1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 1988, 1989, 1992, 1994, 1995, 1996, 1997
3 1998, 1999, 2000 Free Software Foundation, Inc.
5 This file is part of GNU CC.
7 GNU CC is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2, or (at your option)
12 GNU CC is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with GNU CC; see the file COPYING. If not, write to
19 the Free Software Foundation, 59 Temple Place - Suite 330,
20 Boston, MA 02111-1307, USA. */
39 #include "splay-tree.h"
41 /* MULTIBYTE_CHARS support only works for native compilers.
42 ??? Ideally what we want is to model widechar support after
43 the current floating point support. */
45 #undef MULTIBYTE_CHARS
48 #ifdef MULTIBYTE_CHARS
51 #endif /* MULTIBYTE_CHARS */
52 #ifndef GET_ENVIRONMENT
53 #define GET_ENVIRONMENT(ENV_VALUE,ENV_NAME) ((ENV_VALUE) = getenv (ENV_NAME))
56 /* The original file name, before changing "-" to "stdin". */
57 static const char *orig_filename;
59 /* We may keep statistics about how long which files took to compile. */
60 static int header_time, body_time;
61 static splay_tree file_info_tree;
63 /* Cause the `yydebug' variable to be defined. */
66 /* File used for outputting assembler code. */
67 extern FILE *asm_out_file;
69 #undef WCHAR_TYPE_SIZE
70 #define WCHAR_TYPE_SIZE TYPE_PRECISION (wchar_type_node)
72 /* Number of bytes in a wide character. */
73 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
75 int indent_level; /* Number of { minus number of }. */
76 int pending_lang_change; /* If we need to switch languages - C++ only */
77 int c_header_level; /* depth in C headers - C++ only */
79 /* Nonzero tells yylex to ignore \ in string constants. */
80 static int ignore_escape_flag;
82 static const char *readescape PARAMS ((const char *, const char *,
84 static const char *read_ucs PARAMS ((const char *, const char *,
85 unsigned int *, int));
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 char *, unsigned int, int));
90 static void update_header_times PARAMS ((const char *));
91 static int dump_one_header PARAMS ((splay_tree_node, void *));
92 static void cb_ident PARAMS ((cpp_reader *, const cpp_string *));
93 static void cb_file_change PARAMS ((cpp_reader *, const cpp_file_change *));
94 static void cb_def_pragma PARAMS ((cpp_reader *));
95 static void cb_define PARAMS ((cpp_reader *, cpp_hashnode *));
96 static void cb_undef PARAMS ((cpp_reader *, cpp_hashnode *));
100 const char *filename;
102 struct cpp_callbacks *cb;
103 struct c_fileinfo *toplevel;
105 orig_filename = filename;
107 /* Set up filename timing. Must happen before cpp_start_read. */
108 file_info_tree = splay_tree_new ((splay_tree_compare_fn)strcmp,
110 (splay_tree_delete_value_fn)free);
111 toplevel = get_fileinfo ("<top level>");
112 if (flag_detailed_statistics)
115 body_time = get_run_time ();
116 toplevel->time = body_time;
119 #ifdef MULTIBYTE_CHARS
120 /* Change to the native locale for multibyte conversions. */
121 setlocale (LC_CTYPE, "");
122 GET_ENVIRONMENT (literal_codeset, "LANG");
125 cb = cpp_get_callbacks (parse_in);
127 cb->ident = cb_ident;
128 cb->file_change = cb_file_change;
129 cb->def_pragma = cb_def_pragma;
131 /* Set the debug callbacks if we can use them. */
132 if (debug_info_level == DINFO_LEVEL_VERBOSE
133 && (write_symbols == DWARF_DEBUG || write_symbols == DWARF2_DEBUG))
135 cb->define = cb_define;
136 cb->undef = cb_undef;
139 if (filename == 0 || !strcmp (filename, "-"))
148 /* A thin wrapper around the real parser that initializes the
149 integrated preprocessor after debug output has been initialized. */
154 if (! cpp_start_read (parse_in, orig_filename))
155 return 1; /* cpplib has emitted an error. */
165 struct c_fileinfo *fi;
167 n = splay_tree_lookup (file_info_tree, (splay_tree_key) name);
169 return (struct c_fileinfo *) n->value;
171 fi = (struct c_fileinfo *) xmalloc (sizeof (struct c_fileinfo));
173 fi->interface_only = 0;
174 fi->interface_unknown = 1;
175 splay_tree_insert (file_info_tree, (splay_tree_key) name,
176 (splay_tree_value) fi);
181 update_header_times (name)
184 /* Changing files again. This means currently collected time
185 is charged against header time, and body time starts back at 0. */
186 if (flag_detailed_statistics)
188 int this_time = get_run_time ();
189 struct c_fileinfo *file = get_fileinfo (name);
190 header_time += this_time - body_time;
191 file->time += this_time - body_time;
192 body_time = this_time;
197 dump_one_header (n, dummy)
199 void *dummy ATTRIBUTE_UNUSED;
201 print_time ((const char *) n->key,
202 ((struct c_fileinfo *) n->value)->time);
207 dump_time_statistics ()
209 struct c_fileinfo *file = get_fileinfo (input_filename);
210 int this_time = get_run_time ();
211 file->time += this_time - body_time;
213 fprintf (stderr, "\n******\n");
214 print_time ("header files (total)", header_time);
215 print_time ("main file (total)", this_time - body_time);
216 fprintf (stderr, "ratio = %g : 1\n",
217 (double)header_time / (double)(this_time - body_time));
218 fprintf (stderr, "\n******\n");
220 splay_tree_foreach (file_info_tree, dump_one_header, 0);
223 /* Not yet handled: #pragma, #define, #undef.
224 No need to deal with linemarkers under normal conditions. */
227 cb_ident (pfile, str)
228 cpp_reader *pfile ATTRIBUTE_UNUSED;
229 const cpp_string *str ATTRIBUTE_UNUSED;
231 #ifdef ASM_OUTPUT_IDENT
234 /* Convert escapes in the string. */
235 tree value = lex_string ((const char *)str->text, str->len, 0);
236 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (value));
242 cb_file_change (pfile, fc)
243 cpp_reader *pfile ATTRIBUTE_UNUSED;
244 const cpp_file_change *fc;
246 if (fc->reason == FC_ENTER)
248 /* Don't stack the main buffer on the input stack;
249 we already did in compile_file. */
250 if (fc->from.filename)
252 lineno = fc->from.lineno;
253 push_srcloc (fc->to.filename, 1);
254 input_file_stack->indent_level = indent_level;
255 debug_start_source_file (fc->to.filename);
256 #ifndef NO_IMPLICIT_EXTERN_C
259 else if (fc->externc)
262 ++pending_lang_change;
267 main_input_filename = fc->to.filename;
269 else if (fc->reason == FC_LEAVE)
271 /* Popping out of a file. */
272 if (input_file_stack->next)
274 #ifndef NO_IMPLICIT_EXTERN_C
275 if (c_header_level && --c_header_level == 0)
278 warning ("badly nested C headers from preprocessor");
279 --pending_lang_change;
283 if (indent_level != input_file_stack->indent_level)
285 warning_with_file_and_line
286 (input_filename, lineno,
287 "This file contains more '%c's than '%c's.",
288 indent_level > input_file_stack->indent_level ? '{' : '}',
289 indent_level > input_file_stack->indent_level ? '}' : '{');
293 debug_end_source_file (input_file_stack->line);
296 error ("leaving more files than we entered");
299 update_header_times (fc->to.filename);
300 in_system_header = fc->sysp != 0;
301 input_filename = fc->to.filename;
302 lineno = fc->to.lineno; /* Do we need this? */
305 extract_interface_info ();
309 cb_def_pragma (pfile)
312 /* Issue a warning message if we have been asked to do so. Ignore
313 unknown pragmas in system headers unless an explicit
314 -Wunknown-pragmas has been given. */
315 if (warn_unknown_pragmas > in_system_header)
317 const unsigned char *space, *name = 0;
320 cpp_get_token (pfile, &s);
321 space = cpp_token_as_text (pfile, &s);
322 cpp_get_token (pfile, &s);
323 if (s.type == CPP_NAME)
324 name = cpp_token_as_text (pfile, &s);
326 lineno = cpp_get_line (parse_in)->line;
328 warning ("ignoring #pragma %s %s", space, name);
330 warning ("ignoring #pragma %s", space);
334 /* #define callback for DWARF and DWARF2 debug info. */
336 cb_define (pfile, node)
340 debug_define (lineno, (const char *) cpp_macro_definition (pfile, node));
343 /* #undef callback for DWARF and DWARF2 debug info. */
345 cb_undef (pfile, node)
346 cpp_reader *pfile ATTRIBUTE_UNUSED;
349 debug_undef (lineno, (const char *) node->name);
352 /* Parse a '\uNNNN' or '\UNNNNNNNN' sequence.
354 [lex.charset]: The character designated by the universal-character-name
355 \UNNNNNNNN is that character whose character short name in ISO/IEC 10646
356 is NNNNNNNN; the character designated by the universal-character-name
357 \uNNNN is that character whose character short name in ISO/IEC 10646 is
358 0000NNNN. If the hexadecimal value for a universal character name is
359 less than 0x20 or in the range 0x7F-0x9F (inclusive), or if the
360 universal character name designates a character in the basic source
361 character set, then the program is ill-formed.
363 We assume that wchar_t is Unicode, so we don't need to do any
364 mapping. Is this ever wrong? */
367 read_ucs (p, limit, cptr, length)
373 unsigned int code = 0;
376 for (; length; --length)
380 error ("incomplete universal-character-name");
387 error ("non hex digit '%c' in universal-character-name", c);
393 if (c >= 'a' && c <= 'f')
394 code += c - 'a' + 10;
395 if (c >= 'A' && c <= 'F')
396 code += c - 'A' + 10;
397 if (c >= '0' && c <= '9')
402 sorry ("universal-character-name on EBCDIC target");
403 *cptr = 0x3f; /* EBCDIC invalid character */
407 if (code > 0x9f && !(code & 0x80000000))
408 /* True extended character, OK. */;
409 else if (code >= 0x20 && code < 0x7f)
411 /* ASCII printable character. The C character set consists of all of
412 these except $, @ and `. We use hex escapes so that this also
413 works with EBCDIC hosts. */
414 if (code != 0x24 && code != 0x40 && code != 0x60)
415 error ("universal-character-name used for '%c'", code);
418 error ("invalid universal-character-name");
424 /* Read an escape sequence and write its character equivalent into *CPTR.
425 P is the input pointer, which is just after the backslash. LIMIT
426 is how much text we have.
427 Returns the updated input pointer. */
430 readescape (p, limit, cptr)
435 unsigned int c, code, count;
436 unsigned firstdig = 0;
441 /* cpp has already issued an error for this. */
451 if (warn_traditional && !in_system_header)
452 warning ("the meaning of `\\x' varies with -traditional");
454 if (flag_traditional)
472 if (c >= 'a' && c <= 'f')
473 code += c - 'a' + 10;
474 if (c >= 'A' && c <= 'F')
475 code += c - 'A' + 10;
476 if (c >= '0' && c <= '9')
478 if (code != 0 || count != 0)
488 warning ("\\x used with no following hex digits");
493 /* Digits are all 0's. Ok. */
495 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
498 << (TYPE_PRECISION (integer_type_node)
501 pedwarn ("hex escape out of range");
505 case '0': case '1': case '2': case '3': case '4':
506 case '5': case '6': case '7':
508 for (count = 0; count < 3; count++)
510 if (c < '0' || c > '7')
515 code = (code * 8) + (c - '0');
527 case '\\': case '\'': case '"': case '?':
531 case 'n': *cptr = TARGET_NEWLINE; return p;
532 case 't': *cptr = TARGET_TAB; return p;
533 case 'r': *cptr = TARGET_CR; return p;
534 case 'f': *cptr = TARGET_FF; return p;
535 case 'b': *cptr = TARGET_BS; return p;
536 case 'v': *cptr = TARGET_VT; return p;
538 if (warn_traditional && !in_system_header)
539 warning ("the meaning of '\\a' varies with -traditional");
540 *cptr = flag_traditional ? c : TARGET_BELL;
543 /* Warnings and support checks handled by read_ucs(). */
545 if (c_language != clk_cplusplus && !flag_isoc99)
548 if (warn_traditional && !in_system_header)
549 warning ("the meaning of '\\%c' varies with -traditional", c);
551 return read_ucs (p, limit, cptr, c == 'u' ? 4 : 8);
555 pedwarn ("non-ISO-standard escape sequence, '\\%c'", c);
556 *cptr = TARGET_ESC; return p;
558 /* '\(', etc, are used at beginning of line to avoid confusing Emacs.
559 '\%' is used to prevent SCCS from getting confused. */
560 case '(': case '{': case '[': case '%':
562 pedwarn ("unknown escape sequence '\\%c'", c);
568 pedwarn ("unknown escape sequence '\\%c'", c);
570 pedwarn ("unknown escape sequence: '\\' followed by char 0x%x", c);
577 /* Returns nonzero if C is a universal-character-name. Give an error if it
578 is not one which may appear in an identifier, as per [extendid].
580 Note that extended character support in identifiers has not yet been
581 implemented. It is my personal opinion that this is not a desirable
582 feature. Portable code cannot count on support for more than the basic
583 identifier character set. */
596 /* None of the valid chars are outside the Basic Multilingual Plane (the
600 error ("universal-character-name '\\U%08x' not valid in identifier", c);
605 if ((c >= 0x00c0 && c <= 0x00d6)
606 || (c >= 0x00d8 && c <= 0x00f6)
607 || (c >= 0x00f8 && c <= 0x01f5)
608 || (c >= 0x01fa && c <= 0x0217)
609 || (c >= 0x0250 && c <= 0x02a8)
610 || (c >= 0x1e00 && c <= 0x1e9a)
611 || (c >= 0x1ea0 && c <= 0x1ef9))
616 || (c >= 0x0388 && c <= 0x038a)
618 || (c >= 0x038e && c <= 0x03a1)
619 || (c >= 0x03a3 && c <= 0x03ce)
620 || (c >= 0x03d0 && c <= 0x03d6)
625 || (c >= 0x03e2 && c <= 0x03f3)
626 || (c >= 0x1f00 && c <= 0x1f15)
627 || (c >= 0x1f18 && c <= 0x1f1d)
628 || (c >= 0x1f20 && c <= 0x1f45)
629 || (c >= 0x1f48 && c <= 0x1f4d)
630 || (c >= 0x1f50 && c <= 0x1f57)
634 || (c >= 0x1f5f && c <= 0x1f7d)
635 || (c >= 0x1f80 && c <= 0x1fb4)
636 || (c >= 0x1fb6 && c <= 0x1fbc)
637 || (c >= 0x1fc2 && c <= 0x1fc4)
638 || (c >= 0x1fc6 && c <= 0x1fcc)
639 || (c >= 0x1fd0 && c <= 0x1fd3)
640 || (c >= 0x1fd6 && c <= 0x1fdb)
641 || (c >= 0x1fe0 && c <= 0x1fec)
642 || (c >= 0x1ff2 && c <= 0x1ff4)
643 || (c >= 0x1ff6 && c <= 0x1ffc))
647 if ((c >= 0x0401 && c <= 0x040d)
648 || (c >= 0x040f && c <= 0x044f)
649 || (c >= 0x0451 && c <= 0x045c)
650 || (c >= 0x045e && c <= 0x0481)
651 || (c >= 0x0490 && c <= 0x04c4)
652 || (c >= 0x04c7 && c <= 0x04c8)
653 || (c >= 0x04cb && c <= 0x04cc)
654 || (c >= 0x04d0 && c <= 0x04eb)
655 || (c >= 0x04ee && c <= 0x04f5)
656 || (c >= 0x04f8 && c <= 0x04f9))
660 if ((c >= 0x0531 && c <= 0x0556)
661 || (c >= 0x0561 && c <= 0x0587))
665 if ((c >= 0x05d0 && c <= 0x05ea)
666 || (c >= 0x05f0 && c <= 0x05f4))
670 if ((c >= 0x0621 && c <= 0x063a)
671 || (c >= 0x0640 && c <= 0x0652)
672 || (c >= 0x0670 && c <= 0x06b7)
673 || (c >= 0x06ba && c <= 0x06be)
674 || (c >= 0x06c0 && c <= 0x06ce)
675 || (c >= 0x06e5 && c <= 0x06e7))
679 if ((c >= 0x0905 && c <= 0x0939)
680 || (c >= 0x0958 && c <= 0x0962))
684 if ((c >= 0x0985 && c <= 0x098c)
685 || (c >= 0x098f && c <= 0x0990)
686 || (c >= 0x0993 && c <= 0x09a8)
687 || (c >= 0x09aa && c <= 0x09b0)
689 || (c >= 0x09b6 && c <= 0x09b9)
690 || (c >= 0x09dc && c <= 0x09dd)
691 || (c >= 0x09df && c <= 0x09e1)
692 || (c >= 0x09f0 && c <= 0x09f1))
696 if ((c >= 0x0a05 && c <= 0x0a0a)
697 || (c >= 0x0a0f && c <= 0x0a10)
698 || (c >= 0x0a13 && c <= 0x0a28)
699 || (c >= 0x0a2a && c <= 0x0a30)
700 || (c >= 0x0a32 && c <= 0x0a33)
701 || (c >= 0x0a35 && c <= 0x0a36)
702 || (c >= 0x0a38 && c <= 0x0a39)
703 || (c >= 0x0a59 && c <= 0x0a5c)
708 if ((c >= 0x0a85 && c <= 0x0a8b)
710 || (c >= 0x0a8f && c <= 0x0a91)
711 || (c >= 0x0a93 && c <= 0x0aa8)
712 || (c >= 0x0aaa && c <= 0x0ab0)
713 || (c >= 0x0ab2 && c <= 0x0ab3)
714 || (c >= 0x0ab5 && c <= 0x0ab9)
719 if ((c >= 0x0b05 && c <= 0x0b0c)
720 || (c >= 0x0b0f && c <= 0x0b10)
721 || (c >= 0x0b13 && c <= 0x0b28)
722 || (c >= 0x0b2a && c <= 0x0b30)
723 || (c >= 0x0b32 && c <= 0x0b33)
724 || (c >= 0x0b36 && c <= 0x0b39)
725 || (c >= 0x0b5c && c <= 0x0b5d)
726 || (c >= 0x0b5f && c <= 0x0b61))
730 if ((c >= 0x0b85 && c <= 0x0b8a)
731 || (c >= 0x0b8e && c <= 0x0b90)
732 || (c >= 0x0b92 && c <= 0x0b95)
733 || (c >= 0x0b99 && c <= 0x0b9a)
735 || (c >= 0x0b9e && c <= 0x0b9f)
736 || (c >= 0x0ba3 && c <= 0x0ba4)
737 || (c >= 0x0ba8 && c <= 0x0baa)
738 || (c >= 0x0bae && c <= 0x0bb5)
739 || (c >= 0x0bb7 && c <= 0x0bb9))
743 if ((c >= 0x0c05 && c <= 0x0c0c)
744 || (c >= 0x0c0e && c <= 0x0c10)
745 || (c >= 0x0c12 && c <= 0x0c28)
746 || (c >= 0x0c2a && c <= 0x0c33)
747 || (c >= 0x0c35 && c <= 0x0c39)
748 || (c >= 0x0c60 && c <= 0x0c61))
752 if ((c >= 0x0c85 && c <= 0x0c8c)
753 || (c >= 0x0c8e && c <= 0x0c90)
754 || (c >= 0x0c92 && c <= 0x0ca8)
755 || (c >= 0x0caa && c <= 0x0cb3)
756 || (c >= 0x0cb5 && c <= 0x0cb9)
757 || (c >= 0x0ce0 && c <= 0x0ce1))
761 if ((c >= 0x0d05 && c <= 0x0d0c)
762 || (c >= 0x0d0e && c <= 0x0d10)
763 || (c >= 0x0d12 && c <= 0x0d28)
764 || (c >= 0x0d2a && c <= 0x0d39)
765 || (c >= 0x0d60 && c <= 0x0d61))
769 if ((c >= 0x0e01 && c <= 0x0e30)
770 || (c >= 0x0e32 && c <= 0x0e33)
771 || (c >= 0x0e40 && c <= 0x0e46)
772 || (c >= 0x0e4f && c <= 0x0e5b))
776 if ((c >= 0x0e81 && c <= 0x0e82)
782 || (c >= 0x0e94 && c <= 0x0e97)
783 || (c >= 0x0e99 && c <= 0x0e9f)
784 || (c >= 0x0ea1 && c <= 0x0ea3)
789 || (c >= 0x0ead && c <= 0x0eb0)
793 || (c >= 0x0ec0 && c <= 0x0ec4)
798 if ((c >= 0x10a0 && c <= 0x10c5)
799 || (c >= 0x10d0 && c <= 0x10f6))
803 if ((c >= 0x3041 && c <= 0x3094)
804 || (c >= 0x309b && c <= 0x309e))
808 if ((c >= 0x30a1 && c <= 0x30fe))
812 if ((c >= 0x3105 && c <= 0x312c))
816 if ((c >= 0x1100 && c <= 0x1159)
817 || (c >= 0x1161 && c <= 0x11a2)
818 || (c >= 0x11a8 && c <= 0x11f9))
821 /* CJK Unified Ideographs */
822 if ((c >= 0xf900 && c <= 0xfa2d)
823 || (c >= 0xfb1f && c <= 0xfb36)
824 || (c >= 0xfb38 && c <= 0xfb3c)
826 || (c >= 0xfb40 && c <= 0xfb41)
827 || (c >= 0xfb42 && c <= 0xfb44)
828 || (c >= 0xfb46 && c <= 0xfbb1)
829 || (c >= 0xfbd3 && c <= 0xfd3f)
830 || (c >= 0xfd50 && c <= 0xfd8f)
831 || (c >= 0xfd92 && c <= 0xfdc7)
832 || (c >= 0xfdf0 && c <= 0xfdfb)
833 || (c >= 0xfe70 && c <= 0xfe72)
835 || (c >= 0xfe76 && c <= 0xfefc)
836 || (c >= 0xff21 && c <= 0xff3a)
837 || (c >= 0xff41 && c <= 0xff5a)
838 || (c >= 0xff66 && c <= 0xffbe)
839 || (c >= 0xffc2 && c <= 0xffc7)
840 || (c >= 0xffca && c <= 0xffcf)
841 || (c >= 0xffd2 && c <= 0xffd7)
842 || (c >= 0xffda && c <= 0xffdc)
843 || (c >= 0x4e00 && c <= 0x9fa5))
846 error ("universal-character-name '\\u%04x' not valid in identifier", c);
851 /* Add the UTF-8 representation of C to the token_buffer. */
854 utf8_extend_token (c)
864 else if (c <= 0x000007ff)
865 shift = 6, mask = 0xc0;
866 else if (c <= 0x0000ffff)
867 shift = 12, mask = 0xe0;
868 else if (c <= 0x001fffff)
869 shift = 18, mask = 0xf0;
870 else if (c <= 0x03ffffff)
871 shift = 24, mask = 0xf8;
873 shift = 30, mask = 0xfc;
875 extend_token (mask | (c >> shift));
879 extend_token ((unsigned char) (0x80 | (c >> shift)));
894 struct try_type type_sequence[] =
896 { &integer_type_node, 0, 0, 0},
897 { &unsigned_type_node, 1, 0, 0},
898 { &long_integer_type_node, 0, 1, 0},
899 { &long_unsigned_type_node, 1, 1, 0},
900 { &long_long_integer_type_node, 0, 1, 1},
901 { &long_long_unsigned_type_node, 1, 1, 1}
913 int conversion_errno;
914 REAL_VALUE_TYPE value;
922 struct pf_args * args = (struct pf_args *) data;
923 const char *typename;
925 args->conversion_errno = 0;
926 args->type = double_type_node;
929 /* The second argument, machine_mode, of REAL_VALUE_ATOF
930 tells the desired precision of the binary result
931 of decimal-to-binary conversion. */
936 error ("both 'f' and 'l' suffixes on floating constant");
938 args->type = float_type_node;
941 else if (args->lflag)
943 args->type = long_double_type_node;
944 typename = "long double";
946 else if (flag_single_precision_constant)
948 args->type = float_type_node;
953 if (args->base == 16)
954 args->value = REAL_VALUE_HTOF (args->str, TYPE_MODE (args->type));
956 args->value = REAL_VALUE_ATOF (args->str, TYPE_MODE (args->type));
958 args->conversion_errno = errno;
959 /* A diagnostic is required here by some ISO C testsuites.
960 This is not pedwarn, because some people don't want
961 an error for this. */
962 if (REAL_VALUE_ISINF (args->value) && pedantic)
963 warning ("floating point number exceeds range of '%s'", typename);
974 timevar_push (TV_CPP);
975 cpp_get_token (parse_in, &tok);
976 timevar_pop (TV_CPP);
978 /* The C++ front end does horrible things with the current line
979 number. To ensure an accurate line number, we must reset it
980 every time we return a token. */
981 lineno = cpp_get_line (parse_in)->line;
987 case CPP_OPEN_BRACE: indent_level++; break;
988 case CPP_CLOSE_BRACE: indent_level--; break;
990 /* Issue this error here, where we can get at tok.val.c. */
992 if (ISGRAPH (tok.val.c))
993 error ("stray '%c' in program", tok.val.c);
995 error ("stray '\\%#o' in program", tok.val.c);
999 *value = get_identifier ((const char *)tok.val.node->name);
1005 *value = lex_number ((const char *)tok.val.str.text, tok.val.str.len);
1010 *value = lex_charconst ((const char *)tok.val.str.text,
1011 tok.val.str.len, tok.type == CPP_WCHAR);
1017 *value = lex_string ((const char *)tok.val.str.text,
1018 tok.val.str.len, tok.type == CPP_WSTRING);
1021 /* These tokens should not be visible outside cpplib. */
1022 case CPP_HEADER_NAME:
1033 #define ERROR(msgid) do { error(msgid); goto syntax_error; } while(0)
1036 lex_number (str, len)
1042 int largest_digit = 0;
1048 enum anon1 { NOT_FLOAT = 0, AFTER_POINT, AFTER_EXPON } floatflag = NOT_FLOAT;
1050 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
1051 The code below which fills the parts array assumes that a host
1052 int is at least twice as wide as a host char, and that
1053 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
1054 Two HOST_WIDE_INTs is the largest int literal we can store.
1055 In order to detect overflow below, the number of parts (TOTAL_PARTS)
1056 must be exactly the number of parts needed to hold the bits
1057 of two HOST_WIDE_INTs. */
1058 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
1059 unsigned int parts[TOTAL_PARTS];
1061 /* Optimize for most frequent case. */
1065 return integer_zero_node;
1066 else if (*str == '1')
1067 return integer_one_node;
1069 return build_int_2 (*str - '0', 0);
1072 for (count = 0; count < TOTAL_PARTS; count++)
1075 /* len is known to be >1 at this point. */
1078 if (len > 2 && str[0] == '0' && (str[1] == 'x' || str[1] == 'X'))
1083 /* The ISDIGIT check is so we are not confused by a suffix on 0. */
1084 else if (str[0] == '0' && ISDIGIT (str[1]))
1096 if (base == 16 && pedantic && !flag_isoc99)
1097 pedwarn ("floating constant may not be in radix 16");
1098 else if (floatflag == AFTER_POINT)
1099 ERROR ("too many decimal points in floating constant");
1100 else if (floatflag == AFTER_EXPON)
1101 ERROR ("decimal point in exponent - impossible!");
1103 floatflag = AFTER_POINT;
1109 /* Possible future extension: silently ignore _ in numbers,
1110 permitting cosmetic grouping - e.g. 0x8000_0000 == 0x80000000
1111 but somewhat easier to read. Ada has this? */
1112 ERROR ("underscore in number");
1116 /* It is not a decimal point.
1117 It should be a digit (perhaps a hex digit). */
1123 else if (base <= 10 && (c == 'e' || c == 'E'))
1126 floatflag = AFTER_EXPON;
1129 else if (base == 16 && (c == 'p' || c == 'P'))
1131 floatflag = AFTER_EXPON;
1132 break; /* start of exponent */
1134 else if (base == 16 && c >= 'a' && c <= 'f')
1138 else if (base == 16 && c >= 'A' && c <= 'F')
1145 break; /* start of suffix */
1148 if (n >= largest_digit)
1152 for (count = 0; count < TOTAL_PARTS; count++)
1154 parts[count] *= base;
1158 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1160 &= (1 << HOST_BITS_PER_CHAR) - 1;
1166 /* If the highest-order part overflows (gets larger than
1167 a host char will hold) then the whole number has
1168 overflowed. Record this and truncate the highest-order
1170 if (parts[TOTAL_PARTS - 1] >> HOST_BITS_PER_CHAR)
1173 parts[TOTAL_PARTS - 1] &= (1 << HOST_BITS_PER_CHAR) - 1;
1177 while (p < str + len);
1179 /* This can happen on input like `int i = 0x;' */
1181 ERROR ("numeric constant with no digits");
1183 if (largest_digit >= base)
1184 ERROR ("numeric constant contains digits beyond the radix");
1186 if (floatflag != NOT_FLOAT)
1189 int imag, fflag, lflag, conversion_errno;
1190 REAL_VALUE_TYPE real;
1191 struct pf_args args;
1194 if (base == 16 && floatflag != AFTER_EXPON)
1195 ERROR ("hexadecimal floating constant has no exponent");
1197 /* Read explicit exponent if any, and put it in tokenbuf. */
1198 if ((base == 10 && ((c == 'e') || (c == 'E')))
1199 || (base == 16 && (c == 'p' || c == 'P')))
1203 if (p < str + len && (c == '+' || c == '-'))
1205 /* Exponent is decimal, even if string is a hex float. */
1207 ERROR ("floating constant exponent has no digits");
1208 while (p < str + len && ISDIGIT (c))
1214 /* Copy the float constant now; we don't want any suffixes in the
1215 string passed to parse_float. */
1216 copy = alloca (p - str + 1);
1217 memcpy (copy, str, p - str);
1218 copy[p - str] = '\0';
1220 /* Now parse suffixes. */
1221 fflag = lflag = imag = 0;
1222 while (p < str + len)
1227 ERROR ("more than one 'f' suffix on floating constant");
1228 else if (warn_traditional && !in_system_header)
1229 warning ("traditional C rejects the 'f' suffix");
1236 ERROR ("more than one 'l' suffix on floating constant");
1237 else if (warn_traditional && !in_system_header)
1238 warning ("traditional C rejects the 'l' suffix");
1246 ERROR ("more than one 'i' or 'j' suffix on floating constant");
1248 pedwarn ("ISO C forbids imaginary numeric constants");
1253 ERROR ("invalid suffix on floating constant");
1256 /* Setup input for parse_float() */
1262 /* Convert string to a double, checking for overflow. */
1263 if (do_float_handler (parse_float, (PTR) &args))
1265 /* Receive output from parse_float() */
1269 /* We got an exception from parse_float() */
1270 ERROR ("floating constant out of range");
1272 /* Receive output from parse_float() */
1273 conversion_errno = args.conversion_errno;
1277 /* ERANGE is also reported for underflow,
1278 so test the value to distinguish overflow from that. */
1279 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1280 && (REAL_VALUES_LESS (dconst1, real)
1281 || REAL_VALUES_LESS (real, dconstm1)))
1282 warning ("floating point number exceeds range of 'double'");
1285 /* Create a node with determined type and value. */
1287 value = build_complex (NULL_TREE, convert (type, integer_zero_node),
1288 build_real (type, real));
1290 value = build_real (type, real);
1294 tree trad_type, ansi_type, type;
1295 HOST_WIDE_INT high, low;
1296 int spec_unsigned = 0;
1298 int spec_long_long = 0;
1303 trad_type = ansi_type = type = NULL_TREE;
1304 while (p < str + len)
1311 error ("two 'u' suffixes on integer constant");
1312 else if (warn_traditional && !in_system_header)
1313 warning ("traditional C rejects the 'u' suffix");
1324 error ("three 'l' suffixes on integer constant");
1326 error ("'lul' is not a valid integer suffix");
1327 else if (c != spec_long)
1328 error ("'Ll' and 'lL' are not valid integer suffixes");
1329 else if (pedantic && ! flag_isoc99
1330 && ! in_system_header && warn_long_long)
1331 pedwarn ("ISO C89 forbids long long integer constants");
1337 case 'i': case 'I': case 'j': case 'J':
1339 error ("more than one 'i' or 'j' suffix on integer constant");
1341 pedwarn ("ISO C forbids imaginary numeric constants");
1346 ERROR ("invalid suffix on integer constant");
1350 /* If the literal overflowed, pedwarn about it now. */
1354 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT * 2);
1357 /* This is simplified by the fact that our constant
1358 is always positive. */
1362 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1364 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1365 / HOST_BITS_PER_CHAR)]
1366 << (i * HOST_BITS_PER_CHAR));
1367 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1370 value = build_int_2 (low, high);
1371 TREE_TYPE (value) = long_long_unsigned_type_node;
1373 /* If warn_traditional, calculate both the ISO type and the
1374 traditional type, then see if they disagree.
1375 Otherwise, calculate only the type for the dialect in use. */
1376 if (warn_traditional || flag_traditional)
1378 /* Calculate the traditional type. */
1379 /* Traditionally, any constant is signed; but if unsigned is
1380 specified explicitly, obey that. Use the smallest size
1381 with the right number of bits, except for one special
1382 case with decimal constants. */
1383 if (! spec_long && base != 10
1384 && int_fits_type_p (value, unsigned_type_node))
1385 trad_type = spec_unsigned ? unsigned_type_node : integer_type_node;
1386 /* A decimal constant must be long if it does not fit in
1387 type int. I think this is independent of whether the
1388 constant is signed. */
1389 else if (! spec_long && base == 10
1390 && int_fits_type_p (value, integer_type_node))
1391 trad_type = spec_unsigned ? unsigned_type_node : integer_type_node;
1392 else if (! spec_long_long)
1393 trad_type = (spec_unsigned
1394 ? long_unsigned_type_node
1395 : long_integer_type_node);
1396 else if (int_fits_type_p (value,
1398 ? long_long_unsigned_type_node
1399 : long_long_integer_type_node))
1400 trad_type = (spec_unsigned
1401 ? long_long_unsigned_type_node
1402 : long_long_integer_type_node);
1404 trad_type = (spec_unsigned
1405 ? widest_unsigned_literal_type_node
1406 : widest_integer_literal_type_node);
1408 if (warn_traditional || ! flag_traditional)
1410 /* Calculate the ISO type. */
1411 if (! spec_long && ! spec_unsigned
1412 && int_fits_type_p (value, integer_type_node))
1413 ansi_type = integer_type_node;
1414 else if (! spec_long && (base != 10 || spec_unsigned)
1415 && int_fits_type_p (value, unsigned_type_node))
1416 ansi_type = unsigned_type_node;
1417 else if (! spec_unsigned && !spec_long_long
1418 && int_fits_type_p (value, long_integer_type_node))
1419 ansi_type = long_integer_type_node;
1420 else if (! spec_long_long
1421 && int_fits_type_p (value, long_unsigned_type_node))
1422 ansi_type = long_unsigned_type_node;
1423 else if (! spec_unsigned
1424 && int_fits_type_p (value, long_long_integer_type_node))
1425 ansi_type = long_long_integer_type_node;
1426 else if (int_fits_type_p (value, long_long_unsigned_type_node))
1427 ansi_type = long_long_unsigned_type_node;
1428 else if (! spec_unsigned
1429 && int_fits_type_p (value, widest_integer_literal_type_node))
1430 ansi_type = widest_integer_literal_type_node;
1432 ansi_type = widest_unsigned_literal_type_node;
1435 type = flag_traditional ? trad_type : ansi_type;
1437 /* We assume that constants specified in a non-decimal
1438 base are bit patterns, and that the programmer really
1439 meant what they wrote. */
1440 if (warn_traditional && !in_system_header
1441 && base == 10 && trad_type != ansi_type)
1443 if (TYPE_PRECISION (trad_type) != TYPE_PRECISION (ansi_type))
1444 warning ("width of integer constant changes with -traditional");
1445 else if (TREE_UNSIGNED (trad_type) != TREE_UNSIGNED (ansi_type))
1446 warning ("integer constant is unsigned in ISO C, signed with -traditional");
1448 warning ("width of integer constant may change on other systems with -traditional");
1451 if (pedantic && !flag_traditional && (flag_isoc99 || !spec_long_long)
1454 ? TYPE_PRECISION (long_long_integer_type_node)
1455 : TYPE_PRECISION (long_integer_type_node)) < TYPE_PRECISION (type)))
1458 pedwarn ("integer constant larger than the maximum value of %s",
1460 ? (TREE_UNSIGNED (type)
1461 ? "an unsigned long long int"
1462 : "a long long int")
1463 : "an unsigned long int"));
1466 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1467 warning ("decimal constant is so large that it is unsigned");
1471 if (TYPE_PRECISION (type)
1472 <= TYPE_PRECISION (integer_type_node))
1473 value = build_complex (NULL_TREE, integer_zero_node,
1474 convert (integer_type_node, value));
1476 ERROR ("complex integer constant is too wide for 'complex int'");
1478 else if (flag_traditional && !int_fits_type_p (value, type))
1479 /* The traditional constant 0x80000000 is signed
1480 but doesn't fit in the range of int.
1481 This will change it to -0x80000000, which does fit. */
1483 TREE_TYPE (value) = unsigned_type (type);
1484 value = convert (type, value);
1485 TREE_OVERFLOW (value) = TREE_CONSTANT_OVERFLOW (value) = 0;
1488 TREE_TYPE (value) = type;
1490 /* If it's still an integer (not a complex), and it doesn't
1491 fit in the type we choose for it, then pedwarn. */
1494 && TREE_CODE (TREE_TYPE (value)) == INTEGER_TYPE
1495 && ! int_fits_type_p (value, TREE_TYPE (value)))
1496 pedwarn ("integer constant is larger than the maximum value for its type");
1500 error ("missing white space after number '%.*s'", (int) (p - str), str);
1505 return integer_zero_node;
1509 lex_string (str, len, wide)
1515 char *buf = alloca ((len + 1) * (wide ? WCHAR_BYTES : 1));
1517 const char *p = str, *limit = str + len;
1519 unsigned width = wide ? WCHAR_TYPE_SIZE
1520 : TYPE_PRECISION (char_type_node);
1522 #ifdef MULTIBYTE_CHARS
1523 /* Reset multibyte conversion state. */
1524 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1529 #ifdef MULTIBYTE_CHARS
1533 char_len = local_mbtowc (&wc, p, limit - p);
1536 warning ("Ignoring invalid multibyte character");
1549 if (c == '\\' && !ignore_escape_flag)
1551 p = readescape (p, limit, &c);
1552 if (width < HOST_BITS_PER_INT
1553 && (unsigned) c >= ((unsigned)1 << width))
1554 pedwarn ("escape sequence out of range for character");
1557 /* Add this single character into the buffer either as a wchar_t
1558 or as a single byte. */
1561 unsigned charwidth = TYPE_PRECISION (char_type_node);
1562 unsigned bytemask = (1 << charwidth) - 1;
1565 for (byte = 0; byte < WCHAR_BYTES; ++byte)
1568 if (byte >= (int) sizeof (c))
1571 n = (c >> (byte * charwidth)) & bytemask;
1572 if (BYTES_BIG_ENDIAN)
1573 q[WCHAR_BYTES - byte - 1] = n;
1585 /* Terminate the string value, either with a single byte zero
1586 or with a wide zero. */
1590 memset (q, 0, WCHAR_BYTES);
1598 value = build_string (q - buf, buf);
1601 TREE_TYPE (value) = wchar_array_type_node;
1603 TREE_TYPE (value) = char_array_type_node;
1608 lex_charconst (str, len, wide)
1613 const char *limit = str + len;
1617 unsigned width = TYPE_PRECISION (char_type_node);
1622 #ifdef MULTIBYTE_CHARS
1623 int longest_char = local_mb_cur_max ();
1624 (void) local_mbtowc (NULL_PTR, NULL_PTR, 0);
1627 max_chars = TYPE_PRECISION (integer_type_node) / width;
1629 width = WCHAR_TYPE_SIZE;
1633 #ifdef MULTIBYTE_CHARS
1637 char_len = local_mbtowc (&wc, str, limit - str);
1640 warning ("Ignoring invalid multibyte character");
1656 str = readescape (str, limit, &c);
1657 if (width < HOST_BITS_PER_INT
1658 && (unsigned) c >= ((unsigned)1 << width))
1659 pedwarn ("escape sequence out of range for character");
1661 #ifdef MAP_CHARACTER
1663 c = MAP_CHARACTER (c);
1666 /* Merge character into result; ignore excess chars. */
1667 num_chars += (width / TYPE_PRECISION (char_type_node));
1668 if (num_chars < max_chars + 1)
1670 if (width < HOST_BITS_PER_INT)
1671 result = (result << width) | (c & ((1 << width) - 1));
1677 if (chars_seen == 0)
1678 error ("empty character constant");
1679 else if (num_chars > max_chars)
1681 num_chars = max_chars;
1682 error ("character constant too long");
1684 else if (chars_seen != 1 && ! flag_traditional && warn_multichar)
1685 warning ("multi-character character constant");
1687 /* If char type is signed, sign-extend the constant. */
1690 int num_bits = num_chars * width;
1692 /* We already got an error; avoid invalid shift. */
1693 value = build_int_2 (0, 0);
1694 else if (TREE_UNSIGNED (char_type_node)
1695 || ((result >> (num_bits - 1)) & 1) == 0)
1696 value = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
1697 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1700 value = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
1701 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1703 /* In C, a character constant has type 'int'; in C++, 'char'. */
1704 if (chars_seen <= 1 && c_language == clk_cplusplus)
1705 TREE_TYPE (value) = char_type_node;
1707 TREE_TYPE (value) = integer_type_node;
1711 value = build_int_2 (result, 0);
1712 TREE_TYPE (value) = wchar_type_node;