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 if (fc
->from
.filename
)
251 lineno
= fc
->from
.lineno
;
252 push_srcloc (fc
->to
.filename
, 1);
253 input_file_stack
->indent_level
= indent_level
;
254 debug_start_source_file (fc
->to
.filename
);
255 #ifndef NO_IMPLICIT_EXTERN_C
258 else if (fc
->externc
)
261 ++pending_lang_change
;
266 main_input_filename
= fc
->to
.filename
;
268 else if (fc
->reason
== FC_LEAVE
)
270 /* Popping out of a file. */
271 if (input_file_stack
->next
)
273 #ifndef NO_IMPLICIT_EXTERN_C
274 if (c_header_level
&& --c_header_level
== 0)
277 warning ("badly nested C headers from preprocessor");
278 --pending_lang_change
;
282 if (indent_level
!= input_file_stack
->indent_level
)
284 warning_with_file_and_line
285 (input_filename
, lex_lineno
,
286 "This file contains more '%c's than '%c's.",
287 indent_level
> input_file_stack
->indent_level
? '{' : '}',
288 indent_level
> input_file_stack
->indent_level
? '}' : '{');
292 debug_end_source_file (input_file_stack
->line
);
295 error ("leaving more files than we entered");
298 update_header_times (fc
->to
.filename
);
299 in_system_header
= fc
->sysp
!= 0;
300 input_filename
= fc
->to
.filename
;
301 lineno
= fc
->to
.lineno
; /* Do we need this? */
304 extract_interface_info ();
308 cb_def_pragma (pfile
)
311 /* Issue a warning message if we have been asked to do so. Ignore
312 unknown pragmas in system headers unless an explicit
313 -Wunknown-pragmas has been given. */
314 if (warn_unknown_pragmas
> in_system_header
)
316 const unsigned char *space
, *name
= 0;
319 cpp_get_token (pfile
, &s
);
320 space
= cpp_token_as_text (pfile
, &s
);
321 cpp_get_token (pfile
, &s
);
322 if (s
.type
== CPP_NAME
)
323 name
= cpp_token_as_text (pfile
, &s
);
325 lineno
= cpp_get_line (parse_in
)->line
;
327 warning ("ignoring #pragma %s %s", space
, name
);
329 warning ("ignoring #pragma %s", space
);
333 /* #define callback for DWARF and DWARF2 debug info. */
335 cb_define (pfile
, node
)
339 debug_define (lineno
, (const char *) cpp_macro_definition (pfile
, node
));
342 /* #undef callback for DWARF and DWARF2 debug info. */
344 cb_undef (pfile
, node
)
345 cpp_reader
*pfile ATTRIBUTE_UNUSED
;
348 debug_undef (lineno
, (const char *) node
->name
);
351 /* Parse a '\uNNNN' or '\UNNNNNNNN' sequence.
353 [lex.charset]: The character designated by the universal-character-name
354 \UNNNNNNNN is that character whose character short name in ISO/IEC 10646
355 is NNNNNNNN; the character designated by the universal-character-name
356 \uNNNN is that character whose character short name in ISO/IEC 10646 is
357 0000NNNN. If the hexadecimal value for a universal character name is
358 less than 0x20 or in the range 0x7F-0x9F (inclusive), or if the
359 universal character name designates a character in the basic source
360 character set, then the program is ill-formed.
362 We assume that wchar_t is Unicode, so we don't need to do any
363 mapping. Is this ever wrong? */
366 read_ucs (p
, limit
, cptr
, length
)
372 unsigned int code
= 0;
375 for (; length
; --length
)
379 error ("incomplete universal-character-name");
386 error ("non hex digit '%c' in universal-character-name", c
);
392 if (c
>= 'a' && c
<= 'f')
393 code
+= c
- 'a' + 10;
394 if (c
>= 'A' && c
<= 'F')
395 code
+= c
- 'A' + 10;
396 if (c
>= '0' && c
<= '9')
401 sorry ("universal-character-name on EBCDIC target");
402 *cptr
= 0x3f; /* EBCDIC invalid character */
406 if (code
> 0x9f && !(code
& 0x80000000))
407 /* True extended character, OK. */;
408 else if (code
>= 0x20 && code
< 0x7f)
410 /* ASCII printable character. The C character set consists of all of
411 these except $, @ and `. We use hex escapes so that this also
412 works with EBCDIC hosts. */
413 if (code
!= 0x24 && code
!= 0x40 && code
!= 0x60)
414 error ("universal-character-name used for '%c'", code
);
417 error ("invalid universal-character-name");
423 /* Read an escape sequence and write its character equivalent into *CPTR.
424 P is the input pointer, which is just after the backslash. LIMIT
425 is how much text we have.
426 Returns the updated input pointer. */
429 readescape (p
, limit
, cptr
)
434 unsigned int c
, code
, count
;
435 unsigned firstdig
= 0;
440 /* cpp has already issued an error for this. */
450 if (warn_traditional
&& !in_system_header
)
451 warning ("the meaning of `\\x' varies with -traditional");
453 if (flag_traditional
)
471 if (c
>= 'a' && c
<= 'f')
472 code
+= c
- 'a' + 10;
473 if (c
>= 'A' && c
<= 'F')
474 code
+= c
- 'A' + 10;
475 if (c
>= '0' && c
<= '9')
477 if (code
!= 0 || count
!= 0)
487 warning ("\\x used with no following hex digits");
492 /* Digits are all 0's. Ok. */
494 else if ((count
- 1) * 4 >= TYPE_PRECISION (integer_type_node
)
497 << (TYPE_PRECISION (integer_type_node
)
500 pedwarn ("hex escape out of range");
504 case '0': case '1': case '2': case '3': case '4':
505 case '5': case '6': case '7':
507 for (count
= 0; count
< 3; count
++)
509 if (c
< '0' || c
> '7')
514 code
= (code
* 8) + (c
- '0');
526 case '\\': case '\'': case '"': case '?':
530 case 'n': *cptr
= TARGET_NEWLINE
; return p
;
531 case 't': *cptr
= TARGET_TAB
; return p
;
532 case 'r': *cptr
= TARGET_CR
; return p
;
533 case 'f': *cptr
= TARGET_FF
; return p
;
534 case 'b': *cptr
= TARGET_BS
; return p
;
535 case 'v': *cptr
= TARGET_VT
; return p
;
537 if (warn_traditional
&& !in_system_header
)
538 warning ("the meaning of '\\a' varies with -traditional");
539 *cptr
= flag_traditional
? c
: TARGET_BELL
;
542 /* Warnings and support checks handled by read_ucs(). */
544 if (c_language
!= clk_cplusplus
&& !flag_isoc99
)
547 if (warn_traditional
&& !in_system_header
)
548 warning ("the meaning of '\\%c' varies with -traditional", c
);
550 return read_ucs (p
, limit
, cptr
, c
== 'u' ? 4 : 8);
554 pedwarn ("non-ISO-standard escape sequence, '\\%c'", c
);
555 *cptr
= TARGET_ESC
; return p
;
557 /* '\(', etc, are used at beginning of line to avoid confusing Emacs.
558 '\%' is used to prevent SCCS from getting confused. */
559 case '(': case '{': case '[': case '%':
561 pedwarn ("unknown escape sequence '\\%c'", c
);
567 pedwarn ("unknown escape sequence '\\%c'", c
);
569 pedwarn ("unknown escape sequence: '\\' followed by char 0x%x", c
);
576 /* Returns nonzero if C is a universal-character-name. Give an error if it
577 is not one which may appear in an identifier, as per [extendid].
579 Note that extended character support in identifiers has not yet been
580 implemented. It is my personal opinion that this is not a desirable
581 feature. Portable code cannot count on support for more than the basic
582 identifier character set. */
595 /* None of the valid chars are outside the Basic Multilingual Plane (the
599 error ("universal-character-name '\\U%08x' not valid in identifier", c
);
604 if ((c
>= 0x00c0 && c
<= 0x00d6)
605 || (c
>= 0x00d8 && c
<= 0x00f6)
606 || (c
>= 0x00f8 && c
<= 0x01f5)
607 || (c
>= 0x01fa && c
<= 0x0217)
608 || (c
>= 0x0250 && c
<= 0x02a8)
609 || (c
>= 0x1e00 && c
<= 0x1e9a)
610 || (c
>= 0x1ea0 && c
<= 0x1ef9))
615 || (c
>= 0x0388 && c
<= 0x038a)
617 || (c
>= 0x038e && c
<= 0x03a1)
618 || (c
>= 0x03a3 && c
<= 0x03ce)
619 || (c
>= 0x03d0 && c
<= 0x03d6)
624 || (c
>= 0x03e2 && c
<= 0x03f3)
625 || (c
>= 0x1f00 && c
<= 0x1f15)
626 || (c
>= 0x1f18 && c
<= 0x1f1d)
627 || (c
>= 0x1f20 && c
<= 0x1f45)
628 || (c
>= 0x1f48 && c
<= 0x1f4d)
629 || (c
>= 0x1f50 && c
<= 0x1f57)
633 || (c
>= 0x1f5f && c
<= 0x1f7d)
634 || (c
>= 0x1f80 && c
<= 0x1fb4)
635 || (c
>= 0x1fb6 && c
<= 0x1fbc)
636 || (c
>= 0x1fc2 && c
<= 0x1fc4)
637 || (c
>= 0x1fc6 && c
<= 0x1fcc)
638 || (c
>= 0x1fd0 && c
<= 0x1fd3)
639 || (c
>= 0x1fd6 && c
<= 0x1fdb)
640 || (c
>= 0x1fe0 && c
<= 0x1fec)
641 || (c
>= 0x1ff2 && c
<= 0x1ff4)
642 || (c
>= 0x1ff6 && c
<= 0x1ffc))
646 if ((c
>= 0x0401 && c
<= 0x040d)
647 || (c
>= 0x040f && c
<= 0x044f)
648 || (c
>= 0x0451 && c
<= 0x045c)
649 || (c
>= 0x045e && c
<= 0x0481)
650 || (c
>= 0x0490 && c
<= 0x04c4)
651 || (c
>= 0x04c7 && c
<= 0x04c8)
652 || (c
>= 0x04cb && c
<= 0x04cc)
653 || (c
>= 0x04d0 && c
<= 0x04eb)
654 || (c
>= 0x04ee && c
<= 0x04f5)
655 || (c
>= 0x04f8 && c
<= 0x04f9))
659 if ((c
>= 0x0531 && c
<= 0x0556)
660 || (c
>= 0x0561 && c
<= 0x0587))
664 if ((c
>= 0x05d0 && c
<= 0x05ea)
665 || (c
>= 0x05f0 && c
<= 0x05f4))
669 if ((c
>= 0x0621 && c
<= 0x063a)
670 || (c
>= 0x0640 && c
<= 0x0652)
671 || (c
>= 0x0670 && c
<= 0x06b7)
672 || (c
>= 0x06ba && c
<= 0x06be)
673 || (c
>= 0x06c0 && c
<= 0x06ce)
674 || (c
>= 0x06e5 && c
<= 0x06e7))
678 if ((c
>= 0x0905 && c
<= 0x0939)
679 || (c
>= 0x0958 && c
<= 0x0962))
683 if ((c
>= 0x0985 && c
<= 0x098c)
684 || (c
>= 0x098f && c
<= 0x0990)
685 || (c
>= 0x0993 && c
<= 0x09a8)
686 || (c
>= 0x09aa && c
<= 0x09b0)
688 || (c
>= 0x09b6 && c
<= 0x09b9)
689 || (c
>= 0x09dc && c
<= 0x09dd)
690 || (c
>= 0x09df && c
<= 0x09e1)
691 || (c
>= 0x09f0 && c
<= 0x09f1))
695 if ((c
>= 0x0a05 && c
<= 0x0a0a)
696 || (c
>= 0x0a0f && c
<= 0x0a10)
697 || (c
>= 0x0a13 && c
<= 0x0a28)
698 || (c
>= 0x0a2a && c
<= 0x0a30)
699 || (c
>= 0x0a32 && c
<= 0x0a33)
700 || (c
>= 0x0a35 && c
<= 0x0a36)
701 || (c
>= 0x0a38 && c
<= 0x0a39)
702 || (c
>= 0x0a59 && c
<= 0x0a5c)
707 if ((c
>= 0x0a85 && c
<= 0x0a8b)
709 || (c
>= 0x0a8f && c
<= 0x0a91)
710 || (c
>= 0x0a93 && c
<= 0x0aa8)
711 || (c
>= 0x0aaa && c
<= 0x0ab0)
712 || (c
>= 0x0ab2 && c
<= 0x0ab3)
713 || (c
>= 0x0ab5 && c
<= 0x0ab9)
718 if ((c
>= 0x0b05 && c
<= 0x0b0c)
719 || (c
>= 0x0b0f && c
<= 0x0b10)
720 || (c
>= 0x0b13 && c
<= 0x0b28)
721 || (c
>= 0x0b2a && c
<= 0x0b30)
722 || (c
>= 0x0b32 && c
<= 0x0b33)
723 || (c
>= 0x0b36 && c
<= 0x0b39)
724 || (c
>= 0x0b5c && c
<= 0x0b5d)
725 || (c
>= 0x0b5f && c
<= 0x0b61))
729 if ((c
>= 0x0b85 && c
<= 0x0b8a)
730 || (c
>= 0x0b8e && c
<= 0x0b90)
731 || (c
>= 0x0b92 && c
<= 0x0b95)
732 || (c
>= 0x0b99 && c
<= 0x0b9a)
734 || (c
>= 0x0b9e && c
<= 0x0b9f)
735 || (c
>= 0x0ba3 && c
<= 0x0ba4)
736 || (c
>= 0x0ba8 && c
<= 0x0baa)
737 || (c
>= 0x0bae && c
<= 0x0bb5)
738 || (c
>= 0x0bb7 && c
<= 0x0bb9))
742 if ((c
>= 0x0c05 && c
<= 0x0c0c)
743 || (c
>= 0x0c0e && c
<= 0x0c10)
744 || (c
>= 0x0c12 && c
<= 0x0c28)
745 || (c
>= 0x0c2a && c
<= 0x0c33)
746 || (c
>= 0x0c35 && c
<= 0x0c39)
747 || (c
>= 0x0c60 && c
<= 0x0c61))
751 if ((c
>= 0x0c85 && c
<= 0x0c8c)
752 || (c
>= 0x0c8e && c
<= 0x0c90)
753 || (c
>= 0x0c92 && c
<= 0x0ca8)
754 || (c
>= 0x0caa && c
<= 0x0cb3)
755 || (c
>= 0x0cb5 && c
<= 0x0cb9)
756 || (c
>= 0x0ce0 && c
<= 0x0ce1))
760 if ((c
>= 0x0d05 && c
<= 0x0d0c)
761 || (c
>= 0x0d0e && c
<= 0x0d10)
762 || (c
>= 0x0d12 && c
<= 0x0d28)
763 || (c
>= 0x0d2a && c
<= 0x0d39)
764 || (c
>= 0x0d60 && c
<= 0x0d61))
768 if ((c
>= 0x0e01 && c
<= 0x0e30)
769 || (c
>= 0x0e32 && c
<= 0x0e33)
770 || (c
>= 0x0e40 && c
<= 0x0e46)
771 || (c
>= 0x0e4f && c
<= 0x0e5b))
775 if ((c
>= 0x0e81 && c
<= 0x0e82)
781 || (c
>= 0x0e94 && c
<= 0x0e97)
782 || (c
>= 0x0e99 && c
<= 0x0e9f)
783 || (c
>= 0x0ea1 && c
<= 0x0ea3)
788 || (c
>= 0x0ead && c
<= 0x0eb0)
792 || (c
>= 0x0ec0 && c
<= 0x0ec4)
797 if ((c
>= 0x10a0 && c
<= 0x10c5)
798 || (c
>= 0x10d0 && c
<= 0x10f6))
802 if ((c
>= 0x3041 && c
<= 0x3094)
803 || (c
>= 0x309b && c
<= 0x309e))
807 if ((c
>= 0x30a1 && c
<= 0x30fe))
811 if ((c
>= 0x3105 && c
<= 0x312c))
815 if ((c
>= 0x1100 && c
<= 0x1159)
816 || (c
>= 0x1161 && c
<= 0x11a2)
817 || (c
>= 0x11a8 && c
<= 0x11f9))
820 /* CJK Unified Ideographs */
821 if ((c
>= 0xf900 && c
<= 0xfa2d)
822 || (c
>= 0xfb1f && c
<= 0xfb36)
823 || (c
>= 0xfb38 && c
<= 0xfb3c)
825 || (c
>= 0xfb40 && c
<= 0xfb41)
826 || (c
>= 0xfb42 && c
<= 0xfb44)
827 || (c
>= 0xfb46 && c
<= 0xfbb1)
828 || (c
>= 0xfbd3 && c
<= 0xfd3f)
829 || (c
>= 0xfd50 && c
<= 0xfd8f)
830 || (c
>= 0xfd92 && c
<= 0xfdc7)
831 || (c
>= 0xfdf0 && c
<= 0xfdfb)
832 || (c
>= 0xfe70 && c
<= 0xfe72)
834 || (c
>= 0xfe76 && c
<= 0xfefc)
835 || (c
>= 0xff21 && c
<= 0xff3a)
836 || (c
>= 0xff41 && c
<= 0xff5a)
837 || (c
>= 0xff66 && c
<= 0xffbe)
838 || (c
>= 0xffc2 && c
<= 0xffc7)
839 || (c
>= 0xffca && c
<= 0xffcf)
840 || (c
>= 0xffd2 && c
<= 0xffd7)
841 || (c
>= 0xffda && c
<= 0xffdc)
842 || (c
>= 0x4e00 && c
<= 0x9fa5))
845 error ("universal-character-name '\\u%04x' not valid in identifier", c
);
850 /* Add the UTF-8 representation of C to the token_buffer. */
853 utf8_extend_token (c
)
863 else if (c
<= 0x000007ff)
864 shift
= 6, mask
= 0xc0;
865 else if (c
<= 0x0000ffff)
866 shift
= 12, mask
= 0xe0;
867 else if (c
<= 0x001fffff)
868 shift
= 18, mask
= 0xf0;
869 else if (c
<= 0x03ffffff)
870 shift
= 24, mask
= 0xf8;
872 shift
= 30, mask
= 0xfc;
874 extend_token (mask
| (c
>> shift
));
878 extend_token ((unsigned char) (0x80 | (c
>> shift
)));
893 struct try_type type_sequence
[] =
895 { &integer_type_node
, 0, 0, 0},
896 { &unsigned_type_node
, 1, 0, 0},
897 { &long_integer_type_node
, 0, 1, 0},
898 { &long_unsigned_type_node
, 1, 1, 0},
899 { &long_long_integer_type_node
, 0, 1, 1},
900 { &long_long_unsigned_type_node
, 1, 1, 1}
912 int conversion_errno
;
913 REAL_VALUE_TYPE value
;
921 struct pf_args
* args
= (struct pf_args
*) data
;
922 const char *typename
;
924 args
->conversion_errno
= 0;
925 args
->type
= double_type_node
;
928 /* The second argument, machine_mode, of REAL_VALUE_ATOF
929 tells the desired precision of the binary result
930 of decimal-to-binary conversion. */
935 error ("both 'f' and 'l' suffixes on floating constant");
937 args
->type
= float_type_node
;
940 else if (args
->lflag
)
942 args
->type
= long_double_type_node
;
943 typename
= "long double";
945 else if (flag_single_precision_constant
)
947 args
->type
= float_type_node
;
952 if (args
->base
== 16)
953 args
->value
= REAL_VALUE_HTOF (args
->str
, TYPE_MODE (args
->type
));
955 args
->value
= REAL_VALUE_ATOF (args
->str
, TYPE_MODE (args
->type
));
957 args
->conversion_errno
= errno
;
958 /* A diagnostic is required here by some ISO C testsuites.
959 This is not pedwarn, because some people don't want
960 an error for this. */
961 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
962 warning ("floating point number exceeds range of '%s'", typename
);
973 timevar_push (TV_CPP
);
974 cpp_get_token (parse_in
, &tok
);
975 timevar_pop (TV_CPP
);
977 /* The C++ front end does horrible things with the current line
978 number. To ensure an accurate line number, we must reset it
979 every time we return a token. */
980 lineno
= cpp_get_line (parse_in
)->line
;
986 case CPP_OPEN_BRACE
: indent_level
++; break;
987 case CPP_CLOSE_BRACE
: indent_level
--; break;
989 /* Issue this error here, where we can get at tok.val.c. */
991 if (ISGRAPH (tok
.val
.c
))
992 error ("stray '%c' in program", tok
.val
.c
);
994 error ("stray '\\%#o' in program", tok
.val
.c
);
998 *value
= get_identifier ((const char *)tok
.val
.node
->name
);
1004 *value
= lex_number ((const char *)tok
.val
.str
.text
, tok
.val
.str
.len
);
1009 *value
= lex_charconst ((const char *)tok
.val
.str
.text
,
1010 tok
.val
.str
.len
, tok
.type
== CPP_WCHAR
);
1016 *value
= lex_string ((const char *)tok
.val
.str
.text
,
1017 tok
.val
.str
.len
, tok
.type
== CPP_WSTRING
);
1020 /* These tokens should not be visible outside cpplib. */
1021 case CPP_HEADER_NAME
:
1032 #define ERROR(msgid) do { error(msgid); goto syntax_error; } while(0)
1035 lex_number (str
, len
)
1041 int largest_digit
= 0;
1047 enum anon1
{ NOT_FLOAT
= 0, AFTER_POINT
, AFTER_EXPON
} floatflag
= NOT_FLOAT
;
1049 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
1050 The code below which fills the parts array assumes that a host
1051 int is at least twice as wide as a host char, and that
1052 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
1053 Two HOST_WIDE_INTs is the largest int literal we can store.
1054 In order to detect overflow below, the number of parts (TOTAL_PARTS)
1055 must be exactly the number of parts needed to hold the bits
1056 of two HOST_WIDE_INTs. */
1057 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
1058 unsigned int parts
[TOTAL_PARTS
];
1060 /* Optimize for most frequent case. */
1064 return integer_zero_node
;
1065 else if (*str
== '1')
1066 return integer_one_node
;
1068 return build_int_2 (*str
- '0', 0);
1071 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1074 /* len is known to be >1 at this point. */
1077 if (len
> 2 && str
[0] == '0' && (str
[1] == 'x' || str
[1] == 'X'))
1082 /* The ISDIGIT check is so we are not confused by a suffix on 0. */
1083 else if (str
[0] == '0' && ISDIGIT (str
[1]))
1095 if (base
== 16 && pedantic
&& !flag_isoc99
)
1096 pedwarn ("floating constant may not be in radix 16");
1097 else if (floatflag
== AFTER_POINT
)
1098 ERROR ("too many decimal points in floating constant");
1099 else if (floatflag
== AFTER_EXPON
)
1100 ERROR ("decimal point in exponent - impossible!");
1102 floatflag
= AFTER_POINT
;
1108 /* Possible future extension: silently ignore _ in numbers,
1109 permitting cosmetic grouping - e.g. 0x8000_0000 == 0x80000000
1110 but somewhat easier to read. Ada has this? */
1111 ERROR ("underscore in number");
1115 /* It is not a decimal point.
1116 It should be a digit (perhaps a hex digit). */
1122 else if (base
<= 10 && (c
== 'e' || c
== 'E'))
1125 floatflag
= AFTER_EXPON
;
1128 else if (base
== 16 && (c
== 'p' || c
== 'P'))
1130 floatflag
= AFTER_EXPON
;
1131 break; /* start of exponent */
1133 else if (base
== 16 && c
>= 'a' && c
<= 'f')
1137 else if (base
== 16 && c
>= 'A' && c
<= 'F')
1144 break; /* start of suffix */
1147 if (n
>= largest_digit
)
1151 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1153 parts
[count
] *= base
;
1157 += (parts
[count
-1] >> HOST_BITS_PER_CHAR
);
1159 &= (1 << HOST_BITS_PER_CHAR
) - 1;
1165 /* If the highest-order part overflows (gets larger than
1166 a host char will hold) then the whole number has
1167 overflowed. Record this and truncate the highest-order
1169 if (parts
[TOTAL_PARTS
- 1] >> HOST_BITS_PER_CHAR
)
1172 parts
[TOTAL_PARTS
- 1] &= (1 << HOST_BITS_PER_CHAR
) - 1;
1176 while (p
< str
+ len
);
1178 /* This can happen on input like `int i = 0x;' */
1180 ERROR ("numeric constant with no digits");
1182 if (largest_digit
>= base
)
1183 ERROR ("numeric constant contains digits beyond the radix");
1185 if (floatflag
!= NOT_FLOAT
)
1188 int imag
, fflag
, lflag
, conversion_errno
;
1189 REAL_VALUE_TYPE real
;
1190 struct pf_args args
;
1193 if (base
== 16 && floatflag
!= AFTER_EXPON
)
1194 ERROR ("hexadecimal floating constant has no exponent");
1196 /* Read explicit exponent if any, and put it in tokenbuf. */
1197 if ((base
== 10 && ((c
== 'e') || (c
== 'E')))
1198 || (base
== 16 && (c
== 'p' || c
== 'P')))
1202 if (p
< str
+ len
&& (c
== '+' || c
== '-'))
1204 /* Exponent is decimal, even if string is a hex float. */
1206 ERROR ("floating constant exponent has no digits");
1207 while (p
< str
+ len
&& ISDIGIT (c
))
1213 /* Copy the float constant now; we don't want any suffixes in the
1214 string passed to parse_float. */
1215 copy
= alloca (p
- str
+ 1);
1216 memcpy (copy
, str
, p
- str
);
1217 copy
[p
- str
] = '\0';
1219 /* Now parse suffixes. */
1220 fflag
= lflag
= imag
= 0;
1221 while (p
< str
+ len
)
1226 ERROR ("more than one 'f' suffix on floating constant");
1227 else if (warn_traditional
&& !in_system_header
)
1228 warning ("traditional C rejects the 'f' suffix");
1235 ERROR ("more than one 'l' suffix on floating constant");
1236 else if (warn_traditional
&& !in_system_header
)
1237 warning ("traditional C rejects the 'l' suffix");
1245 ERROR ("more than one 'i' or 'j' suffix on floating constant");
1247 pedwarn ("ISO C forbids imaginary numeric constants");
1252 ERROR ("invalid suffix on floating constant");
1255 /* Setup input for parse_float() */
1261 /* Convert string to a double, checking for overflow. */
1262 if (do_float_handler (parse_float
, (PTR
) &args
))
1264 /* Receive output from parse_float() */
1268 /* We got an exception from parse_float() */
1269 ERROR ("floating constant out of range");
1271 /* Receive output from parse_float() */
1272 conversion_errno
= args
.conversion_errno
;
1276 /* ERANGE is also reported for underflow,
1277 so test the value to distinguish overflow from that. */
1278 if (conversion_errno
== ERANGE
&& !flag_traditional
&& pedantic
1279 && (REAL_VALUES_LESS (dconst1
, real
)
1280 || REAL_VALUES_LESS (real
, dconstm1
)))
1281 warning ("floating point number exceeds range of 'double'");
1284 /* Create a node with determined type and value. */
1286 value
= build_complex (NULL_TREE
, convert (type
, integer_zero_node
),
1287 build_real (type
, real
));
1289 value
= build_real (type
, real
);
1293 tree trad_type
, ansi_type
, type
;
1294 HOST_WIDE_INT high
, low
;
1295 int spec_unsigned
= 0;
1297 int spec_long_long
= 0;
1302 trad_type
= ansi_type
= type
= NULL_TREE
;
1303 while (p
< str
+ len
)
1310 error ("two 'u' suffixes on integer constant");
1311 else if (warn_traditional
&& !in_system_header
)
1312 warning ("traditional C rejects the 'u' suffix");
1323 error ("three 'l' suffixes on integer constant");
1325 error ("'lul' is not a valid integer suffix");
1326 else if (c
!= spec_long
)
1327 error ("'Ll' and 'lL' are not valid integer suffixes");
1328 else if (pedantic
&& ! flag_isoc99
1329 && ! in_system_header
&& warn_long_long
)
1330 pedwarn ("ISO C89 forbids long long integer constants");
1336 case 'i': case 'I': case 'j': case 'J':
1338 error ("more than one 'i' or 'j' suffix on integer constant");
1340 pedwarn ("ISO C forbids imaginary numeric constants");
1345 ERROR ("invalid suffix on integer constant");
1349 /* If the literal overflowed, pedwarn about it now. */
1353 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT
* 2);
1356 /* This is simplified by the fact that our constant
1357 is always positive. */
1361 for (i
= 0; i
< HOST_BITS_PER_WIDE_INT
/ HOST_BITS_PER_CHAR
; i
++)
1363 high
|= ((HOST_WIDE_INT
) parts
[i
+ (HOST_BITS_PER_WIDE_INT
1364 / HOST_BITS_PER_CHAR
)]
1365 << (i
* HOST_BITS_PER_CHAR
));
1366 low
|= (HOST_WIDE_INT
) parts
[i
] << (i
* HOST_BITS_PER_CHAR
);
1369 value
= build_int_2 (low
, high
);
1370 TREE_TYPE (value
) = long_long_unsigned_type_node
;
1372 /* If warn_traditional, calculate both the ISO type and the
1373 traditional type, then see if they disagree.
1374 Otherwise, calculate only the type for the dialect in use. */
1375 if (warn_traditional
|| flag_traditional
)
1377 /* Calculate the traditional type. */
1378 /* Traditionally, any constant is signed; but if unsigned is
1379 specified explicitly, obey that. Use the smallest size
1380 with the right number of bits, except for one special
1381 case with decimal constants. */
1382 if (! spec_long
&& base
!= 10
1383 && int_fits_type_p (value
, unsigned_type_node
))
1384 trad_type
= spec_unsigned
? unsigned_type_node
: integer_type_node
;
1385 /* A decimal constant must be long if it does not fit in
1386 type int. I think this is independent of whether the
1387 constant is signed. */
1388 else if (! spec_long
&& base
== 10
1389 && int_fits_type_p (value
, integer_type_node
))
1390 trad_type
= spec_unsigned
? unsigned_type_node
: integer_type_node
;
1391 else if (! spec_long_long
)
1392 trad_type
= (spec_unsigned
1393 ? long_unsigned_type_node
1394 : long_integer_type_node
);
1395 else if (int_fits_type_p (value
,
1397 ? long_long_unsigned_type_node
1398 : long_long_integer_type_node
))
1399 trad_type
= (spec_unsigned
1400 ? long_long_unsigned_type_node
1401 : long_long_integer_type_node
);
1403 trad_type
= (spec_unsigned
1404 ? widest_unsigned_literal_type_node
1405 : widest_integer_literal_type_node
);
1407 if (warn_traditional
|| ! flag_traditional
)
1409 /* Calculate the ISO type. */
1410 if (! spec_long
&& ! spec_unsigned
1411 && int_fits_type_p (value
, integer_type_node
))
1412 ansi_type
= integer_type_node
;
1413 else if (! spec_long
&& (base
!= 10 || spec_unsigned
)
1414 && int_fits_type_p (value
, unsigned_type_node
))
1415 ansi_type
= unsigned_type_node
;
1416 else if (! spec_unsigned
&& !spec_long_long
1417 && int_fits_type_p (value
, long_integer_type_node
))
1418 ansi_type
= long_integer_type_node
;
1419 else if (! spec_long_long
1420 && int_fits_type_p (value
, long_unsigned_type_node
))
1421 ansi_type
= long_unsigned_type_node
;
1422 else if (! spec_unsigned
1423 && int_fits_type_p (value
, long_long_integer_type_node
))
1424 ansi_type
= long_long_integer_type_node
;
1425 else if (int_fits_type_p (value
, long_long_unsigned_type_node
))
1426 ansi_type
= long_long_unsigned_type_node
;
1427 else if (! spec_unsigned
1428 && int_fits_type_p (value
, widest_integer_literal_type_node
))
1429 ansi_type
= widest_integer_literal_type_node
;
1431 ansi_type
= widest_unsigned_literal_type_node
;
1434 type
= flag_traditional
? trad_type
: ansi_type
;
1436 /* We assume that constants specified in a non-decimal
1437 base are bit patterns, and that the programmer really
1438 meant what they wrote. */
1439 if (warn_traditional
&& !in_system_header
1440 && base
== 10 && trad_type
!= ansi_type
)
1442 if (TYPE_PRECISION (trad_type
) != TYPE_PRECISION (ansi_type
))
1443 warning ("width of integer constant changes with -traditional");
1444 else if (TREE_UNSIGNED (trad_type
) != TREE_UNSIGNED (ansi_type
))
1445 warning ("integer constant is unsigned in ISO C, signed with -traditional");
1447 warning ("width of integer constant may change on other systems with -traditional");
1450 if (pedantic
&& !flag_traditional
&& (flag_isoc99
|| !spec_long_long
)
1453 ? TYPE_PRECISION (long_long_integer_type_node
)
1454 : TYPE_PRECISION (long_integer_type_node
)) < TYPE_PRECISION (type
)))
1457 pedwarn ("integer constant larger than the maximum value of %s",
1459 ? (TREE_UNSIGNED (type
)
1460 ? "an unsigned long long int"
1461 : "a long long int")
1462 : "an unsigned long int"));
1465 if (base
== 10 && ! spec_unsigned
&& TREE_UNSIGNED (type
))
1466 warning ("decimal constant is so large that it is unsigned");
1470 if (TYPE_PRECISION (type
)
1471 <= TYPE_PRECISION (integer_type_node
))
1472 value
= build_complex (NULL_TREE
, integer_zero_node
,
1473 convert (integer_type_node
, value
));
1475 ERROR ("complex integer constant is too wide for 'complex int'");
1477 else if (flag_traditional
&& !int_fits_type_p (value
, type
))
1478 /* The traditional constant 0x80000000 is signed
1479 but doesn't fit in the range of int.
1480 This will change it to -0x80000000, which does fit. */
1482 TREE_TYPE (value
) = unsigned_type (type
);
1483 value
= convert (type
, value
);
1484 TREE_OVERFLOW (value
) = TREE_CONSTANT_OVERFLOW (value
) = 0;
1487 TREE_TYPE (value
) = type
;
1489 /* If it's still an integer (not a complex), and it doesn't
1490 fit in the type we choose for it, then pedwarn. */
1493 && TREE_CODE (TREE_TYPE (value
)) == INTEGER_TYPE
1494 && ! int_fits_type_p (value
, TREE_TYPE (value
)))
1495 pedwarn ("integer constant is larger than the maximum value for its type");
1499 error ("missing white space after number '%.*s'", (int) (p
- str
), str
);
1504 return integer_zero_node
;
1508 lex_string (str
, len
, wide
)
1514 char *buf
= alloca ((len
+ 1) * (wide
? WCHAR_BYTES
: 1));
1516 const char *p
= str
, *limit
= str
+ len
;
1518 unsigned width
= wide
? WCHAR_TYPE_SIZE
1519 : TYPE_PRECISION (char_type_node
);
1521 #ifdef MULTIBYTE_CHARS
1522 /* Reset multibyte conversion state. */
1523 (void) local_mbtowc (NULL_PTR
, NULL_PTR
, 0);
1528 #ifdef MULTIBYTE_CHARS
1532 char_len
= local_mbtowc (&wc
, p
, limit
- p
);
1535 warning ("Ignoring invalid multibyte character");
1548 if (c
== '\\' && !ignore_escape_flag
)
1550 p
= readescape (p
, limit
, &c
);
1551 if (width
< HOST_BITS_PER_INT
1552 && (unsigned) c
>= ((unsigned)1 << width
))
1553 pedwarn ("escape sequence out of range for character");
1556 /* Add this single character into the buffer either as a wchar_t
1557 or as a single byte. */
1560 unsigned charwidth
= TYPE_PRECISION (char_type_node
);
1561 unsigned bytemask
= (1 << charwidth
) - 1;
1564 for (byte
= 0; byte
< WCHAR_BYTES
; ++byte
)
1567 if (byte
>= (int) sizeof (c
))
1570 n
= (c
>> (byte
* charwidth
)) & bytemask
;
1571 if (BYTES_BIG_ENDIAN
)
1572 q
[WCHAR_BYTES
- byte
- 1] = n
;
1584 /* Terminate the string value, either with a single byte zero
1585 or with a wide zero. */
1589 memset (q
, 0, WCHAR_BYTES
);
1597 value
= build_string (q
- buf
, buf
);
1600 TREE_TYPE (value
) = wchar_array_type_node
;
1602 TREE_TYPE (value
) = char_array_type_node
;
1607 lex_charconst (str
, len
, wide
)
1612 const char *limit
= str
+ len
;
1616 unsigned width
= TYPE_PRECISION (char_type_node
);
1621 #ifdef MULTIBYTE_CHARS
1622 int longest_char
= local_mb_cur_max ();
1623 (void) local_mbtowc (NULL_PTR
, NULL_PTR
, 0);
1626 max_chars
= TYPE_PRECISION (integer_type_node
) / width
;
1628 width
= WCHAR_TYPE_SIZE
;
1632 #ifdef MULTIBYTE_CHARS
1636 char_len
= local_mbtowc (&wc
, str
, limit
- str
);
1639 warning ("Ignoring invalid multibyte character");
1655 str
= readescape (str
, limit
, &c
);
1656 if (width
< HOST_BITS_PER_INT
1657 && (unsigned) c
>= ((unsigned)1 << width
))
1658 pedwarn ("escape sequence out of range for character");
1660 #ifdef MAP_CHARACTER
1662 c
= MAP_CHARACTER (c
);
1665 /* Merge character into result; ignore excess chars. */
1666 num_chars
+= (width
/ TYPE_PRECISION (char_type_node
));
1667 if (num_chars
< max_chars
+ 1)
1669 if (width
< HOST_BITS_PER_INT
)
1670 result
= (result
<< width
) | (c
& ((1 << width
) - 1));
1676 if (chars_seen
== 0)
1677 error ("empty character constant");
1678 else if (num_chars
> max_chars
)
1680 num_chars
= max_chars
;
1681 error ("character constant too long");
1683 else if (chars_seen
!= 1 && ! flag_traditional
&& warn_multichar
)
1684 warning ("multi-character character constant");
1686 /* If char type is signed, sign-extend the constant. */
1689 int num_bits
= num_chars
* width
;
1691 /* We already got an error; avoid invalid shift. */
1692 value
= build_int_2 (0, 0);
1693 else if (TREE_UNSIGNED (char_type_node
)
1694 || ((result
>> (num_bits
- 1)) & 1) == 0)
1695 value
= build_int_2 (result
& (~(unsigned HOST_WIDE_INT
) 0
1696 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
1699 value
= build_int_2 (result
| ~(~(unsigned HOST_WIDE_INT
) 0
1700 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
1702 /* In C, a character constant has type 'int'; in C++, 'char'. */
1703 if (chars_seen
<= 1 && c_language
== clk_cplusplus
)
1704 TREE_TYPE (value
) = char_type_node
;
1706 TREE_TYPE (value
) = integer_type_node
;
1710 value
= build_int_2 (result
, 0);
1711 TREE_TYPE (value
) = wchar_type_node
;