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 GCC.
7 GCC is free software; you can redistribute it and/or modify it under
8 the terms of the GNU General Public License as published by the Free
9 Software Foundation; either version 2, or (at your option) any later
12 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
13 WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17 You should have received a copy of the GNU General Public License
18 along with GCC; see the file COPYING. If not, write to the Free
19 Software Foundation, 59 Temple Place - Suite 330, Boston, MA
39 #include "splay-tree.h"
42 /* MULTIBYTE_CHARS support only works for native compilers.
43 ??? Ideally what we want is to model widechar support after
44 the current floating point support. */
46 #undef MULTIBYTE_CHARS
49 #ifdef MULTIBYTE_CHARS
52 #endif /* MULTIBYTE_CHARS */
53 #ifndef GET_ENVIRONMENT
54 #define GET_ENVIRONMENT(ENV_VALUE,ENV_NAME) ((ENV_VALUE) = getenv (ENV_NAME))
57 /* The current line map. */
58 static const struct line_map *map;
60 /* The line used to refresh the lineno global variable after each token. */
61 static unsigned int src_lineno;
63 /* We may keep statistics about how long which files took to compile. */
64 static int header_time, body_time;
65 static splay_tree file_info_tree;
67 /* Cause the `yydebug' variable to be defined. */
70 /* File used for outputting assembler code. */
71 extern FILE *asm_out_file;
73 #undef WCHAR_TYPE_SIZE
74 #define WCHAR_TYPE_SIZE TYPE_PRECISION (wchar_type_node)
76 /* Number of bytes in a wide character. */
77 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
79 int indent_level; /* Number of { minus number of }. */
80 int pending_lang_change; /* If we need to switch languages - C++ only */
81 int c_header_level; /* depth in C headers - C++ only */
83 /* Nonzero tells yylex to ignore \ in string constants. */
84 static int ignore_escape_flag;
86 static void parse_float PARAMS ((PTR));
87 static tree lex_number PARAMS ((const char *, unsigned int));
88 static tree lex_string PARAMS ((const char *, unsigned int, int));
89 static tree lex_charconst PARAMS ((const cpp_token *));
90 static void update_header_times PARAMS ((const char *));
91 static int dump_one_header PARAMS ((splay_tree_node, void *));
92 static void cb_line_change PARAMS ((cpp_reader *, const cpp_token *, int));
93 static void cb_ident PARAMS ((cpp_reader *, unsigned int,
95 static void cb_file_change PARAMS ((cpp_reader *, const struct line_map *));
96 static void cb_def_pragma PARAMS ((cpp_reader *, unsigned int));
97 static void cb_define PARAMS ((cpp_reader *, unsigned int,
99 static void cb_undef PARAMS ((cpp_reader *, unsigned int,
103 init_c_lex (filename)
104 const char *filename;
106 struct cpp_callbacks *cb;
107 struct c_fileinfo *toplevel;
109 /* Set up filename timing. Must happen before cpp_read_main_file. */
110 file_info_tree = splay_tree_new ((splay_tree_compare_fn)strcmp,
112 (splay_tree_delete_value_fn)free);
113 toplevel = get_fileinfo ("<top level>");
114 if (flag_detailed_statistics)
117 body_time = get_run_time ();
118 toplevel->time = body_time;
121 #ifdef MULTIBYTE_CHARS
122 /* Change to the native locale for multibyte conversions. */
123 setlocale (LC_CTYPE, "");
124 GET_ENVIRONMENT (literal_codeset, "LANG");
127 cb = cpp_get_callbacks (parse_in);
129 cb->line_change = cb_line_change;
130 cb->ident = cb_ident;
131 cb->file_change = cb_file_change;
132 cb->def_pragma = cb_def_pragma;
134 /* Set the debug callbacks if we can use them. */
135 if (debug_info_level == DINFO_LEVEL_VERBOSE
136 && (write_symbols == DWARF_DEBUG || write_symbols == DWARF2_DEBUG
137 || write_symbols == VMS_AND_DWARF2_DEBUG))
139 cb->define = cb_define;
140 cb->undef = cb_undef;
146 if (filename == NULL || !strcmp (filename, "-"))
149 return cpp_read_main_file (parse_in, filename, ident_hash);
152 /* A thin wrapper around the real parser that initializes the
153 integrated preprocessor after debug output has been initialized.
154 Also, make sure the start_source_file debug hook gets called for
155 the primary source file. */
160 (*debug_hooks->start_source_file) (lineno, input_filename);
161 cpp_finish_options (parse_in);
171 struct c_fileinfo *fi;
173 n = splay_tree_lookup (file_info_tree, (splay_tree_key) name);
175 return (struct c_fileinfo *) n->value;
177 fi = (struct c_fileinfo *) xmalloc (sizeof (struct c_fileinfo));
179 fi->interface_only = 0;
180 fi->interface_unknown = 1;
181 splay_tree_insert (file_info_tree, (splay_tree_key) name,
182 (splay_tree_value) fi);
187 update_header_times (name)
190 /* Changing files again. This means currently collected time
191 is charged against header time, and body time starts back at 0. */
192 if (flag_detailed_statistics)
194 int this_time = get_run_time ();
195 struct c_fileinfo *file = get_fileinfo (name);
196 header_time += this_time - body_time;
197 file->time += this_time - body_time;
198 body_time = this_time;
203 dump_one_header (n, dummy)
205 void *dummy ATTRIBUTE_UNUSED;
207 print_time ((const char *) n->key,
208 ((struct c_fileinfo *) n->value)->time);
213 dump_time_statistics ()
215 struct c_fileinfo *file = get_fileinfo (input_filename);
216 int this_time = get_run_time ();
217 file->time += this_time - body_time;
219 fprintf (stderr, "\n******\n");
220 print_time ("header files (total)", header_time);
221 print_time ("main file (total)", this_time - body_time);
222 fprintf (stderr, "ratio = %g : 1\n",
223 (double)header_time / (double)(this_time - body_time));
224 fprintf (stderr, "\n******\n");
226 splay_tree_foreach (file_info_tree, dump_one_header, 0);
229 /* Not yet handled: #pragma, #define, #undef.
230 No need to deal with linemarkers under normal conditions. */
233 cb_ident (pfile, line, str)
234 cpp_reader *pfile ATTRIBUTE_UNUSED;
235 unsigned int line ATTRIBUTE_UNUSED;
236 const cpp_string *str ATTRIBUTE_UNUSED;
238 #ifdef ASM_OUTPUT_IDENT
241 /* Convert escapes in the string. */
242 tree value = lex_string ((const char *)str->text, str->len, 0);
243 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (value));
248 /* Called at the start of every non-empty line. TOKEN is the first
249 lexed token on the line. Used for diagnostic line numbers. */
251 cb_line_change (pfile, token, parsing_args)
252 cpp_reader *pfile ATTRIBUTE_UNUSED;
253 const cpp_token *token;
254 int parsing_args ATTRIBUTE_UNUSED;
256 src_lineno = SOURCE_LINE (map, token->line);
260 cb_file_change (pfile, new_map)
261 cpp_reader *pfile ATTRIBUTE_UNUSED;
262 const struct line_map *new_map;
264 unsigned int to_line = SOURCE_LINE (new_map, new_map->to_line);
266 if (new_map->reason == LC_ENTER)
268 /* Don't stack the main buffer on the input stack;
269 we already did in compile_file. */
271 main_input_filename = new_map->to_file;
274 lineno = SOURCE_LINE (new_map - 1, new_map->from_line - 1);
275 push_srcloc (new_map->to_file, 1);
276 input_file_stack->indent_level = indent_level;
277 (*debug_hooks->start_source_file) (lineno, new_map->to_file);
278 #ifndef NO_IMPLICIT_EXTERN_C
281 else if (new_map->sysp == 2)
284 ++pending_lang_change;
289 else if (new_map->reason == LC_LEAVE)
291 #ifndef NO_IMPLICIT_EXTERN_C
292 if (c_header_level && --c_header_level == 0)
294 if (new_map->sysp == 2)
295 warning ("badly nested C headers from preprocessor");
296 --pending_lang_change;
300 if (indent_level != input_file_stack->indent_level)
302 warning_with_file_and_line
303 (input_filename, lineno,
304 "this file contains more '%c's than '%c's",
305 indent_level > input_file_stack->indent_level ? '{' : '}',
306 indent_level > input_file_stack->indent_level ? '}' : '{');
311 (*debug_hooks->end_source_file) (to_line);
314 update_header_times (new_map->to_file);
315 in_system_header = new_map->sysp != 0;
316 input_filename = new_map->to_file;
321 extract_interface_info ();
325 cb_def_pragma (pfile, line)
329 /* Issue a warning message if we have been asked to do so. Ignore
330 unknown pragmas in system headers unless an explicit
331 -Wunknown-pragmas has been given. */
332 if (warn_unknown_pragmas > in_system_header)
334 const unsigned char *space, *name = 0;
337 s = cpp_get_token (pfile);
338 space = cpp_token_as_text (pfile, s);
339 s = cpp_get_token (pfile);
340 if (s->type == CPP_NAME)
341 name = cpp_token_as_text (pfile, s);
343 lineno = SOURCE_LINE (map, line);
345 warning ("ignoring #pragma %s %s", space, name);
347 warning ("ignoring #pragma %s", space);
351 /* #define callback for DWARF and DWARF2 debug info. */
353 cb_define (pfile, line, node)
358 (*debug_hooks->define) (SOURCE_LINE (map, line),
359 (const char *) cpp_macro_definition (pfile, node));
362 /* #undef callback for DWARF and DWARF2 debug info. */
364 cb_undef (pfile, line, node)
365 cpp_reader *pfile ATTRIBUTE_UNUSED;
369 (*debug_hooks->undef) (SOURCE_LINE (map, line),
370 (const char *) NODE_NAME (node));
374 /* Returns nonzero if C is a universal-character-name. Give an error if it
375 is not one which may appear in an identifier, as per [extendid].
377 Note that extended character support in identifiers has not yet been
378 implemented. It is my personal opinion that this is not a desirable
379 feature. Portable code cannot count on support for more than the basic
380 identifier character set. */
393 /* None of the valid chars are outside the Basic Multilingual Plane (the
397 error ("universal-character-name '\\U%08x' not valid in identifier", c);
402 if ((c >= 0x00c0 && c <= 0x00d6)
403 || (c >= 0x00d8 && c <= 0x00f6)
404 || (c >= 0x00f8 && c <= 0x01f5)
405 || (c >= 0x01fa && c <= 0x0217)
406 || (c >= 0x0250 && c <= 0x02a8)
407 || (c >= 0x1e00 && c <= 0x1e9a)
408 || (c >= 0x1ea0 && c <= 0x1ef9))
413 || (c >= 0x0388 && c <= 0x038a)
415 || (c >= 0x038e && c <= 0x03a1)
416 || (c >= 0x03a3 && c <= 0x03ce)
417 || (c >= 0x03d0 && c <= 0x03d6)
422 || (c >= 0x03e2 && c <= 0x03f3)
423 || (c >= 0x1f00 && c <= 0x1f15)
424 || (c >= 0x1f18 && c <= 0x1f1d)
425 || (c >= 0x1f20 && c <= 0x1f45)
426 || (c >= 0x1f48 && c <= 0x1f4d)
427 || (c >= 0x1f50 && c <= 0x1f57)
431 || (c >= 0x1f5f && c <= 0x1f7d)
432 || (c >= 0x1f80 && c <= 0x1fb4)
433 || (c >= 0x1fb6 && c <= 0x1fbc)
434 || (c >= 0x1fc2 && c <= 0x1fc4)
435 || (c >= 0x1fc6 && c <= 0x1fcc)
436 || (c >= 0x1fd0 && c <= 0x1fd3)
437 || (c >= 0x1fd6 && c <= 0x1fdb)
438 || (c >= 0x1fe0 && c <= 0x1fec)
439 || (c >= 0x1ff2 && c <= 0x1ff4)
440 || (c >= 0x1ff6 && c <= 0x1ffc))
444 if ((c >= 0x0401 && c <= 0x040d)
445 || (c >= 0x040f && c <= 0x044f)
446 || (c >= 0x0451 && c <= 0x045c)
447 || (c >= 0x045e && c <= 0x0481)
448 || (c >= 0x0490 && c <= 0x04c4)
449 || (c >= 0x04c7 && c <= 0x04c8)
450 || (c >= 0x04cb && c <= 0x04cc)
451 || (c >= 0x04d0 && c <= 0x04eb)
452 || (c >= 0x04ee && c <= 0x04f5)
453 || (c >= 0x04f8 && c <= 0x04f9))
457 if ((c >= 0x0531 && c <= 0x0556)
458 || (c >= 0x0561 && c <= 0x0587))
462 if ((c >= 0x05d0 && c <= 0x05ea)
463 || (c >= 0x05f0 && c <= 0x05f4))
467 if ((c >= 0x0621 && c <= 0x063a)
468 || (c >= 0x0640 && c <= 0x0652)
469 || (c >= 0x0670 && c <= 0x06b7)
470 || (c >= 0x06ba && c <= 0x06be)
471 || (c >= 0x06c0 && c <= 0x06ce)
472 || (c >= 0x06e5 && c <= 0x06e7))
476 if ((c >= 0x0905 && c <= 0x0939)
477 || (c >= 0x0958 && c <= 0x0962))
481 if ((c >= 0x0985 && c <= 0x098c)
482 || (c >= 0x098f && c <= 0x0990)
483 || (c >= 0x0993 && c <= 0x09a8)
484 || (c >= 0x09aa && c <= 0x09b0)
486 || (c >= 0x09b6 && c <= 0x09b9)
487 || (c >= 0x09dc && c <= 0x09dd)
488 || (c >= 0x09df && c <= 0x09e1)
489 || (c >= 0x09f0 && c <= 0x09f1))
493 if ((c >= 0x0a05 && c <= 0x0a0a)
494 || (c >= 0x0a0f && c <= 0x0a10)
495 || (c >= 0x0a13 && c <= 0x0a28)
496 || (c >= 0x0a2a && c <= 0x0a30)
497 || (c >= 0x0a32 && c <= 0x0a33)
498 || (c >= 0x0a35 && c <= 0x0a36)
499 || (c >= 0x0a38 && c <= 0x0a39)
500 || (c >= 0x0a59 && c <= 0x0a5c)
505 if ((c >= 0x0a85 && c <= 0x0a8b)
507 || (c >= 0x0a8f && c <= 0x0a91)
508 || (c >= 0x0a93 && c <= 0x0aa8)
509 || (c >= 0x0aaa && c <= 0x0ab0)
510 || (c >= 0x0ab2 && c <= 0x0ab3)
511 || (c >= 0x0ab5 && c <= 0x0ab9)
516 if ((c >= 0x0b05 && c <= 0x0b0c)
517 || (c >= 0x0b0f && c <= 0x0b10)
518 || (c >= 0x0b13 && c <= 0x0b28)
519 || (c >= 0x0b2a && c <= 0x0b30)
520 || (c >= 0x0b32 && c <= 0x0b33)
521 || (c >= 0x0b36 && c <= 0x0b39)
522 || (c >= 0x0b5c && c <= 0x0b5d)
523 || (c >= 0x0b5f && c <= 0x0b61))
527 if ((c >= 0x0b85 && c <= 0x0b8a)
528 || (c >= 0x0b8e && c <= 0x0b90)
529 || (c >= 0x0b92 && c <= 0x0b95)
530 || (c >= 0x0b99 && c <= 0x0b9a)
532 || (c >= 0x0b9e && c <= 0x0b9f)
533 || (c >= 0x0ba3 && c <= 0x0ba4)
534 || (c >= 0x0ba8 && c <= 0x0baa)
535 || (c >= 0x0bae && c <= 0x0bb5)
536 || (c >= 0x0bb7 && c <= 0x0bb9))
540 if ((c >= 0x0c05 && c <= 0x0c0c)
541 || (c >= 0x0c0e && c <= 0x0c10)
542 || (c >= 0x0c12 && c <= 0x0c28)
543 || (c >= 0x0c2a && c <= 0x0c33)
544 || (c >= 0x0c35 && c <= 0x0c39)
545 || (c >= 0x0c60 && c <= 0x0c61))
549 if ((c >= 0x0c85 && c <= 0x0c8c)
550 || (c >= 0x0c8e && c <= 0x0c90)
551 || (c >= 0x0c92 && c <= 0x0ca8)
552 || (c >= 0x0caa && c <= 0x0cb3)
553 || (c >= 0x0cb5 && c <= 0x0cb9)
554 || (c >= 0x0ce0 && c <= 0x0ce1))
558 if ((c >= 0x0d05 && c <= 0x0d0c)
559 || (c >= 0x0d0e && c <= 0x0d10)
560 || (c >= 0x0d12 && c <= 0x0d28)
561 || (c >= 0x0d2a && c <= 0x0d39)
562 || (c >= 0x0d60 && c <= 0x0d61))
566 if ((c >= 0x0e01 && c <= 0x0e30)
567 || (c >= 0x0e32 && c <= 0x0e33)
568 || (c >= 0x0e40 && c <= 0x0e46)
569 || (c >= 0x0e4f && c <= 0x0e5b))
573 if ((c >= 0x0e81 && c <= 0x0e82)
579 || (c >= 0x0e94 && c <= 0x0e97)
580 || (c >= 0x0e99 && c <= 0x0e9f)
581 || (c >= 0x0ea1 && c <= 0x0ea3)
586 || (c >= 0x0ead && c <= 0x0eb0)
590 || (c >= 0x0ec0 && c <= 0x0ec4)
595 if ((c >= 0x10a0 && c <= 0x10c5)
596 || (c >= 0x10d0 && c <= 0x10f6))
600 if ((c >= 0x3041 && c <= 0x3094)
601 || (c >= 0x309b && c <= 0x309e))
605 if ((c >= 0x30a1 && c <= 0x30fe))
609 if ((c >= 0x3105 && c <= 0x312c))
613 if ((c >= 0x1100 && c <= 0x1159)
614 || (c >= 0x1161 && c <= 0x11a2)
615 || (c >= 0x11a8 && c <= 0x11f9))
618 /* CJK Unified Ideographs */
619 if ((c >= 0xf900 && c <= 0xfa2d)
620 || (c >= 0xfb1f && c <= 0xfb36)
621 || (c >= 0xfb38 && c <= 0xfb3c)
623 || (c >= 0xfb40 && c <= 0xfb41)
624 || (c >= 0xfb42 && c <= 0xfb44)
625 || (c >= 0xfb46 && c <= 0xfbb1)
626 || (c >= 0xfbd3 && c <= 0xfd3f)
627 || (c >= 0xfd50 && c <= 0xfd8f)
628 || (c >= 0xfd92 && c <= 0xfdc7)
629 || (c >= 0xfdf0 && c <= 0xfdfb)
630 || (c >= 0xfe70 && c <= 0xfe72)
632 || (c >= 0xfe76 && c <= 0xfefc)
633 || (c >= 0xff21 && c <= 0xff3a)
634 || (c >= 0xff41 && c <= 0xff5a)
635 || (c >= 0xff66 && c <= 0xffbe)
636 || (c >= 0xffc2 && c <= 0xffc7)
637 || (c >= 0xffca && c <= 0xffcf)
638 || (c >= 0xffd2 && c <= 0xffd7)
639 || (c >= 0xffda && c <= 0xffdc)
640 || (c >= 0x4e00 && c <= 0x9fa5))
643 error ("universal-character-name '\\u%04x' not valid in identifier", c);
648 /* Add the UTF-8 representation of C to the token_buffer. */
651 utf8_extend_token (c)
661 else if (c <= 0x000007ff)
662 shift = 6, mask = 0xc0;
663 else if (c <= 0x0000ffff)
664 shift = 12, mask = 0xe0;
665 else if (c <= 0x001fffff)
666 shift = 18, mask = 0xf0;
667 else if (c <= 0x03ffffff)
668 shift = 24, mask = 0xf8;
670 shift = 30, mask = 0xfc;
672 extend_token (mask | (c >> shift));
676 extend_token ((unsigned char) (0x80 | (c >> shift)));
685 tree *const node_var;
686 const char unsigned_flag;
687 const char long_flag;
688 const char long_long_flag;
691 struct try_type type_sequence[] =
693 { &integer_type_node, 0, 0, 0},
694 { &unsigned_type_node, 1, 0, 0},
695 { &long_integer_type_node, 0, 1, 0},
696 { &long_unsigned_type_node, 1, 1, 0},
697 { &long_long_integer_type_node, 0, 1, 1},
698 { &long_long_unsigned_type_node, 1, 1, 1}
710 int conversion_errno;
711 REAL_VALUE_TYPE value;
719 struct pf_args * args = (struct pf_args *) data;
720 const char *typename;
722 args->conversion_errno = 0;
723 args->type = double_type_node;
726 /* The second argument, machine_mode, of REAL_VALUE_ATOF
727 tells the desired precision of the binary result
728 of decimal-to-binary conversion. */
733 error ("both 'f' and 'l' suffixes on floating constant");
735 args->type = float_type_node;
738 else if (args->lflag)
740 args->type = long_double_type_node;
741 typename = "long double";
743 else if (flag_single_precision_constant)
745 args->type = float_type_node;
750 if (args->base == 16)
751 args->value = REAL_VALUE_HTOF (args->str, TYPE_MODE (args->type));
753 args->value = REAL_VALUE_ATOF (args->str, TYPE_MODE (args->type));
755 args->conversion_errno = errno;
756 /* A diagnostic is required here by some ISO C testsuites.
757 This is not pedwarn, because some people don't want
758 an error for this. */
759 if (REAL_VALUE_ISINF (args->value) && pedantic)
760 warning ("floating point number exceeds range of '%s'", typename);
767 const cpp_token *tok;
770 timevar_push (TV_CPP);
772 tok = cpp_get_token (parse_in);
773 while (tok->type == CPP_PADDING);
774 timevar_pop (TV_CPP);
776 /* The C++ front end does horrible things with the current line
777 number. To ensure an accurate line number, we must reset it
778 every time we return a token. */
784 case CPP_OPEN_BRACE: indent_level++; break;
785 case CPP_CLOSE_BRACE: indent_level--; break;
787 /* Issue this error here, where we can get at tok->val.c. */
789 if (ISGRAPH (tok->val.c))
790 error ("stray '%c' in program", tok->val.c);
792 error ("stray '\\%o' in program", tok->val.c);
796 *value = HT_IDENT_TO_GCC_IDENT (HT_NODE (tok->val.node));
800 *value = lex_number ((const char *)tok->val.str.text, tok->val.str.len);
805 *value = lex_charconst (tok);
810 *value = lex_string ((const char *)tok->val.str.text,
811 tok->val.str.len, tok->type == CPP_WSTRING);
814 /* These tokens should not be visible outside cpplib. */
815 case CPP_HEADER_NAME:
826 #define ERROR(msgid) do { error(msgid); goto syntax_error; } while(0)
829 lex_number (str, len)
835 int largest_digit = 0;
841 enum anon1 { NOT_FLOAT = 0, AFTER_POINT, AFTER_EXPON } floatflag = NOT_FLOAT;
843 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
844 The code below which fills the parts array assumes that a host
845 int is at least twice as wide as a host char, and that
846 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
847 Two HOST_WIDE_INTs is the largest int literal we can store.
848 In order to detect overflow below, the number of parts (TOTAL_PARTS)
849 must be exactly the number of parts needed to hold the bits
850 of two HOST_WIDE_INTs. */
851 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
852 unsigned int parts[TOTAL_PARTS];
854 /* Optimize for most frequent case. */
858 return integer_zero_node;
859 else if (*str == '1')
860 return integer_one_node;
862 return build_int_2 (*str - '0', 0);
865 for (count = 0; count < TOTAL_PARTS; count++)
868 /* len is known to be >1 at this point. */
871 if (len > 2 && str[0] == '0' && (str[1] == 'x' || str[1] == 'X'))
876 /* The ISDIGIT check is so we are not confused by a suffix on 0. */
877 else if (str[0] == '0' && ISDIGIT (str[1]))
889 if (floatflag == AFTER_POINT)
890 ERROR ("too many decimal points in floating constant");
891 else if (floatflag == AFTER_EXPON)
892 ERROR ("decimal point in exponent - impossible!");
894 floatflag = AFTER_POINT;
900 /* Possible future extension: silently ignore _ in numbers,
901 permitting cosmetic grouping - e.g. 0x8000_0000 == 0x80000000
902 but somewhat easier to read. Ada has this? */
903 ERROR ("underscore in number");
907 /* It is not a decimal point.
908 It should be a digit (perhaps a hex digit). */
911 || (base == 16 && ISXDIGIT (c)))
915 else if (base <= 10 && (c == 'e' || c == 'E'))
918 floatflag = AFTER_EXPON;
921 else if (base == 16 && (c == 'p' || c == 'P'))
923 floatflag = AFTER_EXPON;
924 break; /* start of exponent */
929 break; /* start of suffix */
932 if (n >= largest_digit)
936 for (count = 0; count < TOTAL_PARTS; count++)
938 parts[count] *= base;
942 += (parts[count-1] >> HOST_BITS_PER_CHAR);
944 &= (1 << HOST_BITS_PER_CHAR) - 1;
950 /* If the highest-order part overflows (gets larger than
951 a host char will hold) then the whole number has
952 overflowed. Record this and truncate the highest-order
954 if (parts[TOTAL_PARTS - 1] >> HOST_BITS_PER_CHAR)
957 parts[TOTAL_PARTS - 1] &= (1 << HOST_BITS_PER_CHAR) - 1;
961 while (p < str + len);
963 /* This can happen on input like `int i = 0x;' */
965 ERROR ("numeric constant with no digits");
967 if (largest_digit >= base)
968 ERROR ("numeric constant contains digits beyond the radix");
970 if (floatflag != NOT_FLOAT)
973 int imag, fflag, lflag, conversion_errno;
974 REAL_VALUE_TYPE real;
978 if (base == 16 && pedantic && !flag_isoc99)
979 pedwarn ("floating constant may not be in radix 16");
981 if (base == 16 && floatflag != AFTER_EXPON)
982 ERROR ("hexadecimal floating constant has no exponent");
984 /* Read explicit exponent if any, and put it in tokenbuf. */
985 if ((base == 10 && ((c == 'e') || (c == 'E')))
986 || (base == 16 && (c == 'p' || c == 'P')))
990 if (p < str + len && (c == '+' || c == '-'))
992 /* Exponent is decimal, even if string is a hex float. */
994 ERROR ("floating constant exponent has no digits");
995 while (p < str + len && ISDIGIT (c))
1001 /* Copy the float constant now; we don't want any suffixes in the
1002 string passed to parse_float. */
1003 copy = alloca (p - str + 1);
1004 memcpy (copy, str, p - str);
1005 copy[p - str] = '\0';
1007 /* Now parse suffixes. */
1008 fflag = lflag = imag = 0;
1009 while (p < str + len)
1014 ERROR ("more than one 'f' suffix on floating constant");
1015 else if (warn_traditional && !in_system_header
1016 && ! cpp_sys_macro_p (parse_in))
1017 warning ("traditional C rejects the 'f' suffix");
1024 ERROR ("more than one 'l' suffix on floating constant");
1025 else if (warn_traditional && !in_system_header
1026 && ! cpp_sys_macro_p (parse_in))
1027 warning ("traditional C rejects the 'l' suffix");
1035 ERROR ("more than one 'i' or 'j' suffix on floating constant");
1037 pedwarn ("ISO C forbids imaginary numeric constants");
1042 ERROR ("invalid suffix on floating constant");
1045 /* Setup input for parse_float() */
1051 /* Convert string to a double, checking for overflow. */
1052 if (do_float_handler (parse_float, (PTR) &args))
1054 /* Receive output from parse_float() */
1058 /* We got an exception from parse_float() */
1059 ERROR ("floating constant out of range");
1061 /* Receive output from parse_float() */
1062 conversion_errno = args.conversion_errno;
1066 /* ERANGE is also reported for underflow,
1067 so test the value to distinguish overflow from that. */
1068 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1069 && (REAL_VALUES_LESS (dconst1, real)
1070 || REAL_VALUES_LESS (real, dconstm1)))
1071 warning ("floating point number exceeds range of 'double'");
1074 /* Create a node with determined type and value. */
1076 value = build_complex (NULL_TREE, convert (type, integer_zero_node),
1077 build_real (type, real));
1079 value = build_real (type, real);
1083 tree trad_type, ansi_type, type;
1084 HOST_WIDE_INT high, low;
1085 int spec_unsigned = 0;
1087 int spec_long_long = 0;
1092 trad_type = ansi_type = type = NULL_TREE;
1093 while (p < str + len)
1100 error ("two 'u' suffixes on integer constant");
1101 else if (warn_traditional && !in_system_header
1102 && ! cpp_sys_macro_p (parse_in))
1103 warning ("traditional C rejects the 'u' suffix");
1114 error ("three 'l' suffixes on integer constant");
1116 error ("'lul' is not a valid integer suffix");
1117 else if (c != spec_long)
1118 error ("'Ll' and 'lL' are not valid integer suffixes");
1119 else if (pedantic && ! flag_isoc99
1120 && ! in_system_header && warn_long_long)
1121 pedwarn ("ISO C89 forbids long long integer constants");
1127 case 'i': case 'I': case 'j': case 'J':
1129 error ("more than one 'i' or 'j' suffix on integer constant");
1131 pedwarn ("ISO C forbids imaginary numeric constants");
1136 ERROR ("invalid suffix on integer constant");
1140 /* If the literal overflowed, pedwarn about it now. */
1144 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT * 2);
1147 /* This is simplified by the fact that our constant
1148 is always positive. */
1152 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1154 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1155 / HOST_BITS_PER_CHAR)]
1156 << (i * HOST_BITS_PER_CHAR));
1157 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1160 value = build_int_2 (low, high);
1161 TREE_TYPE (value) = long_long_unsigned_type_node;
1163 /* If warn_traditional, calculate both the ISO type and the
1164 traditional type, then see if they disagree.
1165 Otherwise, calculate only the type for the dialect in use. */
1166 if (warn_traditional || flag_traditional)
1168 /* Calculate the traditional type. */
1169 /* Traditionally, any constant is signed; but if unsigned is
1170 specified explicitly, obey that. Use the smallest size
1171 with the right number of bits, except for one special
1172 case with decimal constants. */
1173 if (! spec_long && base != 10
1174 && int_fits_type_p (value, unsigned_type_node))
1175 trad_type = spec_unsigned ? unsigned_type_node : integer_type_node;
1176 /* A decimal constant must be long if it does not fit in
1177 type int. I think this is independent of whether the
1178 constant is signed. */
1179 else if (! spec_long && base == 10
1180 && int_fits_type_p (value, integer_type_node))
1181 trad_type = spec_unsigned ? unsigned_type_node : integer_type_node;
1182 else if (! spec_long_long)
1183 trad_type = (spec_unsigned
1184 ? long_unsigned_type_node
1185 : long_integer_type_node);
1186 else if (int_fits_type_p (value,
1188 ? long_long_unsigned_type_node
1189 : long_long_integer_type_node))
1190 trad_type = (spec_unsigned
1191 ? long_long_unsigned_type_node
1192 : long_long_integer_type_node);
1194 trad_type = (spec_unsigned
1195 ? widest_unsigned_literal_type_node
1196 : widest_integer_literal_type_node);
1198 if (warn_traditional || ! flag_traditional)
1200 /* Calculate the ISO type. */
1201 if (! spec_long && ! spec_unsigned
1202 && int_fits_type_p (value, integer_type_node))
1203 ansi_type = integer_type_node;
1204 else if (! spec_long && (base != 10 || spec_unsigned)
1205 && int_fits_type_p (value, unsigned_type_node))
1206 ansi_type = unsigned_type_node;
1207 else if (! spec_unsigned && !spec_long_long
1208 && int_fits_type_p (value, long_integer_type_node))
1209 ansi_type = long_integer_type_node;
1210 else if (! spec_long_long
1211 && int_fits_type_p (value, long_unsigned_type_node))
1212 ansi_type = long_unsigned_type_node;
1213 else if (! spec_unsigned
1214 && int_fits_type_p (value, long_long_integer_type_node))
1215 ansi_type = long_long_integer_type_node;
1216 else if (int_fits_type_p (value, long_long_unsigned_type_node))
1217 ansi_type = long_long_unsigned_type_node;
1218 else if (! spec_unsigned
1219 && int_fits_type_p (value, widest_integer_literal_type_node))
1220 ansi_type = widest_integer_literal_type_node;
1222 ansi_type = widest_unsigned_literal_type_node;
1225 type = flag_traditional ? trad_type : ansi_type;
1227 /* We assume that constants specified in a non-decimal
1228 base are bit patterns, and that the programmer really
1229 meant what they wrote. */
1230 if (warn_traditional && !in_system_header
1231 && base == 10 && trad_type != ansi_type)
1233 if (TYPE_PRECISION (trad_type) != TYPE_PRECISION (ansi_type))
1234 warning ("width of integer constant changes with -traditional");
1235 else if (TREE_UNSIGNED (trad_type) != TREE_UNSIGNED (ansi_type))
1236 warning ("integer constant is unsigned in ISO C, signed with -traditional");
1238 warning ("width of integer constant may change on other systems with -traditional");
1241 if (pedantic && !flag_traditional && (flag_isoc99 || !spec_long_long)
1244 ? TYPE_PRECISION (long_long_integer_type_node)
1245 : TYPE_PRECISION (long_integer_type_node)) < TYPE_PRECISION (type)))
1248 pedwarn ("integer constant larger than the maximum value of %s",
1250 ? (TREE_UNSIGNED (type)
1251 ? _("an unsigned long long int")
1252 : _("a long long int"))
1253 : _("an unsigned long int")));
1256 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1257 warning ("decimal constant is so large that it is unsigned");
1261 if (TYPE_PRECISION (type)
1262 <= TYPE_PRECISION (integer_type_node))
1263 value = build_complex (NULL_TREE, integer_zero_node,
1264 convert (integer_type_node, value));
1266 ERROR ("complex integer constant is too wide for 'complex int'");
1268 else if (flag_traditional && !int_fits_type_p (value, type))
1269 /* The traditional constant 0x80000000 is signed
1270 but doesn't fit in the range of int.
1271 This will change it to -0x80000000, which does fit. */
1273 TREE_TYPE (value) = unsigned_type (type);
1274 value = convert (type, value);
1275 TREE_OVERFLOW (value) = TREE_CONSTANT_OVERFLOW (value) = 0;
1278 TREE_TYPE (value) = type;
1280 /* If it's still an integer (not a complex), and it doesn't
1281 fit in the type we choose for it, then pedwarn. */
1284 && TREE_CODE (TREE_TYPE (value)) == INTEGER_TYPE
1285 && ! int_fits_type_p (value, TREE_TYPE (value)))
1286 pedwarn ("integer constant is larger than the maximum value for its type");
1290 error ("missing white space after number '%.*s'", (int) (p - str), str);
1295 return integer_zero_node;
1299 lex_string (str, len, wide)
1305 char *buf = alloca ((len + 1) * (wide ? WCHAR_BYTES : 1));
1307 const char *p = str, *limit = str + len;
1309 unsigned width = wide ? WCHAR_TYPE_SIZE
1310 : TYPE_PRECISION (char_type_node);
1312 #ifdef MULTIBYTE_CHARS
1313 /* Reset multibyte conversion state. */
1314 (void) local_mbtowc (NULL, NULL, 0);
1319 #ifdef MULTIBYTE_CHARS
1323 char_len = local_mbtowc (&wc, p, limit - p);
1326 warning ("ignoring invalid multibyte character");
1339 if (c == '\\' && !ignore_escape_flag)
1343 if (width < HOST_BITS_PER_INT)
1344 mask = ((unsigned int) 1 << width) - 1;
1347 c = cpp_parse_escape (parse_in, (const unsigned char **) &p,
1348 (const unsigned char *) limit,
1349 mask, flag_traditional);
1352 /* Add this single character into the buffer either as a wchar_t
1353 or as a single byte. */
1356 unsigned charwidth = TYPE_PRECISION (char_type_node);
1357 unsigned bytemask = (1 << charwidth) - 1;
1360 for (byte = 0; byte < WCHAR_BYTES; ++byte)
1363 if (byte >= (int) sizeof (c))
1366 n = (c >> (byte * charwidth)) & bytemask;
1367 if (BYTES_BIG_ENDIAN)
1368 q[WCHAR_BYTES - byte - 1] = n;
1380 /* Terminate the string value, either with a single byte zero
1381 or with a wide zero. */
1385 memset (q, 0, WCHAR_BYTES);
1393 value = build_string (q - buf, buf);
1396 TREE_TYPE (value) = wchar_array_type_node;
1398 TREE_TYPE (value) = char_array_type_node;
1402 /* Converts a (possibly wide) character constant token into a tree. */
1404 lex_charconst (token)
1405 const cpp_token *token;
1407 HOST_WIDE_INT result;
1409 unsigned int chars_seen;
1411 result = cpp_interpret_charconst (parse_in, token, warn_multichar,
1412 flag_traditional, &chars_seen);
1413 if (token->type == CPP_WCHAR)
1415 value = build_int_2 (result, 0);
1416 TREE_TYPE (value) = wchar_type_node;
1421 value = build_int_2 (result, -1);
1423 value = build_int_2 (result, 0);
1425 /* In C, a character constant has type 'int'.
1426 In C++ 'char', but multi-char charconsts have type 'int'. */
1427 if (c_language == clk_cplusplus && chars_seen <= 1)
1428 TREE_TYPE (value) = char_type_node;
1430 TREE_TYPE (value) = integer_type_node;