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 /* MULTIBYTE_CHARS support only works for native compilers.
40 ??? Ideally what we want is to model widechar support after
41 the current floating point support. */
43 #undef MULTIBYTE_CHARS
46 #ifdef MULTIBYTE_CHARS
49 #endif /* MULTIBYTE_CHARS */
50 #ifndef GET_ENVIRONMENT
51 #define GET_ENVIRONMENT(ENV_VALUE,ENV_NAME) ((ENV_VALUE) = getenv (ENV_NAME))
56 extern cpp_reader parse_in
;
57 extern cpp_options parse_options
;
59 /* Stream for reading from the input file. */
63 extern void yyprint
PARAMS ((FILE *, int, YYSTYPE
));
65 /* Cause the `yydebug' variable to be defined. */
69 extern unsigned char *yy_cur
, *yy_lim
;
70 extern enum cpp_token cpp_token
;
72 extern int yy_get_token ();
74 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
75 #define UNGETC(c) ((c) == EOF ? 0 : yy_cur--)
77 #else /* ! USE_CPPLIB */
79 #define GETC() getch ()
80 #define UNGETC(c) put_back (c)
82 struct putback_buffer
{
83 unsigned char *buffer
;
88 static struct putback_buffer putback
= {NULL
, 0, -1};
90 static inline int getch
PARAMS ((void));
95 if (putback
.index
!= -1)
97 int ch
= putback
.buffer
[putback
.index
];
101 return getc (finput
);
104 static inline void put_back
PARAMS ((int));
112 if (putback
.index
== putback
.buffer_size
- 1)
114 putback
.buffer_size
+= 16;
115 putback
.buffer
= xrealloc (putback
.buffer
, putback
.buffer_size
);
117 putback
.buffer
[++putback
.index
] = ch
;
120 #endif /* ! USE_CPPLIB */
126 /* File used for outputting assembler code. */
127 extern FILE *asm_out_file
;
129 #undef WCHAR_TYPE_SIZE
130 #define WCHAR_TYPE_SIZE TYPE_PRECISION (wchar_type_node)
132 /* Number of bytes in a wide character. */
133 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
135 static int maxtoken
; /* Current nominal length of token buffer. */
136 char *token_buffer
; /* Pointer to token buffer.
137 Actual allocated length is maxtoken + 2.
138 This is not static because objc-parse.y uses it. */
140 static int indent_level
; /* Number of { minus number of }. */
142 /* Nonzero tells yylex to ignore \ in string constants. */
143 static int ignore_escape_flag
;
145 /* Nonzero if end-of-file has been seen on input. */
146 static int end_of_file
;
148 #ifdef HANDLE_GENERIC_PRAGMAS
149 static int handle_generic_pragma
PARAMS ((int));
150 #endif /* HANDLE_GENERIC_PRAGMAS */
151 static int whitespace_cr
PARAMS ((int));
152 static int skip_white_space
PARAMS ((int));
153 static char *extend_token_buffer
PARAMS ((const char *));
154 static int readescape
PARAMS ((int *));
155 static void parse_float
PARAMS ((PTR
));
156 static void extend_token_buffer_to
PARAMS ((int));
157 static int read_line_number
PARAMS ((int *));
159 /* Do not insert generated code into the source, instead, include it.
160 This allows us to build gcc automatically even for targets that
161 need to add or modify the reserved keyword lists. */
164 /* Return something to represent absolute declarators containing a *.
165 TARGET is the absolute declarator that the * contains.
166 TYPE_QUALS is a list of modifiers such as const or volatile
167 to apply to the pointer type, represented as identifiers.
169 We return an INDIRECT_REF whose "contents" are TARGET
170 and whose type is the modifier list. */
173 make_pointer_declarator (type_quals
, target
)
174 tree type_quals
, target
;
176 return build1 (INDIRECT_REF
, type_quals
, target
);
180 forget_protocol_qualifiers ()
182 int i
, n
= sizeof wordlist
/ sizeof (struct resword
);
184 for (i
= 0; i
< n
; i
++)
185 if ((int) wordlist
[i
].rid
>= (int) RID_IN
186 && (int) wordlist
[i
].rid
<= (int) RID_ONEWAY
)
187 wordlist
[i
].name
= "";
191 remember_protocol_qualifiers ()
193 int i
, n
= sizeof wordlist
/ sizeof (struct resword
);
195 for (i
= 0; i
< n
; i
++)
196 if (wordlist
[i
].rid
== RID_IN
)
197 wordlist
[i
].name
= "in";
198 else if (wordlist
[i
].rid
== RID_OUT
)
199 wordlist
[i
].name
= "out";
200 else if (wordlist
[i
].rid
== RID_INOUT
)
201 wordlist
[i
].name
= "inout";
202 else if (wordlist
[i
].rid
== RID_BYCOPY
)
203 wordlist
[i
].name
= "bycopy";
204 else if (wordlist
[i
].rid
== RID_BYREF
)
205 wordlist
[i
].name
= "byref";
206 else if (wordlist
[i
].rid
== RID_ONEWAY
)
207 wordlist
[i
].name
= "oneway";
211 init_parse (filename
)
212 const char *filename
;
215 /* Open input file. */
216 if (filename
== 0 || !strcmp (filename
, "-"))
222 finput
= fopen (filename
, "r");
224 pfatal_with_name (filename
);
226 #ifdef IO_BUFFER_SIZE
227 setvbuf (finput
, (char *) xmalloc (IO_BUFFER_SIZE
), _IOFBF
, IO_BUFFER_SIZE
);
229 #else /* !USE_CPPLIB */
230 parse_in
.show_column
= 1;
231 if (! cpp_start_read (&parse_in
, filename
))
234 if (filename
== 0 || !strcmp (filename
, "-"))
237 /* cpp_start_read always puts at least one line directive into the
238 token buffer. We must arrange to read it out here. */
239 yy_cur
= parse_in
.token_buffer
;
240 yy_lim
= CPP_PWRITTEN (&parse_in
);
241 cpp_token
= CPP_DIRECTIVE
;
256 cpp_finish (&parse_in
);
257 errorcount
+= parse_in
.errors
;
266 /* Make identifier nodes long enough for the language-specific slots. */
267 set_identifier_size (sizeof (struct lang_identifier
));
269 /* Start it at 0, because check_newline is called at the very beginning
270 and will increment it to 1. */
273 #ifdef MULTIBYTE_CHARS
274 /* Change to the native locale for multibyte conversions. */
275 setlocale (LC_CTYPE
, "");
276 GET_ENVIRONMENT (literal_codeset
, "LANG");
280 token_buffer
= (char *) xmalloc (maxtoken
+ 2);
282 ridpointers
= (tree
*) xcalloc ((int) RID_MAX
, sizeof (tree
));
283 ridpointers
[(int) RID_INT
] = get_identifier ("int");
284 ridpointers
[(int) RID_CHAR
] = get_identifier ("char");
285 ridpointers
[(int) RID_VOID
] = get_identifier ("void");
286 ridpointers
[(int) RID_FLOAT
] = get_identifier ("float");
287 ridpointers
[(int) RID_DOUBLE
] = get_identifier ("double");
288 ridpointers
[(int) RID_SHORT
] = get_identifier ("short");
289 ridpointers
[(int) RID_LONG
] = get_identifier ("long");
290 ridpointers
[(int) RID_UNSIGNED
] = get_identifier ("unsigned");
291 ridpointers
[(int) RID_SIGNED
] = get_identifier ("signed");
292 ridpointers
[(int) RID_INLINE
] = get_identifier ("inline");
293 ridpointers
[(int) RID_CONST
] = get_identifier ("const");
294 ridpointers
[(int) RID_RESTRICT
] = get_identifier ("restrict");
295 ridpointers
[(int) RID_VOLATILE
] = get_identifier ("volatile");
296 ridpointers
[(int) RID_BOUNDED
] = get_identifier ("__bounded");
297 ridpointers
[(int) RID_UNBOUNDED
] = get_identifier ("__unbounded");
298 ridpointers
[(int) RID_AUTO
] = get_identifier ("auto");
299 ridpointers
[(int) RID_STATIC
] = get_identifier ("static");
300 ridpointers
[(int) RID_EXTERN
] = get_identifier ("extern");
301 ridpointers
[(int) RID_TYPEDEF
] = get_identifier ("typedef");
302 ridpointers
[(int) RID_REGISTER
] = get_identifier ("register");
303 ridpointers
[(int) RID_ITERATOR
] = get_identifier ("iterator");
304 ridpointers
[(int) RID_COMPLEX
] = get_identifier ("complex");
305 ridpointers
[(int) RID_ID
] = get_identifier ("id");
306 ridpointers
[(int) RID_IN
] = get_identifier ("in");
307 ridpointers
[(int) RID_OUT
] = get_identifier ("out");
308 ridpointers
[(int) RID_INOUT
] = get_identifier ("inout");
309 ridpointers
[(int) RID_BYCOPY
] = get_identifier ("bycopy");
310 ridpointers
[(int) RID_BYREF
] = get_identifier ("byref");
311 ridpointers
[(int) RID_ONEWAY
] = get_identifier ("oneway");
312 forget_protocol_qualifiers();
314 /* Some options inhibit certain reserved words.
315 Clear those words out of the hash table so they won't be recognized. */
316 #define UNSET_RESERVED_WORD(STRING) \
317 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
318 if (s) s->name = ""; } while (0)
320 if (! doing_objc_thang
)
321 UNSET_RESERVED_WORD ("id");
323 if (flag_traditional
)
325 UNSET_RESERVED_WORD ("const");
326 UNSET_RESERVED_WORD ("restrict");
327 UNSET_RESERVED_WORD ("volatile");
328 UNSET_RESERVED_WORD ("typeof");
329 UNSET_RESERVED_WORD ("signed");
330 UNSET_RESERVED_WORD ("inline");
331 UNSET_RESERVED_WORD ("iterator");
332 UNSET_RESERVED_WORD ("complex");
334 else if (!flag_isoc99
)
335 UNSET_RESERVED_WORD ("restrict");
339 UNSET_RESERVED_WORD ("asm");
340 UNSET_RESERVED_WORD ("typeof");
342 UNSET_RESERVED_WORD ("inline");
343 UNSET_RESERVED_WORD ("iterator");
344 UNSET_RESERVED_WORD ("complex");
349 reinit_parse_for_function ()
353 /* Function used when yydebug is set, to print a token in more detail. */
356 yyprint (file
, yychar
, yylval
)
368 if (IDENTIFIER_POINTER (t
))
369 fprintf (file
, " `%s'", IDENTIFIER_POINTER (t
));
374 if (TREE_CODE (t
) == INTEGER_CST
)
376 #if HOST_BITS_PER_WIDE_INT == 64
377 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
380 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
387 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
393 TREE_INT_CST_HIGH (t
), TREE_INT_CST_LOW (t
));
398 /* Iff C is a carriage return, warn about it - if appropriate -
399 and return nonzero. */
405 static int newline_warning
= 0;
409 /* ANSI C says the effects of a carriage return in a source file
411 if (pedantic
&& !newline_warning
)
413 warning ("carriage return in source file");
414 warning ("(we only warn about the first carriage return)");
422 /* If C is not whitespace, return C.
423 Otherwise skip whitespace and return first nonwhite char read. */
433 /* We don't recognize comments here, because
434 cpp output can include / and * consecutively as operators.
435 Also, there's no need, since cpp removes all comments. */
443 c
= check_newline ();
452 /* While processing a # directive we don't get CPP_HSPACE
453 tokens, so we also need to handle whitespace the normal way. */
454 if (cpp_token
== CPP_HSPACE
)
471 error ("stray '\\' in program");
481 /* Skips all of the white space at the current location in the input file. */
484 position_after_white_space ()
490 UNGETC (skip_white_space (c
));
493 /* Make the token buffer longer, preserving the data in it.
494 P should point to just beyond the last valid character in the old buffer.
495 The value we return is a pointer to the new buffer
496 at a place corresponding to P. */
499 extend_token_buffer_to (size
)
503 maxtoken
= maxtoken
* 2 + 10;
504 while (maxtoken
< size
);
505 token_buffer
= (char *) xrealloc (token_buffer
, maxtoken
+ 2);
509 extend_token_buffer (p
)
512 int offset
= p
- token_buffer
;
513 extend_token_buffer_to (offset
);
514 return token_buffer
+ offset
;
517 #if defined HANDLE_PRAGMA
518 /* Local versions of these macros, that can be passed as function pointers. */
534 read_line_number (num
)
537 register int token
= yylex ();
539 if (token
== CONSTANT
540 && TREE_CODE (yylval
.ttype
) == INTEGER_CST
)
542 *num
= TREE_INT_CST_LOW (yylval
.ttype
);
547 if (token
!= END_OF_LINE
)
548 error ("invalid #-line");
553 /* At the beginning of a line, increment the line number
554 and process any #-directive on this line.
555 If the line is a #-directive, read the entire line and return a newline.
556 Otherwise, return the line's first non-whitespace character.
558 Note that in the case of USE_CPPLIB, we get the whole line as one
559 CPP_DIRECTIVE token. */
567 enum { act_none
, act_push
, act_pop
} action
;
568 int old_lineno
, action_number
, l
;
571 /* Read first nonwhite char on the line. */
575 /* In some cases where we're leaving an include file, we can get multiple
576 CPP_HSPACE tokens in a row, so we need to loop. */
577 while (cpp_token
== CPP_HSPACE
)
582 while (c
== ' ' || c
== '\t');
589 /* Sequences of multiple newlines are very common; optimize them. */
593 /* If not #, return it so caller will use it. */
597 /* Don't read beyond this line. */
602 if (cpp_token
== CPP_VSPACE
)
604 /* Format is "<space> <line number> <filename> <newline>".
605 Only the line number is interesting, and even that
606 we can get more efficiently than scanning the line. */
608 lineno
= parse_in
.lineno
- 1;
615 if (token
== IDENTIFIER
)
617 /* If a letter follows, then if the word here is `line', skip
618 it and ignore it; otherwise, ignore the line, with an error
619 if the word isn't `pragma'. */
621 const char *name
= IDENTIFIER_POINTER (yylval
.ttype
);
623 if (!strcmp (name
, "pragma"))
626 if (token
!= IDENTIFIER
627 || TREE_CODE (yylval
.ttype
) != IDENTIFIER_NODE
)
631 /* We invoke HANDLE_PRAGMA before HANDLE_GENERIC_PRAGMAS
632 (if both are defined), in order to give the back
633 end a chance to override the interpretation of
634 SYSV style pragmas. */
635 if (HANDLE_PRAGMA (pragma_getc
, pragma_ungetc
,
636 IDENTIFIER_POINTER (yylval
.ttype
)))
638 #endif /* HANDLE_PRAGMA */
640 #ifdef HANDLE_GENERIC_PRAGMAS
641 if (handle_generic_pragma (token
))
643 #endif /* HANDLE_GENERIC_PRAGMAS */
645 /* Issue a warning message if we have been asked to do so.
646 Ignoring unknown pragmas in system header file unless
647 an explcit -Wunknown-pragmas has been given. */
648 if (warn_unknown_pragmas
> 1
649 || (warn_unknown_pragmas
&& ! in_system_header
))
650 warning ("ignoring pragma: %s", token_buffer
);
654 else if (!strcmp (name
, "define"))
656 debug_define (lineno
, GET_DIRECTIVE_LINE ());
659 else if (!strcmp (name
, "undef"))
661 debug_undef (lineno
, GET_DIRECTIVE_LINE ());
664 else if (!strcmp (name
, "line"))
670 else if (!strcmp (name
, "ident"))
672 /* #ident. The pedantic warning is now in cpp. */
674 /* Here we have just seen `#ident '.
675 A string constant should follow. */
678 if (token
== END_OF_LINE
)
681 || TREE_CODE (yylval
.ttype
) != STRING_CST
)
683 error ("invalid #ident");
689 #ifdef ASM_OUTPUT_IDENT
690 ASM_OUTPUT_IDENT (asm_out_file
,
691 TREE_STRING_POINTER (yylval
.ttype
));
695 /* Skip the rest of this line. */
699 error ("undefined or invalid # directive `%s'", name
);
703 /* If the # is the only nonwhite char on the line,
704 just ignore it. Check the new newline. */
705 if (token
== END_OF_LINE
)
709 /* Here we have either `#line' or `# <nonletter>'.
710 In either case, it should be a line number; a digit should follow. */
712 if (token
!= CONSTANT
713 || TREE_CODE (yylval
.ttype
) != INTEGER_CST
)
715 error ("invalid #-line");
719 /* subtract one, because it is the following line that
720 gets the specified number */
722 l
= TREE_INT_CST_LOW (yylval
.ttype
) - 1;
724 /* More follows: it must be a string constant (filename).
725 It would be neat to use cpplib to quickly process the string, but
726 (1) we don't have a handy tokenization of the string, and
727 (2) I don't know how well that would work in the presense
728 of filenames that contain wide characters. */
732 /* Don't treat \ as special if we are processing #line 1 "...".
733 If you want it to be treated specially, use # 1 "...". */
734 ignore_escape_flag
= 1;
737 /* Read the string constant. */
740 ignore_escape_flag
= 0;
742 if (token
== END_OF_LINE
)
744 /* No more: store the line number and check following line. */
749 if (token
!= STRING
|| TREE_CODE (yylval
.ttype
) != STRING_CST
)
751 error ("invalid #line");
755 input_filename
= TREE_STRING_POINTER (yylval
.ttype
);
757 if (main_input_filename
== 0)
758 main_input_filename
= input_filename
;
765 /* Each change of file name
766 reinitializes whether we are now in a system header. */
767 in_system_header
= 0;
769 if (!read_line_number (&action_number
))
771 /* Update the name in the top element of input_file_stack. */
772 if (input_file_stack
)
773 input_file_stack
->name
= input_filename
;
776 /* `1' after file name means entering new file.
777 `2' after file name means just left a file. */
779 if (action_number
== 1)
782 read_line_number (&action_number
);
784 else if (action_number
== 2)
787 read_line_number (&action_number
);
789 if (action_number
== 3)
791 /* `3' after file name means this is a system header file. */
792 in_system_header
= 1;
793 read_line_number (&action_number
);
796 /* Do the actions implied by the preceding numbers. */
798 if (action
== act_push
)
800 /* Pushing to a new file. */
802 = (struct file_stack
*) xmalloc (sizeof (struct file_stack
));
803 input_file_stack
->line
= old_lineno
;
804 p
->next
= input_file_stack
;
805 p
->name
= input_filename
;
806 p
->indent_level
= indent_level
;
807 input_file_stack
= p
;
808 input_file_stack_tick
++;
809 debug_start_source_file (input_filename
);
811 else if (action
== act_pop
)
813 /* Popping out of a file. */
814 if (input_file_stack
->next
)
816 struct file_stack
*p
= input_file_stack
;
817 if (indent_level
!= p
->indent_level
)
819 warning_with_file_and_line
820 (p
->name
, old_lineno
,
821 "This file contains more `%c's than `%c's.",
822 indent_level
> p
->indent_level
? '{' : '}',
823 indent_level
> p
->indent_level
? '}' : '{');
825 input_file_stack
= p
->next
;
827 input_file_stack_tick
++;
828 debug_end_source_file (input_file_stack
->line
);
831 error ("#-lines for entering and leaving files don't match");
834 /* Now that we've pushed or popped the input stack,
835 update the name in the top element. */
836 if (input_file_stack
)
837 input_file_stack
->name
= input_filename
;
839 /* skip the rest of this line. */
846 while (c
!= '\n' && c
!= EOF
);
850 #ifdef HANDLE_GENERIC_PRAGMAS
852 /* Handle a #pragma directive.
853 TOKEN is the token we read after `#pragma'. Processes the entire input
854 line and return non-zero iff the pragma has been successfully parsed. */
856 /* This function has to be in this file, in order to get at
860 handle_generic_pragma (token
)
871 handle_pragma_token (token_buffer
, yylval
.ttype
);
875 return handle_pragma_token (NULL_PTR
, NULL_TREE
);
878 handle_pragma_token (token_buffer
, NULL
);
885 #endif /* HANDLE_GENERIC_PRAGMAS */
887 #define ENDFILE -1 /* token that represents end-of-file */
889 /* Read an escape sequence, returning its equivalent as a character,
890 or store 1 in *ignore_ptr if it is backslash-newline. */
893 readescape (ignore_ptr
)
896 register int c
= GETC();
898 register unsigned count
;
899 unsigned firstdig
= 0;
905 if (warn_traditional
&& !in_system_header
)
906 warning ("the meaning of `\\x' varies with -traditional");
908 if (flag_traditional
)
923 if (c
>= 'a' && c
<= 'f')
924 code
+= c
- 'a' + 10;
925 if (c
>= 'A' && c
<= 'F')
926 code
+= c
- 'A' + 10;
927 if (c
>= '0' && c
<= '9')
929 if (code
!= 0 || count
!= 0)
939 warning ("\\x used with no following hex digits");
943 /* Digits are all 0's. Ok. */
945 else if ((count
- 1) * 4 >= TYPE_PRECISION (integer_type_node
)
948 << (TYPE_PRECISION (integer_type_node
)
951 pedwarn ("hex escape out of range");
954 case '0': case '1': case '2': case '3': case '4':
955 case '5': case '6': case '7':
958 while ((c
<= '7') && (c
>= '0') && (count
++ < 3))
960 code
= (code
* 8) + (c
- '0');
966 case '\\': case '\'': case '"':
975 return TARGET_NEWLINE
;
990 if (warn_traditional
&& !in_system_header
)
991 warning ("the meaning of `\\a' varies with -traditional");
993 if (flag_traditional
)
998 #if 0 /* Vertical tab is present in common usage compilers. */
999 if (flag_traditional
)
1007 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c
);
1013 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1017 /* `\%' is used to prevent SCCS from getting confused. */
1020 pedwarn ("unknown escape sequence `\\%c'", c
);
1024 pedwarn ("unknown escape sequence `\\%c'", c
);
1026 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c
);
1034 const char *string
= _(msgid
);
1036 /* We can't print string and character constants well
1037 because the token_buffer contains the result of processing escapes. */
1039 error ("%s at end of input", string
);
1040 else if (token_buffer
[0] == 0)
1041 error ("%s at null character", string
);
1042 else if (token_buffer
[0] == '"')
1043 error ("%s before string constant", string
);
1044 else if (token_buffer
[0] == '\'')
1045 error ("%s before character constant", string
);
1046 else if (!ISGRAPH(token_buffer
[0]))
1047 error ("%s before character 0%o", string
, (unsigned char) token_buffer
[0]);
1049 error ("%s before `%s'", string
, token_buffer
);
1059 char long_long_flag
;
1062 struct try_type type_sequence
[] =
1064 { &integer_type_node
, 0, 0, 0},
1065 { &unsigned_type_node
, 1, 0, 0},
1066 { &long_integer_type_node
, 0, 1, 0},
1067 { &long_unsigned_type_node
, 1, 1, 0},
1068 { &long_long_integer_type_node
, 0, 1, 1},
1069 { &long_long_unsigned_type_node
, 1, 1, 1}
1083 int conversion_errno
;
1084 REAL_VALUE_TYPE value
;
1091 struct pf_args
* args
= (struct pf_args
*) data
;
1092 int fflag
= 0, lflag
= 0;
1093 /* Copy token_buffer now, while it has just the number
1094 and not the suffixes; once we add `f' or `i',
1095 REAL_VALUE_ATOF may not work any more. */
1096 char *copy
= (char *) alloca (args
->p
- token_buffer
+ 1);
1097 bcopy (token_buffer
, copy
, args
->p
- token_buffer
+ 1);
1099 args
->conversion_errno
= 0;
1100 args
->type
= double_type_node
;
1106 /* Read the suffixes to choose a data type. */
1111 error ("more than one `f' in numeric constant");
1112 else if (warn_traditional
&& !in_system_header
)
1113 warning ("traditional C rejects the `%c' suffix", args
->c
);
1119 error ("more than one `l' in numeric constant");
1120 else if (warn_traditional
&& !in_system_header
)
1121 warning ("traditional C rejects the `%c' suffix", args
->c
);
1127 error ("more than one `i' or `j' in numeric constant");
1129 pedwarn ("ISO C forbids imaginary numeric constants");
1140 if (args
->p
>= token_buffer
+ maxtoken
- 3)
1141 args
->p
= extend_token_buffer (args
->p
);
1142 *(args
->p
++) = args
->c
;
1147 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1148 tells the desired precision of the binary result
1149 of decimal-to-binary conversion. */
1154 error ("both `f' and `l' in floating constant");
1156 args
->type
= float_type_node
;
1158 if (args
->base
== 16)
1159 args
->value
= REAL_VALUE_HTOF (copy
, TYPE_MODE (args
->type
));
1161 args
->value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (args
->type
));
1162 args
->conversion_errno
= errno
;
1163 /* A diagnostic is required here by some ANSI C testsuites.
1164 This is not pedwarn, because some people don't want
1165 an error for this. */
1166 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
1167 warning ("floating point number exceeds range of `float'");
1171 args
->type
= long_double_type_node
;
1173 if (args
->base
== 16)
1174 args
->value
= REAL_VALUE_HTOF (copy
, TYPE_MODE (args
->type
));
1176 args
->value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (args
->type
));
1177 args
->conversion_errno
= errno
;
1178 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
1179 warning ("floating point number exceeds range of `long double'");
1184 if (flag_single_precision_constant
)
1185 args
->type
= float_type_node
;
1186 if (args
->base
== 16)
1187 args
->value
= REAL_VALUE_HTOF (copy
, TYPE_MODE (args
->type
));
1189 args
->value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (args
->type
));
1190 args
->conversion_errno
= errno
;
1191 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
1192 warning ("floating point number exceeds range of `double'");
1196 /* Get the next character, staying within the current token if possible.
1197 If we're lexing a token, we don't want to look beyond the end of the
1198 token cpplib has prepared for us; otherwise, we end up reading in the
1199 next token, which screws up feed_input. So just return a null
1202 static inline int token_getch
PARAMS ((void));
1208 if (yy_cur
== yy_lim
)
1214 static inline void token_put_back
PARAMS ((int));
1227 /* Read a single token from the input stream, and assign it lexical
1241 /* Effectively do c = skip_white_space (c)
1242 but do it faster in the usual cases. */
1252 if (cpp_token
== CPP_HSPACE
)
1253 c
= yy_get_token ();
1260 /* Call skip_white_space so we can warn if appropriate. */
1265 c
= skip_white_space (c
);
1267 goto found_nonwhite
;
1271 token_buffer
[0] = c
;
1272 token_buffer
[1] = 0;
1274 /* yylloc.first_line = lineno; */
1280 token_buffer
[0] = 0;
1282 value
= END_OF_LINE
;
1289 if (cpp_token
== CPP_NAME
)
1292 /* Capital L may start a wide-string or wide-character constant. */
1294 register int c
= token_getch();
1303 goto string_constant
;
1310 if (!doing_objc_thang
)
1317 /* '@' may start a constant string object. */
1318 register int c
= token_getch ();
1322 goto string_constant
;
1325 /* Fall through to treat '@' as the start of an identifier. */
1328 case 'A': case 'B': case 'C': case 'D': case 'E':
1329 case 'F': case 'G': case 'H': case 'I': case 'J':
1330 case 'K': case 'M': case 'N': case 'O':
1331 case 'P': case 'Q': case 'R': case 'S': case 'T':
1332 case 'U': case 'V': case 'W': case 'X': case 'Y':
1334 case 'a': case 'b': case 'c': case 'd': case 'e':
1335 case 'f': case 'g': case 'h': case 'i': case 'j':
1336 case 'k': case 'l': case 'm': case 'n': case 'o':
1337 case 'p': case 'q': case 'r': case 's': case 't':
1338 case 'u': case 'v': case 'w': case 'x': case 'y':
1344 if (cpp_token
== CPP_NAME
)
1346 /* Note that one character has already been read from
1347 yy_cur into token_buffer. Also, cpplib complains about
1348 $ in identifiers, so we don't have to. */
1350 int len
= yy_lim
- yy_cur
+ 1;
1351 if (len
>= maxtoken
)
1352 extend_token_buffer_to (len
+ 1);
1353 memcpy (token_buffer
+ 1, yy_cur
, len
);
1354 p
= token_buffer
+ len
;
1361 while (ISALNUM (c
) || c
== '_' || c
== '$' || c
== '@')
1363 /* Make sure this char really belongs in an identifier. */
1366 if (! dollars_in_ident
)
1367 error ("`$' in identifier");
1369 pedwarn ("`$' in identifier");
1372 if (p
>= token_buffer
+ maxtoken
)
1373 p
= extend_token_buffer (p
);
1386 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1389 register struct resword
*ptr
;
1391 if ((ptr
= is_reserved_word (token_buffer
, p
- token_buffer
)))
1394 yylval
.ttype
= ridpointers
[(int) ptr
->rid
];
1395 value
= (int) ptr
->token
;
1397 /* Only return OBJECTNAME if it is a typedef. */
1398 if (doing_objc_thang
&& value
== OBJECTNAME
)
1400 tree decl
= lookup_name(yylval
.ttype
);
1402 if (decl
== NULL_TREE
1403 || TREE_CODE (decl
) != TYPE_DECL
)
1407 /* Even if we decided to recognize asm, still perhaps warn. */
1409 && (value
== ASM_KEYWORD
|| value
== TYPEOF
1410 || (ptr
->rid
== RID_INLINE
&& ! flag_isoc99
))
1411 && token_buffer
[0] != '_')
1412 pedwarn ("ANSI does not permit the keyword `%s'",
1417 /* If we did not find a keyword, look for an identifier
1420 if (value
== IDENTIFIER
)
1424 if (token_buffer
[0] == '@')
1425 error("invalid identifier `%s'", token_buffer
);
1427 yylval
.ttype
= get_identifier (token_buffer
);
1428 decl
= lookup_name (yylval
.ttype
);
1430 if (decl
!= 0 && TREE_CODE (decl
) == TYPE_DECL
)
1432 /* A user-invisible read-only initialized variable
1433 should be replaced by its value.
1434 We handle only strings since that's the only case used in C. */
1435 else if (decl
!= 0 && TREE_CODE (decl
) == VAR_DECL
1436 && DECL_IGNORED_P (decl
)
1437 && TREE_READONLY (decl
)
1438 && DECL_INITIAL (decl
) != 0
1439 && TREE_CODE (DECL_INITIAL (decl
)) == STRING_CST
)
1441 tree stringval
= DECL_INITIAL (decl
);
1443 /* Copy the string value so that we won't clobber anything
1444 if we put something in the TREE_CHAIN of this one. */
1445 yylval
.ttype
= build_string (TREE_STRING_LENGTH (stringval
),
1446 TREE_STRING_POINTER (stringval
));
1449 else if (doing_objc_thang
)
1451 tree objc_interface_decl
= is_class_name (yylval
.ttype
);
1453 if (objc_interface_decl
)
1456 yylval
.ttype
= objc_interface_decl
;
1465 if (yy_cur
< yy_lim
)
1468 /* It's hard to preserve tokenization on '.' because
1469 it could be a symbol by itself, or it could be the
1470 start of a floating point number and cpp won't tell us. */
1471 register int c1
= token_getch ();
1472 token_buffer
[1] = c1
;
1475 c1
= token_getch ();
1478 token_buffer
[2] = c1
;
1479 token_buffer
[3] = 0;
1483 error ("parse error at `..'");
1487 token_put_back (c1
);
1490 token_put_back (c1
);
1493 token_buffer
[1] = 0;
1497 /* Optimize for most frequent case. */
1502 cond
= (yy_cur
== yy_lim
);
1504 register int c1
= token_getch ();
1505 token_put_back (c1
);
1506 cond
= (! ISALNUM (c1
) && c1
!= '.');
1510 yylval
.ttype
= (c
== '0') ? integer_zero_node
: integer_one_node
;
1516 case '2': case '3': case '4':
1517 case '5': case '6': case '7': case '8': case '9':
1522 int largest_digit
= 0;
1526 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
1527 The code below which fills the parts array assumes that a host
1528 int is at least twice as wide as a host char, and that
1529 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
1530 Two HOST_WIDE_INTs is the largest int literal we can store.
1531 In order to detect overflow below, the number of parts (TOTAL_PARTS)
1532 must be exactly the number of parts needed to hold the bits
1533 of two HOST_WIDE_INTs. */
1534 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
1535 unsigned int parts
[TOTAL_PARTS
];
1537 enum anon1
{ NOT_FLOAT
, AFTER_POINT
, TOO_MANY_POINTS
, AFTER_EXPON
}
1538 floatflag
= NOT_FLOAT
;
1540 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1548 *p
++ = (c
= token_getch());
1549 if ((c
== 'x') || (c
== 'X'))
1552 *p
++ = (c
= token_getch());
1554 /* Leading 0 forces octal unless the 0 is the only digit. */
1555 else if (c
>= '0' && c
<= '9')
1564 /* Read all the digits-and-decimal-points. */
1567 || (ISALNUM (c
) && c
!= 'l' && c
!= 'L'
1568 && c
!= 'u' && c
!= 'U'
1569 && c
!= 'i' && c
!= 'I' && c
!= 'j' && c
!= 'J'
1570 && (floatflag
== NOT_FLOAT
1571 || ((base
!= 16) && (c
!= 'f') && (c
!= 'F'))
1576 if (base
== 16 && pedantic
&& !flag_isoc99
)
1577 pedwarn ("floating constant may not be in radix 16");
1578 if (floatflag
== TOO_MANY_POINTS
)
1579 /* We have already emitted an error. Don't need another. */
1581 else if (floatflag
== AFTER_POINT
|| floatflag
== AFTER_EXPON
)
1583 error ("malformed floating constant");
1584 floatflag
= TOO_MANY_POINTS
;
1585 /* Avoid another error from atof by forcing all characters
1586 from here on to be ignored. */
1590 floatflag
= AFTER_POINT
;
1594 *p
++ = c
= token_getch();
1595 /* Accept '.' as the start of a floating-point number
1596 only when it is followed by a digit. */
1597 if (p
== token_buffer
+ 2 && !ISDIGIT (c
))
1602 /* It is not a decimal point.
1603 It should be a digit (perhaps a hex digit). */
1609 else if (base
<= 10)
1611 if (c
== 'e' || c
== 'E')
1614 floatflag
= AFTER_EXPON
;
1615 break; /* start of exponent */
1617 error ("nondigits in number and not hexadecimal");
1620 else if (base
== 16 && (c
== 'p' || c
== 'P'))
1622 floatflag
= AFTER_EXPON
;
1623 break; /* start of exponent */
1625 else if (c
>= 'a' && c
<= 'f')
1633 if (c
>= largest_digit
)
1637 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1639 parts
[count
] *= base
;
1643 += (parts
[count
-1] >> HOST_BITS_PER_CHAR
);
1645 &= (1 << HOST_BITS_PER_CHAR
) - 1;
1651 /* If the highest-order part overflows (gets larger than
1652 a host char will hold) then the whole number has
1653 overflowed. Record this and truncate the highest-order
1655 if (parts
[TOTAL_PARTS
- 1] >> HOST_BITS_PER_CHAR
)
1658 parts
[TOTAL_PARTS
- 1] &= (1 << HOST_BITS_PER_CHAR
) - 1;
1661 if (p
>= token_buffer
+ maxtoken
- 3)
1662 p
= extend_token_buffer (p
);
1663 *p
++ = (c
= token_getch());
1667 /* This can happen on input like `int i = 0x;' */
1669 error ("numeric constant with no digits");
1671 if (largest_digit
>= base
)
1672 error ("numeric constant contains digits beyond the radix");
1674 /* Remove terminating char from the token buffer and delimit the
1678 if (floatflag
!= NOT_FLOAT
)
1681 int imag
, conversion_errno
;
1682 REAL_VALUE_TYPE value
;
1683 struct pf_args args
;
1685 /* Read explicit exponent if any, and put it in tokenbuf. */
1687 if ((base
== 10 && ((c
== 'e') || (c
== 'E')))
1688 || (base
== 16 && (c
== 'p' || c
== 'P')))
1690 if (p
>= token_buffer
+ maxtoken
- 3)
1691 p
= extend_token_buffer (p
);
1694 if ((c
== '+') || (c
== '-'))
1699 /* Exponent is decimal, even if string is a hex float. */
1701 error ("floating constant exponent has no digits");
1704 if (p
>= token_buffer
+ maxtoken
- 3)
1705 p
= extend_token_buffer (p
);
1710 if (base
== 16 && floatflag
!= AFTER_EXPON
)
1711 error ("hexadecimal floating constant has no exponent");
1715 /* Setup input for parse_float() */
1720 /* Convert string to a double, checking for overflow. */
1721 if (do_float_handler (parse_float
, (PTR
) &args
))
1723 /* Receive output from parse_float() */
1728 /* We got an exception from parse_float() */
1729 error ("floating constant out of range");
1733 /* Receive output from parse_float() */
1737 conversion_errno
= args
.conversion_errno
;
1740 /* ERANGE is also reported for underflow,
1741 so test the value to distinguish overflow from that. */
1742 if (conversion_errno
== ERANGE
&& !flag_traditional
&& pedantic
1743 && (REAL_VALUES_LESS (dconst1
, value
)
1744 || REAL_VALUES_LESS (value
, dconstm1
)))
1745 warning ("floating point number exceeds range of `double'");
1748 /* If the result is not a number, assume it must have been
1749 due to some error message above, so silently convert
1751 if (REAL_VALUE_ISNAN (value
))
1754 /* Create a node with determined type and value. */
1756 yylval
.ttype
= build_complex (NULL_TREE
,
1757 convert (type
, integer_zero_node
),
1758 build_real (type
, value
));
1760 yylval
.ttype
= build_real (type
, value
);
1764 tree traditional_type
, ansi_type
, type
;
1765 HOST_WIDE_INT high
, low
;
1766 int spec_unsigned
= 0;
1768 int spec_long_long
= 0;
1773 traditional_type
= ansi_type
= type
= NULL_TREE
;
1776 if (c
== 'u' || c
== 'U')
1779 error ("two `u's in integer constant");
1780 else if (warn_traditional
&& !in_system_header
)
1781 warning ("traditional C rejects the `%c' suffix", c
);
1786 else if (c
== 'l' || c
== 'L')
1791 error ("three `l's in integer constant");
1793 error ("`LUL' is not a valid integer suffix");
1794 else if (c
!= spec_long
)
1795 error ("`Ll' and `lL' are not valid integer suffixes");
1796 else if (pedantic
&& ! flag_isoc99
1797 && ! in_system_header
&& warn_long_long
)
1798 pedwarn ("ISO C89 forbids long long integer constants");
1803 else if (c
== 'i' || c
== 'j' || c
== 'I' || c
== 'J')
1806 error ("more than one `i' or `j' in numeric constant");
1808 pedwarn ("ISO C forbids imaginary numeric constants");
1813 if (p
>= token_buffer
+ maxtoken
- 3)
1814 p
= extend_token_buffer (p
);
1819 /* If the literal overflowed, pedwarn about it now. */
1823 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT
* 2);
1826 /* This is simplified by the fact that our constant
1827 is always positive. */
1831 for (i
= 0; i
< HOST_BITS_PER_WIDE_INT
/ HOST_BITS_PER_CHAR
; i
++)
1833 high
|= ((HOST_WIDE_INT
) parts
[i
+ (HOST_BITS_PER_WIDE_INT
1834 / HOST_BITS_PER_CHAR
)]
1835 << (i
* HOST_BITS_PER_CHAR
));
1836 low
|= (HOST_WIDE_INT
) parts
[i
] << (i
* HOST_BITS_PER_CHAR
);
1839 yylval
.ttype
= build_int_2 (low
, high
);
1840 TREE_TYPE (yylval
.ttype
) = long_long_unsigned_type_node
;
1842 /* If warn_traditional, calculate both the ANSI type and the
1843 traditional type, then see if they disagree.
1844 Otherwise, calculate only the type for the dialect in use. */
1845 if (warn_traditional
|| flag_traditional
)
1847 /* Calculate the traditional type. */
1848 /* Traditionally, any constant is signed;
1849 but if unsigned is specified explicitly, obey that.
1850 Use the smallest size with the right number of bits,
1851 except for one special case with decimal constants. */
1852 if (! spec_long
&& base
!= 10
1853 && int_fits_type_p (yylval
.ttype
, unsigned_type_node
))
1854 traditional_type
= (spec_unsigned
? unsigned_type_node
1855 : integer_type_node
);
1856 /* A decimal constant must be long
1857 if it does not fit in type int.
1858 I think this is independent of whether
1859 the constant is signed. */
1860 else if (! spec_long
&& base
== 10
1861 && int_fits_type_p (yylval
.ttype
, integer_type_node
))
1862 traditional_type
= (spec_unsigned
? unsigned_type_node
1863 : integer_type_node
);
1864 else if (! spec_long_long
)
1865 traditional_type
= (spec_unsigned
? long_unsigned_type_node
1866 : long_integer_type_node
);
1867 else if (int_fits_type_p (yylval
.ttype
,
1869 ? long_long_unsigned_type_node
1870 : long_long_integer_type_node
))
1871 traditional_type
= (spec_unsigned
1872 ? long_long_unsigned_type_node
1873 : long_long_integer_type_node
);
1875 traditional_type
= (spec_unsigned
1876 ? widest_unsigned_literal_type_node
1877 : widest_integer_literal_type_node
);
1879 if (warn_traditional
|| ! flag_traditional
)
1881 /* Calculate the ANSI type. */
1882 if (! spec_long
&& ! spec_unsigned
1883 && int_fits_type_p (yylval
.ttype
, integer_type_node
))
1884 ansi_type
= integer_type_node
;
1885 else if (! spec_long
&& (base
!= 10 || spec_unsigned
)
1886 && int_fits_type_p (yylval
.ttype
, unsigned_type_node
))
1887 ansi_type
= unsigned_type_node
;
1888 else if (! spec_unsigned
&& !spec_long_long
1889 && int_fits_type_p (yylval
.ttype
, long_integer_type_node
))
1890 ansi_type
= long_integer_type_node
;
1891 else if (! spec_long_long
1892 && int_fits_type_p (yylval
.ttype
,
1893 long_unsigned_type_node
))
1894 ansi_type
= long_unsigned_type_node
;
1895 else if (! spec_unsigned
1896 && int_fits_type_p (yylval
.ttype
,
1897 long_long_integer_type_node
))
1898 ansi_type
= long_long_integer_type_node
;
1899 else if (int_fits_type_p (yylval
.ttype
,
1900 long_long_unsigned_type_node
))
1901 ansi_type
= long_long_unsigned_type_node
;
1902 else if (! spec_unsigned
1903 && int_fits_type_p (yylval
.ttype
,
1904 widest_integer_literal_type_node
))
1905 ansi_type
= widest_integer_literal_type_node
;
1907 ansi_type
= widest_unsigned_literal_type_node
;
1910 type
= flag_traditional
? traditional_type
: ansi_type
;
1912 /* We assume that constants specified in a non-decimal
1913 base are bit patterns, and that the programmer really
1914 meant what they wrote. */
1915 if (warn_traditional
&& !in_system_header
&& base
== 10
1916 && traditional_type
!= ansi_type
)
1918 if (TYPE_PRECISION (traditional_type
)
1919 != TYPE_PRECISION (ansi_type
))
1920 warning ("width of integer constant changes with -traditional");
1921 else if (TREE_UNSIGNED (traditional_type
)
1922 != TREE_UNSIGNED (ansi_type
))
1923 warning ("integer constant is unsigned in ISO C, signed with -traditional");
1925 warning ("width of integer constant may change on other systems with -traditional");
1928 if (pedantic
&& !flag_traditional
&& !spec_long_long
&& !warn
1929 && (TYPE_PRECISION (long_integer_type_node
)
1930 < TYPE_PRECISION (type
)))
1933 pedwarn ("integer constant larger than the maximum value of an unsigned long int");
1936 if (base
== 10 && ! spec_unsigned
&& TREE_UNSIGNED (type
))
1937 warning ("decimal constant is so large that it is unsigned");
1941 if (TYPE_PRECISION (type
)
1942 <= TYPE_PRECISION (integer_type_node
))
1944 = build_complex (NULL_TREE
, integer_zero_node
,
1945 convert (integer_type_node
,
1948 error ("complex integer constant is too wide for `complex int'");
1950 else if (flag_traditional
&& !int_fits_type_p (yylval
.ttype
, type
))
1951 /* The traditional constant 0x80000000 is signed
1952 but doesn't fit in the range of int.
1953 This will change it to -0x80000000, which does fit. */
1955 TREE_TYPE (yylval
.ttype
) = unsigned_type (type
);
1956 yylval
.ttype
= convert (type
, yylval
.ttype
);
1957 TREE_OVERFLOW (yylval
.ttype
)
1958 = TREE_CONSTANT_OVERFLOW (yylval
.ttype
) = 0;
1961 TREE_TYPE (yylval
.ttype
) = type
;
1964 /* If it's still an integer (not a complex), and it doesn't
1965 fit in the type we choose for it, then pedwarn. */
1968 && TREE_CODE (TREE_TYPE (yylval
.ttype
)) == INTEGER_TYPE
1969 && ! int_fits_type_p (yylval
.ttype
, TREE_TYPE (yylval
.ttype
)))
1970 pedwarn ("integer constant is larger than the maximum value for its type");
1976 if (ISALNUM (c
) || c
== '.' || c
== '_' || c
== '$'
1977 || (!flag_traditional
&& (c
== '-' || c
== '+')
1978 && (p
[-1] == 'e' || p
[-1] == 'E')))
1979 error ("missing white space after number `%s'", token_buffer
);
1981 value
= CONSTANT
; break;
1987 register int result
= 0;
1988 register int num_chars
= 0;
1990 unsigned width
= TYPE_PRECISION (char_type_node
);
1992 #ifdef MULTIBYTE_CHARS
1993 int longest_char
= local_mb_cur_max ();
1994 (void) local_mbtowc (NULL_PTR
, NULL_PTR
, 0);
1997 max_chars
= TYPE_PRECISION (integer_type_node
) / width
;
1999 width
= WCHAR_TYPE_SIZE
;
2006 if (c
== '\'' || c
== EOF
)
2013 c
= readescape (&ignore
);
2016 if (width
< HOST_BITS_PER_INT
2017 && (unsigned) c
>= ((unsigned)1 << width
))
2018 pedwarn ("escape sequence out of range for character");
2019 #ifdef MAP_CHARACTER
2021 c
= MAP_CHARACTER (c
);
2027 pedwarn ("ISO C forbids newline in character constant");
2032 #ifdef MULTIBYTE_CHARS
2036 for (i
= 1; i
<= longest_char
; ++i
)
2038 if (i
> maxtoken
- 4)
2039 extend_token_buffer (token_buffer
);
2041 token_buffer
[i
] = c
;
2042 char_len
= local_mbtowc (& wc
,
2051 /* mbtowc sometimes needs an extra char before accepting */
2056 /* Merge character into result; ignore excess chars. */
2057 for (i
= 1; i
<= char_len
; ++i
)
2061 if (width
< HOST_BITS_PER_INT
)
2062 result
= (result
<< width
)
2064 & ((1 << width
) - 1));
2066 result
= token_buffer
[i
];
2068 num_chars
+= char_len
;
2077 warning ("Ignoring invalid multibyte character");
2078 /* Replace all but the first byte. */
2079 for (--i
; i
> 1; --i
)
2080 token_put_back (token_buffer
[i
]);
2081 wc
= token_buffer
[1];
2083 #ifdef MAP_CHARACTER
2084 c
= MAP_CHARACTER (wc
);
2089 #else /* ! MULTIBYTE_CHARS */
2090 #ifdef MAP_CHARACTER
2091 c
= MAP_CHARACTER (c
);
2093 #endif /* ! MULTIBYTE_CHARS */
2098 if (chars_seen
== 1) /* only keep the first one */
2103 /* Merge character into result; ignore excess chars. */
2104 num_chars
+= (width
/ TYPE_PRECISION (char_type_node
));
2105 if (num_chars
< max_chars
+ 1)
2107 if (width
< HOST_BITS_PER_INT
)
2108 result
= (result
<< width
) | (c
& ((1 << width
) - 1));
2115 error ("malformed character constant");
2116 else if (chars_seen
== 0)
2117 error ("empty character constant");
2118 else if (num_chars
> max_chars
)
2120 num_chars
= max_chars
;
2121 error ("character constant too long");
2123 else if (chars_seen
!= 1 && ! flag_traditional
&& warn_multichar
)
2124 warning ("multi-character character constant");
2126 /* If char type is signed, sign-extend the constant. */
2129 int num_bits
= num_chars
* width
;
2131 /* We already got an error; avoid invalid shift. */
2132 yylval
.ttype
= build_int_2 (0, 0);
2133 else if (TREE_UNSIGNED (char_type_node
)
2134 || ((result
>> (num_bits
- 1)) & 1) == 0)
2136 = build_int_2 (result
& (~(unsigned HOST_WIDE_INT
) 0
2137 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
2141 = build_int_2 (result
| ~(~(unsigned HOST_WIDE_INT
) 0
2142 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
2144 TREE_TYPE (yylval
.ttype
) = integer_type_node
;
2148 yylval
.ttype
= build_int_2 (result
, 0);
2149 TREE_TYPE (yylval
.ttype
) = wchar_type_node
;
2159 unsigned width
= wide_flag
? WCHAR_TYPE_SIZE
2160 : TYPE_PRECISION (char_type_node
);
2161 #ifdef MULTIBYTE_CHARS
2162 int longest_char
= local_mb_cur_max ();
2163 (void) local_mbtowc (NULL_PTR
, NULL_PTR
, 0);
2166 p
= token_buffer
+ 1;
2168 while (c
!= '"' && c
!= EOF
)
2170 /* ignore_escape_flag is set for reading the filename in #line. */
2171 if (!ignore_escape_flag
&& c
== '\\')
2174 c
= readescape (&ignore
);
2177 if (width
< HOST_BITS_PER_INT
2178 && (unsigned) c
>= ((unsigned)1 << width
))
2179 pedwarn ("escape sequence out of range for character");
2184 pedwarn ("ISO C forbids newline in string constant");
2189 #ifdef MULTIBYTE_CHARS
2193 for (i
= 0; i
< longest_char
; ++i
)
2195 if (p
+ i
>= token_buffer
+ maxtoken
)
2196 p
= extend_token_buffer (p
);
2199 char_len
= local_mbtowc (& wc
, p
, i
+ 1);
2206 warning ("Ignoring invalid multibyte character");
2207 /* Replace all except the first byte. */
2209 for (--i
; i
> 0; --i
)
2210 token_put_back (p
[i
]);
2213 /* mbtowc sometimes needs an extra char before accepting */
2223 #endif /* MULTIBYTE_CHARS */
2226 /* Add this single character into the buffer either as a wchar_t
2227 or as a single byte. */
2230 unsigned width
= TYPE_PRECISION (char_type_node
);
2231 unsigned bytemask
= (1 << width
) - 1;
2234 if (p
+ WCHAR_BYTES
> token_buffer
+ maxtoken
)
2235 p
= extend_token_buffer (p
);
2237 for (byte
= 0; byte
< WCHAR_BYTES
; ++byte
)
2240 if (byte
>= (int) sizeof (c
))
2243 value
= (c
>> (byte
* width
)) & bytemask
;
2244 if (BYTES_BIG_ENDIAN
)
2245 p
[WCHAR_BYTES
- byte
- 1] = value
;
2253 if (p
>= token_buffer
+ maxtoken
)
2254 p
= extend_token_buffer (p
);
2262 /* Terminate the string value, either with a single byte zero
2263 or with a wide zero. */
2266 if (p
+ WCHAR_BYTES
> token_buffer
+ maxtoken
)
2267 p
= extend_token_buffer (p
);
2268 bzero (p
, WCHAR_BYTES
);
2273 if (p
>= token_buffer
+ maxtoken
)
2274 p
= extend_token_buffer (p
);
2279 error ("Unterminated string constant");
2281 /* We have read the entire constant.
2282 Construct a STRING_CST for the result. */
2284 yylval
.ttype
= build_string (p
- (token_buffer
+ 1), token_buffer
+ 1);
2287 TREE_TYPE (yylval
.ttype
) = wchar_array_type_node
;
2292 TREE_TYPE (yylval
.ttype
) = char_array_type_node
;
2293 value
= OBJC_STRING
;
2297 TREE_TYPE (yylval
.ttype
) = char_array_type_node
;
2325 yylval
.code
= PLUS_EXPR
; break;
2327 yylval
.code
= MINUS_EXPR
; break;
2329 yylval
.code
= BIT_AND_EXPR
; break;
2331 yylval
.code
= BIT_IOR_EXPR
; break;
2333 yylval
.code
= MULT_EXPR
; break;
2335 yylval
.code
= TRUNC_DIV_EXPR
; break;
2337 yylval
.code
= TRUNC_MOD_EXPR
; break;
2339 yylval
.code
= BIT_XOR_EXPR
; break;
2341 yylval
.code
= LSHIFT_EXPR
; break;
2343 yylval
.code
= RSHIFT_EXPR
; break;
2345 yylval
.code
= LT_EXPR
; break;
2347 yylval
.code
= GT_EXPR
; break;
2350 token_buffer
[1] = c1
= token_getch();
2351 token_buffer
[2] = 0;
2358 value
= ARITHCOMPARE
; yylval
.code
= LE_EXPR
; goto done
;
2360 value
= ARITHCOMPARE
; yylval
.code
= GE_EXPR
; goto done
;
2362 value
= EQCOMPARE
; yylval
.code
= NE_EXPR
; goto done
;
2364 value
= EQCOMPARE
; yylval
.code
= EQ_EXPR
; goto done
;
2366 value
= ASSIGN
; goto done
;
2372 value
= PLUSPLUS
; goto done
;
2374 value
= MINUSMINUS
; goto done
;
2376 value
= ANDAND
; goto done
;
2378 value
= OROR
; goto done
;
2391 { value
= POINTSAT
; goto done
; }
2396 if (c1
== '>' && flag_digraphs
)
2397 { value
= ']'; goto done
; }
2403 { value
= '{'; indent_level
++; goto done
; }
2405 { value
= '['; goto done
; }
2409 if (c1
== '>' && flag_digraphs
)
2410 { value
= '}'; indent_level
--; goto done
; }
2414 token_put_back (c1
);
2415 token_buffer
[1] = 0;
2417 if ((c
== '<') || (c
== '>'))
2418 value
= ARITHCOMPARE
;
2424 /* Don't make yyparse think this is eof. */
2443 /* yylloc.last_line = lineno; */
2448 /* Sets the value of the 'yydebug' variable to VALUE.
2449 This is a function so we don't have to have YYDEBUG defined
2450 in order to build the compiler. */
2459 warning ("YYDEBUG not defined.");