1 /* Lexical analyzer for C and Objective C.
2 Copyright (C) 1987, 88, 89, 92, 94-97, 1998 Free Software Foundation, Inc.
4 This file is part of GNU CC.
6 GNU CC is free software; you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation; either version 2, or (at your option)
11 GNU CC is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with GNU CC; see the file COPYING. If not, write to
18 the Free Software Foundation, 59 Temple Place - Suite 330,
19 Boston, MA 02111-1307, USA. */
36 /* MULTIBYTE_CHARS support only works for native compilers.
37 ??? Ideally what we want is to model widechar support after
38 the current floating point support. */
40 #undef MULTIBYTE_CHARS
43 #ifdef MULTIBYTE_CHARS
49 extern cpp_reader parse_in
;
50 extern cpp_options parse_options
;
52 /* Stream for reading from the input file. */
56 extern void yyprint
PROTO((FILE *, int, YYSTYPE
));
58 /* The elements of `ridpointers' are identifier nodes
59 for the reserved type names and storage classes.
60 It is indexed by a RID_... value. */
61 tree ridpointers
[(int) RID_MAX
];
63 /* Cause the `yydebug' variable to be defined. */
67 extern unsigned char *yy_cur
, *yy_lim
;
69 extern int yy_get_token ();
71 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
72 #define UNGETC(c) ((c), yy_cur--)
74 #define GETC() getc (finput)
75 #define UNGETC(c) ungetc (c, finput)
78 /* the declaration found for the last IDENTIFIER token read in.
79 yylex must look this up to detect typedefs, which get token type TYPENAME,
80 so it is left around in case the identifier is not a typedef but is
81 used in a context which makes it a reference to a variable. */
84 /* Nonzero enables objc features. */
90 /* File used for outputting assembler code. */
91 extern FILE *asm_out_file
;
93 #ifndef WCHAR_TYPE_SIZE
95 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
97 #define WCHAR_TYPE_SIZE BITS_PER_WORD
101 /* Number of bytes in a wide character. */
102 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
104 static int maxtoken
; /* Current nominal length of token buffer. */
105 char *token_buffer
; /* Pointer to token buffer.
106 Actual allocated length is maxtoken + 2.
107 This is not static because objc-parse.y uses it. */
109 static int indent_level
= 0; /* Number of { minus number of }. */
111 /* Nonzero if end-of-file has been seen on input. */
112 static int end_of_file
;
115 /* Buffered-back input character; faster than using ungetc. */
116 static int nextchar
= -1;
119 #ifdef HANDLE_SYSV_PRAGMA
120 static int handle_sysv_pragma
PROTO((int));
121 #endif /* HANDLE_SYSV_PRAGMA */
122 static int whitespace_cr
PROTO((int));
123 static int skip_white_space
PROTO((int));
124 static int skip_white_space_on_line
PROTO((void));
125 static char *extend_token_buffer
PROTO((char *));
126 static int readescape
PROTO((int *));
128 /* Do not insert generated code into the source, instead, include it.
129 This allows us to build gcc automatically even for targets that
130 need to add or modify the reserved keyword lists. */
133 /* Return something to represent absolute declarators containing a *.
134 TARGET is the absolute declarator that the * contains.
135 TYPE_QUALS is a list of modifiers such as const or volatile
136 to apply to the pointer type, represented as identifiers.
138 We return an INDIRECT_REF whose "contents" are TARGET
139 and whose type is the modifier list. */
142 make_pointer_declarator (type_quals
, target
)
143 tree type_quals
, target
;
145 return build1 (INDIRECT_REF
, type_quals
, target
);
149 forget_protocol_qualifiers ()
151 int i
, n
= sizeof wordlist
/ sizeof (struct resword
);
153 for (i
= 0; i
< n
; i
++)
154 if ((int) wordlist
[i
].rid
>= (int) RID_IN
155 && (int) wordlist
[i
].rid
<= (int) RID_ONEWAY
)
156 wordlist
[i
].name
= "";
160 remember_protocol_qualifiers ()
162 int i
, n
= sizeof wordlist
/ sizeof (struct resword
);
164 for (i
= 0; i
< n
; i
++)
165 if (wordlist
[i
].rid
== RID_IN
)
166 wordlist
[i
].name
= "in";
167 else if (wordlist
[i
].rid
== RID_OUT
)
168 wordlist
[i
].name
= "out";
169 else if (wordlist
[i
].rid
== RID_INOUT
)
170 wordlist
[i
].name
= "inout";
171 else if (wordlist
[i
].rid
== RID_BYCOPY
)
172 wordlist
[i
].name
= "bycopy";
173 else if (wordlist
[i
].rid
== RID_ONEWAY
)
174 wordlist
[i
].name
= "oneway";
178 init_parse (filename
)
182 /* Open input file. */
183 if (filename
== 0 || !strcmp (filename
, "-"))
189 finput
= fopen (filename
, "r");
191 pfatal_with_name (filename
);
193 #ifdef IO_BUFFER_SIZE
194 setvbuf (finput
, (char *) xmalloc (IO_BUFFER_SIZE
), _IOFBF
, IO_BUFFER_SIZE
);
196 #endif /* !USE_CPPLIB */
204 parse_in
.show_column
= 1;
205 if (! cpp_start_read (&parse_in
, filename
))
216 cpp_finish (&parse_in
);
225 /* Make identifier nodes long enough for the language-specific slots. */
226 set_identifier_size (sizeof (struct lang_identifier
));
228 /* Start it at 0, because check_newline is called at the very beginning
229 and will increment it to 1. */
232 #ifdef MULTIBYTE_CHARS
233 /* Change to the native locale for multibyte conversions. */
234 setlocale (LC_CTYPE
, "");
238 token_buffer
= (char *) xmalloc (maxtoken
+ 2);
240 ridpointers
[(int) RID_INT
] = get_identifier ("int");
241 ridpointers
[(int) RID_CHAR
] = get_identifier ("char");
242 ridpointers
[(int) RID_VOID
] = get_identifier ("void");
243 ridpointers
[(int) RID_FLOAT
] = get_identifier ("float");
244 ridpointers
[(int) RID_DOUBLE
] = get_identifier ("double");
245 ridpointers
[(int) RID_SHORT
] = get_identifier ("short");
246 ridpointers
[(int) RID_LONG
] = get_identifier ("long");
247 ridpointers
[(int) RID_UNSIGNED
] = get_identifier ("unsigned");
248 ridpointers
[(int) RID_SIGNED
] = get_identifier ("signed");
249 ridpointers
[(int) RID_INLINE
] = get_identifier ("inline");
250 ridpointers
[(int) RID_CONST
] = get_identifier ("const");
251 ridpointers
[(int) RID_VOLATILE
] = get_identifier ("volatile");
252 ridpointers
[(int) RID_AUTO
] = get_identifier ("auto");
253 ridpointers
[(int) RID_STATIC
] = get_identifier ("static");
254 ridpointers
[(int) RID_EXTERN
] = get_identifier ("extern");
255 ridpointers
[(int) RID_TYPEDEF
] = get_identifier ("typedef");
256 ridpointers
[(int) RID_REGISTER
] = get_identifier ("register");
257 ridpointers
[(int) RID_ITERATOR
] = get_identifier ("iterator");
258 ridpointers
[(int) RID_COMPLEX
] = get_identifier ("complex");
259 ridpointers
[(int) RID_ID
] = get_identifier ("id");
260 ridpointers
[(int) RID_IN
] = get_identifier ("in");
261 ridpointers
[(int) RID_OUT
] = get_identifier ("out");
262 ridpointers
[(int) RID_INOUT
] = get_identifier ("inout");
263 ridpointers
[(int) RID_BYCOPY
] = get_identifier ("bycopy");
264 ridpointers
[(int) RID_ONEWAY
] = get_identifier ("oneway");
265 forget_protocol_qualifiers();
267 /* Some options inhibit certain reserved words.
268 Clear those words out of the hash table so they won't be recognized. */
269 #define UNSET_RESERVED_WORD(STRING) \
270 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
271 if (s) s->name = ""; } while (0)
273 if (! doing_objc_thang
)
274 UNSET_RESERVED_WORD ("id");
276 if (flag_traditional
)
278 UNSET_RESERVED_WORD ("const");
279 UNSET_RESERVED_WORD ("volatile");
280 UNSET_RESERVED_WORD ("typeof");
281 UNSET_RESERVED_WORD ("signed");
282 UNSET_RESERVED_WORD ("inline");
283 UNSET_RESERVED_WORD ("iterator");
284 UNSET_RESERVED_WORD ("complex");
288 UNSET_RESERVED_WORD ("asm");
289 UNSET_RESERVED_WORD ("typeof");
290 UNSET_RESERVED_WORD ("inline");
291 UNSET_RESERVED_WORD ("iterator");
292 UNSET_RESERVED_WORD ("complex");
297 reinit_parse_for_function ()
301 /* Function used when yydebug is set, to print a token in more detail. */
304 yyprint (file
, yychar
, yylval
)
316 if (IDENTIFIER_POINTER (t
))
317 fprintf (file
, " `%s'", IDENTIFIER_POINTER (t
));
322 if (TREE_CODE (t
) == INTEGER_CST
)
324 #if HOST_BITS_PER_WIDE_INT == 64
325 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
328 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
335 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
341 TREE_INT_CST_HIGH (t
), TREE_INT_CST_LOW (t
));
346 /* Iff C is a carriage return, warn about it - if appropriate -
347 and return nonzero. */
352 static int newline_warning
= 0;
356 /* ANSI C says the effects of a carriage return in a source file
358 if (pedantic
&& !newline_warning
)
360 warning ("carriage return in source file");
361 warning ("(we only warn about the first carriage return)");
369 /* If C is not whitespace, return C.
370 Otherwise skip whitespace and return first nonwhite char read. */
380 /* We don't recognize comments here, because
381 cpp output can include / and * consecutively as operators.
382 Also, there's no need, since cpp removes all comments. */
385 c
= check_newline ();
406 error ("stray '\\' in program");
416 /* Skips all of the white space at the current location in the input file.
417 Must use and reset nextchar if it has the next character. */
420 position_after_white_space ()
426 c
= nextchar
, nextchar
= -1;
431 UNGETC (skip_white_space (c
));
434 /* Like skip_white_space, but don't advance beyond the end of line.
435 Moreover, we don't get passed a character to start with. */
437 skip_white_space_on_line ()
466 /* Make the token buffer longer, preserving the data in it.
467 P should point to just beyond the last valid character in the old buffer.
468 The value we return is a pointer to the new buffer
469 at a place corresponding to P. */
472 extend_token_buffer (p
)
475 int offset
= p
- token_buffer
;
477 maxtoken
= maxtoken
* 2 + 10;
478 token_buffer
= (char *) xrealloc (token_buffer
, maxtoken
+ 2);
480 return token_buffer
+ offset
;
483 /* At the beginning of a line, increment the line number
484 and process any #-directive on this line.
485 If the line is a #-directive, read the entire line and return a newline.
486 Otherwise, return the line's first non-whitespace character. */
496 /* Read first nonwhite char on the line. */
499 while (c
== ' ' || c
== '\t')
504 /* If not #, return it so caller will use it. */
508 /* Read first nonwhite char after the `#'. */
511 while (c
== ' ' || c
== '\t')
514 /* If a letter follows, then if the word here is `line', skip
515 it and ignore it; otherwise, ignore the line, with an error
516 if the word isn't `pragma', `ident', `define', or `undef'. */
518 if ((c
>= 'a' && c
<= 'z') || (c
>= 'A' && c
<= 'Z'))
527 && ((c
= GETC()) == ' ' || c
== '\t' || c
== '\n'
528 || whitespace_cr (c
) ))
530 while (c
== ' ' || c
== '\t' || whitespace_cr (c
))
534 #ifdef HANDLE_SYSV_PRAGMA
537 if (token
!= IDENTIFIER
)
539 return handle_sysv_pragma (token
);
540 #else /* !HANDLE_SYSV_PRAGMA */
545 if (token
!= IDENTIFIER
)
548 c
= nextchar
, nextchar
= -1;
552 if (HANDLE_PRAGMA (finput
, yylval
.ttype
))
558 ??? do not know what to
do ???;
559 #endif /* !USE_CPPLIB */
560 #endif /* HANDLE_PRAGMA */
561 #endif /* !HANDLE_SYSV_PRAGMA */
573 && ((c
= GETC()) == ' ' || c
== '\t' || c
== '\n'))
576 debug_define (lineno
, GET_DIRECTIVE_LINE ());
586 && ((c
= GETC()) == ' ' || c
== '\t' || c
== '\n'))
589 debug_undef (lineno
, GET_DIRECTIVE_LINE ());
598 && ((c
= GETC()) == ' ' || c
== '\t'))
607 && ((c
= GETC()) == ' ' || c
== '\t'))
609 /* #ident. The pedantic warning is now in cccp.c. */
611 /* Here we have just seen `#ident '.
612 A string constant should follow. */
614 c
= skip_white_space_on_line ();
616 /* If no argument, ignore the line. */
623 || TREE_CODE (yylval
.ttype
) != STRING_CST
)
625 error ("invalid #ident");
631 #ifdef ASM_OUTPUT_IDENT
632 ASM_OUTPUT_IDENT (asm_out_file
, TREE_STRING_POINTER (yylval
.ttype
));
636 /* Skip the rest of this line. */
641 error ("undefined or invalid # directive");
646 /* Here we have either `#line' or `# <nonletter>'.
647 In either case, it should be a line number; a digit should follow. */
649 /* Can't use skip_white_space here, but must handle all whitespace
650 that is not '\n', lest we get a recursion for '\r' '\n' when
653 c
= skip_white_space_on_line ();
655 /* If the # is the only nonwhite char on the line,
656 just ignore it. Check the new newline. */
660 /* Something follows the #; read a token. */
665 if (token
== CONSTANT
666 && TREE_CODE (yylval
.ttype
) == INTEGER_CST
)
668 int old_lineno
= lineno
;
670 /* subtract one, because it is the following line that
671 gets the specified number */
673 int l
= TREE_INT_CST_LOW (yylval
.ttype
) - 1;
675 /* Is this the last nonwhite stuff on the line? */
676 c
= skip_white_space_on_line ();
679 /* No more: store the line number and check following line. */
685 /* More follows: it must be a string constant (filename). */
687 /* Read the string constant. */
690 if (token
!= STRING
|| TREE_CODE (yylval
.ttype
) != STRING_CST
)
692 error ("invalid #line");
697 = (char *) permalloc (TREE_STRING_LENGTH (yylval
.ttype
) + 1);
698 strcpy (input_filename
, TREE_STRING_POINTER (yylval
.ttype
));
701 /* Each change of file name
702 reinitializes whether we are now in a system header. */
703 in_system_header
= 0;
705 if (main_input_filename
== 0)
706 main_input_filename
= input_filename
;
708 /* Is this the last nonwhite stuff on the line? */
709 c
= skip_white_space_on_line ();
712 /* Update the name in the top element of input_file_stack. */
713 if (input_file_stack
)
714 input_file_stack
->name
= input_filename
;
723 /* `1' after file name means entering new file.
724 `2' after file name means just left a file. */
726 if (token
== CONSTANT
727 && TREE_CODE (yylval
.ttype
) == INTEGER_CST
)
729 if (TREE_INT_CST_LOW (yylval
.ttype
) == 1)
731 /* Pushing to a new file. */
733 = (struct file_stack
*) xmalloc (sizeof (struct file_stack
));
734 input_file_stack
->line
= old_lineno
;
735 p
->next
= input_file_stack
;
736 p
->name
= input_filename
;
737 p
->indent_level
= indent_level
;
738 input_file_stack
= p
;
739 input_file_stack_tick
++;
740 debug_start_source_file (input_filename
);
743 else if (TREE_INT_CST_LOW (yylval
.ttype
) == 2)
745 /* Popping out of a file. */
746 if (input_file_stack
->next
)
748 struct file_stack
*p
= input_file_stack
;
749 if (indent_level
!= p
->indent_level
)
751 warning_with_file_and_line
752 (p
->name
, old_lineno
,
753 "This file contains more `%c's than `%c's.",
754 indent_level
> p
->indent_level
? '{' : '}',
755 indent_level
> p
->indent_level
? '}' : '{');
757 input_file_stack
= p
->next
;
759 input_file_stack_tick
++;
760 debug_end_source_file (input_file_stack
->line
);
763 error ("#-lines for entering and leaving files don't match");
769 /* Now that we've pushed or popped the input stack,
770 update the name in the top element. */
771 if (input_file_stack
)
772 input_file_stack
->name
= input_filename
;
774 /* If we have handled a `1' or a `2',
775 see if there is another number to read. */
778 /* Is this the last nonwhite stuff on the line? */
779 c
= skip_white_space_on_line ();
788 /* `3' after file name means this is a system header file. */
790 if (token
== CONSTANT
791 && TREE_CODE (yylval
.ttype
) == INTEGER_CST
792 && TREE_INT_CST_LOW (yylval
.ttype
) == 3)
793 in_system_header
= 1, used_up
= 1;
797 /* Is this the last nonwhite stuff on the line? */
798 c
= skip_white_space_on_line ();
804 warning ("unrecognized text at end of #line");
807 error ("invalid #-line");
809 /* skip the rest of this line. */
812 if (c
!= '\n' && c
!= EOF
&& nextchar
>= 0)
813 c
= nextchar
, nextchar
= -1;
815 while (c
!= '\n' && c
!= EOF
)
820 #ifdef HANDLE_SYSV_PRAGMA
822 /* Handle a #pragma directive.
823 TOKEN is the token we read after `#pragma'. Processes the entire input
824 line and returns a character for the caller to reread: either \n or EOF. */
826 /* This function has to be in this file, in order to get at
830 handle_sysv_pragma (token
)
843 handle_pragma_token (token_buffer
, yylval
.ttype
);
846 handle_pragma_token (token_buffer
, 0);
850 c
= nextchar
, nextchar
= -1;
855 while (c
== ' ' || c
== '\t')
857 if (c
== '\n' || c
== EOF
)
859 handle_pragma_token (0, 0);
867 #endif /* HANDLE_SYSV_PRAGMA */
869 #define ENDFILE -1 /* token that represents end-of-file */
871 /* Read an escape sequence, returning its equivalent as a character,
872 or store 1 in *ignore_ptr if it is backslash-newline. */
875 readescape (ignore_ptr
)
878 register int c
= GETC();
880 register unsigned count
;
881 unsigned firstdig
= 0;
887 if (warn_traditional
)
888 warning ("the meaning of `\\x' varies with -traditional");
890 if (flag_traditional
)
899 if (!(c
>= 'a' && c
<= 'f')
900 && !(c
>= 'A' && c
<= 'F')
901 && !(c
>= '0' && c
<= '9'))
907 if (c
>= 'a' && c
<= 'f')
908 code
+= c
- 'a' + 10;
909 if (c
>= 'A' && c
<= 'F')
910 code
+= c
- 'A' + 10;
911 if (c
>= '0' && c
<= '9')
913 if (code
!= 0 || count
!= 0)
922 error ("\\x used with no following hex digits");
924 /* Digits are all 0's. Ok. */
926 else if ((count
- 1) * 4 >= TYPE_PRECISION (integer_type_node
)
928 && ((1 << (TYPE_PRECISION (integer_type_node
) - (count
- 1) * 4))
930 pedwarn ("hex escape out of range");
933 case '0': case '1': case '2': case '3': case '4':
934 case '5': case '6': case '7':
937 while ((c
<= '7') && (c
>= '0') && (count
++ < 3))
939 code
= (code
* 8) + (c
- '0');
945 case '\\': case '\'': case '"':
954 return TARGET_NEWLINE
;
969 if (warn_traditional
)
970 warning ("the meaning of `\\a' varies with -traditional");
972 if (flag_traditional
)
977 #if 0 /* Vertical tab is present in common usage compilers. */
978 if (flag_traditional
)
986 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c
);
992 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
996 /* `\%' is used to prevent SCCS from getting confused. */
999 pedwarn ("non-ANSI escape sequence `\\%c'", c
);
1002 if (c
>= 040 && c
< 0177)
1003 pedwarn ("unknown escape sequence `\\%c'", c
);
1005 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c
);
1015 strcpy (buf
, string
);
1017 /* We can't print string and character constants well
1018 because the token_buffer contains the result of processing escapes. */
1020 strcat (buf
, " at end of input");
1021 else if (token_buffer
[0] == 0)
1022 strcat (buf
, " at null character");
1023 else if (token_buffer
[0] == '"')
1024 strcat (buf
, " before string constant");
1025 else if (token_buffer
[0] == '\'')
1026 strcat (buf
, " before character constant");
1027 else if (token_buffer
[0] < 040 || (unsigned char) token_buffer
[0] >= 0177)
1028 sprintf (buf
+ strlen (buf
), " before character 0%o",
1029 (unsigned char) token_buffer
[0]);
1031 strcat (buf
, " before `%s'");
1033 error (buf
, token_buffer
);
1043 char long_long_flag
;
1046 struct try_type type_sequence
[] =
1048 { &integer_type_node
, 0, 0, 0},
1049 { &unsigned_type_node
, 1, 0, 0},
1050 { &long_integer_type_node
, 0, 1, 0},
1051 { &long_unsigned_type_node
, 1, 1, 0},
1052 { &long_long_integer_type_node
, 0, 1, 1},
1053 { &long_long_unsigned_type_node
, 1, 1, 1}
1068 c
= nextchar
, nextchar
= -1;
1073 /* Effectively do c = skip_white_space (c)
1074 but do it faster in the usual cases. */
1087 /* Call skip_white_space so we can warn if appropriate. */
1092 c
= skip_white_space (c
);
1094 goto found_nonwhite
;
1098 token_buffer
[0] = c
;
1099 token_buffer
[1] = 0;
1101 /* yylloc.first_line = lineno; */
1107 token_buffer
[0] = 0;
1112 /* Capital L may start a wide-string or wide-character constant. */
1114 register int c
= GETC();
1123 goto string_constant
;
1130 if (!doing_objc_thang
)
1137 /* '@' may start a constant string object. */
1138 register int c
= GETC ();
1142 goto string_constant
;
1145 /* Fall through to treat '@' as the start of an identifier. */
1148 case 'A': case 'B': case 'C': case 'D': case 'E':
1149 case 'F': case 'G': case 'H': case 'I': case 'J':
1150 case 'K': case 'M': case 'N': case 'O':
1151 case 'P': case 'Q': case 'R': case 'S': case 'T':
1152 case 'U': case 'V': case 'W': case 'X': case 'Y':
1154 case 'a': case 'b': case 'c': case 'd': case 'e':
1155 case 'f': case 'g': case 'h': case 'i': case 'j':
1156 case 'k': case 'l': case 'm': case 'n': case 'o':
1157 case 'p': case 'q': case 'r': case 's': case 't':
1158 case 'u': case 'v': case 'w': case 'x': case 'y':
1164 while (ISALNUM (c
) || c
== '_' || c
== '$' || c
== '@')
1166 /* Make sure this char really belongs in an identifier. */
1167 if (c
== '@' && ! doing_objc_thang
)
1171 if (! dollars_in_ident
)
1172 error ("`$' in identifier");
1174 pedwarn ("`$' in identifier");
1177 if (p
>= token_buffer
+ maxtoken
)
1178 p
= extend_token_buffer (p
);
1194 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1197 register struct resword
*ptr
;
1199 if ((ptr
= is_reserved_word (token_buffer
, p
- token_buffer
)))
1202 yylval
.ttype
= ridpointers
[(int) ptr
->rid
];
1203 value
= (int) ptr
->token
;
1205 /* Only return OBJECTNAME if it is a typedef. */
1206 if (doing_objc_thang
&& value
== OBJECTNAME
)
1208 lastiddecl
= lookup_name(yylval
.ttype
);
1210 if (lastiddecl
== NULL_TREE
1211 || TREE_CODE (lastiddecl
) != TYPE_DECL
)
1215 /* Even if we decided to recognize asm, still perhaps warn. */
1217 && (value
== ASM_KEYWORD
|| value
== TYPEOF
1218 || ptr
->rid
== RID_INLINE
)
1219 && token_buffer
[0] != '_')
1220 pedwarn ("ANSI does not permit the keyword `%s'",
1225 /* If we did not find a keyword, look for an identifier
1228 if (value
== IDENTIFIER
)
1230 if (token_buffer
[0] == '@')
1231 error("invalid identifier `%s'", token_buffer
);
1233 yylval
.ttype
= get_identifier (token_buffer
);
1234 lastiddecl
= lookup_name (yylval
.ttype
);
1236 if (lastiddecl
!= 0 && TREE_CODE (lastiddecl
) == TYPE_DECL
)
1238 /* A user-invisible read-only initialized variable
1239 should be replaced by its value.
1240 We handle only strings since that's the only case used in C. */
1241 else if (lastiddecl
!= 0 && TREE_CODE (lastiddecl
) == VAR_DECL
1242 && DECL_IGNORED_P (lastiddecl
)
1243 && TREE_READONLY (lastiddecl
)
1244 && DECL_INITIAL (lastiddecl
) != 0
1245 && TREE_CODE (DECL_INITIAL (lastiddecl
)) == STRING_CST
)
1247 tree stringval
= DECL_INITIAL (lastiddecl
);
1249 /* Copy the string value so that we won't clobber anything
1250 if we put something in the TREE_CHAIN of this one. */
1251 yylval
.ttype
= build_string (TREE_STRING_LENGTH (stringval
),
1252 TREE_STRING_POINTER (stringval
));
1255 else if (doing_objc_thang
)
1257 tree objc_interface_decl
= is_class_name (yylval
.ttype
);
1259 if (objc_interface_decl
)
1262 yylval
.ttype
= objc_interface_decl
;
1272 /* Check first for common special case: single-digit 0 or 1. */
1275 UNGETC (next_c
); /* Always undo this lookahead. */
1276 if (!ISALNUM (next_c
) && next_c
!= '.')
1278 token_buffer
[0] = (char)c
, token_buffer
[1] = '\0';
1279 yylval
.ttype
= (c
== '0') ? integer_zero_node
: integer_one_node
;
1285 case '2': case '3': case '4':
1286 case '5': case '6': case '7': case '8': case '9':
1291 int largest_digit
= 0;
1293 /* for multi-precision arithmetic,
1294 we actually store only HOST_BITS_PER_CHAR bits in each part.
1295 The number of parts is chosen so as to be sufficient to hold
1296 the enough bits to fit into the two HOST_WIDE_INTs that contain
1297 the integer value (this is always at least as many bits as are
1298 in a target `long long' value, but may be wider). */
1299 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1300 int parts
[TOTAL_PARTS
];
1303 enum anon1
{ NOT_FLOAT
, AFTER_POINT
, TOO_MANY_POINTS
} floatflag
1306 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1314 *p
++ = (c
= GETC());
1315 if ((c
== 'x') || (c
== 'X'))
1318 *p
++ = (c
= GETC());
1320 /* Leading 0 forces octal unless the 0 is the only digit. */
1321 else if (c
>= '0' && c
<= '9')
1330 /* Read all the digits-and-decimal-points. */
1333 || (ISALNUM (c
) && c
!= 'l' && c
!= 'L'
1334 && c
!= 'u' && c
!= 'U'
1335 && c
!= 'i' && c
!= 'I' && c
!= 'j' && c
!= 'J'
1336 && (floatflag
== NOT_FLOAT
|| ((c
!= 'f') && (c
!= 'F')))))
1341 error ("floating constant may not be in radix 16");
1342 if (floatflag
== TOO_MANY_POINTS
)
1343 /* We have already emitted an error. Don't need another. */
1345 else if (floatflag
== AFTER_POINT
)
1347 error ("malformed floating constant");
1348 floatflag
= TOO_MANY_POINTS
;
1349 /* Avoid another error from atof by forcing all characters
1350 from here on to be ignored. */
1354 floatflag
= AFTER_POINT
;
1358 /* Accept '.' as the start of a floating-point number
1359 only when it is followed by a digit.
1360 Otherwise, unread the following non-digit
1361 and use the '.' as a structural token. */
1362 if (p
== token_buffer
+ 2 && !ISDIGIT (c
))
1373 error ("parse error at `..'");
1376 token_buffer
[1] = 0;
1383 /* It is not a decimal point.
1384 It should be a digit (perhaps a hex digit). */
1390 else if (base
<= 10)
1392 if (c
== 'e' || c
== 'E')
1395 floatflag
= AFTER_POINT
;
1396 break; /* start of exponent */
1398 error ("nondigits in number and not hexadecimal");
1409 if (c
>= largest_digit
)
1413 for (count
= 0; count
< TOTAL_PARTS
; count
++)
1415 parts
[count
] *= base
;
1419 += (parts
[count
-1] >> HOST_BITS_PER_CHAR
);
1421 &= (1 << HOST_BITS_PER_CHAR
) - 1;
1427 /* If the extra highest-order part ever gets anything in it,
1428 the number is certainly too big. */
1429 if (parts
[TOTAL_PARTS
- 1] != 0)
1432 if (p
>= token_buffer
+ maxtoken
- 3)
1433 p
= extend_token_buffer (p
);
1434 *p
++ = (c
= GETC());
1439 error ("numeric constant with no digits");
1441 if (largest_digit
>= base
)
1442 error ("numeric constant contains digits beyond the radix");
1444 /* Remove terminating char from the token buffer and delimit the string */
1447 if (floatflag
!= NOT_FLOAT
)
1449 tree type
= double_type_node
;
1451 int conversion_errno
= 0;
1452 REAL_VALUE_TYPE value
;
1455 /* Read explicit exponent if any, and put it in tokenbuf. */
1457 if ((c
== 'e') || (c
== 'E'))
1459 if (p
>= token_buffer
+ maxtoken
- 3)
1460 p
= extend_token_buffer (p
);
1463 if ((c
== '+') || (c
== '-'))
1469 error ("floating constant exponent has no digits");
1472 if (p
>= token_buffer
+ maxtoken
- 3)
1473 p
= extend_token_buffer (p
);
1481 /* Convert string to a double, checking for overflow. */
1482 if (setjmp (handler
))
1484 error ("floating constant out of range");
1489 int fflag
= 0, lflag
= 0;
1490 /* Copy token_buffer now, while it has just the number
1491 and not the suffixes; once we add `f' or `i',
1492 REAL_VALUE_ATOF may not work any more. */
1493 char *copy
= (char *) alloca (p
- token_buffer
+ 1);
1494 bcopy (token_buffer
, copy
, p
- token_buffer
+ 1);
1496 set_float_handler (handler
);
1502 /* Read the suffixes to choose a data type. */
1507 error ("more than one `f' in numeric constant");
1513 error ("more than one `l' in numeric constant");
1519 error ("more than one `i' or `j' in numeric constant");
1521 pedwarn ("ANSI C forbids imaginary numeric constants");
1532 if (p
>= token_buffer
+ maxtoken
- 3)
1533 p
= extend_token_buffer (p
);
1539 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1540 tells the desired precision of the binary result
1541 of decimal-to-binary conversion. */
1546 error ("both `f' and `l' in floating constant");
1548 type
= float_type_node
;
1550 value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (type
));
1551 conversion_errno
= errno
;
1552 /* A diagnostic is required here by some ANSI C testsuites.
1553 This is not pedwarn, become some people don't want
1554 an error for this. */
1555 if (REAL_VALUE_ISINF (value
) && pedantic
)
1556 warning ("floating point number exceeds range of `float'");
1560 type
= long_double_type_node
;
1562 value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (type
));
1563 conversion_errno
= errno
;
1564 if (REAL_VALUE_ISINF (value
) && pedantic
)
1565 warning ("floating point number exceeds range of `long double'");
1570 value
= REAL_VALUE_ATOF (copy
, TYPE_MODE (type
));
1571 conversion_errno
= errno
;
1572 if (REAL_VALUE_ISINF (value
) && pedantic
)
1573 warning ("floating point number exceeds range of `double'");
1576 set_float_handler (NULL_PTR
);
1579 /* ERANGE is also reported for underflow,
1580 so test the value to distinguish overflow from that. */
1581 if (conversion_errno
== ERANGE
&& !flag_traditional
&& pedantic
1582 && (REAL_VALUES_LESS (dconst1
, value
)
1583 || REAL_VALUES_LESS (value
, dconstm1
)))
1584 warning ("floating point number exceeds range of `double'");
1587 /* If the result is not a number, assume it must have been
1588 due to some error message above, so silently convert
1590 if (REAL_VALUE_ISNAN (value
))
1593 /* Create a node with determined type and value. */
1595 yylval
.ttype
= build_complex (NULL_TREE
,
1596 convert (type
, integer_zero_node
),
1597 build_real (type
, value
));
1599 yylval
.ttype
= build_real (type
, value
);
1603 tree traditional_type
, ansi_type
, type
;
1604 HOST_WIDE_INT high
, low
;
1605 int spec_unsigned
= 0;
1607 int spec_long_long
= 0;
1611 traditional_type
= ansi_type
= type
= NULL_TREE
;
1614 if (c
== 'u' || c
== 'U')
1617 error ("two `u's in integer constant");
1620 else if (c
== 'l' || c
== 'L')
1625 error ("three `l's in integer constant");
1627 pedwarn ("ANSI C forbids long long integer constants");
1632 else if (c
== 'i' || c
== 'j' || c
== 'I' || c
== 'J')
1635 error ("more than one `i' or `j' in numeric constant");
1637 pedwarn ("ANSI C forbids imaginary numeric constants");
1642 if (p
>= token_buffer
+ maxtoken
- 3)
1643 p
= extend_token_buffer (p
);
1648 /* If the constant won't fit in an unsigned long long,
1649 then warn that the constant is out of range. */
1651 /* ??? This assumes that long long and long integer types are
1652 a multiple of 8 bits. This better than the original code
1653 though which assumed that long was exactly 32 bits and long
1654 long was exactly 64 bits. */
1656 bytes
= TYPE_PRECISION (long_long_integer_type_node
) / 8;
1659 for (i
= bytes
; i
< TOTAL_PARTS
; i
++)
1663 pedwarn ("integer constant out of range");
1665 /* This is simplified by the fact that our constant
1666 is always positive. */
1670 for (i
= 0; i
< HOST_BITS_PER_WIDE_INT
/ HOST_BITS_PER_CHAR
; i
++)
1672 high
|= ((HOST_WIDE_INT
) parts
[i
+ (HOST_BITS_PER_WIDE_INT
1673 / HOST_BITS_PER_CHAR
)]
1674 << (i
* HOST_BITS_PER_CHAR
));
1675 low
|= (HOST_WIDE_INT
) parts
[i
] << (i
* HOST_BITS_PER_CHAR
);
1678 yylval
.ttype
= build_int_2 (low
, high
);
1679 TREE_TYPE (yylval
.ttype
) = long_long_unsigned_type_node
;
1681 /* If warn_traditional, calculate both the ANSI type and the
1682 traditional type, then see if they disagree.
1683 Otherwise, calculate only the type for the dialect in use. */
1684 if (warn_traditional
|| flag_traditional
)
1686 /* Calculate the traditional type. */
1687 /* Traditionally, any constant is signed;
1688 but if unsigned is specified explicitly, obey that.
1689 Use the smallest size with the right number of bits,
1690 except for one special case with decimal constants. */
1691 if (! spec_long
&& base
!= 10
1692 && int_fits_type_p (yylval
.ttype
, unsigned_type_node
))
1693 traditional_type
= (spec_unsigned
? unsigned_type_node
1694 : integer_type_node
);
1695 /* A decimal constant must be long
1696 if it does not fit in type int.
1697 I think this is independent of whether
1698 the constant is signed. */
1699 else if (! spec_long
&& base
== 10
1700 && int_fits_type_p (yylval
.ttype
, integer_type_node
))
1701 traditional_type
= (spec_unsigned
? unsigned_type_node
1702 : integer_type_node
);
1703 else if (! spec_long_long
)
1704 traditional_type
= (spec_unsigned
? long_unsigned_type_node
1705 : long_integer_type_node
);
1707 traditional_type
= (spec_unsigned
1708 ? long_long_unsigned_type_node
1709 : long_long_integer_type_node
);
1711 if (warn_traditional
|| ! flag_traditional
)
1713 /* Calculate the ANSI type. */
1714 if (! spec_long
&& ! spec_unsigned
1715 && int_fits_type_p (yylval
.ttype
, integer_type_node
))
1716 ansi_type
= integer_type_node
;
1717 else if (! spec_long
&& (base
!= 10 || spec_unsigned
)
1718 && int_fits_type_p (yylval
.ttype
, unsigned_type_node
))
1719 ansi_type
= unsigned_type_node
;
1720 else if (! spec_unsigned
&& !spec_long_long
1721 && int_fits_type_p (yylval
.ttype
, long_integer_type_node
))
1722 ansi_type
= long_integer_type_node
;
1723 else if (! spec_long_long
1724 && int_fits_type_p (yylval
.ttype
,
1725 long_unsigned_type_node
))
1726 ansi_type
= long_unsigned_type_node
;
1727 else if (! spec_unsigned
1728 && int_fits_type_p (yylval
.ttype
,
1729 long_long_integer_type_node
))
1730 ansi_type
= long_long_integer_type_node
;
1732 ansi_type
= long_long_unsigned_type_node
;
1735 type
= flag_traditional
? traditional_type
: ansi_type
;
1737 if (warn_traditional
&& traditional_type
!= ansi_type
)
1739 if (TYPE_PRECISION (traditional_type
)
1740 != TYPE_PRECISION (ansi_type
))
1741 warning ("width of integer constant changes with -traditional");
1742 else if (TREE_UNSIGNED (traditional_type
)
1743 != TREE_UNSIGNED (ansi_type
))
1744 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1746 warning ("width of integer constant may change on other systems with -traditional");
1749 if (pedantic
&& !flag_traditional
&& !spec_long_long
&& !warn
1750 && (TYPE_PRECISION (long_integer_type_node
)
1751 < TYPE_PRECISION (type
)))
1752 pedwarn ("integer constant out of range");
1754 if (base
== 10 && ! spec_unsigned
&& TREE_UNSIGNED (type
))
1755 warning ("decimal constant is so large that it is unsigned");
1759 if (TYPE_PRECISION (type
)
1760 <= TYPE_PRECISION (integer_type_node
))
1762 = build_complex (NULL_TREE
, integer_zero_node
,
1763 convert (integer_type_node
,
1766 error ("complex integer constant is too wide for `complex int'");
1768 else if (flag_traditional
&& !int_fits_type_p (yylval
.ttype
, type
))
1769 /* The traditional constant 0x80000000 is signed
1770 but doesn't fit in the range of int.
1771 This will change it to -0x80000000, which does fit. */
1773 TREE_TYPE (yylval
.ttype
) = unsigned_type (type
);
1774 yylval
.ttype
= convert (type
, yylval
.ttype
);
1775 TREE_OVERFLOW (yylval
.ttype
)
1776 = TREE_CONSTANT_OVERFLOW (yylval
.ttype
) = 0;
1779 TREE_TYPE (yylval
.ttype
) = type
;
1785 if (ISALNUM (c
) || c
== '.' || c
== '_' || c
== '$'
1786 || (!flag_traditional
&& (c
== '-' || c
== '+')
1787 && (p
[-1] == 'e' || p
[-1] == 'E')))
1788 error ("missing white space after number `%s'", token_buffer
);
1790 value
= CONSTANT
; break;
1796 register int result
= 0;
1797 register int num_chars
= 0;
1798 unsigned width
= TYPE_PRECISION (char_type_node
);
1803 width
= WCHAR_TYPE_SIZE
;
1804 #ifdef MULTIBYTE_CHARS
1805 max_chars
= MB_CUR_MAX
;
1811 max_chars
= TYPE_PRECISION (integer_type_node
) / width
;
1819 if (c
== '\'' || c
== EOF
)
1825 c
= readescape (&ignore
);
1828 if (width
< HOST_BITS_PER_INT
1829 && (unsigned) c
>= (1 << width
))
1830 pedwarn ("escape sequence out of range for character");
1831 #ifdef MAP_CHARACTER
1833 c
= MAP_CHARACTER (c
);
1839 pedwarn ("ANSI C forbids newline in character constant");
1842 #ifdef MAP_CHARACTER
1844 c
= MAP_CHARACTER (c
);
1848 if (num_chars
> maxtoken
- 4)
1849 extend_token_buffer (token_buffer
);
1851 token_buffer
[num_chars
] = c
;
1853 /* Merge character into result; ignore excess chars. */
1854 if (num_chars
< max_chars
+ 1)
1856 if (width
< HOST_BITS_PER_INT
)
1857 result
= (result
<< width
) | (c
& ((1 << width
) - 1));
1863 token_buffer
[num_chars
+ 1] = '\'';
1864 token_buffer
[num_chars
+ 2] = 0;
1867 error ("malformatted character constant");
1868 else if (num_chars
== 0)
1869 error ("empty character constant");
1870 else if (num_chars
> max_chars
)
1872 num_chars
= max_chars
;
1873 error ("character constant too long");
1875 else if (num_chars
!= 1 && ! flag_traditional
)
1876 warning ("multi-character character constant");
1878 /* If char type is signed, sign-extend the constant. */
1881 int num_bits
= num_chars
* width
;
1883 /* We already got an error; avoid invalid shift. */
1884 yylval
.ttype
= build_int_2 (0, 0);
1885 else if (TREE_UNSIGNED (char_type_node
)
1886 || ((result
>> (num_bits
- 1)) & 1) == 0)
1888 = build_int_2 (result
& (~(unsigned HOST_WIDE_INT
) 0
1889 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
1893 = build_int_2 (result
| ~(~(unsigned HOST_WIDE_INT
) 0
1894 >> (HOST_BITS_PER_WIDE_INT
- num_bits
)),
1896 TREE_TYPE (yylval
.ttype
) = integer_type_node
;
1900 #ifdef MULTIBYTE_CHARS
1901 /* Set the initial shift state and convert the next sequence. */
1903 /* In all locales L'\0' is zero and mbtowc will return zero,
1906 || (num_chars
== 1 && token_buffer
[1] != '\0'))
1909 (void) mbtowc (NULL_PTR
, NULL_PTR
, 0);
1910 if (mbtowc (& wc
, token_buffer
+ 1, num_chars
) == num_chars
)
1913 warning ("Ignoring invalid multibyte character");
1916 yylval
.ttype
= build_int_2 (result
, 0);
1917 TREE_TYPE (yylval
.ttype
) = wchar_type_node
;
1928 p
= token_buffer
+ 1;
1930 while (c
!= '"' && c
>= 0)
1935 c
= readescape (&ignore
);
1939 && TYPE_PRECISION (char_type_node
) < HOST_BITS_PER_INT
1940 && c
>= (1 << TYPE_PRECISION (char_type_node
)))
1941 pedwarn ("escape sequence out of range for character");
1946 pedwarn ("ANSI C forbids newline in string constant");
1950 if (p
== token_buffer
+ maxtoken
)
1951 p
= extend_token_buffer (p
);
1960 error ("Unterminated string constant");
1962 /* We have read the entire constant.
1963 Construct a STRING_CST for the result. */
1967 /* If this is a L"..." wide-string, convert the multibyte string
1968 to a wide character string. */
1969 char *widep
= (char *) alloca ((p
- token_buffer
) * WCHAR_BYTES
);
1972 #ifdef MULTIBYTE_CHARS
1973 len
= mbstowcs ((wchar_t *) widep
, token_buffer
+ 1, p
- token_buffer
);
1974 if (len
< 0 || len
>= (p
- token_buffer
))
1976 warning ("Ignoring invalid multibyte string");
1979 bzero (widep
+ (len
* WCHAR_BYTES
), WCHAR_BYTES
);
1984 wp
= widep
+ (BYTES_BIG_ENDIAN
? WCHAR_BYTES
- 1 : 0);
1985 bzero (widep
, (p
- token_buffer
) * WCHAR_BYTES
);
1986 for (cp
= token_buffer
+ 1; cp
< p
; cp
++)
1987 *wp
= *cp
, wp
+= WCHAR_BYTES
;
1988 len
= p
- token_buffer
- 1;
1991 yylval
.ttype
= build_string ((len
+ 1) * WCHAR_BYTES
, widep
);
1992 TREE_TYPE (yylval
.ttype
) = wchar_array_type_node
;
1997 /* Return an Objective-C @"..." constant string object. */
1998 yylval
.ttype
= build_objc_string (p
- token_buffer
,
2000 TREE_TYPE (yylval
.ttype
) = char_array_type_node
;
2001 value
= OBJC_STRING
;
2005 yylval
.ttype
= build_string (p
- token_buffer
, token_buffer
+ 1);
2006 TREE_TYPE (yylval
.ttype
) = char_array_type_node
;
2037 yylval
.code
= PLUS_EXPR
; break;
2039 yylval
.code
= MINUS_EXPR
; break;
2041 yylval
.code
= BIT_AND_EXPR
; break;
2043 yylval
.code
= BIT_IOR_EXPR
; break;
2045 yylval
.code
= MULT_EXPR
; break;
2047 yylval
.code
= TRUNC_DIV_EXPR
; break;
2049 yylval
.code
= TRUNC_MOD_EXPR
; break;
2051 yylval
.code
= BIT_XOR_EXPR
; break;
2053 yylval
.code
= LSHIFT_EXPR
; break;
2055 yylval
.code
= RSHIFT_EXPR
; break;
2057 yylval
.code
= LT_EXPR
; break;
2059 yylval
.code
= GT_EXPR
; break;
2062 token_buffer
[1] = c1
= GETC();
2063 token_buffer
[2] = 0;
2070 value
= ARITHCOMPARE
; yylval
.code
= LE_EXPR
; goto done
;
2072 value
= ARITHCOMPARE
; yylval
.code
= GE_EXPR
; goto done
;
2074 value
= EQCOMPARE
; yylval
.code
= NE_EXPR
; goto done
;
2076 value
= EQCOMPARE
; yylval
.code
= EQ_EXPR
; goto done
;
2078 value
= ASSIGN
; goto done
;
2084 value
= PLUSPLUS
; goto done
;
2086 value
= MINUSMINUS
; goto done
;
2088 value
= ANDAND
; goto done
;
2090 value
= OROR
; goto done
;
2103 { value
= POINTSAT
; goto done
; }
2107 { value
= ']'; goto done
; }
2111 { value
= '{'; indent_level
++; goto done
; }
2113 { value
= '['; goto done
; }
2117 { value
= '}'; indent_level
--; goto done
; }
2121 token_buffer
[1] = 0;
2123 if ((c
== '<') || (c
== '>'))
2124 value
= ARITHCOMPARE
;
2130 /* Don't make yyparse think this is eof. */
2149 /* yylloc.last_line = lineno; */
2154 /* Sets the value of the 'yydebug' variable to VALUE.
2155 This is a function so we don't have to have YYDEBUG defined
2156 in order to build the compiler. */
2165 warning ("YYDEBUG not defined.");