
2182 #ifdef MULTIBYTE_CHARS
2183 wchar_t wc;
2184 int i;
2185 int char_len = -1;
2186 for (i = 0; i < longest_char; ++i)
2188 if (p + i >= token_buffer + maxtoken)
2189 p = extend_token_buffer (p);
2190 p[i] = c;
2192 char_len = local_mbtowc (& wc, p, i + 1);
2193 if (char_len != -1)
2194 break;
2195 c = token_getch ();
2197 if (char_len == -1)
2199 warning ("Ignoring invalid multibyte character");
2200 /* Replace all except the first byte. */
2201 token_put_back (c);
2202 for (--i; i > 0; --i)
2203 token_put_back (p[i]);
2204 char_len = 1;
2206 /* mbtowc sometimes needs an extra char before accepting */
2207 if (char_len <= i)
2208 token_put_back (c);
2209 if (! wide_flag)
2211 p += (i + 1);
2212 c = token_getch ();
2213 continue;
2215 c = wc;
2216 #endif /* MULTIBYTE_CHARS */
2219 /* Add this single character into the buffer either as a wchar_t
2220 or as a single byte. */
2221 if (wide_flag)
2223 unsigned width = TYPE_PRECISION (char_type_node);
2224 unsigned bytemask = (1 << width) - 1;
2225 int byte;
2227 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2228 p = extend_token_buffer (p);
2230 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2232 int value;
2233 if (byte >= (int) sizeof (c))
2234 value = 0;
2235 else
2236 value = (c >> (byte * width)) & bytemask;
2237 if (BYTES_BIG_ENDIAN)
2238 p[WCHAR_BYTES - byte - 1] = value;
2239 else
2240 p[byte] = value;
2242 p += WCHAR_BYTES;
2244 else
2246 if (p >= token_buffer + maxtoken)
2247 p = extend_token_buffer (p);
2248 *p++ = c;
2251 skipnewline:
2252 c = token_getch ();
2255 /* Terminate the string value, either with a single byte zero
2256 or with a wide zero. */
2257 if (wide_flag)
2259 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2260 p = extend_token_buffer (p);
2261 bzero (p, WCHAR_BYTES);
2262 p += WCHAR_BYTES;
2264 else
2266 if (p >= token_buffer + maxtoken)
2267 p = extend_token_buffer (p);
2268 *p++ = 0;
2271 if (c == EOF)
2272 error ("Unterminated string constant");
2274 /* We have read the entire constant.
2275 Construct a STRING_CST for the result. */
2277 if (wide_flag)
2279 yylval.ttype = build_string (p - (token_buffer + 1),
2280 token_buffer + 1);
2281 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2282 value = STRING;
2284 else if (objc_flag)
2286 /* Return an Objective-C @"..." constant string object. */
2287 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2288 token_buffer + 1);
2289 TREE_TYPE (yylval.ttype) = char_array_type_node;
2290 value = OBJC_STRING;
2292 else
2294 yylval.ttype = build_string (p - (token_buffer + 1),
2295 token_buffer + 1);
2296 TREE_TYPE (yylval.ttype) = char_array_type_node;
2297 value = STRING;
2300 break;
2303 case '+':
2304 case '-':
2305 case '&':
2306 case '|':
2307 case ':':
2308 case '<':
2309 case '>':
2310 case '*':
2311 case '/':
2312 case '%':
2313 case '^':
2314 case '!':
2315 case '=':
2317 register int c1;
2319 combine:
2321 switch (c)
2323 case '+':
2324 yylval.code = PLUS_EXPR; break;
2325 case '-':
2326 yylval.code = MINUS_EXPR; break;
2327 case '&':
2328 yylval.code = BIT_AND_EXPR; break;
2329 case '|':
2330 yylval.code = BIT_IOR_EXPR; break;
2331 case '*':
2332 yylval.code = MULT_EXPR; break;
2333 case '/':
2334 yylval.code = TRUNC_DIV_EXPR; break;
2335 case '%':
2336 yylval.code = TRUNC_MOD_EXPR; break;
2337 case '^':
2338 yylval.code = BIT_XOR_EXPR; break;
2339 case LSHIFT:
2340 yylval.code = LSHIFT_EXPR; break;
2341 case RSHIFT:
2342 yylval.code = RSHIFT_EXPR; break;
2343 case '<':
2344 yylval.code = LT_EXPR; break;
2345 case '>':
2346 yylval.code = GT_EXPR; break;
2349 token_buffer[1] = c1 = token_getch();
2350 token_buffer[2] = 0;
2352 if (c1 == '=')
2354 switch (c)
2356 case '<':
2357 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2358 case '>':
2359 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2360 case '!':
2361 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2362 case '=':
2363 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2365 value = ASSIGN; goto done;
2367 else if (c == c1)
2368 switch (c)
2370 case '+':
2371 value = PLUSPLUS; goto done;
2372 case '-':
2373 value = MINUSMINUS; goto done;
2374 case '&':
2375 value = ANDAND; goto done;
2376 case '|':
2377 value = OROR; goto done;
2378 case '<':
2379 c = LSHIFT;
2380 goto combine;
2381 case '>':
2382 c = RSHIFT;
2383 goto combine;
2385 else
2386 switch (c)
2388 case '-':
2389 if (c1 == '>')
2390 { value = POINTSAT; goto done; }
2391 break;
2393 /* digraphs */
2394 case ':':
2395 if (c1 == '>')
2396 { value = ']'; goto done; }
2397 break;
2398 case '<':
2399 if (c1 == '%')
2400 { value = '{'; indent_level++; goto done; }
2401 if (c1 == ':')
2402 { value = '['; goto done; }
2403 break;
2404 case '%':
2405 if (c1 == '>')
2406 { value = '}'; indent_level--; goto done; }
2407 break;
2410 token_put_back (c1);
2411 token_buffer[1] = 0;
2413 if ((c == '<') || (c == '>'))
2414 value = ARITHCOMPARE;
2415 else value = c;
2416 break;
2419 case 0:
2420 /* Don't make yyparse think this is eof. */
2421 value = 1;
2422 break;
2424 case '{':
2425 indent_level++;
2426 value = c;
2427 break;
2429 case '}':
2430 indent_level--;
2431 value = c;
2432 break;
2434 default:
2435 value = c;
2438 done:
2439 /* yylloc.last_line = lineno; */
2441 return value;
2444 /* Sets the value of the 'yydebug' variable to VALUE.
2445 This is a function so we don't have to have YYDEBUG defined
2446 in order to build the compiler. */
2448 void
2449 set_yydebug (value)
2450 int value;
2452 #if YYDEBUG != 0
2453 yydebug = value;
2454 #else
2455 warning ("YYDEBUG not defined.");
2456 #endif