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. */
34 /* MULTIBYTE_CHARS support only works for native compilers.
35 ??? Ideally what we want is to model widechar support after
36 the current floating point support. */
38 #undef MULTIBYTE_CHARS
41 #ifdef MULTIBYTE_CHARS
48 cpp_options parse_options;
49 static enum cpp_token cpp_token;
51 /* Stream for reading from the input file. */
55 /* The elements of `ridpointers' are identifier nodes
56 for the reserved type names and storage classes.
57 It is indexed by a RID_... value. */
58 tree ridpointers[(int) RID_MAX];
60 /* Cause the `yydebug' variable to be defined. */
64 static unsigned char *yy_cur, *yy_lim;
71 parse_in.limit = parse_in.token_buffer;
72 cpp_token = cpp_get_token (&parse_in);
73 if (cpp_token == CPP_EOF)
75 yy_lim = CPP_PWRITTEN (&parse_in);
76 yy_cur = parse_in.token_buffer;
82 #define GETC() (yy_cur < yy_lim ? *yy_cur++ : yy_get_token ())
83 #define UNGETC(c) ((c), yy_cur--)
85 #define GETC() getc (finput)
86 #define UNGETC(c) ungetc (c, finput)
89 /* the declaration found for the last IDENTIFIER token read in.
90 yylex must look this up to detect typedefs, which get token type TYPENAME,
91 so it is left around in case the identifier is not a typedef but is
92 used in a context which makes it a reference to a variable. */
95 /* Nonzero enables objc features. */
101 /* File used for outputting assembler code. */
102 extern FILE *asm_out_file;
104 #ifndef WCHAR_TYPE_SIZE
106 #define WCHAR_TYPE_SIZE INT_TYPE_SIZE
108 #define WCHAR_TYPE_SIZE BITS_PER_WORD
112 /* Number of bytes in a wide character. */
113 #define WCHAR_BYTES (WCHAR_TYPE_SIZE / BITS_PER_UNIT)
115 static int maxtoken; /* Current nominal length of token buffer. */
116 char *token_buffer; /* Pointer to token buffer.
117 Actual allocated length is maxtoken + 2.
118 This is not static because objc-parse.y uses it. */
120 static int indent_level = 0; /* Number of { minus number of }. */
122 /* Nonzero if end-of-file has been seen on input. */
123 static int end_of_file;
126 /* Buffered-back input character; faster than using ungetc. */
127 static int nextchar = -1;
130 #ifdef HANDLE_SYSV_PRAGMA
131 static int handle_sysv_pragma PROTO((int));
132 #endif /* HANDLE_SYSV_PRAGMA */
133 static int whitespace_cr PROTO((int));
134 static int skip_white_space PROTO((int));
135 static int skip_white_space_on_line PROTO((void));
136 static char *extend_token_buffer PROTO((char *));
137 static int readescape PROTO((int *));
138 int check_newline ();
140 /* Do not insert generated code into the source, instead, include it.
141 This allows us to build gcc automatically even for targets that
142 need to add or modify the reserved keyword lists. */
145 /* Return something to represent absolute declarators containing a *.
146 TARGET is the absolute declarator that the * contains.
147 TYPE_QUALS is a list of modifiers such as const or volatile
148 to apply to the pointer type, represented as identifiers.
150 We return an INDIRECT_REF whose "contents" are TARGET
151 and whose type is the modifier list. */
154 make_pointer_declarator (type_quals, target)
155 tree type_quals, target;
157 return build1 (INDIRECT_REF, type_quals, target);
161 forget_protocol_qualifiers ()
163 int i, n = sizeof wordlist / sizeof (struct resword);
165 for (i = 0; i < n; i++)
166 if ((int) wordlist[i].rid >= (int) RID_IN
167 && (int) wordlist[i].rid <= (int) RID_ONEWAY)
168 wordlist[i].name = "";
172 remember_protocol_qualifiers ()
174 int i, n = sizeof wordlist / sizeof (struct resword);
176 for (i = 0; i < n; i++)
177 if (wordlist[i].rid == RID_IN)
178 wordlist[i].name = "in";
179 else if (wordlist[i].rid == RID_OUT)
180 wordlist[i].name = "out";
181 else if (wordlist[i].rid == RID_INOUT)
182 wordlist[i].name = "inout";
183 else if (wordlist[i].rid == RID_BYCOPY)
184 wordlist[i].name = "bycopy";
185 else if (wordlist[i].rid == RID_ONEWAY)
186 wordlist[i].name = "oneway";
190 init_parse (filename)
194 /* Open input file. */
195 if (filename == 0 || !strcmp (filename, "-"))
201 finput = fopen (filename, "r");
203 pfatal_with_name (filename);
205 #ifdef IO_BUFFER_SIZE
206 setvbuf (finput, (char *) xmalloc (IO_BUFFER_SIZE), _IOFBF, IO_BUFFER_SIZE);
208 #endif /* !USE_CPPLIB */
216 cpp_reader_init (&parse_in);
217 parse_in.data = &parse_options;
218 cpp_options_init (&parse_options);
219 cpp_handle_options (&parse_in, 0, NULL); /* FIXME */
220 parse_in.show_column = 1;
221 if (! cpp_start_read (&parse_in, filename))
232 cpp_finish (&parse_in);
241 /* Make identifier nodes long enough for the language-specific slots. */
242 set_identifier_size (sizeof (struct lang_identifier));
244 /* Start it at 0, because check_newline is called at the very beginning
245 and will increment it to 1. */
248 #ifdef MULTIBYTE_CHARS
249 /* Change to the native locale for multibyte conversions. */
250 setlocale (LC_CTYPE, "");
254 token_buffer = (char *) xmalloc (maxtoken + 2);
256 ridpointers[(int) RID_INT] = get_identifier ("int");
257 ridpointers[(int) RID_CHAR] = get_identifier ("char");
258 ridpointers[(int) RID_VOID] = get_identifier ("void");
259 ridpointers[(int) RID_FLOAT] = get_identifier ("float");
260 ridpointers[(int) RID_DOUBLE] = get_identifier ("double");
261 ridpointers[(int) RID_SHORT] = get_identifier ("short");
262 ridpointers[(int) RID_LONG] = get_identifier ("long");
263 ridpointers[(int) RID_UNSIGNED] = get_identifier ("unsigned");
264 ridpointers[(int) RID_SIGNED] = get_identifier ("signed");
265 ridpointers[(int) RID_INLINE] = get_identifier ("inline");
266 ridpointers[(int) RID_CONST] = get_identifier ("const");
267 ridpointers[(int) RID_VOLATILE] = get_identifier ("volatile");
268 ridpointers[(int) RID_AUTO] = get_identifier ("auto");
269 ridpointers[(int) RID_STATIC] = get_identifier ("static");
270 ridpointers[(int) RID_EXTERN] = get_identifier ("extern");
271 ridpointers[(int) RID_TYPEDEF] = get_identifier ("typedef");
272 ridpointers[(int) RID_REGISTER] = get_identifier ("register");
273 ridpointers[(int) RID_ITERATOR] = get_identifier ("iterator");
274 ridpointers[(int) RID_COMPLEX] = get_identifier ("complex");
275 ridpointers[(int) RID_ID] = get_identifier ("id");
276 ridpointers[(int) RID_IN] = get_identifier ("in");
277 ridpointers[(int) RID_OUT] = get_identifier ("out");
278 ridpointers[(int) RID_INOUT] = get_identifier ("inout");
279 ridpointers[(int) RID_BYCOPY] = get_identifier ("bycopy");
280 ridpointers[(int) RID_ONEWAY] = get_identifier ("oneway");
281 forget_protocol_qualifiers();
283 /* Some options inhibit certain reserved words.
284 Clear those words out of the hash table so they won't be recognized. */
285 #define UNSET_RESERVED_WORD(STRING) \
286 do { struct resword *s = is_reserved_word (STRING, sizeof (STRING) - 1); \
287 if (s) s->name = ""; } while (0)
289 if (! doing_objc_thang)
290 UNSET_RESERVED_WORD ("id");
292 if (flag_traditional)
294 UNSET_RESERVED_WORD ("const");
295 UNSET_RESERVED_WORD ("volatile");
296 UNSET_RESERVED_WORD ("typeof");
297 UNSET_RESERVED_WORD ("signed");
298 UNSET_RESERVED_WORD ("inline");
299 UNSET_RESERVED_WORD ("iterator");
300 UNSET_RESERVED_WORD ("complex");
304 UNSET_RESERVED_WORD ("asm");
305 UNSET_RESERVED_WORD ("typeof");
306 UNSET_RESERVED_WORD ("inline");
307 UNSET_RESERVED_WORD ("iterator");
308 UNSET_RESERVED_WORD ("complex");
313 reinit_parse_for_function ()
317 /* Function used when yydebug is set, to print a token in more detail. */
320 yyprint (file, yychar, yylval)
332 if (IDENTIFIER_POINTER (t))
333 fprintf (file, " `%s'", IDENTIFIER_POINTER (t));
338 if (TREE_CODE (t) == INTEGER_CST)
340 #if HOST_BITS_PER_WIDE_INT == 64
341 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_INT
344 #if HOST_BITS_PER_WIDE_INT == HOST_BITS_PER_LONG
351 #if HOST_BITS_PER_WIDE_INT != HOST_BITS_PER_INT
357 TREE_INT_CST_HIGH (t), TREE_INT_CST_LOW (t));
362 /* Iff C is a carriage return, warn about it - if appropriate -
363 and return nonzero. */
368 static int newline_warning = 0;
372 /* ANSI C says the effects of a carriage return in a source file
374 if (pedantic && !newline_warning)
376 warning ("carriage return in source file");
377 warning ("(we only warn about the first carriage return)");
385 /* If C is not whitespace, return C.
386 Otherwise skip whitespace and return first nonwhite char read. */
396 /* We don't recognize comments here, because
397 cpp output can include / and * consecutively as operators.
398 Also, there's no need, since cpp removes all comments. */
401 c = check_newline ();
422 error ("stray '\\' in program");
432 /* Skips all of the white space at the current location in the input file.
433 Must use and reset nextchar if it has the next character. */
436 position_after_white_space ()
442 c = nextchar, nextchar = -1;
447 UNGETC (skip_white_space (c));
450 /* Like skip_white_space, but don't advance beyond the end of line.
451 Moreover, we don't get passed a character to start with. */
453 skip_white_space_on_line ()
482 /* Make the token buffer longer, preserving the data in it.
483 P should point to just beyond the last valid character in the old buffer.
484 The value we return is a pointer to the new buffer
485 at a place corresponding to P. */
488 extend_token_buffer (p)
491 int offset = p - token_buffer;
493 maxtoken = maxtoken * 2 + 10;
494 token_buffer = (char *) xrealloc (token_buffer, maxtoken + 2);
496 return token_buffer + offset;
500 #define GET_DIRECTIVE_LINE() get_directive_line (finput)
501 #else /* USE_CPPLIB */
502 /* Read the rest of a #-directive from input stream FINPUT.
503 In normal use, the directive name and the white space after it
504 have already been read, so they won't be included in the result.
505 We allow for the fact that the directive line may contain
506 a newline embedded within a character or string literal which forms
507 a part of the directive.
509 The value is a string in a reusable buffer. It remains valid
510 only until the next time this function is called. */
513 GET_DIRECTIVE_LINE ()
515 static char *directive_buffer = NULL;
516 static unsigned buffer_length = 0;
518 register char *buffer_limit;
519 register int looking_for = 0;
520 register int char_escaped = 0;
522 if (buffer_length == 0)
524 directive_buffer = (char *)xmalloc (128);
528 buffer_limit = &directive_buffer[buffer_length];
530 for (p = directive_buffer; ; )
534 /* Make buffer bigger if it is full. */
535 if (p >= buffer_limit)
537 register unsigned bytes_used = (p - directive_buffer);
541 = (char *)xrealloc (directive_buffer, buffer_length);
542 p = &directive_buffer[bytes_used];
543 buffer_limit = &directive_buffer[buffer_length];
548 /* Discard initial whitespace. */
549 if ((c == ' ' || c == '\t') && p == directive_buffer)
552 /* Detect the end of the directive. */
553 if (c == '\n' && looking_for == 0)
562 return directive_buffer;
564 /* Handle string and character constant syntax. */
567 if (looking_for == c && !char_escaped)
568 looking_for = 0; /* Found terminator... stop looking. */
571 if (c == '\'' || c == '"')
572 looking_for = c; /* Don't stop buffering until we see another
573 another one of these (or an EOF). */
575 /* Handle backslash. */
576 char_escaped = (c == '\\' && ! char_escaped);
579 #endif /* USE_CPPLIB */
581 /* At the beginning of a line, increment the line number
582 and process any #-directive on this line.
583 If the line is a #-directive, read the entire line and return a newline.
584 Otherwise, return the line's first non-whitespace character. */
594 /* Read first nonwhite char on the line. */
597 while (c == ' ' || c == '\t')
602 /* If not #, return it so caller will use it. */
606 /* Read first nonwhite char after the `#'. */
609 while (c == ' ' || c == '\t')
612 /* If a letter follows, then if the word here is `line', skip
613 it and ignore it; otherwise, ignore the line, with an error
614 if the word isn't `pragma', `ident', `define', or `undef'. */
616 if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
625 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'
626 || whitespace_cr (c) ))
628 while (c == ' ' || c == '\t' || whitespace_cr (c))
632 #ifdef HANDLE_SYSV_PRAGMA
635 if (token != IDENTIFIER)
637 return handle_sysv_pragma (token);
638 #else /* !HANDLE_SYSV_PRAGMA */
643 if (token != IDENTIFIER)
645 if (HANDLE_PRAGMA (finput, yylval.ttype))
651 ??? do not know what to do ???;
652 #endif /* !USE_CPPLIB */
653 #endif /* HANDLE_PRAGMA */
654 #endif /* !HANDLE_SYSV_PRAGMA */
666 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
669 debug_define (lineno, GET_DIRECTIVE_LINE ());
679 && ((c = GETC()) == ' ' || c == '\t' || c == '\n'))
682 debug_undef (lineno, GET_DIRECTIVE_LINE ());
691 && ((c = GETC()) == ' ' || c == '\t'))
700 && ((c = GETC()) == ' ' || c == '\t'))
702 /* #ident. The pedantic warning is now in cccp.c. */
704 /* Here we have just seen `#ident '.
705 A string constant should follow. */
707 c = skip_white_space_on_line ();
709 /* If no argument, ignore the line. */
716 || TREE_CODE (yylval.ttype) != STRING_CST)
718 error ("invalid #ident");
724 #ifdef ASM_OUTPUT_IDENT
725 ASM_OUTPUT_IDENT (asm_out_file, TREE_STRING_POINTER (yylval.ttype));
729 /* Skip the rest of this line. */
734 error ("undefined or invalid # directive");
739 /* Here we have either `#line' or `# <nonletter>'.
740 In either case, it should be a line number; a digit should follow. */
742 /* Can't use skip_white_space here, but must handle all whitespace
743 that is not '\n', lest we get a recursion for '\r' '\n' when
746 c = skip_white_space_on_line ();
748 /* If the # is the only nonwhite char on the line,
749 just ignore it. Check the new newline. */
753 /* Something follows the #; read a token. */
758 if (token == CONSTANT
759 && TREE_CODE (yylval.ttype) == INTEGER_CST)
761 int old_lineno = lineno;
763 /* subtract one, because it is the following line that
764 gets the specified number */
766 int l = TREE_INT_CST_LOW (yylval.ttype) - 1;
768 /* Is this the last nonwhite stuff on the line? */
769 c = skip_white_space_on_line ();
772 /* No more: store the line number and check following line. */
778 /* More follows: it must be a string constant (filename). */
780 /* Read the string constant. */
783 if (token != STRING || TREE_CODE (yylval.ttype) != STRING_CST)
785 error ("invalid #line");
790 = (char *) permalloc (TREE_STRING_LENGTH (yylval.ttype) + 1);
791 strcpy (input_filename, TREE_STRING_POINTER (yylval.ttype));
794 /* Each change of file name
795 reinitializes whether we are now in a system header. */
796 in_system_header = 0;
798 if (main_input_filename == 0)
799 main_input_filename = input_filename;
801 /* Is this the last nonwhite stuff on the line? */
802 c = skip_white_space_on_line ();
805 /* Update the name in the top element of input_file_stack. */
806 if (input_file_stack)
807 input_file_stack->name = input_filename;
816 /* `1' after file name means entering new file.
817 `2' after file name means just left a file. */
819 if (token == CONSTANT
820 && TREE_CODE (yylval.ttype) == INTEGER_CST)
822 if (TREE_INT_CST_LOW (yylval.ttype) == 1)
824 /* Pushing to a new file. */
826 = (struct file_stack *) xmalloc (sizeof (struct file_stack));
827 input_file_stack->line = old_lineno;
828 p->next = input_file_stack;
829 p->name = input_filename;
830 p->indent_level = indent_level;
831 input_file_stack = p;
832 input_file_stack_tick++;
833 debug_start_source_file (input_filename);
836 else if (TREE_INT_CST_LOW (yylval.ttype) == 2)
838 /* Popping out of a file. */
839 if (input_file_stack->next)
841 struct file_stack *p = input_file_stack;
842 if (indent_level != p->indent_level)
844 warning_with_file_and_line
845 (p->name, old_lineno,
846 "This file contains more `%c's than `%c's.",
847 indent_level > p->indent_level ? '{' : '}',
848 indent_level > p->indent_level ? '}' : '{');
850 input_file_stack = p->next;
852 input_file_stack_tick++;
853 debug_end_source_file (input_file_stack->line);
856 error ("#-lines for entering and leaving files don't match");
862 /* Now that we've pushed or popped the input stack,
863 update the name in the top element. */
864 if (input_file_stack)
865 input_file_stack->name = input_filename;
867 /* If we have handled a `1' or a `2',
868 see if there is another number to read. */
871 /* Is this the last nonwhite stuff on the line? */
872 c = skip_white_space_on_line ();
881 /* `3' after file name means this is a system header file. */
883 if (token == CONSTANT
884 && TREE_CODE (yylval.ttype) == INTEGER_CST
885 && TREE_INT_CST_LOW (yylval.ttype) == 3)
886 in_system_header = 1, used_up = 1;
890 /* Is this the last nonwhite stuff on the line? */
891 c = skip_white_space_on_line ();
897 warning ("unrecognized text at end of #line");
900 error ("invalid #-line");
902 /* skip the rest of this line. */
905 if (c != '\n' && c != EOF && nextchar >= 0)
906 c = nextchar, nextchar = -1;
908 while (c != '\n' && c != EOF)
913 #ifdef HANDLE_SYSV_PRAGMA
915 /* Handle a #pragma directive.
916 TOKEN is the token we read after `#pragma'. Processes the entire input
917 line and returns a character for the caller to reread: either \n or EOF. */
919 /* This function has to be in this file, in order to get at
923 handle_sysv_pragma (token)
936 handle_pragma_token (token_buffer, yylval.ttype);
939 handle_pragma_token (token_buffer, 0);
943 c = nextchar, nextchar = -1;
948 while (c == ' ' || c == '\t')
950 if (c == '\n' || c == EOF)
952 handle_pragma_token (0, 0);
960 #endif /* HANDLE_SYSV_PRAGMA */
962 #define ENDFILE -1 /* token that represents end-of-file */
964 /* Read an escape sequence, returning its equivalent as a character,
965 or store 1 in *ignore_ptr if it is backslash-newline. */
968 readescape (ignore_ptr)
971 register int c = GETC();
973 register unsigned count;
974 unsigned firstdig = 0;
980 if (warn_traditional)
981 warning ("the meaning of `\\x' varies with -traditional");
983 if (flag_traditional)
992 if (!(c >= 'a' && c <= 'f')
993 && !(c >= 'A' && c <= 'F')
994 && !(c >= '0' && c <= '9'))
1000 if (c >= 'a' && c <= 'f')
1001 code += c - 'a' + 10;
1002 if (c >= 'A' && c <= 'F')
1003 code += c - 'A' + 10;
1004 if (c >= '0' && c <= '9')
1006 if (code != 0 || count != 0)
1015 error ("\\x used with no following hex digits");
1016 else if (count == 0)
1017 /* Digits are all 0's. Ok. */
1019 else if ((count - 1) * 4 >= TYPE_PRECISION (integer_type_node)
1021 && ((1 << (TYPE_PRECISION (integer_type_node) - (count - 1) * 4))
1023 pedwarn ("hex escape out of range");
1026 case '0': case '1': case '2': case '3': case '4':
1027 case '5': case '6': case '7':
1030 while ((c <= '7') && (c >= '0') && (count++ < 3))
1032 code = (code * 8) + (c - '0');
1038 case '\\': case '\'': case '"':
1047 return TARGET_NEWLINE;
1062 if (warn_traditional)
1063 warning ("the meaning of `\\a' varies with -traditional");
1065 if (flag_traditional)
1070 #if 0 /* Vertical tab is present in common usage compilers. */
1071 if (flag_traditional)
1079 pedwarn ("non-ANSI-standard escape sequence, `\\%c'", c);
1085 /* `\(', etc, are used at beginning of line to avoid confusing Emacs. */
1089 /* `\%' is used to prevent SCCS from getting confused. */
1092 pedwarn ("non-ANSI escape sequence `\\%c'", c);
1095 if (c >= 040 && c < 0177)
1096 pedwarn ("unknown escape sequence `\\%c'", c);
1098 pedwarn ("unknown escape sequence: `\\' followed by char code 0x%x", c);
1108 strcpy (buf, string);
1110 /* We can't print string and character constants well
1111 because the token_buffer contains the result of processing escapes. */
1113 strcat (buf, " at end of input");
1114 else if (token_buffer[0] == 0)
1115 strcat (buf, " at null character");
1116 else if (token_buffer[0] == '"')
1117 strcat (buf, " before string constant");
1118 else if (token_buffer[0] == '\'')
1119 strcat (buf, " before character constant");
1120 else if (token_buffer[0] < 040 || (unsigned char) token_buffer[0] >= 0177)
1121 sprintf (buf + strlen (buf), " before character 0%o",
1122 (unsigned char) token_buffer[0]);
1124 strcat (buf, " before `%s'");
1126 error (buf, token_buffer);
1136 char long_long_flag;
1139 struct try_type type_sequence[] =
1141 { &integer_type_node, 0, 0, 0},
1142 { &unsigned_type_node, 1, 0, 0},
1143 { &long_integer_type_node, 0, 1, 0},
1144 { &long_unsigned_type_node, 1, 1, 0},
1145 { &long_long_integer_type_node, 0, 1, 1},
1146 { &long_long_unsigned_type_node, 1, 1, 1}
1161 c = nextchar, nextchar = -1;
1166 /* Effectively do c = skip_white_space (c)
1167 but do it faster in the usual cases. */
1180 /* Call skip_white_space so we can warn if appropriate. */
1185 c = skip_white_space (c);
1187 goto found_nonwhite;
1191 token_buffer[0] = c;
1192 token_buffer[1] = 0;
1194 /* yylloc.first_line = lineno; */
1200 token_buffer[0] = 0;
1205 /* Capital L may start a wide-string or wide-character constant. */
1207 register int c = GETC();
1216 goto string_constant;
1223 if (!doing_objc_thang)
1230 /* '@' may start a constant string object. */
1231 register int c = GETC ();
1235 goto string_constant;
1238 /* Fall through to treat '@' as the start of an identifier. */
1241 case 'A': case 'B': case 'C': case 'D': case 'E':
1242 case 'F': case 'G': case 'H': case 'I': case 'J':
1243 case 'K': case 'M': case 'N': case 'O':
1244 case 'P': case 'Q': case 'R': case 'S': case 'T':
1245 case 'U': case 'V': case 'W': case 'X': case 'Y':
1247 case 'a': case 'b': case 'c': case 'd': case 'e':
1248 case 'f': case 'g': case 'h': case 'i': case 'j':
1249 case 'k': case 'l': case 'm': case 'n': case 'o':
1250 case 'p': case 'q': case 'r': case 's': case 't':
1251 case 'u': case 'v': case 'w': case 'x': case 'y':
1257 while (isalnum (c) || c == '_' || c == '$' || c == '@')
1259 /* Make sure this char really belongs in an identifier. */
1260 if (c == '@' && ! doing_objc_thang)
1264 if (! dollars_in_ident)
1265 error ("`$' in identifier");
1267 pedwarn ("`$' in identifier");
1270 if (p >= token_buffer + maxtoken)
1271 p = extend_token_buffer (p);
1287 /* Try to recognize a keyword. Uses minimum-perfect hash function */
1290 register struct resword *ptr;
1292 if ((ptr = is_reserved_word (token_buffer, p - token_buffer)))
1295 yylval.ttype = ridpointers[(int) ptr->rid];
1296 value = (int) ptr->token;
1298 /* Only return OBJECTNAME if it is a typedef. */
1299 if (doing_objc_thang && value == OBJECTNAME)
1301 lastiddecl = lookup_name(yylval.ttype);
1303 if (lastiddecl == NULL_TREE
1304 || TREE_CODE (lastiddecl) != TYPE_DECL)
1308 /* Even if we decided to recognize asm, still perhaps warn. */
1310 && (value == ASM_KEYWORD || value == TYPEOF
1311 || ptr->rid == RID_INLINE)
1312 && token_buffer[0] != '_')
1313 pedwarn ("ANSI does not permit the keyword `%s'",
1318 /* If we did not find a keyword, look for an identifier
1321 if (value == IDENTIFIER)
1323 if (token_buffer[0] == '@')
1324 error("invalid identifier `%s'", token_buffer);
1326 yylval.ttype = get_identifier (token_buffer);
1327 lastiddecl = lookup_name (yylval.ttype);
1329 if (lastiddecl != 0 && TREE_CODE (lastiddecl) == TYPE_DECL)
1331 /* A user-invisible read-only initialized variable
1332 should be replaced by its value.
1333 We handle only strings since that's the only case used in C. */
1334 else if (lastiddecl != 0 && TREE_CODE (lastiddecl) == VAR_DECL
1335 && DECL_IGNORED_P (lastiddecl)
1336 && TREE_READONLY (lastiddecl)
1337 && DECL_INITIAL (lastiddecl) != 0
1338 && TREE_CODE (DECL_INITIAL (lastiddecl)) == STRING_CST)
1340 tree stringval = DECL_INITIAL (lastiddecl);
1342 /* Copy the string value so that we won't clobber anything
1343 if we put something in the TREE_CHAIN of this one. */
1344 yylval.ttype = build_string (TREE_STRING_LENGTH (stringval),
1345 TREE_STRING_POINTER (stringval));
1348 else if (doing_objc_thang)
1350 tree objc_interface_decl = is_class_name (yylval.ttype);
1352 if (objc_interface_decl)
1355 yylval.ttype = objc_interface_decl;
1365 /* Check first for common special case: single-digit 0 or 1. */
1368 UNGETC (next_c); /* Always undo this lookahead. */
1369 if (!isalnum (next_c) && next_c != '.')
1371 token_buffer[0] = (char)c, token_buffer[1] = '\0';
1372 yylval.ttype = (c == '0') ? integer_zero_node : integer_one_node;
1378 case '2': case '3': case '4':
1379 case '5': case '6': case '7': case '8': case '9':
1384 int largest_digit = 0;
1386 /* for multi-precision arithmetic,
1387 we actually store only HOST_BITS_PER_CHAR bits in each part.
1388 The number of parts is chosen so as to be sufficient to hold
1389 the enough bits to fit into the two HOST_WIDE_INTs that contain
1390 the integer value (this is always at least as many bits as are
1391 in a target `long long' value, but may be wider). */
1392 #define TOTAL_PARTS ((HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR) * 2 + 2)
1393 int parts[TOTAL_PARTS];
1396 enum anon1 { NOT_FLOAT, AFTER_POINT, TOO_MANY_POINTS} floatflag
1399 for (count = 0; count < TOTAL_PARTS; count++)
1407 *p++ = (c = GETC());
1408 if ((c == 'x') || (c == 'X'))
1411 *p++ = (c = GETC());
1413 /* Leading 0 forces octal unless the 0 is the only digit. */
1414 else if (c >= '0' && c <= '9')
1423 /* Read all the digits-and-decimal-points. */
1426 || (isalnum (c) && c != 'l' && c != 'L'
1427 && c != 'u' && c != 'U'
1428 && c != 'i' && c != 'I' && c != 'j' && c != 'J'
1429 && (floatflag == NOT_FLOAT || ((c != 'f') && (c != 'F')))))
1434 error ("floating constant may not be in radix 16");
1435 if (floatflag == TOO_MANY_POINTS)
1436 /* We have already emitted an error. Don't need another. */
1438 else if (floatflag == AFTER_POINT)
1440 error ("malformed floating constant");
1441 floatflag = TOO_MANY_POINTS;
1442 /* Avoid another error from atof by forcing all characters
1443 from here on to be ignored. */
1447 floatflag = AFTER_POINT;
1451 /* Accept '.' as the start of a floating-point number
1452 only when it is followed by a digit.
1453 Otherwise, unread the following non-digit
1454 and use the '.' as a structural token. */
1455 if (p == token_buffer + 2 && !isdigit (c))
1466 error ("parse error at `..'");
1469 token_buffer[1] = 0;
1476 /* It is not a decimal point.
1477 It should be a digit (perhaps a hex digit). */
1483 else if (base <= 10)
1485 if (c == 'e' || c == 'E')
1488 floatflag = AFTER_POINT;
1489 break; /* start of exponent */
1491 error ("nondigits in number and not hexadecimal");
1502 if (c >= largest_digit)
1506 for (count = 0; count < TOTAL_PARTS; count++)
1508 parts[count] *= base;
1512 += (parts[count-1] >> HOST_BITS_PER_CHAR);
1514 &= (1 << HOST_BITS_PER_CHAR) - 1;
1520 /* If the extra highest-order part ever gets anything in it,
1521 the number is certainly too big. */
1522 if (parts[TOTAL_PARTS - 1] != 0)
1525 if (p >= token_buffer + maxtoken - 3)
1526 p = extend_token_buffer (p);
1527 *p++ = (c = GETC());
1532 error ("numeric constant with no digits");
1534 if (largest_digit >= base)
1535 error ("numeric constant contains digits beyond the radix");
1537 /* Remove terminating char from the token buffer and delimit the string */
1540 if (floatflag != NOT_FLOAT)
1542 tree type = double_type_node;
1544 int conversion_errno = 0;
1545 REAL_VALUE_TYPE value;
1548 /* Read explicit exponent if any, and put it in tokenbuf. */
1550 if ((c == 'e') || (c == 'E'))
1552 if (p >= token_buffer + maxtoken - 3)
1553 p = extend_token_buffer (p);
1556 if ((c == '+') || (c == '-'))
1562 error ("floating constant exponent has no digits");
1565 if (p >= token_buffer + maxtoken - 3)
1566 p = extend_token_buffer (p);
1574 /* Convert string to a double, checking for overflow. */
1575 if (setjmp (handler))
1577 error ("floating constant out of range");
1582 int fflag = 0, lflag = 0;
1583 /* Copy token_buffer now, while it has just the number
1584 and not the suffixes; once we add `f' or `i',
1585 REAL_VALUE_ATOF may not work any more. */
1586 char *copy = (char *) alloca (p - token_buffer + 1);
1587 bcopy (token_buffer, copy, p - token_buffer + 1);
1589 set_float_handler (handler);
1595 /* Read the suffixes to choose a data type. */
1600 error ("more than one `f' in numeric constant");
1606 error ("more than one `l' in numeric constant");
1612 error ("more than one `i' or `j' in numeric constant");
1614 pedwarn ("ANSI C forbids imaginary numeric constants");
1625 if (p >= token_buffer + maxtoken - 3)
1626 p = extend_token_buffer (p);
1632 /* The second argument, machine_mode, of REAL_VALUE_ATOF
1633 tells the desired precision of the binary result
1634 of decimal-to-binary conversion. */
1639 error ("both `f' and `l' in floating constant");
1641 type = float_type_node;
1643 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1644 conversion_errno = errno;
1645 /* A diagnostic is required here by some ANSI C testsuites.
1646 This is not pedwarn, become some people don't want
1647 an error for this. */
1648 if (REAL_VALUE_ISINF (value) && pedantic)
1649 warning ("floating point number exceeds range of `float'");
1653 type = long_double_type_node;
1655 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1656 conversion_errno = errno;
1657 if (REAL_VALUE_ISINF (value) && pedantic)
1658 warning ("floating point number exceeds range of `long double'");
1663 value = REAL_VALUE_ATOF (copy, TYPE_MODE (type));
1664 conversion_errno = errno;
1665 if (REAL_VALUE_ISINF (value) && pedantic)
1666 warning ("floating point number exceeds range of `double'");
1669 set_float_handler (NULL_PTR);
1672 /* ERANGE is also reported for underflow,
1673 so test the value to distinguish overflow from that. */
1674 if (conversion_errno == ERANGE && !flag_traditional && pedantic
1675 && (REAL_VALUES_LESS (dconst1, value)
1676 || REAL_VALUES_LESS (value, dconstm1)))
1677 warning ("floating point number exceeds range of `double'");
1680 /* If the result is not a number, assume it must have been
1681 due to some error message above, so silently convert
1683 if (REAL_VALUE_ISNAN (value))
1686 /* Create a node with determined type and value. */
1688 yylval.ttype = build_complex (NULL_TREE,
1689 convert (type, integer_zero_node),
1690 build_real (type, value));
1692 yylval.ttype = build_real (type, value);
1696 tree traditional_type, ansi_type, type;
1697 HOST_WIDE_INT high, low;
1698 int spec_unsigned = 0;
1700 int spec_long_long = 0;
1704 traditional_type = ansi_type = type = NULL_TREE;
1707 if (c == 'u' || c == 'U')
1710 error ("two `u's in integer constant");
1713 else if (c == 'l' || c == 'L')
1718 error ("three `l's in integer constant");
1720 pedwarn ("ANSI C forbids long long integer constants");
1725 else if (c == 'i' || c == 'j' || c == 'I' || c == 'J')
1728 error ("more than one `i' or `j' in numeric constant");
1730 pedwarn ("ANSI C forbids imaginary numeric constants");
1735 if (p >= token_buffer + maxtoken - 3)
1736 p = extend_token_buffer (p);
1741 /* If the constant won't fit in an unsigned long long,
1742 then warn that the constant is out of range. */
1744 /* ??? This assumes that long long and long integer types are
1745 a multiple of 8 bits. This better than the original code
1746 though which assumed that long was exactly 32 bits and long
1747 long was exactly 64 bits. */
1749 bytes = TYPE_PRECISION (long_long_integer_type_node) / 8;
1752 for (i = bytes; i < TOTAL_PARTS; i++)
1756 pedwarn ("integer constant out of range");
1758 /* This is simplified by the fact that our constant
1759 is always positive. */
1763 for (i = 0; i < HOST_BITS_PER_WIDE_INT / HOST_BITS_PER_CHAR; i++)
1765 high |= ((HOST_WIDE_INT) parts[i + (HOST_BITS_PER_WIDE_INT
1766 / HOST_BITS_PER_CHAR)]
1767 << (i * HOST_BITS_PER_CHAR));
1768 low |= (HOST_WIDE_INT) parts[i] << (i * HOST_BITS_PER_CHAR);
1771 yylval.ttype = build_int_2 (low, high);
1772 TREE_TYPE (yylval.ttype) = long_long_unsigned_type_node;
1774 /* If warn_traditional, calculate both the ANSI type and the
1775 traditional type, then see if they disagree.
1776 Otherwise, calculate only the type for the dialect in use. */
1777 if (warn_traditional || flag_traditional)
1779 /* Calculate the traditional type. */
1780 /* Traditionally, any constant is signed;
1781 but if unsigned is specified explicitly, obey that.
1782 Use the smallest size with the right number of bits,
1783 except for one special case with decimal constants. */
1784 if (! spec_long && base != 10
1785 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1786 traditional_type = (spec_unsigned ? unsigned_type_node
1787 : integer_type_node);
1788 /* A decimal constant must be long
1789 if it does not fit in type int.
1790 I think this is independent of whether
1791 the constant is signed. */
1792 else if (! spec_long && base == 10
1793 && int_fits_type_p (yylval.ttype, integer_type_node))
1794 traditional_type = (spec_unsigned ? unsigned_type_node
1795 : integer_type_node);
1796 else if (! spec_long_long)
1797 traditional_type = (spec_unsigned ? long_unsigned_type_node
1798 : long_integer_type_node);
1800 traditional_type = (spec_unsigned
1801 ? long_long_unsigned_type_node
1802 : long_long_integer_type_node);
1804 if (warn_traditional || ! flag_traditional)
1806 /* Calculate the ANSI type. */
1807 if (! spec_long && ! spec_unsigned
1808 && int_fits_type_p (yylval.ttype, integer_type_node))
1809 ansi_type = integer_type_node;
1810 else if (! spec_long && (base != 10 || spec_unsigned)
1811 && int_fits_type_p (yylval.ttype, unsigned_type_node))
1812 ansi_type = unsigned_type_node;
1813 else if (! spec_unsigned && !spec_long_long
1814 && int_fits_type_p (yylval.ttype, long_integer_type_node))
1815 ansi_type = long_integer_type_node;
1816 else if (! spec_long_long
1817 && int_fits_type_p (yylval.ttype,
1818 long_unsigned_type_node))
1819 ansi_type = long_unsigned_type_node;
1820 else if (! spec_unsigned
1821 && int_fits_type_p (yylval.ttype,
1822 long_long_integer_type_node))
1823 ansi_type = long_long_integer_type_node;
1825 ansi_type = long_long_unsigned_type_node;
1828 type = flag_traditional ? traditional_type : ansi_type;
1830 if (warn_traditional && traditional_type != ansi_type)
1832 if (TYPE_PRECISION (traditional_type)
1833 != TYPE_PRECISION (ansi_type))
1834 warning ("width of integer constant changes with -traditional");
1835 else if (TREE_UNSIGNED (traditional_type)
1836 != TREE_UNSIGNED (ansi_type))
1837 warning ("integer constant is unsigned in ANSI C, signed with -traditional");
1839 warning ("width of integer constant may change on other systems with -traditional");
1842 if (pedantic && !flag_traditional && !spec_long_long && !warn
1843 && (TYPE_PRECISION (long_integer_type_node)
1844 < TYPE_PRECISION (type)))
1845 pedwarn ("integer constant out of range");
1847 if (base == 10 && ! spec_unsigned && TREE_UNSIGNED (type))
1848 warning ("decimal constant is so large that it is unsigned");
1852 if (TYPE_PRECISION (type)
1853 <= TYPE_PRECISION (integer_type_node))
1855 = build_complex (NULL_TREE, integer_zero_node,
1856 convert (integer_type_node,
1859 error ("complex integer constant is too wide for `complex int'");
1861 else if (flag_traditional && !int_fits_type_p (yylval.ttype, type))
1862 /* The traditional constant 0x80000000 is signed
1863 but doesn't fit in the range of int.
1864 This will change it to -0x80000000, which does fit. */
1866 TREE_TYPE (yylval.ttype) = unsigned_type (type);
1867 yylval.ttype = convert (type, yylval.ttype);
1868 TREE_OVERFLOW (yylval.ttype)
1869 = TREE_CONSTANT_OVERFLOW (yylval.ttype) = 0;
1872 TREE_TYPE (yylval.ttype) = type;
1878 if (isalnum (c) || c == '.' || c == '_' || c == '$'
1879 || (!flag_traditional && (c == '-' || c == '+')
1880 && (p[-1] == 'e' || p[-1] == 'E')))
1881 error ("missing white space after number `%s'", token_buffer);
1883 value = CONSTANT; break;
1889 register int result = 0;
1890 register int num_chars = 0;
1891 unsigned width = TYPE_PRECISION (char_type_node);
1896 width = WCHAR_TYPE_SIZE;
1897 #ifdef MULTIBYTE_CHARS
1898 max_chars = MB_CUR_MAX;
1904 max_chars = TYPE_PRECISION (integer_type_node) / width;
1912 if (c == '\'' || c == EOF)
1918 c = readescape (&ignore);
1921 if (width < HOST_BITS_PER_INT
1922 && (unsigned) c >= (1 << width))
1923 pedwarn ("escape sequence out of range for character");
1924 #ifdef MAP_CHARACTER
1926 c = MAP_CHARACTER (c);
1932 pedwarn ("ANSI C forbids newline in character constant");
1935 #ifdef MAP_CHARACTER
1937 c = MAP_CHARACTER (c);
1941 if (num_chars > maxtoken - 4)
1942 extend_token_buffer (token_buffer);
1944 token_buffer[num_chars] = c;
1946 /* Merge character into result; ignore excess chars. */
1947 if (num_chars < max_chars + 1)
1949 if (width < HOST_BITS_PER_INT)
1950 result = (result << width) | (c & ((1 << width) - 1));
1956 token_buffer[num_chars + 1] = '\'';
1957 token_buffer[num_chars + 2] = 0;
1960 error ("malformatted character constant");
1961 else if (num_chars == 0)
1962 error ("empty character constant");
1963 else if (num_chars > max_chars)
1965 num_chars = max_chars;
1966 error ("character constant too long");
1968 else if (num_chars != 1 && ! flag_traditional)
1969 warning ("multi-character character constant");
1971 /* If char type is signed, sign-extend the constant. */
1974 int num_bits = num_chars * width;
1976 /* We already got an error; avoid invalid shift. */
1977 yylval.ttype = build_int_2 (0, 0);
1978 else if (TREE_UNSIGNED (char_type_node)
1979 || ((result >> (num_bits - 1)) & 1) == 0)
1981 = build_int_2 (result & (~(unsigned HOST_WIDE_INT) 0
1982 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1986 = build_int_2 (result | ~(~(unsigned HOST_WIDE_INT) 0
1987 >> (HOST_BITS_PER_WIDE_INT - num_bits)),
1989 TREE_TYPE (yylval.ttype) = integer_type_node;
1993 #ifdef MULTIBYTE_CHARS
1994 /* Set the initial shift state and convert the next sequence. */
1996 /* In all locales L'\0' is zero and mbtowc will return zero,
1999 || (num_chars == 1 && token_buffer[1] != '\0'))
2002 (void) mbtowc (NULL_PTR, NULL_PTR, 0);
2003 if (mbtowc (& wc, token_buffer + 1, num_chars) == num_chars)
2006 warning ("Ignoring invalid multibyte character");
2009 yylval.ttype = build_int_2 (result, 0);
2010 TREE_TYPE (yylval.ttype) = wchar_type_node;
2021 p = token_buffer + 1;
2023 while (c != '"' && c >= 0)
2028 c = readescape (&ignore);
2032 && TYPE_PRECISION (char_type_node) < HOST_BITS_PER_INT
2033 && c >= (1 << TYPE_PRECISION (char_type_node)))
2034 pedwarn ("escape sequence out of range for character");
2039 pedwarn ("ANSI C forbids newline in string constant");
2043 if (p == token_buffer + maxtoken)
2044 p = extend_token_buffer (p);
2053 error ("Unterminated string constant");
2055 /* We have read the entire constant.
2056 Construct a STRING_CST for the result. */
2060 /* If this is a L"..." wide-string, convert the multibyte string
2061 to a wide character string. */
2062 char *widep = (char *) alloca ((p - token_buffer) * WCHAR_BYTES);
2065 #ifdef MULTIBYTE_CHARS
2066 len = mbstowcs ((wchar_t *) widep, token_buffer + 1, p - token_buffer);
2067 if (len < 0 || len >= (p - token_buffer))
2069 warning ("Ignoring invalid multibyte string");
2072 bzero (widep + (len * WCHAR_BYTES), WCHAR_BYTES);
2077 wp = widep + (BYTES_BIG_ENDIAN ? WCHAR_BYTES - 1 : 0);
2078 bzero (widep, (p - token_buffer) * WCHAR_BYTES);
2079 for (cp = token_buffer + 1; cp < p; cp++)
2080 *wp = *cp, wp += WCHAR_BYTES;
2081 len = p - token_buffer - 1;
2084 yylval.ttype = build_string ((len + 1) * WCHAR_BYTES, widep);
2085 TREE_TYPE (yylval.ttype) = wchar_array_type_node;
2090 extern tree build_objc_string();
2091 /* Return an Objective-C @"..." constant string object. */
2092 yylval.ttype = build_objc_string (p - token_buffer,
2094 TREE_TYPE (yylval.ttype) = char_array_type_node;
2095 value = OBJC_STRING;
2099 yylval.ttype = build_string (p - token_buffer, token_buffer + 1);
2100 TREE_TYPE (yylval.ttype) = char_array_type_node;
2131 yylval.code = PLUS_EXPR; break;
2133 yylval.code = MINUS_EXPR; break;
2135 yylval.code = BIT_AND_EXPR; break;
2137 yylval.code = BIT_IOR_EXPR; break;
2139 yylval.code = MULT_EXPR; break;
2141 yylval.code = TRUNC_DIV_EXPR; break;
2143 yylval.code = TRUNC_MOD_EXPR; break;
2145 yylval.code = BIT_XOR_EXPR; break;
2147 yylval.code = LSHIFT_EXPR; break;
2149 yylval.code = RSHIFT_EXPR; break;
2151 yylval.code = LT_EXPR; break;
2153 yylval.code = GT_EXPR; break;
2156 token_buffer[1] = c1 = GETC();
2157 token_buffer[2] = 0;
2164 value = ARITHCOMPARE; yylval.code = LE_EXPR; goto done;
2166 value = ARITHCOMPARE; yylval.code = GE_EXPR; goto done;
2168 value = EQCOMPARE; yylval.code = NE_EXPR; goto done;
2170 value = EQCOMPARE; yylval.code = EQ_EXPR; goto done;
2172 value = ASSIGN; goto done;
2178 value = PLUSPLUS; goto done;
2180 value = MINUSMINUS; goto done;
2182 value = ANDAND; goto done;
2184 value = OROR; goto done;
2197 { value = POINTSAT; goto done; }
2201 { value = ']'; goto done; }
2205 { value = '{'; indent_level++; goto done; }
2207 { value = '['; goto done; }
2211 { value = '}'; indent_level--; goto done; }
2215 token_buffer[1] = 0;
2217 if ((c == '<') || (c == '>'))
2218 value = ARITHCOMPARE;
2224 /* Don't make yyparse think this is eof. */
2243 /* yylloc.last_line = lineno; */
2248 /* Sets the value of the 'yydebug' variable to VALUE.
2249 This is a function so we don't have to have YYDEBUG defined
2250 in order to build the compiler. */
2259 warning ("YYDEBUG not defined.");