2 * This file is part of cparser.
3 * Copyright (C) 2007-2009 Matthias Braun <matze@braunis.de>
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License
7 * as published by the Free Software Foundation; either version 2
8 * of the License, or (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
22 #include "adt/strutil.h"
24 #include "diagnostic.h"
28 #include "symbol_table_t.h"
29 #include "adt/error.h"
30 #include "adt/strset.h"
36 #include "lang_features.h"
48 #define MAX_PUTBACK 16 // 3 would be enough, but 16 gives a nicer alignment
51 static input_t *input;
52 static utf32 input_buf[BUF_SIZE + MAX_PUTBACK];
53 static const utf32 *bufpos;
54 static const utf32 *bufend;
56 static source_position_t lexer_pos;
58 static symbol_t *symbol_L;
59 static strset_t stringset;
60 bool allow_dollar_in_symbol = true;
63 * Prints a parse error message at the current token.
65 * @param msg the error message
67 static void parse_error(const char *msg)
69 errorf(&lexer_pos, "%s", msg);
73 * Prints an internal error message at the current token.
75 * @param msg the error message
77 static NORETURN internal_error(const char *msg)
79 internal_errorf(&lexer_pos, "%s", msg);
82 static inline void next_real_char(void)
84 assert(bufpos <= bufend);
85 if (bufpos >= bufend) {
86 size_t n = decode(input, input_buf+MAX_PUTBACK, BUF_SIZE);
91 bufpos = input_buf + MAX_PUTBACK;
99 * Put a character back into the buffer.
101 * @param pc the character to put back
103 static inline void put_back(utf32 const pc)
105 *(--bufpos - input_buf + input_buf) = pc;
109 static inline void next_char(void);
111 #define MATCH_NEWLINE(code) \
118 lexer_pos.lineno++; \
119 lexer_pos.colno = 1; \
122 #define eat(c_type) (assert(c == c_type), next_char())
124 static void maybe_concat_lines(void)
129 MATCH_NEWLINE(return;)
140 * Set c to the next input character, ie.
141 * after expanding trigraphs.
143 static inline void next_char(void)
147 /* filter trigraphs */
148 if (UNLIKELY(c == '\\')) {
149 maybe_concat_lines();
153 if (LIKELY(c != '?'))
157 if (LIKELY(c != '?')) {
165 case '=': c = '#'; break;
166 case '(': c = '['; break;
167 case '/': c = '\\'; maybe_concat_lines(); break;
168 case ')': c = ']'; break;
169 case '\'': c = '^'; break;
170 case '<': c = '{'; break;
171 case '!': c = '|'; break;
172 case '>': c = '}'; break;
173 case '-': c = '~'; break;
182 #define SYMBOL_CHARS \
183 case '$': if (!allow_dollar_in_symbol) goto dollar_sign; \
251 * Read a symbol from the input and build
254 static void parse_symbol(void)
256 obstack_1grow(&symbol_obstack, (char) c);
263 obstack_1grow(&symbol_obstack, (char) c);
274 obstack_1grow(&symbol_obstack, '\0');
276 char *string = obstack_finish(&symbol_obstack);
277 symbol_t *symbol = symbol_table_insert(string);
279 lexer_token.kind = symbol->ID;
280 lexer_token.identifier.symbol = symbol;
282 if (symbol->string != string) {
283 obstack_free(&symbol_obstack, string);
287 static string_t identify_string(char *string, size_t len)
291 const char *result = strset_insert(&stringset, concat);
292 if (result != concat) {
293 obstack_free(&symbol_obstack, concat);
296 const char *result = string;
298 return (string_t) {result, len};
302 * parse suffixes like 'LU' or 'f' after numbers
304 static void parse_number_suffix(void)
306 assert(obstack_object_size(&symbol_obstack) == 0);
310 obstack_1grow(&symbol_obstack, (char) c);
319 if (obstack_object_size(&symbol_obstack) == 0) {
320 lexer_token.number.suffix.begin = NULL;
321 lexer_token.number.suffix.size = 0;
325 obstack_1grow(&symbol_obstack, '\0');
326 size_t size = obstack_object_size(&symbol_obstack) - 1;
327 char *string = obstack_finish(&symbol_obstack);
329 lexer_token.number.suffix = identify_string(string, size);
333 * Parses a hex number including hex floats and set the
336 static void parse_number_hex(void)
338 bool is_float = false;
339 bool has_digits = false;
341 assert(obstack_object_size(&symbol_obstack) == 0);
342 while (isxdigit(c)) {
344 obstack_1grow(&symbol_obstack, (char) c);
350 obstack_1grow(&symbol_obstack, (char) c);
353 while (isxdigit(c)) {
355 obstack_1grow(&symbol_obstack, (char) c);
359 if (c == 'p' || c == 'P') {
361 obstack_1grow(&symbol_obstack, (char) c);
364 if (c == '-' || c == '+') {
365 obstack_1grow(&symbol_obstack, (char) c);
369 while (isxdigit(c)) {
370 obstack_1grow(&symbol_obstack, (char) c);
373 } else if (is_float) {
374 errorf(&lexer_token.base.source_position,
375 "hexadecimal floatingpoint constant requires an exponent");
377 obstack_1grow(&symbol_obstack, '\0');
379 size_t size = obstack_object_size(&symbol_obstack) - 1;
380 char *string = obstack_finish(&symbol_obstack);
381 lexer_token.number.number = identify_string(string, size);
384 is_float ? T_FLOATINGPOINT_HEXADECIMAL : T_INTEGER_HEXADECIMAL;
387 errorf(&lexer_token.base.source_position,
388 "invalid number literal '0x%S'", &lexer_token.number.number);
389 lexer_token.number.number.begin = "0";
390 lexer_token.number.number.size = 1;
393 parse_number_suffix();
397 * Returns true if the given char is a octal digit.
399 * @param char the character to check
401 static bool is_octal_digit(utf32 chr)
403 return '0' <= chr && chr <= '7';
407 * Parses a number and sets the lexer_token.
409 static void parse_number(void)
411 bool is_float = false;
412 bool has_digits = false;
414 assert(obstack_object_size(&symbol_obstack) == 0);
417 if (c == 'x' || c == 'X') {
424 obstack_1grow(&symbol_obstack, '0');
429 obstack_1grow(&symbol_obstack, (char) c);
435 obstack_1grow(&symbol_obstack, '.');
440 obstack_1grow(&symbol_obstack, (char) c);
444 if (c == 'e' || c == 'E') {
446 obstack_1grow(&symbol_obstack, 'e');
449 if (c == '-' || c == '+') {
450 obstack_1grow(&symbol_obstack, (char) c);
455 obstack_1grow(&symbol_obstack, (char) c);
460 obstack_1grow(&symbol_obstack, '\0');
461 size_t size = obstack_object_size(&symbol_obstack) - 1;
462 char *string = obstack_finish(&symbol_obstack);
463 lexer_token.number.number = identify_string(string, size);
465 /* is it an octal number? */
467 lexer_token.kind = T_FLOATINGPOINT;
468 } else if (string[0] == '0') {
469 lexer_token.kind = T_INTEGER_OCTAL;
471 /* check for invalid octal digits */
472 for (size_t i= 0; i < size; ++i) {
475 errorf(&lexer_token.base.source_position,
476 "invalid digit '%c' in octal number", t);
479 lexer_token.kind = T_INTEGER;
483 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
484 &lexer_token.number.number);
487 parse_number_suffix();
491 * Returns the value of a digit.
492 * The only portable way to do it ...
494 static int digit_value(utf32 const digit)
520 internal_error("wrong character given");
525 * Parses an octal character sequence.
527 * @param first_digit the already read first digit
529 static utf32 parse_octal_sequence(utf32 const first_digit)
531 assert(is_octal_digit(first_digit));
532 utf32 value = digit_value(first_digit);
533 if (!is_octal_digit(c)) return value;
534 value = 8 * value + digit_value(c);
536 if (!is_octal_digit(c)) return value;
537 value = 8 * value + digit_value(c);
543 * Parses a hex character sequence.
545 static utf32 parse_hex_sequence(void)
548 while (isxdigit(c)) {
549 value = 16 * value + digit_value(c);
556 * Parse an escape sequence.
558 static utf32 parse_escape_sequence(void)
566 case '"': return '"';
567 case '\'': return '\'';
568 case '\\': return '\\';
569 case '?': return '\?';
570 case 'a': return '\a';
571 case 'b': return '\b';
572 case 'f': return '\f';
573 case 'n': return '\n';
574 case 'r': return '\r';
575 case 't': return '\t';
576 case 'v': return '\v';
578 return parse_hex_sequence();
587 return parse_octal_sequence(ec);
589 parse_error("reached end of file while parsing escape sequence");
591 /* \E is not documented, but handled, by GCC. It is acceptable according
592 * to §6.11.4, whereas \e is not. */
596 return 27; /* hopefully 27 is ALWAYS the code for ESCAPE */
600 parse_error("universal character parsing not implemented yet");
605 /* §6.4.4.4:8 footnote 64 */
606 parse_error("unknown escape sequence");
611 * Concatenate two strings.
613 string_t concat_strings(const string_t *const s1, const string_t *const s2)
615 const size_t len1 = s1->size - 1;
616 const size_t len2 = s2->size - 1;
618 char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
619 memcpy(concat, s1->begin, len1);
620 memcpy(concat + len1, s2->begin, len2 + 1);
622 return identify_string(concat, len1 + len2 + 1);
625 string_t make_string(const char *string)
627 size_t len = strlen(string) + 1;
628 char *const space = obstack_alloc(&symbol_obstack, len);
629 memcpy(space, string, len);
631 return identify_string(space, len);
635 * Parse a string literal and set lexer_token.
637 static void parse_string_literal(void)
644 utf32 const tc = parse_escape_sequence();
646 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
648 obstack_1grow(&symbol_obstack, tc);
653 errorf(&lexer_token.base.source_position, "string has no end");
654 lexer_token.kind = T_ERROR;
663 obstack_grow_symbol(&symbol_obstack, c);
671 /* TODO: concatenate multiple strings separated by whitespace... */
673 /* add finishing 0 to the string */
674 obstack_1grow(&symbol_obstack, '\0');
675 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
676 char *string = obstack_finish(&symbol_obstack);
678 lexer_token.kind = T_STRING_LITERAL;
679 lexer_token.string.string = identify_string(string, size);
683 * Parse a wide character constant and set lexer_token.
685 static void parse_wide_character_constant(void)
692 const utf32 tc = parse_escape_sequence();
693 obstack_grow_symbol(&symbol_obstack, tc);
698 parse_error("newline while parsing character constant");
704 goto end_of_wide_char_constant;
707 errorf(&lexer_token.base.source_position,
708 "EOF while parsing character constant");
709 lexer_token.kind = T_ERROR;
714 obstack_grow_symbol(&symbol_obstack, c);
720 end_of_wide_char_constant:;
721 obstack_1grow(&symbol_obstack, '\0');
722 size_t size = (size_t) obstack_object_size(&symbol_obstack) - 1;
723 char *string = obstack_finish(&symbol_obstack);
725 lexer_token.kind = T_WIDE_CHARACTER_CONSTANT;
726 lexer_token.string.string = identify_string(string, size);
729 errorf(&lexer_token.base.source_position, "empty character constant");
734 * Parse a wide string literal and set lexer_token.
736 static void parse_wide_string_literal(void)
738 parse_string_literal();
739 if (lexer_token.kind == T_STRING_LITERAL)
740 lexer_token.kind = T_WIDE_STRING_LITERAL;
744 * Parse a character constant and set lexer_token.
746 static void parse_character_constant(void)
753 utf32 const tc = parse_escape_sequence();
755 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
757 obstack_1grow(&symbol_obstack, tc);
762 parse_error("newline while parsing character constant");
768 goto end_of_char_constant;
771 errorf(&lexer_token.base.source_position,
772 "EOF while parsing character constant");
773 lexer_token.kind = T_ERROR;
778 obstack_grow_symbol(&symbol_obstack, c);
785 end_of_char_constant:;
786 obstack_1grow(&symbol_obstack, '\0');
787 const size_t size = (size_t)obstack_object_size(&symbol_obstack)-1;
788 char *const string = obstack_finish(&symbol_obstack);
790 lexer_token.kind = T_CHARACTER_CONSTANT;
791 lexer_token.string.string = identify_string(string, size);
794 errorf(&lexer_token.base.source_position, "empty character constant");
799 * Skip a multiline comment.
801 static void skip_multiline_comment(void)
808 /* nested comment, warn here */
809 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
820 MATCH_NEWLINE(break;)
823 errorf(&lexer_token.base.source_position,
824 "at end of file while looking for comment end");
836 * Skip a single line comment.
838 static void skip_line_comment(void)
851 if (c == '\n' || c == '\r') {
852 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
864 /** The current preprocessor token. */
865 static token_t pp_token;
868 * Read the next preprocessor token.
870 static inline void next_pp_token(void)
872 lexer_next_preprocessing_token();
873 pp_token = lexer_token;
877 * Eat all preprocessor tokens until newline.
879 static void eat_until_newline(void)
881 while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
887 * Parse the line directive.
889 static void parse_line_directive(void)
891 if (pp_token.kind != T_INTEGER) {
892 parse_error("expected integer");
894 /* use offset -1 as this is about the next line */
895 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
898 if (pp_token.kind == T_STRING_LITERAL) {
899 lexer_pos.input_name = pp_token.string.string.begin;
900 lexer_pos.is_system_header = false;
903 /* attempt to parse numeric flags as outputted by gcc preprocessor */
904 while (pp_token.kind == T_INTEGER) {
906 * 1 - indicates start of a new file
907 * 2 - indicates return from a file
908 * 3 - indicates system header
909 * 4 - indicates implicit extern "C" in C++ mode
911 * currently we're only interested in "3"
913 if (streq(pp_token.number.number.begin, "3")) {
914 lexer_pos.is_system_header = true;
926 typedef enum stdc_pragma_kind_t {
930 STDC_CX_LIMITED_RANGE
931 } stdc_pragma_kind_t;
934 * STDC pragma values.
936 typedef enum stdc_pragma_value_kind_t {
941 } stdc_pragma_value_kind_t;
944 * Parse a pragma directive.
946 static void parse_pragma(void)
948 bool unknown_pragma = true;
951 if (pp_token.kind != T_IDENTIFIER) {
952 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
953 "expected identifier after #pragma");
958 symbol_t *symbol = pp_token.identifier.symbol;
959 if (symbol->pp_ID == TP_STDC) {
960 stdc_pragma_kind_t kind = STDC_UNKNOWN;
965 switch (pp_token.identifier.symbol->pp_ID) {
967 kind = STDC_FP_CONTRACT;
970 kind = STDC_FENV_ACCESS;
972 case TP_CX_LIMITED_RANGE:
973 kind = STDC_CX_LIMITED_RANGE;
978 if (kind != STDC_UNKNOWN) {
979 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
981 switch (pp_token.identifier.symbol->pp_ID) {
983 value = STDC_VALUE_ON;
986 value = STDC_VALUE_OFF;
989 value = STDC_VALUE_DEFAULT;
994 if (value != STDC_VALUE_UNKNOWN) {
995 unknown_pragma = false;
997 errorf(&pp_token.base.source_position,
998 "bad STDC pragma argument");
1003 unknown_pragma = true;
1005 eat_until_newline();
1006 if (unknown_pragma) {
1007 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
1008 "encountered unknown #pragma");
1013 * Parse a preprocessor non-null directive.
1015 static void parse_preprocessor_identifier(void)
1017 assert(pp_token.kind == T_IDENTIFIER);
1018 symbol_t *symbol = pp_token.identifier.symbol;
1020 switch (symbol->pp_ID) {
1023 parse_line_directive();
1029 /* TODO; output the rest of the line */
1030 parse_error("#error directive");
1036 * Parse a preprocessor directive.
1038 static void parse_preprocessor_directive(void)
1042 switch (pp_token.kind) {
1044 parse_preprocessor_identifier();
1047 parse_line_directive();
1050 /* NULL directive, see §6.10.7 */
1053 parse_error("invalid preprocessor directive");
1054 eat_until_newline();
1059 #define MAYBE_PROLOG \
1064 #define MAYBE(ch, set_type) \
1067 lexer_token.kind = set_type; \
1070 /* must use this as last thing */
1071 #define MAYBE_MODE(ch, set_type, mode) \
1073 if (c_mode & mode) { \
1075 lexer_token.kind = set_type; \
1080 #define ELSE_CODE(code) \
1085 } /* end of while (true) */ \
1087 #define ELSE(set_type) \
1089 lexer_token.kind = set_type; \
1092 void lexer_next_preprocessing_token(void)
1095 lexer_token.base.source_position = lexer_pos;
1104 lexer_token.kind = '\n';
1110 /* might be a wide string ( L"string" ) */
1111 if (lexer_token.identifier.symbol == symbol_L) {
1113 case '"': parse_wide_string_literal(); break;
1114 case '\'': parse_wide_character_constant(); break;
1124 parse_string_literal();
1128 parse_character_constant();
1141 MAYBE('.', T_DOTDOTDOT)
1145 lexer_token.kind = '.';
1150 MAYBE('&', T_ANDAND)
1151 MAYBE('=', T_ANDEQUAL)
1155 MAYBE('=', T_ASTERISKEQUAL)
1159 MAYBE('+', T_PLUSPLUS)
1160 MAYBE('=', T_PLUSEQUAL)
1164 MAYBE('>', T_MINUSGREATER)
1165 MAYBE('-', T_MINUSMINUS)
1166 MAYBE('=', T_MINUSEQUAL)
1170 MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1174 MAYBE('=', T_SLASHEQUAL)
1177 skip_multiline_comment();
1178 lexer_next_preprocessing_token();
1182 skip_line_comment();
1183 lexer_next_preprocessing_token();
1189 MAYBE('=', T_PERCENTEQUAL)
1194 MAYBE(':', T_HASHHASH)
1198 lexer_token.kind = '#';
1206 MAYBE('=', T_LESSEQUAL)
1209 MAYBE('=', T_LESSLESSEQUAL)
1214 MAYBE('=', T_GREATEREQUAL)
1217 MAYBE('=', T_GREATERGREATEREQUAL)
1218 ELSE(T_GREATERGREATER)
1222 MAYBE('=', T_CARETEQUAL)
1226 MAYBE('=', T_PIPEEQUAL)
1227 MAYBE('|', T_PIPEPIPE)
1232 MAYBE_MODE(':', T_COLONCOLON, _CXX)
1236 MAYBE('=', T_EQUALEQUAL)
1240 MAYBE('#', T_HASHHASH)
1254 lexer_token.kind = c;
1259 lexer_token.kind = T_EOF;
1264 errorf(&lexer_pos, "unknown character '%c' found", c);
1266 lexer_token.kind = T_ERROR;
1272 void lexer_next_token(void)
1274 lexer_next_preprocessing_token();
1276 while (lexer_token.kind == '\n') {
1278 lexer_next_preprocessing_token();
1281 if (lexer_token.kind == '#') {
1282 parse_preprocessor_directive();
1287 void init_lexer(void)
1289 strset_init(&stringset);
1290 symbol_L = symbol_table_insert("L");
1293 static void input_error(unsigned delta_lines, unsigned delta_cols,
1294 const char *message)
1296 lexer_pos.lineno += delta_lines;
1297 lexer_pos.colno += delta_cols;
1298 errorf(&lexer_pos, "%s", message);
1301 void lexer_switch_input(input_t *new_input, const char *input_name)
1303 lexer_pos.lineno = 0;
1304 lexer_pos.colno = 0;
1305 lexer_pos.input_name = input_name;
1307 set_input_error_callback(input_error);
1312 /* place a virtual \n at the beginning so the lexer knows that we're
1313 * at the beginning of a line */
1317 void exit_lexer(void)
1319 strset_destroy(&stringset);
1322 static __attribute__((unused))
1323 void dbg_pos(const source_position_t source_position)
1325 fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1326 source_position.lineno, (unsigned)source_position.colno);