oops - omitted from previous delta fixing UNIQUE_SECTION
[official-gcc.git] / gcc / c-lex.c
blob1d271b0e954e6f248ecb85f5ed06f51e5ad774e1
1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 88, 89, 92, 94-99, 2000 Free Software Foundation, Inc.
4 This file is part of GNU CC.
6 GNU CC is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2, or (at your option)
9 any later version.
11 GNU CC is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with GNU CC; see the file COPYING. If not, write to
18 the Free Software Foundation, 59 Temple Place - Suite 330,
19 Boston, MA 02111-1307, USA. */
21 #include "config.h"
22 #include "system.h"
24 #include "rtl.h"
25 #include "tree.h"
26 #include "input.h"
27 #include "output.h"
28 #include "c-lex.h"
29 #include "c-tree.h"
30 #include "flags.h"
31 #include "c-parse.h"
32 #include "c-pragma.h"
33 #include "toplev.h"
34 #include "intl.h"
35 #include "ggc.h"
36 #include "tm_p.h"
38 /* MULTIBYTE_CHARS support only works for native compilers.
39 ??? Ideally what we want is to model widechar support after
40 the current floating point support. */
41 #ifdef CROSS_COMPILE
42 #undef MULTIBYTE_CHARS
43 #endif
45 #ifdef MULTIBYTE_CHARS
46 #include "mbchar.h"
47 #include <locale.h>
48 #endif /* MULTIBYTE_CHARS */
49 #ifndef GET_ENVIRONMENT
50 #define GET_ENVIRONMENT(ENV_VALUE,ENV_NAME) ((ENV_VALUE) = getenv (ENV_NAME))
51 #endif
53 #if USE_CPPLIB
54 #include "cpplib.h"
55 extern cpp_reader parse_in;
56 extern cpp_options parse_options;
57 #else
58 /* Stream for reading from the input file. */
59 FILE *finput;
60 #endif
62 extern void yyprint PARAMS ((FILE *, int, YYSTYPE));
64 /* The elements of `ridpointers' are identifier nodes
65 for the reserved type names and storage classes.
66 It is indexed by a RID_... value. */
67 tree ridpointers[(int) RID_MAX];
69 /* Cause the `yydebug' variable to be defined. */
70 #define YYDEBUG 1
72 #if USE_CPPLIB
73 extern unsigned char *yy_cur, *yy_lim;
74 extern enum cpp_token cpp_token;
76 extern int yy_get_token ();
78 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
79 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
81 #else /* ! USE_CPPLIB */
83 #define GETC() getch ()
84 #define UNGETC(c) put_back (c)
86 struct putback_buffer {
87 char *buffer;
88 int buffer_size;
89 int index;
92 static struct putback_buffer putback = {NULL, 0, -1};
94 static inline int getch PARAMS ((void));
96 static inline int
97 getch ()
99 if (putback.index != -1)
101 int ch = putback.buffer[putback.index];
102 --putback.index;
103 return ch;
105 return getc (finput);
108 static inline void put_back PARAMS ((int));
110 static inline void
111 put_back (ch)
112 int ch;
114 if (ch != EOF)
116 if (putback.index == putback.buffer_size - 1)
118 putback.buffer_size += 16;
119 putback.buffer = xrealloc (putback.buffer, putback.buffer_size);
121 putback.buffer[++putback.index] = ch;
124 #endif /* ! USE_CPPLIB */
126 int linemode;
128 /* the declaration found for the last IDENTIFIER token read in.
129 yylex must look this up to detect typedefs, which get token type TYPENAME,
130 so it is left around in case the identifier is not a typedef but is
131 used in a context which makes it a reference to a variable. */
132 tree lastiddecl;
134 /* Nonzero enables objc features. */
136 int doing_objc_thang;
138 extern int yydebug;
140 /* File used for outputting assembler code. */
141 extern FILE *asm_out_file;
143 #undef WCHAR_TYPE_SIZE
144 #define WCHAR_TYPE_SIZE TYPE_PRECISION (wchar_type_node)
146 /* Number of bytes in a wide character. */
147 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
149 static int maxtoken; /* Current nominal length of token buffer. */
150 char *token_buffer; /* Pointer to token buffer.
151 Actual allocated length is maxtoken + 2.
152 This is not static because objc-parse.y uses it. */
154 static int indent_level; /* Number of { minus number of }. */
156 /* Nonzero tells yylex to ignore \ in string constants. */
157 static int ignore_escape_flag;
159 /* Nonzero if end-of-file has been seen on input. */
160 static int end_of_file;
162 #ifdef HANDLE_GENERIC_PRAGMAS
163 static int handle_generic_pragma PARAMS ((int));
164 #endif /* HANDLE_GENERIC_PRAGMAS */
165 static int whitespace_cr PARAMS ((int));
166 static int skip_white_space PARAMS ((int));
167 static char *extend_token_buffer PARAMS ((const char *));
168 static int readescape PARAMS ((int *));
169 static void parse_float PARAMS ((PTR));
170 static void extend_token_buffer_to PARAMS ((int));
171 static int read_line_number PARAMS ((int *));
173 /* Do not insert generated code into the source, instead, include it.
174 This allows us to build gcc automatically even for targets that
175 need to add or modify the reserved keyword lists. */
176 #include "c-gperf.h"
178 /* Return something to represent absolute declarators containing a *.
179 TARGET is the absolute declarator that the * contains.
180 TYPE_QUALS is a list of modifiers such as const or volatile
181 to apply to the pointer type, represented as identifiers.
183 We return an INDIRECT_REF whose "contents" are TARGET
184 and whose type is the modifier list. */
186 tree
187 make_pointer_declarator (type_quals, target)
188 tree type_quals, target;
190 return build1 (INDIRECT_REF, type_quals, target);
193 void
194 forget_protocol_qualifiers ()
196 int i, n = sizeof wordlist / sizeof (struct resword);
198 for (i = 0; i < n; i++)
199 if ((int) wordlist[i].rid >= (int) RID_IN
200 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
201 wordlist[i].name = "";
204 void
205 remember_protocol_qualifiers ()
207 int i, n = sizeof wordlist / sizeof (struct resword);
209 for (i = 0; i < n; i++)
210 if (wordlist[i].rid == RID_IN)
211 wordlist[i].name = "in";
212 else if (wordlist[i].rid == RID_OUT)
213 wordlist[i].name = "out";
214 else if (wordlist[i].rid == RID_INOUT)
215 wordlist[i].name = "inout";
216 else if (wordlist[i].rid == RID_BYCOPY)
217 wordlist[i].name = "bycopy";
218 else if (wordlist[i].rid == RID_BYREF)
219 wordlist[i].name = "byref";
220 else if (wordlist[i].rid == RID_ONEWAY)
221 wordlist[i].name = "oneway";
224 char *
225 init_parse (filename)
226 char *filename;
228 #if !USE_CPPLIB
229 /* Open input file. */
230 if (filename == 0 || !strcmp (filename, "-"))
232 finput = stdin;
233 filename = "stdin";
235 else
236 finput = fopen (filename, "r");
237 if (finput == 0)
238 pfatal_with_name (filename);
240 #ifdef IO_BUFFER_SIZE
241 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
242 #endif
243 #else /* !USE_CPPLIB */
244 parse_in.show_column = 1;
245 if (! cpp_start_read (&parse_in, filename))
246 abort ();
248 if (filename == 0 || !strcmp (filename, "-"))
249 filename = "stdin";
251 /* cpp_start_read always puts at least one line directive into the
252 token buffer. We must arrange to read it out here. */
253 yy_cur = parse_in.token_buffer;
254 yy_lim = CPP_PWRITTEN (&parse_in);
255 cpp_token = CPP_DIRECTIVE;
256 #endif
258 init_lex ();
259 init_pragma ();
261 return filename;
264 void
265 finish_parse ()
267 #if USE_CPPLIB
268 cpp_finish (&parse_in);
269 errorcount += parse_in.errors;
270 #else
271 fclose (finput);
272 #endif
275 void
276 init_lex ()
278 /* Make identifier nodes long enough for the language-specific slots. */
279 set_identifier_size (sizeof (struct lang_identifier));
281 /* Start it at 0, because check_newline is called at the very beginning
282 and will increment it to 1. */
283 lineno = 0;
285 #ifdef MULTIBYTE_CHARS
286 /* Change to the native locale for multibyte conversions. */
287 setlocale (LC_CTYPE, "");
288 GET_ENVIRONMENT (literal_codeset, "LANG");
289 #endif
291 maxtoken = 40;
292 token_buffer = (char *) xmalloc (maxtoken + 2);
294 ridpointers[(int) RID_INT] = get_identifier ("int");
295 ridpointers[(int) RID_CHAR] = get_identifier ("char");
296 ridpointers[(int) RID_VOID] = get_identifier ("void");
297 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
298 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
299 ridpointers[(int) RID_SHORT] = get_identifier ("short");
300 ridpointers[(int) RID_LONG] = get_identifier ("long");
301 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
302 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
303 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
304 ridpointers[(int) RID_CONST] = get_identifier ("const");
305 ridpointers[(int) RID_RESTRICT] = get_identifier ("restrict");
306 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
307 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
308 ridpointers[(int) RID_STATIC] = get_identifier ("static");
309 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
310 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
311 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
312 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
313 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
314 ridpointers[(int) RID_ID] = get_identifier ("id");
315 ridpointers[(int) RID_IN] = get_identifier ("in");
316 ridpointers[(int) RID_OUT] = get_identifier ("out");
317 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
318 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
319 ridpointers[(int) RID_BYREF] = get_identifier ("byref");
320 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
321 forget_protocol_qualifiers();
323 /* Some options inhibit certain reserved words.
324 Clear those words out of the hash table so they won't be recognized. */
325 #define UNSET_RESERVED_WORD(STRING) \
326 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
327 if (s) s->name = ""; } while (0)
329 if (! doing_objc_thang)
330 UNSET_RESERVED_WORD ("id");
332 if (flag_traditional)
334 UNSET_RESERVED_WORD ("const");
335 UNSET_RESERVED_WORD ("restrict");
336 UNSET_RESERVED_WORD ("volatile");
337 UNSET_RESERVED_WORD ("typeof");
338 UNSET_RESERVED_WORD ("signed");
339 UNSET_RESERVED_WORD ("inline");
340 UNSET_RESERVED_WORD ("iterator");
341 UNSET_RESERVED_WORD ("complex");
343 else if (!flag_isoc99)
344 UNSET_RESERVED_WORD ("restrict");
346 if (flag_no_asm)
348 UNSET_RESERVED_WORD ("asm");
349 UNSET_RESERVED_WORD ("typeof");
350 UNSET_RESERVED_WORD ("inline");
351 UNSET_RESERVED_WORD ("iterator");
352 UNSET_RESERVED_WORD ("complex");
356 void
357 reinit_parse_for_function ()
361 /* Function used when yydebug is set, to print a token in more detail. */
363 void
364 yyprint (file, yychar, yylval)
365 FILE *file;
366 int yychar;
367 YYSTYPE yylval;
369 tree t;
370 switch (yychar)
372 case IDENTIFIER:
373 case TYPENAME:
374 case OBJECTNAME:
375 t = yylval.ttype;
376 if (IDENTIFIER_POINTER (t))
377 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
378 break;
380 case CONSTANT:
381 t = yylval.ttype;
382 if (TREE_CODE (t) == INTEGER_CST)
383 fprintf (file,
384 #if HOST_BITS_PER_WIDE_INT == 64
385 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
386 " 0x%x%016x",
387 #else
388 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
389 " 0x%lx%016lx",
390 #else
391 " 0x%llx%016llx",
392 #endif
393 #endif
394 #else
395 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
396 " 0x%lx%08lx",
397 #else
398 " 0x%x%08x",
399 #endif
400 #endif
401 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
402 break;
406 /* Iff C is a carriage return, warn about it - if appropriate -
407 and return nonzero. */
408 static int
409 whitespace_cr (c)
410 int c;
412 static int newline_warning = 0;
414 if (c == '\r')
416 /* ANSI C says the effects of a carriage return in a source file
417 are undefined. */
418 if (pedantic && !newline_warning)
420 warning ("carriage return in source file");
421 warning ("(we only warn about the first carriage return)");
422 newline_warning = 1;
424 return 1;
426 return 0;
429 /* If C is not whitespace, return C.
430 Otherwise skip whitespace and return first nonwhite char read. */
432 static int
433 skip_white_space (c)
434 register int c;
436 for (;;)
438 switch (c)
440 /* We don't recognize comments here, because
441 cpp output can include / and * consecutively as operators.
442 Also, there's no need, since cpp removes all comments. */
444 case '\n':
445 if (linemode)
447 UNGETC (c);
448 return EOF;
450 c = check_newline ();
451 break;
453 case ' ':
454 case '\t':
455 case '\f':
456 case '\v':
457 case '\b':
458 #if USE_CPPLIB
459 /* While processing a # directive we don't get CPP_HSPACE
460 tokens, so we also need to handle whitespace the normal way. */
461 if (cpp_token == CPP_HSPACE)
462 c = yy_get_token ();
463 else
464 #endif
465 c = GETC();
466 break;
468 case '\r':
469 whitespace_cr (c);
470 c = GETC();
471 break;
473 case '\\':
474 c = GETC();
475 if (c == '\n')
476 lineno++;
477 else
478 error ("stray '\\' in program");
479 c = GETC();
480 break;
482 default:
483 return (c);
488 /* Skips all of the white space at the current location in the input file. */
490 void
491 position_after_white_space ()
493 register int c;
495 c = GETC();
497 UNGETC (skip_white_space (c));
500 /* Make the token buffer longer, preserving the data in it.
501 P should point to just beyond the last valid character in the old buffer.
502 The value we return is a pointer to the new buffer
503 at a place corresponding to P. */
505 static void
506 extend_token_buffer_to (size)
507 int size;
510 maxtoken = maxtoken * 2 + 10;
511 while (maxtoken < size);
512 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
515 static char *
516 extend_token_buffer (p)
517 const char *p;
519 int offset = p - token_buffer;
520 extend_token_buffer_to (offset);
521 return token_buffer + offset;
524 #if defined HANDLE_PRAGMA
525 /* Local versions of these macros, that can be passed as function pointers. */
526 static int
527 pragma_getc ()
529 return GETC ();
532 static void
533 pragma_ungetc (arg)
534 int arg;
536 UNGETC (arg);
538 #endif
540 static int
541 read_line_number (num)
542 int *num;
544 register int token = yylex ();
546 if (token == CONSTANT
547 && TREE_CODE (yylval.ttype) == INTEGER_CST)
549 *num = TREE_INT_CST_LOW (yylval.ttype);
550 return 1;
552 else
554 if (token != END_OF_LINE)
555 error ("invalid #-line");
556 return 0;
560 /* At the beginning of a line, increment the line number
561 and process any #-directive on this line.
562 If the line is a #-directive, read the entire line and return a newline.
563 Otherwise, return the line's first non-whitespace character.
565 Note that in the case of USE_CPPLIB, we get the whole line as one
566 CPP_DIRECTIVE token. */
569 check_newline ()
571 register int c;
572 register int token;
573 int saw_line;
574 enum { act_none, act_push, act_pop } action;
575 int old_lineno, action_number, l;
577 restart:
578 /* Read first nonwhite char on the line. */
580 #ifdef USE_CPPLIB
581 c = GETC ();
582 /* In some cases where we're leaving an include file, we can get multiple
583 CPP_HSPACE tokens in a row, so we need to loop. */
584 while (cpp_token == CPP_HSPACE)
585 c = yy_get_token ();
586 #else
588 c = GETC ();
589 while (c == ' ' || c == '\t');
590 #endif
592 lineno++;
594 if (c != '#')
596 /* Sequences of multiple newlines are very common; optimize them. */
597 if (c == '\n')
598 goto restart;
600 /* If not #, return it so caller will use it. */
601 return c;
604 /* Don't read beyond this line. */
605 saw_line = 0;
606 linemode = 1;
608 #if USE_CPPLIB
609 if (cpp_token == CPP_VSPACE)
611 /* Format is "<space> <line number> <filename> <newline>".
612 Only the line number is interesting, and even that
613 we can get more efficiently than scanning the line. */
614 yy_cur = yy_lim - 1;
615 lineno = parse_in.lineno - 1;
616 goto skipline;
618 #endif
620 token = yylex ();
622 if (token == IDENTIFIER)
624 /* If a letter follows, then if the word here is `line', skip
625 it and ignore it; otherwise, ignore the line, with an error
626 if the word isn't `pragma'. */
628 const char *name = IDENTIFIER_POINTER (yylval.ttype);
630 if (!strcmp (name, "pragma"))
632 token = yylex ();
633 if (token != IDENTIFIER
634 || TREE_CODE (yylval.ttype) != IDENTIFIER_NODE)
635 goto skipline;
637 #ifdef HANDLE_PRAGMA
638 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS
639 (if both are defined), in order to give the back
640 end a chance to override the interpretation of
641 SYSV style pragmas. */
642 if (HANDLE_PRAGMA (pragma_getc, pragma_ungetc,
643 IDENTIFIER_POINTER (yylval.ttype)))
644 goto skipline;
645 #endif /* HANDLE_PRAGMA */
647 #ifdef HANDLE_GENERIC_PRAGMAS
648 if (handle_generic_pragma (token))
649 goto skipline;
650 #endif /* HANDLE_GENERIC_PRAGMAS */
652 /* Issue a warning message if we have been asked to do so.
653 Ignoring unknown pragmas in system header file unless
654 an explcit -Wunknown-pragmas has been given. */
655 if (warn_unknown_pragmas > 1
656 || (warn_unknown_pragmas && ! in_system_header))
657 warning ("ignoring pragma: %s", token_buffer);
659 goto skipline;
661 else if (!strcmp (name, "define"))
663 debug_define (lineno, GET_DIRECTIVE_LINE ());
664 goto skipline;
666 else if (!strcmp (name, "undef"))
668 debug_undef (lineno, GET_DIRECTIVE_LINE ());
669 goto skipline;
671 else if (!strcmp (name, "line"))
673 saw_line = 1;
674 token = yylex ();
675 goto linenum;
677 else if (!strcmp (name, "ident"))
679 /* #ident. The pedantic warning is now in cccp.c. */
681 /* Here we have just seen `#ident '.
682 A string constant should follow. */
684 token = yylex ();
685 if (token == END_OF_LINE)
686 goto skipline;
687 if (token != STRING
688 || TREE_CODE (yylval.ttype) != STRING_CST)
690 error ("invalid #ident");
691 goto skipline;
694 if (! flag_no_ident)
696 #ifdef ASM_OUTPUT_IDENT
697 ASM_OUTPUT_IDENT (asm_out_file,
698 TREE_STRING_POINTER (yylval.ttype));
699 #endif
702 /* Skip the rest of this line. */
703 goto skipline;
706 error ("undefined or invalid # directive `%s'", name);
707 goto skipline;
710 /* If the # is the only nonwhite char on the line,
711 just ignore it. Check the new newline. */
712 if (token == END_OF_LINE)
713 goto skipline;
715 linenum:
716 /* Here we have either `#line' or `# <nonletter>'.
717 In either case, it should be a line number; a digit should follow. */
719 if (token != CONSTANT
720 || TREE_CODE (yylval.ttype) != INTEGER_CST)
722 error ("invalid #-line");
723 goto skipline;
726 /* subtract one, because it is the following line that
727 gets the specified number */
729 l = TREE_INT_CST_LOW (yylval.ttype) - 1;
731 /* More follows: it must be a string constant (filename).
732 It would be neat to use cpplib to quickly process the string, but
733 (1) we don't have a handy tokenization of the string, and
734 (2) I don't know how well that would work in the presense
735 of filenames that contain wide characters. */
737 if (saw_line)
739 /* Don't treat \ as special if we are processing #line 1 "...".
740 If you want it to be treated specially, use # 1 "...". */
741 ignore_escape_flag = 1;
744 /* Read the string constant. */
745 token = yylex ();
747 ignore_escape_flag = 0;
749 if (token == END_OF_LINE)
751 /* No more: store the line number and check following line. */
752 lineno = l;
753 goto skipline;
756 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
758 error ("invalid #line");
759 goto skipline;
762 input_filename = TREE_STRING_POINTER (yylval.ttype);
764 if (main_input_filename == 0)
765 main_input_filename = input_filename;
767 old_lineno = lineno;
768 action = act_none;
769 action_number = 0;
770 lineno = l;
772 /* Each change of file name
773 reinitializes whether we are now in a system header. */
774 in_system_header = 0;
776 if (!read_line_number (&action_number))
778 /* Update the name in the top element of input_file_stack. */
779 if (input_file_stack)
780 input_file_stack->name = input_filename;
783 /* `1' after file name means entering new file.
784 `2' after file name means just left a file. */
786 if (action_number == 1)
788 action = act_push;
789 read_line_number (&action_number);
791 else if (action_number == 2)
793 action = act_pop;
794 read_line_number (&action_number);
796 if (action_number == 3)
798 /* `3' after file name means this is a system header file. */
799 in_system_header = 1;
800 read_line_number (&action_number);
803 /* Do the actions implied by the preceding numbers. */
805 if (action == act_push)
807 /* Pushing to a new file. */
808 struct file_stack *p
809 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
810 input_file_stack->line = old_lineno;
811 p->next = input_file_stack;
812 p->name = input_filename;
813 p->indent_level = indent_level;
814 input_file_stack = p;
815 input_file_stack_tick++;
816 debug_start_source_file (input_filename);
818 else if (action == act_pop)
820 /* Popping out of a file. */
821 if (input_file_stack->next)
823 struct file_stack *p = input_file_stack;
824 if (indent_level != p->indent_level)
826 warning_with_file_and_line
827 (p->name, old_lineno,
828 "This file contains more `%c's than `%c's.",
829 indent_level > p->indent_level ? '{' : '}',
830 indent_level > p->indent_level ? '}' : '{');
832 input_file_stack = p->next;
833 free (p);
834 input_file_stack_tick++;
835 debug_end_source_file (input_file_stack->line);
837 else
838 error ("#-lines for entering and leaving files don't match");
841 /* Now that we've pushed or popped the input stack,
842 update the name in the top element. */
843 if (input_file_stack)
844 input_file_stack->name = input_filename;
846 /* skip the rest of this line. */
847 skipline:
848 linemode = 0;
849 end_of_file = 0;
852 c = GETC();
853 while (c != '\n' && c != EOF);
854 return c;
857 #ifdef HANDLE_GENERIC_PRAGMAS
859 /* Handle a #pragma directive.
860 TOKEN is the token we read after `#pragma'. Processes the entire input
861 line and return non-zero iff the pragma has been successfully parsed. */
863 /* This function has to be in this file, in order to get at
864 the token types. */
866 static int
867 handle_generic_pragma (token)
868 register int token;
870 for (;;)
872 switch (token)
874 case IDENTIFIER:
875 case TYPENAME:
876 case STRING:
877 case CONSTANT:
878 handle_pragma_token (token_buffer, yylval.ttype);
879 break;
881 case END_OF_LINE:
882 return handle_pragma_token (NULL_PTR, NULL_TREE);
884 default:
885 handle_pragma_token (token_buffer, NULL);
888 token = yylex ();
892 #endif /* HANDLE_GENERIC_PRAGMAS */
894 #define ENDFILE -1 /* token that represents end-of-file */
896 /* Read an escape sequence, returning its equivalent as a character,
897 or store 1 in *ignore_ptr if it is backslash-newline. */
899 static int
900 readescape (ignore_ptr)
901 int *ignore_ptr;
903 register int c = GETC();
904 register int code;
905 register unsigned count;
906 unsigned firstdig = 0;
907 int nonnull;
909 switch (c)
911 case 'x':
912 if (warn_traditional)
913 warning ("the meaning of `\\x' varies with -traditional");
915 if (flag_traditional)
916 return c;
918 code = 0;
919 count = 0;
920 nonnull = 0;
921 while (1)
923 c = GETC();
924 if (! ISXDIGIT (c))
926 UNGETC (c);
927 break;
929 code *= 16;
930 if (c >= 'a' && c <= 'f')
931 code += c - 'a' + 10;
932 if (c >= 'A' && c <= 'F')
933 code += c - 'A' + 10;
934 if (c >= '0' && c <= '9')
935 code += c - '0';
936 if (code != 0 || count != 0)
938 if (count == 0)
939 firstdig = code;
940 count++;
942 nonnull = 1;
944 if (! nonnull)
945 error ("\\x used with no following hex digits");
946 else if (count == 0)
947 /* Digits are all 0's. Ok. */
949 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
950 || (count > 1
951 && (((unsigned)1
952 << (TYPE_PRECISION (integer_type_node)
953 - (count - 1) * 4))
954 <= firstdig)))
955 pedwarn ("hex escape out of range");
956 return code;
958 case '0': case '1': case '2': case '3': case '4':
959 case '5': case '6': case '7':
960 code = 0;
961 count = 0;
962 while ((c <= '7') && (c >= '0') && (count++ < 3))
964 code = (code * 8) + (c - '0');
965 c = GETC();
967 UNGETC (c);
968 return code;
970 case '\\': case '\'': case '"':
971 return c;
973 case '\n':
974 lineno++;
975 *ignore_ptr = 1;
976 return 0;
978 case 'n':
979 return TARGET_NEWLINE;
981 case 't':
982 return TARGET_TAB;
984 case 'r':
985 return TARGET_CR;
987 case 'f':
988 return TARGET_FF;
990 case 'b':
991 return TARGET_BS;
993 case 'a':
994 if (warn_traditional)
995 warning ("the meaning of `\\a' varies with -traditional");
997 if (flag_traditional)
998 return c;
999 return TARGET_BELL;
1001 case 'v':
1002 #if 0 /* Vertical tab is present in common usage compilers. */
1003 if (flag_traditional)
1004 return c;
1005 #endif
1006 return TARGET_VT;
1008 case 'e':
1009 case 'E':
1010 if (pedantic)
1011 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1012 return TARGET_ESC;
1014 case '?':
1015 return c;
1017 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1018 case '(':
1019 case '{':
1020 case '[':
1021 /* `\%' is used to prevent SCCS from getting confused. */
1022 case '%':
1023 if (pedantic)
1024 pedwarn ("unknown escape sequence `\\%c'", c);
1025 return c;
1027 if (ISGRAPH (c))
1028 pedwarn ("unknown escape sequence `\\%c'", c);
1029 else
1030 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1031 return c;
1034 void
1035 yyerror (msgid)
1036 const char *msgid;
1038 const char *string = _(msgid);
1040 /* We can't print string and character constants well
1041 because the token_buffer contains the result of processing escapes. */
1042 if (end_of_file)
1043 error ("%s at end of input", string);
1044 else if (token_buffer[0] == 0)
1045 error ("%s at null character", string);
1046 else if (token_buffer[0] == '"')
1047 error ("%s before string constant", string);
1048 else if (token_buffer[0] == '\'')
1049 error ("%s before character constant", string);
1050 else if (!ISGRAPH(token_buffer[0]))
1051 error ("%s before character 0%o", string, (unsigned char) token_buffer[0]);
1052 else
1053 error ("%s before `%s'", string, token_buffer);
1056 #if 0
1058 struct try_type
1060 tree *node_var;
1061 char unsigned_flag;
1062 char long_flag;
1063 char long_long_flag;
1066 struct try_type type_sequence[] =
1068 { &integer_type_node, 0, 0, 0},
1069 { &unsigned_type_node, 1, 0, 0},
1070 { &long_integer_type_node, 0, 1, 0},
1071 { &long_unsigned_type_node, 1, 1, 0},
1072 { &long_long_integer_type_node, 0, 1, 1},
1073 { &long_long_unsigned_type_node, 1, 1, 1}
1075 #endif /* 0 */
1077 struct pf_args
1079 /* Input */
1080 int base;
1081 char * p;
1082 /* I/O */
1083 int c;
1084 /* Output */
1085 int imag;
1086 tree type;
1087 int conversion_errno;
1088 REAL_VALUE_TYPE value;
1091 static void
1092 parse_float (data)
1093 PTR data;
1095 struct pf_args * args = (struct pf_args *) data;
1096 int fflag = 0, lflag = 0;
1097 /* Copy token_buffer now, while it has just the number
1098 and not the suffixes; once we add `f' or `i',
1099 REAL_VALUE_ATOF may not work any more. */
1100 char *copy = (char *) alloca (args->p - token_buffer + 1);
1101 bcopy (token_buffer, copy, args->p - token_buffer + 1);
1102 args->imag = 0;
1103 args->conversion_errno = 0;
1104 args->type = double_type_node;
1106 while (1)
1108 int lose = 0;
1110 /* Read the suffixes to choose a data type. */
1111 switch (args->c)
1113 case 'f': case 'F':
1114 if (fflag)
1115 error ("more than one `f' in numeric constant");
1116 fflag = 1;
1117 break;
1119 case 'l': case 'L':
1120 if (lflag)
1121 error ("more than one `l' in numeric constant");
1122 lflag = 1;
1123 break;
1125 case 'i': case 'I':
1126 if (args->imag)
1127 error ("more than one `i' or `j' in numeric constant");
1128 else if (pedantic)
1129 pedwarn ("ANSI C forbids imaginary numeric constants");
1130 args->imag = 1;
1131 break;
1133 default:
1134 lose = 1;
1137 if (lose)
1138 break;
1140 if (args->p >= token_buffer + maxtoken - 3)
1141 args->p = extend_token_buffer (args->p);
1142 *(args->p++) = args->c;
1143 *(args->p) = 0;
1144 args->c = GETC();
1147 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1148 tells the desired precision of the binary result
1149 of decimal-to-binary conversion. */
1151 if (fflag)
1153 if (lflag)
1154 error ("both `f' and `l' in floating constant");
1156 args->type = float_type_node;
1157 errno = 0;
1158 if (args->base == 16)
1159 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1160 else
1161 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1162 args->conversion_errno = errno;
1163 /* A diagnostic is required here by some ANSI C testsuites.
1164 This is not pedwarn, because some people don't want
1165 an error for this. */
1166 if (REAL_VALUE_ISINF (args->value) && pedantic)
1167 warning ("floating point number exceeds range of `float'");
1169 else if (lflag)
1171 args->type = long_double_type_node;
1172 errno = 0;
1173 if (args->base == 16)
1174 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1175 else
1176 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1177 args->conversion_errno = errno;
1178 if (REAL_VALUE_ISINF (args->value) && pedantic)
1179 warning ("floating point number exceeds range of `long double'");
1181 else
1183 errno = 0;
1184 if (args->base == 16)
1185 args->value = REAL_VALUE_HTOF (copy, TYPE_MODE (args->type));
1186 else
1187 args->value = REAL_VALUE_ATOF (copy, TYPE_MODE (args->type));
1188 args->conversion_errno = errno;
1189 if (REAL_VALUE_ISINF (args->value) && pedantic)
1190 warning ("floating point number exceeds range of `double'");
1194 /* Get the next character, staying within the current token if possible.
1195 If we're lexing a token, we don't want to look beyond the end of the
1196 token cpplib has prepared for us; otherwise, we end up reading in the
1197 next token, which screws up feed_input. So just return a null
1198 character. */
1200 static inline int token_getch PARAMS ((void));
1202 static inline int
1203 token_getch ()
1205 #if USE_CPPLIB
1206 if (yy_cur == yy_lim)
1207 return '\0';
1208 #endif
1209 return GETC ();
1212 static inline void token_put_back PARAMS ((int));
1214 static inline void
1215 token_put_back (ch)
1216 int ch;
1218 #if USE_CPPLIB
1219 if (ch == '\0')
1220 return;
1221 #endif
1222 UNGETC (ch);
1225 /* Read a single token from the input stream, and assign it lexical
1226 semantics. */
1229 yylex ()
1231 register int c;
1232 register char *p;
1233 register int value;
1234 int wide_flag = 0;
1235 int objc_flag = 0;
1237 c = GETC();
1239 /* Effectively do c = skip_white_space (c)
1240 but do it faster in the usual cases. */
1241 while (1)
1242 switch (c)
1244 case ' ':
1245 case '\t':
1246 case '\f':
1247 case '\v':
1248 case '\b':
1249 #if USE_CPPLIB
1250 if (cpp_token == CPP_HSPACE)
1251 c = yy_get_token ();
1252 else
1253 #endif
1254 c = GETC();
1255 break;
1257 case '\r':
1258 /* Call skip_white_space so we can warn if appropriate. */
1260 case '\n':
1261 case '/':
1262 case '\\':
1263 c = skip_white_space (c);
1264 default:
1265 goto found_nonwhite;
1267 found_nonwhite:
1269 token_buffer[0] = c;
1270 token_buffer[1] = 0;
1272 /* yylloc.first_line = lineno; */
1274 switch (c)
1276 case EOF:
1277 end_of_file = 1;
1278 token_buffer[0] = 0;
1279 if (linemode)
1280 value = END_OF_LINE;
1281 else
1282 value = ENDFILE;
1283 break;
1285 case 'L':
1286 #if USE_CPPLIB
1287 if (cpp_token == CPP_NAME)
1288 goto letter;
1289 #endif
1290 /* Capital L may start a wide-string or wide-character constant. */
1292 register int c = token_getch();
1293 if (c == '\'')
1295 wide_flag = 1;
1296 goto char_constant;
1298 if (c == '"')
1300 wide_flag = 1;
1301 goto string_constant;
1303 token_put_back (c);
1305 goto letter;
1307 case '@':
1308 if (!doing_objc_thang)
1310 value = c;
1311 break;
1313 else
1315 /* '@' may start a constant string object. */
1316 register int c = token_getch ();
1317 if (c == '"')
1319 objc_flag = 1;
1320 goto string_constant;
1322 token_put_back (c);
1323 /* Fall through to treat '@' as the start of an identifier. */
1326 case 'A': case 'B': case 'C': case 'D': case 'E':
1327 case 'F': case 'G': case 'H': case 'I': case 'J':
1328 case 'K': case 'M': case 'N': case 'O':
1329 case 'P': case 'Q': case 'R': case 'S': case 'T':
1330 case 'U': case 'V': case 'W': case 'X': case 'Y':
1331 case 'Z':
1332 case 'a': case 'b': case 'c': case 'd': case 'e':
1333 case 'f': case 'g': case 'h': case 'i': case 'j':
1334 case 'k': case 'l': case 'm': case 'n': case 'o':
1335 case 'p': case 'q': case 'r': case 's': case 't':
1336 case 'u': case 'v': case 'w': case 'x': case 'y':
1337 case 'z':
1338 case '_':
1339 case '$':
1340 letter:
1341 #if USE_CPPLIB
1342 if (cpp_token == CPP_NAME)
1344 /* Note that one character has already been read from
1345 yy_cur into token_buffer. Also, cpplib complains about
1346 $ in identifiers, so we don't have to. */
1348 int len = yy_lim - yy_cur + 1;
1349 if (len >= maxtoken)
1350 extend_token_buffer_to (len + 1);
1351 memcpy (token_buffer + 1, yy_cur, len);
1352 p = token_buffer + len;
1353 yy_cur = yy_lim;
1355 else
1356 #endif
1358 p = token_buffer;
1359 while (ISALNUM (c) || c == '_' || c == '$' || c == '@')
1361 /* Make sure this char really belongs in an identifier. */
1362 if (c == '$')
1364 if (! dollars_in_ident)
1365 error ("`$' in identifier");
1366 else if (pedantic)
1367 pedwarn ("`$' in identifier");
1370 if (p >= token_buffer + maxtoken)
1371 p = extend_token_buffer (p);
1373 *p++ = c;
1374 c = token_getch();
1377 *p = 0;
1378 token_put_back (c);
1381 value = IDENTIFIER;
1382 yylval.itype = 0;
1384 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1387 register struct resword *ptr;
1389 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1391 if (ptr->rid)
1392 yylval.ttype = ridpointers[(int) ptr->rid];
1393 value = (int) ptr->token;
1395 /* Only return OBJECTNAME if it is a typedef. */
1396 if (doing_objc_thang && value == OBJECTNAME)
1398 lastiddecl = lookup_name(yylval.ttype);
1400 if (lastiddecl == NULL_TREE
1401 || TREE_CODE (lastiddecl) != TYPE_DECL)
1402 value = IDENTIFIER;
1405 /* Even if we decided to recognize asm, still perhaps warn. */
1406 if (pedantic
1407 && (value == ASM_KEYWORD || value == TYPEOF
1408 || ptr->rid == RID_INLINE)
1409 && token_buffer[0] != '_')
1410 pedwarn ("ANSI does not permit the keyword `%s'",
1411 token_buffer);
1415 /* If we did not find a keyword, look for an identifier
1416 (or a typename). */
1418 if (value == IDENTIFIER)
1420 if (token_buffer[0] == '@')
1421 error("invalid identifier `%s'", token_buffer);
1423 yylval.ttype = get_identifier (token_buffer);
1424 lastiddecl = lookup_name (yylval.ttype);
1426 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == 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 (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1432 && DECL_IGNORED_P (lastiddecl)
1433 && TREE_READONLY (lastiddecl)
1434 && DECL_INITIAL (lastiddecl) != 0
1435 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1437 tree stringval = DECL_INITIAL (lastiddecl);
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)
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')
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 && ! in_system_header && warn_long_long)
1784 pedwarn ("ANSI C forbids long long integer constants");
1785 spec_long_long = 1;
1787 spec_long = 1;
1789 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1791 if (spec_imag)
1792 error ("more than one `i' or `j' in numeric constant");
1793 else if (pedantic)
1794 pedwarn ("ANSI C forbids imaginary numeric constants");
1795 spec_imag = 1;
1797 else
1798 break;
1799 if (p >= token_buffer + maxtoken - 3)
1800 p = extend_token_buffer (p);
1801 *p++ = c;
1802 c = token_getch();
1805 /* If the literal overflowed, pedwarn about it now. */
1806 if (overflow)
1808 warn = 1;
1809 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT * 2);
1812 /* This is simplified by the fact that our constant
1813 is always positive. */
1815 high = low = 0;
1817 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1819 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1820 / HOST_BITS_PER_CHAR)]
1821 << (i * HOST_BITS_PER_CHAR));
1822 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1825 yylval.ttype = build_int_2 (low, high);
1826 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1828 /* If warn_traditional, calculate both the ANSI type and the
1829 traditional type, then see if they disagree.
1830 Otherwise, calculate only the type for the dialect in use. */
1831 if (warn_traditional || flag_traditional)
1833 /* Calculate the traditional type. */
1834 /* Traditionally, any constant is signed;
1835 but if unsigned is specified explicitly, obey that.
1836 Use the smallest size with the right number of bits,
1837 except for one special case with decimal constants. */
1838 if (! spec_long && base != 10
1839 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1840 traditional_type = (spec_unsigned ? unsigned_type_node
1841 : integer_type_node);
1842 /* A decimal constant must be long
1843 if it does not fit in type int.
1844 I think this is independent of whether
1845 the constant is signed. */
1846 else if (! spec_long && base == 10
1847 && int_fits_type_p (yylval.ttype, integer_type_node))
1848 traditional_type = (spec_unsigned ? unsigned_type_node
1849 : integer_type_node);
1850 else if (! spec_long_long)
1851 traditional_type = (spec_unsigned ? long_unsigned_type_node
1852 : long_integer_type_node);
1853 else if (int_fits_type_p (yylval.ttype,
1854 spec_unsigned
1855 ? long_long_unsigned_type_node
1856 : long_long_integer_type_node))
1857 traditional_type = (spec_unsigned
1858 ? long_long_unsigned_type_node
1859 : long_long_integer_type_node);
1860 else
1861 traditional_type = (spec_unsigned
1862 ? widest_unsigned_literal_type_node
1863 : widest_integer_literal_type_node);
1865 if (warn_traditional || ! flag_traditional)
1867 /* Calculate the ANSI type. */
1868 if (! spec_long && ! spec_unsigned
1869 && int_fits_type_p (yylval.ttype, integer_type_node))
1870 ansi_type = integer_type_node;
1871 else if (! spec_long && (base != 10 || spec_unsigned)
1872 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1873 ansi_type = unsigned_type_node;
1874 else if (! spec_unsigned && !spec_long_long
1875 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1876 ansi_type = long_integer_type_node;
1877 else if (! spec_long_long
1878 && int_fits_type_p (yylval.ttype,
1879 long_unsigned_type_node))
1880 ansi_type = long_unsigned_type_node;
1881 else if (! spec_unsigned
1882 && int_fits_type_p (yylval.ttype,
1883 long_long_integer_type_node))
1884 ansi_type = long_long_integer_type_node;
1885 else if (int_fits_type_p (yylval.ttype,
1886 long_long_unsigned_type_node))
1887 ansi_type = long_long_unsigned_type_node;
1888 else if (! spec_unsigned
1889 && int_fits_type_p (yylval.ttype,
1890 widest_integer_literal_type_node))
1891 ansi_type = widest_integer_literal_type_node;
1892 else
1893 ansi_type = widest_unsigned_literal_type_node;
1896 type = flag_traditional ? traditional_type : ansi_type;
1898 /* We assume that constants specified in a non-decimal
1899 base are bit patterns, and that the programmer really
1900 meant what they wrote. */
1901 if (warn_traditional && base == 10
1902 && traditional_type != ansi_type)
1904 if (TYPE_PRECISION (traditional_type)
1905 != TYPE_PRECISION (ansi_type))
1906 warning ("width of integer constant changes with -traditional");
1907 else if (TREE_UNSIGNED (traditional_type)
1908 != TREE_UNSIGNED (ansi_type))
1909 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1910 else
1911 warning ("width of integer constant may change on other systems with -traditional");
1914 if (pedantic && !flag_traditional && !spec_long_long && !warn
1915 && (TYPE_PRECISION (long_integer_type_node)
1916 < TYPE_PRECISION (type)))
1918 warn = 1;
1919 pedwarn ("integer constant larger than the maximum value of an unsigned long int");
1922 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1923 warning ("decimal constant is so large that it is unsigned");
1925 if (spec_imag)
1927 if (TYPE_PRECISION (type)
1928 <= TYPE_PRECISION (integer_type_node))
1929 yylval.ttype
1930 = build_complex (NULL_TREE, integer_zero_node,
1931 convert (integer_type_node,
1932 yylval.ttype));
1933 else
1934 error ("complex integer constant is too wide for `complex int'");
1936 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1937 /* The traditional constant 0x80000000 is signed
1938 but doesn't fit in the range of int.
1939 This will change it to -0x80000000, which does fit. */
1941 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1942 yylval.ttype = convert (type, yylval.ttype);
1943 TREE_OVERFLOW (yylval.ttype)
1944 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1946 else
1947 TREE_TYPE (yylval.ttype) = type;
1950 /* If it's still an integer (not a complex), and it doesn't
1951 fit in the type we choose for it, then pedwarn. */
1953 if (! warn
1954 && TREE_CODE (TREE_TYPE (yylval.ttype)) == INTEGER_TYPE
1955 && ! int_fits_type_p (yylval.ttype, TREE_TYPE (yylval.ttype)))
1956 pedwarn ("integer constant is larger than the maximum value for its type");
1959 token_put_back (c);
1960 *p = 0;
1962 if (ISALNUM (c) || c == '.' || c == '_' || c == '$'
1963 || (!flag_traditional && (c == '-' || c == '+')
1964 && (p[-1] == 'e' || p[-1] == 'E')))
1965 error ("missing white space after number `%s'", token_buffer);
1967 value = CONSTANT; break;
1970 case '\'':
1971 char_constant:
1973 register int result = 0;
1974 register int num_chars = 0;
1975 int chars_seen = 0;
1976 unsigned width = TYPE_PRECISION (char_type_node);
1977 int max_chars;
1978 #ifdef MULTIBYTE_CHARS
1979 int longest_char = local_mb_cur_max ();
1980 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1981 #endif
1983 max_chars = TYPE_PRECISION (integer_type_node) / width;
1984 if (wide_flag)
1985 width = WCHAR_TYPE_SIZE;
1987 while (1)
1989 tryagain:
1990 c = token_getch();
1992 if (c == '\'' || c == EOF)
1993 break;
1995 ++chars_seen;
1996 if (c == '\\')
1998 int ignore = 0;
1999 c = readescape (&ignore);
2000 if (ignore)
2001 goto tryagain;
2002 if (width < HOST_BITS_PER_INT
2003 && (unsigned) c >= ((unsigned)1 << width))
2004 pedwarn ("escape sequence out of range for character");
2005 #ifdef MAP_CHARACTER
2006 if (ISPRINT (c))
2007 c = MAP_CHARACTER (c);
2008 #endif
2010 else if (c == '\n')
2012 if (pedantic)
2013 pedwarn ("ANSI C forbids newline in character constant");
2014 lineno++;
2016 else
2018 #ifdef MULTIBYTE_CHARS
2019 wchar_t wc;
2020 int i;
2021 int char_len = -1;
2022 for (i = 1; i <= longest_char; ++i)
2024 if (i > maxtoken - 4)
2025 extend_token_buffer (token_buffer);
2027 token_buffer[i] = c;
2028 char_len = local_mbtowc (& wc,
2029 token_buffer + 1,
2031 if (char_len != -1)
2032 break;
2033 c = token_getch ();
2035 if (char_len > 1)
2037 /* mbtowc sometimes needs an extra char before accepting */
2038 if (char_len < i)
2039 token_put_back (c);
2040 if (! wide_flag)
2042 /* Merge character into result; ignore excess chars. */
2043 for (i = 1; i <= char_len; ++i)
2045 if (i > max_chars)
2046 break;
2047 if (width < HOST_BITS_PER_INT)
2048 result = (result << width)
2049 | (token_buffer[i]
2050 & ((1 << width) - 1));
2051 else
2052 result = token_buffer[i];
2054 num_chars += char_len;
2055 goto tryagain;
2057 c = wc;
2059 else
2061 if (char_len == -1)
2063 warning ("Ignoring invalid multibyte character");
2064 /* Replace all but the first byte. */
2065 for (--i; i > 1; --i)
2066 token_put_back (token_buffer[i]);
2067 wc = token_buffer[1];
2069 #ifdef MAP_CHARACTER
2070 c = MAP_CHARACTER (wc);
2071 #else
2072 c = wc;
2073 #endif
2075 #else /* ! MULTIBYTE_CHARS */
2076 #ifdef MAP_CHARACTER
2077 c = MAP_CHARACTER (c);
2078 #endif
2079 #endif /* ! MULTIBYTE_CHARS */
2082 if (wide_flag)
2084 if (chars_seen == 1) /* only keep the first one */
2085 result = c;
2086 goto tryagain;
2089 /* Merge character into result; ignore excess chars. */
2090 num_chars += (width / TYPE_PRECISION (char_type_node));
2091 if (num_chars < max_chars + 1)
2093 if (width < HOST_BITS_PER_INT)
2094 result = (result << width) | (c & ((1 << width) - 1));
2095 else
2096 result = c;
2100 if (c != '\'')
2101 error ("malformed character constant");
2102 else if (chars_seen == 0)
2103 error ("empty character constant");
2104 else if (num_chars > max_chars)
2106 num_chars = max_chars;
2107 error ("character constant too long");
2109 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
2110 warning ("multi-character character constant");
2112 /* If char type is signed, sign-extend the constant. */
2113 if (! wide_flag)
2115 int num_bits = num_chars * width;
2116 if (num_bits == 0)
2117 /* We already got an error; avoid invalid shift. */
2118 yylval.ttype = build_int_2 (0, 0);
2119 else if (TREE_UNSIGNED (char_type_node)
2120 || ((result >> (num_bits - 1)) & 1) == 0)
2121 yylval.ttype
2122 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
2123 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2125 else
2126 yylval.ttype
2127 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
2128 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
2129 -1);
2130 TREE_TYPE (yylval.ttype) = integer_type_node;
2132 else
2134 yylval.ttype = build_int_2 (result, 0);
2135 TREE_TYPE (yylval.ttype) = wchar_type_node;
2138 value = CONSTANT;
2139 break;
2142 case '"':
2143 string_constant:
2145 unsigned width = wide_flag ? WCHAR_TYPE_SIZE
2146 : TYPE_PRECISION (char_type_node);
2147 #ifdef MULTIBYTE_CHARS
2148 int longest_char = local_mb_cur_max ();
2149 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
2150 #endif
2151 c = token_getch ();
2152 p = token_buffer + 1;
2154 while (c != '"' && c != EOF)
2156 /* ignore_escape_flag is set for reading the filename in #line. */
2157 if (!ignore_escape_flag && c == '\\')
2159 int ignore = 0;
2160 c = readescape (&ignore);
2161 if (ignore)
2162 goto skipnewline;
2163 if (width < HOST_BITS_PER_INT
2164 && (unsigned) c >= ((unsigned)1 << width))
2165 pedwarn ("escape sequence out of range for character");
2167 else if (c == '\n')
2169 if (pedantic)
2170 pedwarn ("ANSI C forbids newline in string constant");
2171 lineno++;
2173 else
2175 #ifdef MULTIBYTE_CHARS
2176 wchar_t wc;
2177 int i;
2178 int char_len = -1;
2179 for (i = 0; i < longest_char; ++i)
2181 if (p + i >= token_buffer + maxtoken)
2182 p = extend_token_buffer (p);
2183 p[i] = c;
2185 char_len = local_mbtowc (& wc, p, i + 1);
2186 if (char_len != -1)
2187 break;
2188 c = token_getch ();
2190 if (char_len == -1)
2192 warning ("Ignoring invalid multibyte character");
2193 /* Replace all except the first byte. */
2194 token_put_back (c);
2195 for (--i; i > 0; --i)
2196 token_put_back (p[i]);
2197 char_len = 1;
2199 /* mbtowc sometimes needs an extra char before accepting */
2200 if (char_len <= i)
2201 token_put_back (c);
2202 if (! wide_flag)
2204 p += (i + 1);
2205 c = token_getch ();
2206 continue;
2208 c = wc;
2209 #endif /* MULTIBYTE_CHARS */
2212 /* Add this single character into the buffer either as a wchar_t
2213 or as a single byte. */
2214 if (wide_flag)
2216 unsigned width = TYPE_PRECISION (char_type_node);
2217 unsigned bytemask = (1 << width) - 1;
2218 int byte;
2220 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2221 p = extend_token_buffer (p);
2223 for (byte = 0; byte < WCHAR_BYTES; ++byte)
2225 int value;
2226 if (byte >= (int) sizeof (c))
2227 value = 0;
2228 else
2229 value = (c >> (byte * width)) & bytemask;
2230 if (BYTES_BIG_ENDIAN)
2231 p[WCHAR_BYTES - byte - 1] = value;
2232 else
2233 p[byte] = value;
2235 p += WCHAR_BYTES;
2237 else
2239 if (p >= token_buffer + maxtoken)
2240 p = extend_token_buffer (p);
2241 *p++ = c;
2244 skipnewline:
2245 c = token_getch ();
2248 /* Terminate the string value, either with a single byte zero
2249 or with a wide zero. */
2250 if (wide_flag)
2252 if (p + WCHAR_BYTES > token_buffer + maxtoken)
2253 p = extend_token_buffer (p);
2254 bzero (p, WCHAR_BYTES);
2255 p += WCHAR_BYTES;
2257 else
2259 if (p >= token_buffer + maxtoken)
2260 p = extend_token_buffer (p);
2261 *p++ = 0;
2264 if (c == EOF)
2265 error ("Unterminated string constant");
2267 /* We have read the entire constant.
2268 Construct a STRING_CST for the result. */
2270 if (wide_flag)
2272 yylval.ttype = build_string (p - (token_buffer + 1),
2273 token_buffer + 1);
2274 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2275 value = STRING;
2277 else if (objc_flag)
2279 /* Return an Objective-C @"..." constant string object. */
2280 yylval.ttype = build_objc_string (p - (token_buffer + 1),
2281 token_buffer + 1);
2282 TREE_TYPE (yylval.ttype) = char_array_type_node;
2283 value = OBJC_STRING;
2285 else
2287 yylval.ttype = build_string (p - (token_buffer + 1),
2288 token_buffer + 1);
2289 TREE_TYPE (yylval.ttype) = char_array_type_node;
2290 value = STRING;
2293 break;
2296 case '+':
2297 case '-':
2298 case '&':
2299 case '|':
2300 case ':':
2301 case '<':
2302 case '>':
2303 case '*':
2304 case '/':
2305 case '%':
2306 case '^':
2307 case '!':
2308 case '=':
2310 register int c1;
2312 combine:
2314 switch (c)
2316 case '+':
2317 yylval.code = PLUS_EXPR; break;
2318 case '-':
2319 yylval.code = MINUS_EXPR; break;
2320 case '&':
2321 yylval.code = BIT_AND_EXPR; break;
2322 case '|':
2323 yylval.code = BIT_IOR_EXPR; break;
2324 case '*':
2325 yylval.code = MULT_EXPR; break;
2326 case '/':
2327 yylval.code = TRUNC_DIV_EXPR; break;
2328 case '%':
2329 yylval.code = TRUNC_MOD_EXPR; break;
2330 case '^':
2331 yylval.code = BIT_XOR_EXPR; break;
2332 case LSHIFT:
2333 yylval.code = LSHIFT_EXPR; break;
2334 case RSHIFT:
2335 yylval.code = RSHIFT_EXPR; break;
2336 case '<':
2337 yylval.code = LT_EXPR; break;
2338 case '>':
2339 yylval.code = GT_EXPR; break;
2342 token_buffer[1] = c1 = token_getch();
2343 token_buffer[2] = 0;
2345 if (c1 == '=')
2347 switch (c)
2349 case '<':
2350 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2351 case '>':
2352 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2353 case '!':
2354 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2355 case '=':
2356 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2358 value = ASSIGN; goto done;
2360 else if (c == c1)
2361 switch (c)
2363 case '+':
2364 value = PLUSPLUS; goto done;
2365 case '-':
2366 value = MINUSMINUS; goto done;
2367 case '&':
2368 value = ANDAND; goto done;
2369 case '|':
2370 value = OROR; goto done;
2371 case '<':
2372 c = LSHIFT;
2373 goto combine;
2374 case '>':
2375 c = RSHIFT;
2376 goto combine;
2378 else
2379 switch (c)
2381 case '-':
2382 if (c1 == '>')
2383 { value = POINTSAT; goto done; }
2384 break;
2386 /* digraphs */
2387 case ':':
2388 if (c1 == '>')
2389 { value = ']'; goto done; }
2390 break;
2391 case '<':
2392 if (c1 == '%')
2393 { value = '{'; indent_level++; goto done; }
2394 if (c1 == ':')
2395 { value = '['; goto done; }
2396 break;
2397 case '%':
2398 if (c1 == '>')
2399 { value = '}'; indent_level--; goto done; }
2400 break;
2403 token_put_back (c1);
2404 token_buffer[1] = 0;
2406 if ((c == '<') || (c == '>'))
2407 value = ARITHCOMPARE;
2408 else value = c;
2409 break;
2412 case 0:
2413 /* Don't make yyparse think this is eof. */
2414 value = 1;
2415 break;
2417 case '{':
2418 indent_level++;
2419 value = c;
2420 break;
2422 case '}':
2423 indent_level--;
2424 value = c;
2425 break;
2427 default:
2428 value = c;
2431 done:
2432 /* yylloc.last_line = lineno; */
2434 return value;
2437 /* Sets the value of the 'yydebug' variable to VALUE.
2438 This is a function so we don't have to have YYDEBUG defined
2439 in order to build the compiler. */
2441 void
2442 set_yydebug (value)
2443 int value;
2445 #if YYDEBUG != 0
2446 yydebug = value;
2447 #else
2448 warning ("YYDEBUG not defined.");
2449 #endif