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
)
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
)
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");
1117 error ("more than one `l' in numeric constant");
1123 error ("more than one `i' or `j' in numeric constant");
1125 pedwarn ("ANSI C forbids imaginary numeric constants");
1136 if (args
->p
>= token_buffer
+ maxtoken
- 3)
1137 args
->p
= extend_token_buffer (args
->p
);
1138 *(args
->p
++) = args
->c
;
1143 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1144 tells the desired precision of the binary result
1145 of decimal-to-binary conversion. */
1150 error ("both `f' and `l' in floating constant");
1152 args
->type
= float_type_node
;
1154 if (args
->base
== 16)
1155 args
->value
= REAL_VALUE_HTOF (copy
, TYPE_MODE (args
->type
));
1157 args
->value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (args
->type
));
1158 args
->conversion_errno
= errno
;
1159 /* A diagnostic is required here by some ANSI C testsuites.
1160 This is not pedwarn, because some people don't want
1161 an error for this. */
1162 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
1163 warning ("floating point number exceeds range of `float'");
1167 args
->type
= long_double_type_node
;
1169 if (args
->base
== 16)
1170 args
->value
= REAL_VALUE_HTOF (copy
, TYPE_MODE (args
->type
));
1172 args
->value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (args
->type
));
1173 args
->conversion_errno
= errno
;
1174 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
1175 warning ("floating point number exceeds range of `long double'");
1180 if (flag_single_precision_constant
)
1181 args
->type
= float_type_node
;
1182 if (args
->base
== 16)
1183 args
->value
= REAL_VALUE_HTOF (copy
, TYPE_MODE (args
->type
));
1185 args
->value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (args
->type
));
1186 args
->conversion_errno
= errno
;
1187 if (REAL_VALUE_ISINF (args
->value
) && pedantic
)
1188 warning ("floating point number exceeds range of `double'");
1192 /* Get the next character, staying within the current token if possible.
1193 If we're lexing a token, we don't want to look beyond the end of the
1194 token cpplib has prepared for us; otherwise, we end up reading in the
1195 next token, which screws up feed_input. So just return a null
1198 static inline int token_getch
PARAMS ((void));
1204 if (yy_cur
== yy_lim
)
1210 static inline void token_put_back
PARAMS ((int));
1223 /* Read a single token from the input stream, and assign it lexical
1237 /* Effectively do c = skip_white_space (c)
1238 but do it faster in the usual cases. */
1248 if (cpp_token
== CPP_HSPACE
)
1249 c
= yy_get_token ();
1256 /* Call skip_white_space so we can warn if appropriate. */
1261 c
= skip_white_space (c
);
1263 goto found_nonwhite
;
1267 token_buffer
[0] = c
;
1268 token_buffer
[1] = 0;
1270 /* yylloc.first_line = lineno; */
1276 token_buffer
[0] = 0;
1278 value
= END_OF_LINE
;
1285 if (cpp_token
== CPP_NAME
)
1288 /* Capital L may start a wide-string or wide-character constant. */
1290 register int c
= token_getch();
1299 goto string_constant
;
1306 if (!doing_objc_thang
)
1313 /* '@' may start a constant string object. */
1314 register int c
= token_getch ();
1318 goto string_constant
;
1321 /* Fall through to treat '@' as the start of an identifier. */
1324 case 'A': case 'B': case 'C': case 'D': case 'E':
1325 case 'F': case 'G': case 'H': case 'I': case 'J':
1326 case 'K': case 'M': case 'N': case 'O':
1327 case 'P': case 'Q': case 'R': case 'S': case 'T':
1328 case 'U': case 'V': case 'W': case 'X': case 'Y':
1330 case 'a': case 'b': case 'c': case 'd': case 'e':
1331 case 'f': case 'g': case 'h': case 'i': case 'j':
1332 case 'k': case 'l': case 'm': case 'n': case 'o':
1333 case 'p': case 'q': case 'r': case 's': case 't':
1334 case 'u': case 'v': case 'w': case 'x': case 'y':
1340 if (cpp_token
== CPP_NAME
)
1342 /* Note that one character has already been read from
1343 yy_cur into token_buffer. Also, cpplib complains about
1344 $ in identifiers, so we don't have to. */
1346 int len
= yy_lim
- yy_cur
+ 1;
1347 if (len
>= maxtoken
)
1348 extend_token_buffer_to (len
+ 1);
1349 memcpy (token_buffer
+ 1, yy_cur
, len
);
1350 p
= token_buffer
+ len
;
1357 while (ISALNUM (c
) || c
== '_' || c
== '$' || c
== '@')
1359 /* Make sure this char really belongs in an identifier. */
1362 if (! dollars_in_ident
)
1363 error ("`$' in identifier");
1365 pedwarn ("`$' in identifier");
1368 if (p
>= token_buffer
+ maxtoken
)
1369 p
= extend_token_buffer (p
);
1382 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1385 register struct resword
*ptr
;
1387 if ((ptr
= is_reserved_word (token_buffer
, p
- token_buffer
)))
1390 yylval
.ttype
= ridpointers
[(int) ptr
->rid
];
1391 value
= (int) ptr
->token
;
1393 /* Only return OBJECTNAME if it is a typedef. */
1394 if (doing_objc_thang
&& value
== OBJECTNAME
)
1396 tree decl
= lookup_name(yylval
.ttype
);
1398 if (decl
== NULL_TREE
1399 || TREE_CODE (decl
) != TYPE_DECL
)
1403 /* Even if we decided to recognize asm, still perhaps warn. */
1405 && (value
== ASM_KEYWORD
|| value
== TYPEOF
1406 || (ptr
->rid
== RID_INLINE
&& ! flag_isoc99
))
1407 && token_buffer
[0] != '_')
1408 pedwarn ("ANSI does not permit the keyword `%s'",
1413 /* If we did not find a keyword, look for an identifier
1416 if (value
== IDENTIFIER
)
1420 if (token_buffer
[0] == '@')
1421 error("invalid identifier `%s'", token_buffer
);
1423 yylval
.ttype
= get_identifier (token_buffer
);
1424 decl
= lookup_name (yylval
.ttype
);
1426 if (decl
!= 0 && TREE_CODE (decl
) == TYPE_DECL
)
1428 /* A user-invisible read-only initialized variable
1429 should be replaced by its value.
1430 We handle only strings since that's the only case used in C. */
1431 else if (decl
!= 0 && TREE_CODE (decl
) == VAR_DECL
1432 && DECL_IGNORED_P (decl
)
1433 && TREE_READONLY (decl
)
1434 && DECL_INITIAL (decl
) != 0
1435 && TREE_CODE (DECL_INITIAL (decl
)) == STRING_CST
)
1437 tree stringval
= DECL_INITIAL (decl
);
1439 /* Copy the string value so that we won't clobber anything
1440 if we put something in the TREE_CHAIN of this one. */
1441 yylval
.ttype
= build_string (TREE_STRING_LENGTH (stringval
),
1442 TREE_STRING_POINTER (stringval
));
1445 else if (doing_objc_thang
)
1447 tree objc_interface_decl
= is_class_name (yylval
.ttype
);
1449 if (objc_interface_decl
)
1452 yylval
.ttype
= objc_interface_decl
;
1461 if (yy_cur
< yy_lim
)
1464 /* It's hard to preserve tokenization on '.' because
1465 it could be a symbol by itself, or it could be the
1466 start of a floating point number and cpp won't tell us. */
1467 register int c1
= token_getch ();
1468 token_buffer
[1] = c1
;
1471 c1
= token_getch ();
1474 token_buffer
[2] = c1
;
1475 token_buffer
[3] = 0;
1479 error ("parse error at `..'");
1483 token_put_back (c1
);
1486 token_put_back (c1
);
1489 token_buffer
[1] = 0;
1493 /* Optimize for most frequent case. */
1498 cond
= (yy_cur
== yy_lim
);
1500 register int c1
= token_getch ();
1501 token_put_back (c1
);
1502 cond
= (! ISALNUM (c1
) && c1
!= '.');
1506 yylval
.ttype
= (c
== '0') ? integer_zero_node
: integer_one_node
;
1512 case '2': case '3': case '4':
1513 case '5': case '6': case '7': case '8': case '9':
1518 int largest_digit
= 0;
1522 /* We actually store only HOST_BITS_PER_CHAR bits in each part.
1523 The code below which fills the parts array assumes that a host
1524 int is at least twice as wide as a host char, and that
1525 HOST_BITS_PER_WIDE_INT is an even multiple of HOST_BITS_PER_CHAR.
1526 Two HOST_WIDE_INTs is the largest int literal we can store.
1527 In order to detect overflow below, the number of parts (TOTAL_PARTS)
1528 must be exactly the number of parts needed to hold the bits
1529 of two HOST_WIDE_INTs. */
1530 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2)
1531 unsigned int parts
[TOTAL_PARTS
];
1533 enum anon1
{ NOT_FLOAT
, AFTER_POINT
, TOO_MANY_POINTS
, AFTER_EXPON
}
1534 floatflag
= NOT_FLOAT
;
1536 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1544 *p
++ = (c
= token_getch());
1545 if ((c
== 'x') || (c
== 'X'))
1548 *p
++ = (c
= token_getch());
1550 /* Leading 0 forces octal unless the 0 is the only digit. */
1551 else if (c
>= '0' && c
<= '9')
1560 /* Read all the digits-and-decimal-points. */
1563 || (ISALNUM (c
) && c
!= 'l' && c
!= 'L'
1564 && c
!= 'u' && c
!= 'U'
1565 && c
!= 'i' && c
!= 'I' && c
!= 'j' && c
!= 'J'
1566 && (floatflag
== NOT_FLOAT
1567 || ((base
!= 16) && (c
!= 'f') && (c
!= 'F'))
1572 if (base
== 16 && pedantic
&& !flag_isoc99
)
1573 pedwarn ("floating constant may not be in radix 16");
1574 if (floatflag
== TOO_MANY_POINTS
)
1575 /* We have already emitted an error. Don't need another. */
1577 else if (floatflag
== AFTER_POINT
|| floatflag
== AFTER_EXPON
)
1579 error ("malformed floating constant");
1580 floatflag
= TOO_MANY_POINTS
;
1581 /* Avoid another error from atof by forcing all characters
1582 from here on to be ignored. */
1586 floatflag
= AFTER_POINT
;
1590 *p
++ = c
= token_getch();
1591 /* Accept '.' as the start of a floating-point number
1592 only when it is followed by a digit. */
1593 if (p
== token_buffer
+ 2 && !ISDIGIT (c
))
1598 /* It is not a decimal point.
1599 It should be a digit (perhaps a hex digit). */
1605 else if (base
<= 10)
1607 if (c
== 'e' || c
== 'E')
1610 floatflag
= AFTER_EXPON
;
1611 break; /* start of exponent */
1613 error ("nondigits in number and not hexadecimal");
1616 else if (base
== 16 && (c
== 'p' || c
== 'P'))
1618 floatflag
= AFTER_EXPON
;
1619 break; /* start of exponent */
1621 else if (c
>= 'a' && c
<= 'f')
1629 if (c
>= largest_digit
)
1633 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1635 parts
[count
] *= base
;
1639 += (parts
[count
-1] >> HOST_BITS_PER_CHAR
);
1641 &= (1 << HOST_BITS_PER_CHAR
) - 1;
1647 /* If the highest-order part overflows (gets larger than
1648 a host char will hold) then the whole number has
1649 overflowed. Record this and truncate the highest-order
1651 if (parts
[TOTAL_PARTS
- 1] >> HOST_BITS_PER_CHAR
)
1654 parts
[TOTAL_PARTS
- 1] &= (1 << HOST_BITS_PER_CHAR
) - 1;
1657 if (p
>= token_buffer
+ maxtoken
- 3)
1658 p
= extend_token_buffer (p
);
1659 *p
++ = (c
= token_getch());
1663 /* This can happen on input like `int i = 0x;' */
1665 error ("numeric constant with no digits");
1667 if (largest_digit
>= base
)
1668 error ("numeric constant contains digits beyond the radix");
1670 /* Remove terminating char from the token buffer and delimit the
1674 if (floatflag
!= NOT_FLOAT
)
1677 int imag
, conversion_errno
;
1678 REAL_VALUE_TYPE value
;
1679 struct pf_args args
;
1681 /* Read explicit exponent if any, and put it in tokenbuf. */
1683 if ((base
== 10 && ((c
== 'e') || (c
== 'E')))
1684 || (base
== 16 && (c
== 'p' || c
== 'P')))
1686 if (p
>= token_buffer
+ maxtoken
- 3)
1687 p
= extend_token_buffer (p
);
1690 if ((c
== '+') || (c
== '-'))
1695 /* Exponent is decimal, even if string is a hex float. */
1697 error ("floating constant exponent has no digits");
1700 if (p
>= token_buffer
+ maxtoken
- 3)
1701 p
= extend_token_buffer (p
);
1706 if (base
== 16 && floatflag
!= AFTER_EXPON
)
1707 error ("hexadecimal floating constant has no exponent");
1711 /* Setup input for parse_float() */
1716 /* Convert string to a double, checking for overflow. */
1717 if (do_float_handler (parse_float
, (PTR
) &args
))
1719 /* Receive output from parse_float() */
1724 /* We got an exception from parse_float() */
1725 error ("floating constant out of range");
1729 /* Receive output from parse_float() */
1733 conversion_errno
= args
.conversion_errno
;
1736 /* ERANGE is also reported for underflow,
1737 so test the value to distinguish overflow from that. */
1738 if (conversion_errno
== ERANGE
&& !flag_traditional
&& pedantic
1739 && (REAL_VALUES_LESS (dconst1
, value
)
1740 || REAL_VALUES_LESS (value
, dconstm1
)))
1741 warning ("floating point number exceeds range of `double'");
1744 /* If the result is not a number, assume it must have been
1745 due to some error message above, so silently convert
1747 if (REAL_VALUE_ISNAN (value
))
1750 /* Create a node with determined type and value. */
1752 yylval
.ttype
= build_complex (NULL_TREE
,
1753 convert (type
, integer_zero_node
),
1754 build_real (type
, value
));
1756 yylval
.ttype
= build_real (type
, value
);
1760 tree traditional_type
, ansi_type
, type
;
1761 HOST_WIDE_INT high
, low
;
1762 int spec_unsigned
= 0;
1764 int spec_long_long
= 0;
1768 traditional_type
= ansi_type
= type
= NULL_TREE
;
1771 if (c
== 'u' || c
== 'U')
1774 error ("two `u's in integer constant");
1777 else if (c
== 'l' || c
== 'L')
1782 error ("three `l's in integer constant");
1783 else if (pedantic
&& ! flag_isoc99
1784 && ! in_system_header
&& warn_long_long
)
1785 pedwarn ("ANSI C forbids long long integer constants");
1790 else if (c
== 'i' || c
== 'j' || c
== 'I' || c
== 'J')
1793 error ("more than one `i' or `j' in numeric constant");
1795 pedwarn ("ANSI C forbids imaginary numeric constants");
1800 if (p
>= token_buffer
+ maxtoken
- 3)
1801 p
= extend_token_buffer (p
);
1806 /* If the literal overflowed, pedwarn about it now. */
1810 pedwarn ("integer constant is too large for this configuration of the compiler - truncated to %d bits", HOST_BITS_PER_WIDE_INT
* 2);
1813 /* This is simplified by the fact that our constant
1814 is always positive. */
1818 for (i
= 0; i
< HOST_BITS_PER_WIDE_INT
/ HOST_BITS_PER_CHAR
; i
++)
1820 high
|= ((HOST_WIDE_INT
) parts
[i
+ (HOST_BITS_PER_WIDE_INT
1821 / HOST_BITS_PER_CHAR
)]
1822 << (i
* HOST_BITS_PER_CHAR
));
1823 low
|= (HOST_WIDE_INT
) parts
[i
] << (i
* HOST_BITS_PER_CHAR
);
1826 yylval
.ttype
= build_int_2 (low
, high
);
1827 TREE_TYPE (yylval
.ttype
) = long_long_unsigned_type_node
;
1829 /* If warn_traditional, calculate both the ANSI type and the
1830 traditional type, then see if they disagree.
1831 Otherwise, calculate only the type for the dialect in use. */
1832 if (warn_traditional
|| flag_traditional
)
1834 /* Calculate the traditional type. */
1835 /* Traditionally, any constant is signed;
1836 but if unsigned is specified explicitly, obey that.
1837 Use the smallest size with the right number of bits,
1838 except for one special case with decimal constants. */
1839 if (! spec_long
&& base
!= 10
1840 && int_fits_type_p (yylval
.ttype
, unsigned_type_node
))
1841 traditional_type
= (spec_unsigned
? unsigned_type_node
1842 : integer_type_node
);
1843 /* A decimal constant must be long
1844 if it does not fit in type int.
1845 I think this is independent of whether
1846 the constant is signed. */
1847 else if (! spec_long
&& base
== 10
1848 && int_fits_type_p (yylval
.ttype
, integer_type_node
))
1849 traditional_type
= (spec_unsigned
? unsigned_type_node
1850 : integer_type_node
);
1851 else if (! spec_long_long
)
1852 traditional_type
= (spec_unsigned
? long_unsigned_type_node
1853 : long_integer_type_node
);
1854 else if (int_fits_type_p (yylval
.ttype
,
1856 ? long_long_unsigned_type_node
1857 : long_long_integer_type_node
))
1858 traditional_type
= (spec_unsigned
1859 ? long_long_unsigned_type_node
1860 : long_long_integer_type_node
);
1862 traditional_type
= (spec_unsigned
1863 ? widest_unsigned_literal_type_node
1864 : widest_integer_literal_type_node
);
1866 if (warn_traditional
|| ! flag_traditional
)
1868 /* Calculate the ANSI type. */
1869 if (! spec_long
&& ! spec_unsigned
1870 && int_fits_type_p (yylval
.ttype
, integer_type_node
))
1871 ansi_type
= integer_type_node
;
1872 else if (! spec_long
&& (base
!= 10 || spec_unsigned
)
1873 && int_fits_type_p (yylval
.ttype
, unsigned_type_node
))
1874 ansi_type
= unsigned_type_node
;
1875 else if (! spec_unsigned
&& !spec_long_long
1876 && int_fits_type_p (yylval
.ttype
, long_integer_type_node
))
1877 ansi_type
= long_integer_type_node
;
1878 else if (! spec_long_long
1879 && int_fits_type_p (yylval
.ttype
,
1880 long_unsigned_type_node
))
1881 ansi_type
= long_unsigned_type_node
;
1882 else if (! spec_unsigned
1883 && int_fits_type_p (yylval
.ttype
,
1884 long_long_integer_type_node
))
1885 ansi_type
= long_long_integer_type_node
;
1886 else if (int_fits_type_p (yylval
.ttype
,
1887 long_long_unsigned_type_node
))
1888 ansi_type
= long_long_unsigned_type_node
;
1889 else if (! spec_unsigned
1890 && int_fits_type_p (yylval
.ttype
,
1891 widest_integer_literal_type_node
))
1892 ansi_type
= widest_integer_literal_type_node
;
1894 ansi_type
= widest_unsigned_literal_type_node
;
1897 type
= flag_traditional
? traditional_type
: ansi_type
;
1899 /* We assume that constants specified in a non-decimal
1900 base are bit patterns, and that the programmer really
1901 meant what they wrote. */
1902 if (warn_traditional
&& base
== 10
1903 && traditional_type
!= ansi_type
)
1905 if (TYPE_PRECISION (traditional_type
)
1906 != TYPE_PRECISION (ansi_type
))
1907 warning ("width of integer constant changes with -traditional");
1908 else if (TREE_UNSIGNED (traditional_type
)
1909 != TREE_UNSIGNED (ansi_type
))
1910 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1912 warning ("width of integer constant may change on other systems with -traditional");
1915 if (pedantic
&& !flag_traditional
&& !spec_long_long
&& !warn
1916 && (TYPE_PRECISION (long_integer_type_node
)
1917 < TYPE_PRECISION (type
)))
1920 pedwarn ("integer constant larger than the maximum value of an unsigned long int");
1923 if (base
== 10 && ! spec_unsigned
&& TREE_UNSIGNED (type
))
1924 warning ("decimal constant is so large that it is unsigned");
1928 if (TYPE_PRECISION (type
)
1929 <= TYPE_PRECISION (integer_type_node
))
1931 = build_complex (NULL_TREE
, integer_zero_node
,
1932 convert (integer_type_node
,
1935 error ("complex integer constant is too wide for `complex int'");
1937 else if (flag_traditional
&& !int_fits_type_p (yylval
.ttype
, type
))
1938 /* The traditional constant 0x80000000 is signed
1939 but doesn't fit in the range of int.
1940 This will change it to -0x80000000, which does fit. */
1942 TREE_TYPE (yylval
.ttype
) = unsigned_type (type
);
1943 yylval
.ttype
= convert (type
, yylval
.ttype
);
1944 TREE_OVERFLOW (yylval
.ttype
)
1945 = TREE_CONSTANT_OVERFLOW (yylval
.ttype
) = 0;
1948 TREE_TYPE (yylval
.ttype
) = type
;
1951 /* If it's still an integer (not a complex), and it doesn't
1952 fit in the type we choose for it, then pedwarn. */
1955 && TREE_CODE (TREE_TYPE (yylval
.ttype
)) == INTEGER_TYPE
1956 && ! int_fits_type_p (yylval
.ttype
, TREE_TYPE (yylval
.ttype
)))
1957 pedwarn ("integer constant is larger than the maximum value for its type");
1963 if (ISALNUM (c
) || c
== '.' || c
== '_' || c
== '$'
1964 || (!flag_traditional
&& (c
== '-' || c
== '+')
1965 && (p
[-1] == 'e' || p
[-1] == 'E')))
1966 error ("missing white space after number `%s'", token_buffer
);
1968 value
= CONSTANT
; break;
1974 register int result
= 0;
1975 register int num_chars
= 0;
1977 unsigned width
= TYPE_PRECISION (char_type_node
);
1979 #ifdef MULTIBYTE_CHARS
1980 int longest_char
= local_mb_cur_max ();
1981 (void) local_mbtowc (NULL_PTR
, NULL_PTR
, 0);
1984 max_chars
= TYPE_PRECISION (integer_type_node
) / width
;
1986 width
= WCHAR_TYPE_SIZE
;
1993 if (c
== '\'' || c
== EOF
)
2000 c
= readescape (&ignore
);
2003 if (width
< HOST_BITS_PER_INT
2004 && (unsigned) c
>= ((unsigned)1 << width
))
2005 pedwarn ("escape sequence out of range for character");
2006 #ifdef MAP_CHARACTER
2008 c
= MAP_CHARACTER (c
);
2014 pedwarn ("ANSI C forbids newline in character constant");
2019 #ifdef MULTIBYTE_CHARS
2023 for (i
= 1; i
<= longest_char
; ++i
)
2025 if (i
> maxtoken
- 4)
2026 extend_token_buffer (token_buffer
);
2028 token_buffer
[i
] = c
;
2029 char_len
= local_mbtowc (& wc
,
2038 /* mbtowc sometimes needs an extra char before accepting */
2043 /* Merge character into result; ignore excess chars. */
2044 for (i
= 1; i
<= char_len
; ++i
)
2048 if (width
< HOST_BITS_PER_INT
)
2049 result
= (result
<< width
)
2051 & ((1 << width
) - 1));
2053 result
= token_buffer
[i
];
2055 num_chars
+= char_len
;
2064 warning ("Ignoring invalid multibyte character");
2065 /* Replace all but the first byte. */
2066 for (--i
; i
> 1; --i
)
2067 token_put_back (token_buffer
[i
]);
2068 wc
= token_buffer
[1];
2070 #ifdef MAP_CHARACTER
2071 c
= MAP_CHARACTER (wc
);
2076 #else /* ! MULTIBYTE_CHARS */
2077 #ifdef MAP_CHARACTER
2078 c
= MAP_CHARACTER (c
);
2080 #endif /* ! MULTIBYTE_CHARS */
2085 if (chars_seen
== 1) /* only keep the first one */
2090 /* Merge character into result; ignore excess chars. */
2091 num_chars
+= (width
/ TYPE_PRECISION (char_type_node
));
2092 if (num_chars
< max_chars
+ 1)
2094 if (width
< HOST_BITS_PER_INT
)
2095 result
= (result
<< width
) | (c
& ((1 << width
) - 1));
2102 error ("malformed character constant");
2103 else if (chars_seen
== 0)
2104 error ("empty character constant");
2105 else if (num_chars
> max_chars
)
2107 num_chars
= max_chars
;
2108 error ("character constant too long");
2110 else if (chars_seen
!= 1 && ! flag_traditional
&& warn_multichar
)
2111 warning ("multi-character character constant");
2113 /* If char type is signed, sign-extend the constant. */
2116 int num_bits
= num_chars
* width
;
2118 /* We already got an error; avoid invalid shift. */
2119 yylval
.ttype
= build_int_2 (0, 0);
2120 else if (TREE_UNSIGNED (char_type_node
)
2121 || ((result
>> (num_bits
- 1)) & 1) == 0)
2123 = build_int_2 (result
& (~(unsigned HOST_WIDE_INT
) 0
2124 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
2128 = build_int_2 (result
| ~(~(unsigned HOST_WIDE_INT
) 0
2129 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
2131 TREE_TYPE (yylval
.ttype
) = integer_type_node
;
2135 yylval
.ttype
= build_int_2 (result
, 0);
2136 TREE_TYPE (yylval
.ttype
) = wchar_type_node
;
2146 unsigned width
= wide_flag
? WCHAR_TYPE_SIZE
2147 : TYPE_PRECISION (char_type_node
);
2148 #ifdef MULTIBYTE_CHARS
2149 int longest_char
= local_mb_cur_max ();
2150 (void) local_mbtowc (NULL_PTR
, NULL_PTR
, 0);
2153 p
= token_buffer
+ 1;
2155 while (c
!= '"' && c
!= EOF
)
2157 /* ignore_escape_flag is set for reading the filename in #line. */
2158 if (!ignore_escape_flag
&& c
== '\\')
2161 c
= readescape (&ignore
);
2164 if (width
< HOST_BITS_PER_INT
2165 && (unsigned) c
>= ((unsigned)1 << width
))
2166 pedwarn ("escape sequence out of range for character");
2171 pedwarn ("ANSI C forbids newline in string constant");
2176 #ifdef MULTIBYTE_CHARS
2180 for (i
= 0; i
< longest_char
; ++i
)
2182 if (p
+ i
>= token_buffer
+ maxtoken
)
2183 p
= extend_token_buffer (p
);
2186 char_len
= local_mbtowc (& wc
, p
, i
+ 1);
2193 warning ("Ignoring invalid multibyte character");
2194 /* Replace all except the first byte. */
2196 for (--i
; i
> 0; --i
)
2197 token_put_back (p
[i
]);
2200 /* mbtowc sometimes needs an extra char before accepting */
2210 #endif /* MULTIBYTE_CHARS */
2213 /* Add this single character into the buffer either as a wchar_t
2214 or as a single byte. */
2217 unsigned width
= TYPE_PRECISION (char_type_node
);
2218 unsigned bytemask
= (1 << width
) - 1;
2221 if (p
+ WCHAR_BYTES
> token_buffer
+ maxtoken
)
2222 p
= extend_token_buffer (p
);
2224 for (byte
= 0; byte
< WCHAR_BYTES
; ++byte
)
2227 if (byte
>= (int) sizeof (c
))
2230 value
= (c
>> (byte
* width
)) & bytemask
;
2231 if (BYTES_BIG_ENDIAN
)
2232 p
[WCHAR_BYTES
- byte
- 1] = value
;
2240 if (p
>= token_buffer
+ maxtoken
)
2241 p
= extend_token_buffer (p
);
2249 /* Terminate the string value, either with a single byte zero
2250 or with a wide zero. */
2253 if (p
+ WCHAR_BYTES
> token_buffer
+ maxtoken
)
2254 p
= extend_token_buffer (p
);
2255 bzero (p
, WCHAR_BYTES
);
2260 if (p
>= token_buffer
+ maxtoken
)
2261 p
= extend_token_buffer (p
);
2266 error ("Unterminated string constant");
2268 /* We have read the entire constant.
2269 Construct a STRING_CST for the result. */
2273 yylval
.ttype
= build_string (p
- (token_buffer
+ 1),
2275 TREE_TYPE (yylval
.ttype
) = wchar_array_type_node
;
2280 /* Return an Objective-C @"..." constant string object. */
2281 yylval
.ttype
= build_objc_string (p
- (token_buffer
+ 1),
2283 TREE_TYPE (yylval
.ttype
) = char_array_type_node
;
2284 value
= OBJC_STRING
;
2288 yylval
.ttype
= build_string (p
- (token_buffer
+ 1),
2290 TREE_TYPE (yylval
.ttype
) = char_array_type_node
;
2318 yylval
.code
= PLUS_EXPR
; break;
2320 yylval
.code
= MINUS_EXPR
; break;
2322 yylval
.code
= BIT_AND_EXPR
; break;
2324 yylval
.code
= BIT_IOR_EXPR
; break;
2326 yylval
.code
= MULT_EXPR
; break;
2328 yylval
.code
= TRUNC_DIV_EXPR
; break;
2330 yylval
.code
= TRUNC_MOD_EXPR
; break;
2332 yylval
.code
= BIT_XOR_EXPR
; break;
2334 yylval
.code
= LSHIFT_EXPR
; break;
2336 yylval
.code
= RSHIFT_EXPR
; break;
2338 yylval
.code
= LT_EXPR
; break;
2340 yylval
.code
= GT_EXPR
; break;
2343 token_buffer
[1] = c1
= token_getch();
2344 token_buffer
[2] = 0;
2351 value
= ARITHCOMPARE
; yylval
.code
= LE_EXPR
; goto done
;
2353 value
= ARITHCOMPARE
; yylval
.code
= GE_EXPR
; goto done
;
2355 value
= EQCOMPARE
; yylval
.code
= NE_EXPR
; goto done
;
2357 value
= EQCOMPARE
; yylval
.code
= EQ_EXPR
; goto done
;
2359 value
= ASSIGN
; goto done
;
2365 value
= PLUSPLUS
; goto done
;
2367 value
= MINUSMINUS
; goto done
;
2369 value
= ANDAND
; goto done
;
2371 value
= OROR
; goto done
;
2384 { value
= POINTSAT
; goto done
; }
2389 if (c1
== '>' && flag_digraphs
)
2390 { value
= ']'; goto done
; }
2396 { value
= '{'; indent_level
++; goto done
; }
2398 { value
= '['; goto done
; }
2402 if (c1
== '>' && flag_digraphs
)
2403 { value
= '}'; indent_level
--; goto done
; }
2407 token_put_back (c1
);
2408 token_buffer
[1] = 0;
2410 if ((c
== '<') || (c
== '>'))
2411 value
= ARITHCOMPARE
;
2417 /* Don't make yyparse think this is eof. */
2436 /* yylloc.last_line = lineno; */
2441 /* Sets the value of the 'yydebug' variable to VALUE.
2442 This is a function so we don't have to have YYDEBUG defined
2443 in order to build the compiler. */
2452 warning ("YYDEBUG not defined.");