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))
138 cb->define = cb_define;
139 cb->undef = cb_undef;
145 if (filename == NULL)
148 return cpp_read_main_file (parse_in, filename, ident_hash);
151 /* A thin wrapper around the real parser that initializes the
152 integrated preprocessor after debug output has been initialized. */
157 cpp_finish_options (parse_in);
167 struct c_fileinfo *fi;
169 n = splay_tree_lookup (file_info_tree, (splay_tree_key) name);
171 return (struct c_fileinfo *) n->value;
173 fi = (struct c_fileinfo *) xmalloc (sizeof (struct c_fileinfo));
175 fi->interface_only = 0;
176 fi->interface_unknown = 1;
177 splay_tree_insert (file_info_tree, (splay_tree_key) name,
178 (splay_tree_value) fi);
183 update_header_times (name)
186 /* Changing files again. This means currently collected time
187 is charged against header time, and body time starts back at 0. */
188 if (flag_detailed_statistics)
190 int this_time = get_run_time ();
191 struct c_fileinfo *file = get_fileinfo (name);
192 header_time += this_time - body_time;
193 file->time += this_time - body_time;
194 body_time = this_time;
199 dump_one_header (n, dummy)
201 void *dummy ATTRIBUTE_UNUSED;
203 print_time ((const char *) n->key,
204 ((struct c_fileinfo *) n->value)->time);
209 dump_time_statistics ()
211 struct c_fileinfo *file = get_fileinfo (input_filename);
212 int this_time = get_run_time ();
213 file->time += this_time - body_time;
215 fprintf (stderr, "\n******\n");
216 print_time ("header files (total)", header_time);
217 print_time ("main file (total)", this_time - body_time);
218 fprintf (stderr, "ratio = %g : 1\n",
219 (double)header_time / (double)(this_time - body_time));
220 fprintf (stderr, "\n******\n");
222 splay_tree_foreach (file_info_tree, dump_one_header, 0);
225 /* Not yet handled: #pragma, #define, #undef.
226 No need to deal with linemarkers under normal conditions. */
229 cb_ident (pfile, line, str)
230 cpp_reader *pfile ATTRIBUTE_UNUSED;
231 unsigned int line ATTRIBUTE_UNUSED;
232 const cpp_string *str ATTRIBUTE_UNUSED;
234 #ifdef ASM_OUTPUT_IDENT
237 /* Convert escapes in the string. */
238 tree value = lex_string ((const char *)str->text, str->len, 0);
239 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (value));
244 /* Called at the start of every non-empty line. TOKEN is the first
245 lexed token on the line. Used for diagnostic line numbers. */
247 cb_line_change (pfile, token, parsing_args)
248 cpp_reader *pfile ATTRIBUTE_UNUSED;
249 const cpp_token *token;
250 int parsing_args ATTRIBUTE_UNUSED;
252 src_lineno = SOURCE_LINE (map, token->line);
256 cb_file_change (pfile, new_map)
257 cpp_reader *pfile ATTRIBUTE_UNUSED;
258 const struct line_map *new_map;
260 unsigned int to_line = SOURCE_LINE (new_map, new_map->to_line);
262 if (new_map->reason == LC_ENTER)
264 /* Don't stack the main buffer on the input stack;
265 we already did in compile_file. */
267 main_input_filename = new_map->to_file;
270 lineno = SOURCE_LINE (new_map - 1, new_map->from_line - 1);
271 push_srcloc (new_map->to_file, 1);
272 input_file_stack->indent_level = indent_level;
273 (*debug_hooks->start_source_file) (lineno, new_map->to_file);
274 #ifndef NO_IMPLICIT_EXTERN_C
277 else if (new_map->sysp == 2)
280 ++pending_lang_change;
285 else if (new_map->reason == LC_LEAVE)
287 #ifndef NO_IMPLICIT_EXTERN_C
288 if (c_header_level && --c_header_level == 0)
290 if (new_map->sysp == 2)
291 warning ("badly nested C headers from preprocessor");
292 --pending_lang_change;
296 if (indent_level != input_file_stack->indent_level)
298 warning_with_file_and_line
299 (input_filename, lineno,
300 "This file contains more '%c's than '%c's.",
301 indent_level > input_file_stack->indent_level ? '{' : '}',
302 indent_level > input_file_stack->indent_level ? '}' : '{');
307 (*debug_hooks->end_source_file) (to_line);
310 update_header_times (new_map->to_file);
311 in_system_header = new_map->sysp != 0;
312 input_filename = new_map->to_file;
317 extract_interface_info ();
321 cb_def_pragma (pfile, line)
325 /* Issue a warning message if we have been asked to do so. Ignore
326 unknown pragmas in system headers unless an explicit
327 -Wunknown-pragmas has been given. */
328 if (warn_unknown_pragmas > in_system_header)
330 const unsigned char *space, *name = 0;
333 s = cpp_get_token (pfile);
334 space = cpp_token_as_text (pfile, s);
335 s = cpp_get_token (pfile);
336 if (s->type == CPP_NAME)
337 name = cpp_token_as_text (pfile, s);
339 lineno = SOURCE_LINE (map, line);
341 warning ("ignoring #pragma %s %s", space, name);
343 warning ("ignoring #pragma %s", space);
347 /* #define callback for DWARF and DWARF2 debug info. */
349 cb_define (pfile, line, node)
354 (*debug_hooks->define) (SOURCE_LINE (map, line),
355 (const char *) cpp_macro_definition (pfile, node));
358 /* #undef callback for DWARF and DWARF2 debug info. */
360 cb_undef (pfile, line, node)
361 cpp_reader *pfile ATTRIBUTE_UNUSED;
365 (*debug_hooks->undef) (SOURCE_LINE (map, line),
366 (const char *) NODE_NAME (node));
370 /* Returns nonzero if C is a universal-character-name. Give an error if it
371 is not one which may appear in an identifier, as per [extendid].
373 Note that extended character support in identifiers has not yet been
374 implemented. It is my personal opinion that this is not a desirable
375 feature. Portable code cannot count on support for more than the basic
376 identifier character set. */
389 /* None of the valid chars are outside the Basic Multilingual Plane (the
393 error ("universal-character-name '\\U%08x' not valid in identifier", c);
398 if ((c >= 0x00c0 && c <= 0x00d6)
399 || (c >= 0x00d8 && c <= 0x00f6)
400 || (c >= 0x00f8 && c <= 0x01f5)
401 || (c >= 0x01fa && c <= 0x0217)
402 || (c >= 0x0250 && c <= 0x02a8)
403 || (c >= 0x1e00 && c <= 0x1e9a)
404 || (c >= 0x1ea0 && c <= 0x1ef9))
409 || (c >= 0x0388 && c <= 0x038a)
411 || (c >= 0x038e && c <= 0x03a1)
412 || (c >= 0x03a3 && c <= 0x03ce)
413 || (c >= 0x03d0 && c <= 0x03d6)
418 || (c >= 0x03e2 && c <= 0x03f3)
419 || (c >= 0x1f00 && c <= 0x1f15)
420 || (c >= 0x1f18 && c <= 0x1f1d)
421 || (c >= 0x1f20 && c <= 0x1f45)
422 || (c >= 0x1f48 && c <= 0x1f4d)
423 || (c >= 0x1f50 && c <= 0x1f57)
427 || (c >= 0x1f5f && c <= 0x1f7d)
428 || (c >= 0x1f80 && c <= 0x1fb4)
429 || (c >= 0x1fb6 && c <= 0x1fbc)
430 || (c >= 0x1fc2 && c <= 0x1fc4)
431 || (c >= 0x1fc6 && c <= 0x1fcc)
432 || (c >= 0x1fd0 && c <= 0x1fd3)
433 || (c >= 0x1fd6 && c <= 0x1fdb)
434 || (c >= 0x1fe0 && c <= 0x1fec)
435 || (c >= 0x1ff2 && c <= 0x1ff4)
436 || (c >= 0x1ff6 && c <= 0x1ffc))
440 if ((c >= 0x0401 && c <= 0x040d)
441 || (c >= 0x040f && c <= 0x044f)
442 || (c >= 0x0451 && c <= 0x045c)
443 || (c >= 0x045e && c <= 0x0481)
444 || (c >= 0x0490 && c <= 0x04c4)
445 || (c >= 0x04c7 && c <= 0x04c8)
446 || (c >= 0x04cb && c <= 0x04cc)
447 || (c >= 0x04d0 && c <= 0x04eb)
448 || (c >= 0x04ee && c <= 0x04f5)
449 || (c >= 0x04f8 && c <= 0x04f9))
453 if ((c >= 0x0531 && c <= 0x0556)
454 || (c >= 0x0561 && c <= 0x0587))
458 if ((c >= 0x05d0 && c <= 0x05ea)
459 || (c >= 0x05f0 && c <= 0x05f4))
463 if ((c >= 0x0621 && c <= 0x063a)
464 || (c >= 0x0640 && c <= 0x0652)
465 || (c >= 0x0670 && c <= 0x06b7)
466 || (c >= 0x06ba && c <= 0x06be)
467 || (c >= 0x06c0 && c <= 0x06ce)
468 || (c >= 0x06e5 && c <= 0x06e7))
472 if ((c >= 0x0905 && c <= 0x0939)
473 || (c >= 0x0958 && c <= 0x0962))
477 if ((c >= 0x0985 && c <= 0x098c)
478 || (c >= 0x098f && c <= 0x0990)
479 || (c >= 0x0993 && c <= 0x09a8)
480 || (c >= 0x09aa && c <= 0x09b0)
482 || (c >= 0x09b6 && c <= 0x09b9)
483 || (c >= 0x09dc && c <= 0x09dd)
484 || (c >= 0x09df && c <= 0x09e1)
485 || (c >= 0x09f0 && c <= 0x09f1))
489 if ((c >= 0x0a05 && c <= 0x0a0a)
490 || (c >= 0x0a0f && c <= 0x0a10)
491 || (c >= 0x0a13 && c <= 0x0a28)
492 || (c >= 0x0a2a && c <= 0x0a30)
493 || (c >= 0x0a32 && c <= 0x0a33)
494 || (c >= 0x0a35 && c <= 0x0a36)
495 || (c >= 0x0a38 && c <= 0x0a39)
496 || (c >= 0x0a59 && c <= 0x0a5c)
501 if ((c >= 0x0a85 && c <= 0x0a8b)
503 || (c >= 0x0a8f && c <= 0x0a91)
504 || (c >= 0x0a93 && c <= 0x0aa8)
505 || (c >= 0x0aaa && c <= 0x0ab0)
506 || (c >= 0x0ab2 && c <= 0x0ab3)
507 || (c >= 0x0ab5 && c <= 0x0ab9)
512 if ((c >= 0x0b05 && c <= 0x0b0c)
513 || (c >= 0x0b0f && c <= 0x0b10)
514 || (c >= 0x0b13 && c <= 0x0b28)
515 || (c >= 0x0b2a && c <= 0x0b30)
516 || (c >= 0x0b32 && c <= 0x0b33)
517 || (c >= 0x0b36 && c <= 0x0b39)
518 || (c >= 0x0b5c && c <= 0x0b5d)
519 || (c >= 0x0b5f && c <= 0x0b61))
523 if ((c >= 0x0b85 && c <= 0x0b8a)
524 || (c >= 0x0b8e && c <= 0x0b90)
525 || (c >= 0x0b92 && c <= 0x0b95)
526 || (c >= 0x0b99 && c <= 0x0b9a)
528 || (c >= 0x0b9e && c <= 0x0b9f)
529 || (c >= 0x0ba3 && c <= 0x0ba4)
530 || (c >= 0x0ba8 && c <= 0x0baa)
531 || (c >= 0x0bae && c <= 0x0bb5)
532 || (c >= 0x0bb7 && c <= 0x0bb9))
536 if ((c >= 0x0c05 && c <= 0x0c0c)
537 || (c >= 0x0c0e && c <= 0x0c10)
538 || (c >= 0x0c12 && c <= 0x0c28)
539 || (c >= 0x0c2a && c <= 0x0c33)
540 || (c >= 0x0c35 && c <= 0x0c39)
541 || (c >= 0x0c60 && c <= 0x0c61))
545 if ((c >= 0x0c85 && c <= 0x0c8c)
546 || (c >= 0x0c8e && c <= 0x0c90)
547 || (c >= 0x0c92 && c <= 0x0ca8)
548 || (c >= 0x0caa && c <= 0x0cb3)
549 || (c >= 0x0cb5 && c <= 0x0cb9)
550 || (c >= 0x0ce0 && c <= 0x0ce1))
554 if ((c >= 0x0d05 && c <= 0x0d0c)
555 || (c >= 0x0d0e && c <= 0x0d10)
556 || (c >= 0x0d12 && c <= 0x0d28)
557 || (c >= 0x0d2a && c <= 0x0d39)
558 || (c >= 0x0d60 && c <= 0x0d61))
562 if ((c >= 0x0e01 && c <= 0x0e30)
563 || (c >= 0x0e32 && c <= 0x0e33)
564 || (c >= 0x0e40 && c <= 0x0e46)
565 || (c >= 0x0e4f && c <= 0x0e5b))
569 if ((c >= 0x0e81 && c <= 0x0e82)
575 || (c >= 0x0e94 && c <= 0x0e97)
576 || (c >= 0x0e99 && c <= 0x0e9f)
577 || (c >= 0x0ea1 && c <= 0x0ea3)
582 || (c >= 0x0ead && c <= 0x0eb0)
586 || (c >= 0x0ec0 && c <= 0x0ec4)
591 if ((c >= 0x10a0 && c <= 0x10c5)
592 || (c >= 0x10d0 && c <= 0x10f6))
596 if ((c >= 0x3041 && c <= 0x3094)
597 || (c >= 0x309b && c <= 0x309e))
601 if ((c >= 0x30a1 && c <= 0x30fe))
605 if ((c >= 0x3105 && c <= 0x312c))
609 if ((c >= 0x1100 && c <= 0x1159)
610 || (c >= 0x1161 && c <= 0x11a2)
611 || (c >= 0x11a8 && c <= 0x11f9))
614 /* CJK Unified Ideographs */
615 if ((c >= 0xf900 && c <= 0xfa2d)
616 || (c >= 0xfb1f && c <= 0xfb36)
617 || (c >= 0xfb38 && c <= 0xfb3c)
619 || (c >= 0xfb40 && c <= 0xfb41)
620 || (c >= 0xfb42 && c <= 0xfb44)
621 || (c >= 0xfb46 && c <= 0xfbb1)
622 || (c >= 0xfbd3 && c <= 0xfd3f)
623 || (c >= 0xfd50 && c <= 0xfd8f)
624 || (c >= 0xfd92 && c <= 0xfdc7)
625 || (c >= 0xfdf0 && c <= 0xfdfb)
626 || (c >= 0xfe70 && c <= 0xfe72)
628 || (c >= 0xfe76 && c <= 0xfefc)
629 || (c >= 0xff21 && c <= 0xff3a)
630 || (c >= 0xff41 && c <= 0xff5a)
631 || (c >= 0xff66 && c <= 0xffbe)
632 || (c >= 0xffc2 && c <= 0xffc7)
633 || (c >= 0xffca && c <= 0xffcf)
634 || (c >= 0xffd2 && c <= 0xffd7)
635 || (c >= 0xffda && c <= 0xffdc)
636 || (c >= 0x4e00 && c <= 0x9fa5))
639 error ("universal-character-name '\\u%04x' not valid in identifier", c);
644 /* Add the UTF-8 representation of C to the token_buffer. */
647 utf8_extend_token (c)
657 else if (c <= 0x000007ff)
658 shift = 6, mask = 0xc0;
659 else if (c <= 0x0000ffff)
660 shift = 12, mask = 0xe0;
661 else if (c <= 0x001fffff)
662 shift = 18, mask = 0xf0;
663 else if (c <= 0x03ffffff)
664 shift = 24, mask = 0xf8;
666 shift = 30, mask = 0xfc;
668 extend_token (mask | (c >> shift));
672 extend_token ((unsigned char) (0x80 | (c >> shift)));
681 tree *const node_var;
682 const char unsigned_flag;
683 const char long_flag;
684 const char long_long_flag;
687 struct try_type type_sequence[] =
689 { &integer_type_node, 0, 0, 0},
690 { &unsigned_type_node, 1, 0, 0},
691 { &long_integer_type_node, 0, 1, 0},
692 { &long_unsigned_type_node, 1, 1, 0},
693 { &long_long_integer_type_node, 0, 1, 1},
694 { &long_long_unsigned_type_node, 1, 1, 1}
706 int conversion_errno;
707 REAL_VALUE_TYPE value;
715 struct pf_args * args = (struct pf_args *) data;
716 const char *typename;
718 args->conversion_errno = 0;
719 args->type = double_type_node;
722 /* The second argument, machine_mode, of REAL_VALUE_ATOF
723 tells the desired precision of the binary result
724 of decimal-to-binary conversion. */
729 error ("both 'f' and 'l' suffixes on floating constant");
731 args->type = float_type_node;
734 else if (args->lflag)
736 args->type = long_double_type_node;
737 typename = "long double";
739 else if (flag_single_precision_constant)
741 args->type = float_type_node;
746 if (args->base == 16)
747 args->value = REAL_VALUE_HTOF (args->str, TYPE_MODE (args->type));
749 args->value = REAL_VALUE_ATOF (args->str, TYPE_MODE (args->type));
751 args->conversion_errno = errno;
752 /* A diagnostic is required here by some ISO C testsuites.
753 This is not pedwarn, because some people don't want
754 an error for this. */
755 if (REAL_VALUE_ISINF (args->value) && pedantic)
756 warning ("floating point number exceeds range of '%s'", typename);
763 const cpp_token *tok;
766 timevar_push (TV_CPP);
768 tok = cpp_get_token (parse_in);
769 while (tok->type == CPP_PADDING);
770 timevar_pop (TV_CPP);
772 /* The C++ front end does horrible things with the current line
773 number. To ensure an accurate line number, we must reset it
774 every time we return a token. */
780 case CPP_OPEN_BRACE: indent_level++; break;
781 case CPP_CLOSE_BRACE: indent_level--; break;
783 /* Issue this error here, where we can get at tok->val.c. */
785 if (ISGRAPH (tok->val.c))
786 error ("stray '%c' in program", tok->val.c);
788 error ("stray '\\%o' in program", tok->val.c);
792 *value = HT_IDENT_TO_GCC_IDENT (HT_NODE (tok->val.node));
796 *value = lex_number ((const char *)tok->val.str.text, tok->val.str.len);
801 *value = lex_charconst (tok);
806 *value = lex_string ((const char *)tok->val.str.text,
807 tok->val.str.len, tok->type == CPP_WSTRING);
810 /* These tokens should not be visible outside cpplib. */
811 case CPP_HEADER_NAME:
822 #define ERROR(msgid) do { error(msgid); goto syntax_error; } while(0)
825 lex_number (str, len)
831 int largest_digit = 0;
837 enum anon1 { NOT_FLOAT = 0, AFTER_POINT, AFTER_EXPON } floatflag = NOT_FLOAT;
839 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
840 The code below which fills the parts array assumes that a host
841 int is at least twice as wide as a host char, and that
842 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
843 Two HOST_WIDE_INTs is the largest int literal we can store.
844 In order to detect overflow below, the number of parts (TOTAL_PARTS)
845 must be exactly the number of parts needed to hold the bits
846 of two HOST_WIDE_INTs. */
847 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
848 unsigned int parts[TOTAL_PARTS];
850 /* Optimize for most frequent case. */
854 return integer_zero_node;
855 else if (*str == '1')
856 return integer_one_node;
858 return build_int_2 (*str - '0', 0);
861 for (count = 0; count < TOTAL_PARTS; count++)
864 /* len is known to be >1 at this point. */
867 if (len > 2 && str[0] == '0' && (str[1] == 'x' || str[1] == 'X'))
872 /* The ISDIGIT check is so we are not confused by a suffix on 0. */
873 else if (str[0] == '0' && ISDIGIT (str[1]))
885 if (floatflag == AFTER_POINT)
886 ERROR ("too many decimal points in floating constant");
887 else if (floatflag == AFTER_EXPON)
888 ERROR ("decimal point in exponent - impossible!");
890 floatflag = AFTER_POINT;
896 /* Possible future extension: silently ignore _ in numbers,
897 permitting cosmetic grouping - e.g. 0x8000_0000 == 0x80000000
898 but somewhat easier to read. Ada has this? */
899 ERROR ("underscore in number");
903 /* It is not a decimal point.
904 It should be a digit (perhaps a hex digit). */
910 else if (base <= 10 && (c == 'e' || c == 'E'))
913 floatflag = AFTER_EXPON;
916 else if (base == 16 && (c == 'p' || c == 'P'))
918 floatflag = AFTER_EXPON;
919 break; /* start of exponent */
921 else if (base == 16 && c >= 'a' && c <= 'f')
925 else if (base == 16 && c >= 'A' && c <= 'F')
932 break; /* start of suffix */
935 if (n >= largest_digit)
939 for (count = 0; count < TOTAL_PARTS; count++)
941 parts[count] *= base;
945 += (parts[count-1] >> HOST_BITS_PER_CHAR);
947 &= (1 << HOST_BITS_PER_CHAR) - 1;
953 /* If the highest-order part overflows (gets larger than
954 a host char will hold) then the whole number has
955 overflowed. Record this and truncate the highest-order
957 if (parts[TOTAL_PARTS - 1] >> HOST_BITS_PER_CHAR)
960 parts[TOTAL_PARTS - 1] &= (1 << HOST_BITS_PER_CHAR) - 1;
964 while (p < str + len);
966 /* This can happen on input like `int i = 0x;' */
968 ERROR ("numeric constant with no digits");
970 if (largest_digit >= base)
971 ERROR ("numeric constant contains digits beyond the radix");
973 if (floatflag != NOT_FLOAT)
976 int imag, fflag, lflag, conversion_errno;
977 REAL_VALUE_TYPE real;
981 if (base == 16 && pedantic && !flag_isoc99)
982 pedwarn ("floating constant may not be in radix 16");
984 if (base == 16 && floatflag != AFTER_EXPON)
985 ERROR ("hexadecimal floating constant has no exponent");
987 /* Read explicit exponent if any, and put it in tokenbuf. */
988 if ((base == 10 && ((c == 'e') || (c == 'E')))
989 || (base == 16 && (c == 'p' || c == 'P')))
993 if (p < str + len && (c == '+' || c == '-'))
995 /* Exponent is decimal, even if string is a hex float. */
997 ERROR ("floating constant exponent has no digits");
998 while (p < str + len && ISDIGIT (c))
1004 /* Copy the float constant now; we don't want any suffixes in the
1005 string passed to parse_float. */
1006 copy = alloca (p - str + 1);
1007 memcpy (copy, str, p - str);
1008 copy[p - str] = '\0';
1010 /* Now parse suffixes. */
1011 fflag = lflag = imag = 0;
1012 while (p < str + len)
1017 ERROR ("more than one 'f' suffix on floating constant");
1018 else if (warn_traditional && !in_system_header
1019 && ! cpp_sys_macro_p (parse_in))
1020 warning ("traditional C rejects the 'f' suffix");
1027 ERROR ("more than one 'l' suffix on floating constant");
1028 else if (warn_traditional && !in_system_header
1029 && ! cpp_sys_macro_p (parse_in))
1030 warning ("traditional C rejects the 'l' suffix");
1038 ERROR ("more than one 'i' or 'j' suffix on floating constant");
1040 pedwarn ("ISO C forbids imaginary numeric constants");
1045 ERROR ("invalid suffix on floating constant");
1048 /* Setup input for parse_float() */
1054 /* Convert string to a double, checking for overflow. */
1055 if (do_float_handler (parse_float, (PTR) &args))
1057 /* Receive output from parse_float() */
1061 /* We got an exception from parse_float() */
1062 ERROR ("floating constant out of range");
1064 /* Receive output from parse_float() */
1065 conversion_errno = args.conversion_errno;
1069 /* ERANGE is also reported for underflow,
1070 so test the value to distinguish overflow from that. */
1071 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1072 && (REAL_VALUES_LESS (dconst1, real)
1073 || REAL_VALUES_LESS (real, dconstm1)))
1074 warning ("floating point number exceeds range of 'double'");
1077 /* Create a node with determined type and value. */
1079 value = build_complex (NULL_TREE, convert (type, integer_zero_node),
1080 build_real (type, real));
1082 value = build_real (type, real);
1086 tree trad_type, ansi_type, type;
1087 HOST_WIDE_INT high, low;
1088 int spec_unsigned = 0;
1090 int spec_long_long = 0;
1095 trad_type = ansi_type = type = NULL_TREE;
1096 while (p < str + len)
1103 error ("two 'u' suffixes on integer constant");
1104 else if (warn_traditional && !in_system_header
1105 && ! cpp_sys_macro_p (parse_in))
1106 warning ("traditional C rejects the 'u' suffix");
1117 error ("three 'l' suffixes on integer constant");
1119 error ("'lul' is not a valid integer suffix");
1120 else if (c != spec_long)
1121 error ("'Ll' and 'lL' are not valid integer suffixes");
1122 else if (pedantic && ! flag_isoc99
1123 && ! in_system_header && warn_long_long)
1124 pedwarn ("ISO C89 forbids long long integer constants");
1130 case 'i': case 'I': case 'j': case 'J':
1132 error ("more than one 'i' or 'j' suffix on integer constant");
1134 pedwarn ("ISO C forbids imaginary numeric constants");
1139 ERROR ("invalid suffix on integer constant");
1143 /* If the literal overflowed, pedwarn about it now. */
1147 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT * 2);
1150 /* This is simplified by the fact that our constant
1151 is always positive. */
1155 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1157 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1158 / HOST_BITS_PER_CHAR)]
1159 << (i * HOST_BITS_PER_CHAR));
1160 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1163 value = build_int_2 (low, high);
1164 TREE_TYPE (value) = long_long_unsigned_type_node;
1166 /* If warn_traditional, calculate both the ISO type and the
1167 traditional type, then see if they disagree.
1168 Otherwise, calculate only the type for the dialect in use. */
1169 if (warn_traditional || flag_traditional)
1171 /* Calculate the traditional type. */
1172 /* Traditionally, any constant is signed; but if unsigned is
1173 specified explicitly, obey that. Use the smallest size
1174 with the right number of bits, except for one special
1175 case with decimal constants. */
1176 if (! spec_long && base != 10
1177 && int_fits_type_p (value, unsigned_type_node))
1178 trad_type = spec_unsigned ? unsigned_type_node : integer_type_node;
1179 /* A decimal constant must be long if it does not fit in
1180 type int. I think this is independent of whether the
1181 constant is signed. */
1182 else if (! spec_long && base == 10
1183 && int_fits_type_p (value, integer_type_node))
1184 trad_type = spec_unsigned ? unsigned_type_node : integer_type_node;
1185 else if (! spec_long_long)
1186 trad_type = (spec_unsigned
1187 ? long_unsigned_type_node
1188 : long_integer_type_node);
1189 else if (int_fits_type_p (value,
1191 ? long_long_unsigned_type_node
1192 : long_long_integer_type_node))
1193 trad_type = (spec_unsigned
1194 ? long_long_unsigned_type_node
1195 : long_long_integer_type_node);
1197 trad_type = (spec_unsigned
1198 ? widest_unsigned_literal_type_node
1199 : widest_integer_literal_type_node);
1201 if (warn_traditional || ! flag_traditional)
1203 /* Calculate the ISO type. */
1204 if (! spec_long && ! spec_unsigned
1205 && int_fits_type_p (value, integer_type_node))
1206 ansi_type = integer_type_node;
1207 else if (! spec_long && (base != 10 || spec_unsigned)
1208 && int_fits_type_p (value, unsigned_type_node))
1209 ansi_type = unsigned_type_node;
1210 else if (! spec_unsigned && !spec_long_long
1211 && int_fits_type_p (value, long_integer_type_node))
1212 ansi_type = long_integer_type_node;
1213 else if (! spec_long_long
1214 && int_fits_type_p (value, long_unsigned_type_node))
1215 ansi_type = long_unsigned_type_node;
1216 else if (! spec_unsigned
1217 && int_fits_type_p (value, long_long_integer_type_node))
1218 ansi_type = long_long_integer_type_node;
1219 else if (int_fits_type_p (value, long_long_unsigned_type_node))
1220 ansi_type = long_long_unsigned_type_node;
1221 else if (! spec_unsigned
1222 && int_fits_type_p (value, widest_integer_literal_type_node))
1223 ansi_type = widest_integer_literal_type_node;
1225 ansi_type = widest_unsigned_literal_type_node;
1228 type = flag_traditional ? trad_type : ansi_type;
1230 /* We assume that constants specified in a non-decimal
1231 base are bit patterns, and that the programmer really
1232 meant what they wrote. */
1233 if (warn_traditional && !in_system_header
1234 && base == 10 && trad_type != ansi_type)
1236 if (TYPE_PRECISION (trad_type) != TYPE_PRECISION (ansi_type))
1237 warning ("width of integer constant changes with -traditional");
1238 else if (TREE_UNSIGNED (trad_type) != TREE_UNSIGNED (ansi_type))
1239 warning ("integer constant is unsigned in ISO C, signed with -traditional");
1241 warning ("width of integer constant may change on other systems with -traditional");
1244 if (pedantic && !flag_traditional && (flag_isoc99 || !spec_long_long)
1247 ? TYPE_PRECISION (long_long_integer_type_node)
1248 : TYPE_PRECISION (long_integer_type_node)) < TYPE_PRECISION (type)))
1251 pedwarn ("integer constant larger than the maximum value of %s",
1253 ? (TREE_UNSIGNED (type)
1254 ? _("an unsigned long long int")
1255 : _("a long long int"))
1256 : _("an unsigned long int")));
1259 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1260 warning ("decimal constant is so large that it is unsigned");
1264 if (TYPE_PRECISION (type)
1265 <= TYPE_PRECISION (integer_type_node))
1266 value = build_complex (NULL_TREE, integer_zero_node,
1267 convert (integer_type_node, value));
1269 ERROR ("complex integer constant is too wide for 'complex int'");
1271 else if (flag_traditional && !int_fits_type_p (value, type))
1272 /* The traditional constant 0x80000000 is signed
1273 but doesn't fit in the range of int.
1274 This will change it to -0x80000000, which does fit. */
1276 TREE_TYPE (value) = unsigned_type (type);
1277 value = convert (type, value);
1278 TREE_OVERFLOW (value) = TREE_CONSTANT_OVERFLOW (value) = 0;
1281 TREE_TYPE (value) = type;
1283 /* If it's still an integer (not a complex), and it doesn't
1284 fit in the type we choose for it, then pedwarn. */
1287 && TREE_CODE (TREE_TYPE (value)) == INTEGER_TYPE
1288 && ! int_fits_type_p (value, TREE_TYPE (value)))
1289 pedwarn ("integer constant is larger than the maximum value for its type");
1293 error ("missing white space after number '%.*s'", (int) (p - str), str);
1298 return integer_zero_node;
1302 lex_string (str, len, wide)
1308 char *buf = alloca ((len + 1) * (wide ? WCHAR_BYTES : 1));
1310 const char *p = str, *limit = str + len;
1312 unsigned width = wide ? WCHAR_TYPE_SIZE
1313 : TYPE_PRECISION (char_type_node);
1315 #ifdef MULTIBYTE_CHARS
1316 /* Reset multibyte conversion state. */
1317 (void) local_mbtowc (NULL, NULL, 0);
1322 #ifdef MULTIBYTE_CHARS
1326 char_len = local_mbtowc (&wc, p, limit - p);
1329 warning ("ignoring invalid multibyte character");
1342 if (c == '\\' && !ignore_escape_flag)
1346 if (width < HOST_BITS_PER_INT)
1347 mask = ((unsigned int) 1 << width) - 1;
1350 c = cpp_parse_escape (parse_in, (const unsigned char **) &p,
1351 (const unsigned char *) limit,
1352 mask, flag_traditional);
1355 /* Add this single character into the buffer either as a wchar_t
1356 or as a single byte. */
1359 unsigned charwidth = TYPE_PRECISION (char_type_node);
1360 unsigned bytemask = (1 << charwidth) - 1;
1363 for (byte = 0; byte < WCHAR_BYTES; ++byte)
1366 if (byte >= (int) sizeof (c))
1369 n = (c >> (byte * charwidth)) & bytemask;
1370 if (BYTES_BIG_ENDIAN)
1371 q[WCHAR_BYTES - byte - 1] = n;
1383 /* Terminate the string value, either with a single byte zero
1384 or with a wide zero. */
1388 memset (q, 0, WCHAR_BYTES);
1396 value = build_string (q - buf, buf);
1399 TREE_TYPE (value) = wchar_array_type_node;
1401 TREE_TYPE (value) = char_array_type_node;
1405 /* Converts a (possibly wide) character constant token into a tree. */
1407 lex_charconst (token)
1408 const cpp_token *token;
1410 HOST_WIDE_INT result;
1412 unsigned int chars_seen;
1414 result = cpp_interpret_charconst (parse_in, token, warn_multichar,
1415 flag_traditional, &chars_seen);
1416 if (token->type == CPP_WCHAR)
1418 value = build_int_2 (result, 0);
1419 TREE_TYPE (value) = wchar_type_node;
1424 value = build_int_2 (result, -1);
1426 value = build_int_2 (result, 0);
1428 /* In C, a character constant has type 'int'.
1429 In C++ 'char', but multi-char charconsts have type 'int'. */
1430 if (c_language == clk_cplusplus && chars_seen <= 1)
1431 TREE_TYPE (value) = char_type_node;
1433 TREE_TYPE (value) = integer_type_node;