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);
332 static void parse_exponent(void)
334 if (c == '-' || c == '+') {
335 obstack_1grow(&symbol_obstack, (char)c);
341 obstack_1grow(&symbol_obstack, (char)c);
343 } while (isdigit(c));
345 errorf(&lexer_token.base.source_position, "exponent has no digits");
350 * Parses a hex number including hex floats and set the
353 static void parse_number_hex(void)
355 bool is_float = false;
356 bool has_digits = false;
358 while (isxdigit(c)) {
360 obstack_1grow(&symbol_obstack, (char) c);
366 obstack_1grow(&symbol_obstack, (char) c);
369 while (isxdigit(c)) {
371 obstack_1grow(&symbol_obstack, (char) c);
375 if (c == 'p' || c == 'P') {
377 obstack_1grow(&symbol_obstack, (char) c);
380 } else if (is_float) {
381 errorf(&lexer_token.base.source_position,
382 "hexadecimal floatingpoint constant requires an exponent");
384 obstack_1grow(&symbol_obstack, '\0');
386 size_t size = obstack_object_size(&symbol_obstack) - 1;
387 char *string = obstack_finish(&symbol_obstack);
388 lexer_token.number.number = identify_string(string, size);
391 is_float ? T_FLOATINGPOINT_HEXADECIMAL : T_INTEGER_HEXADECIMAL;
394 errorf(&lexer_token.base.source_position, "invalid number literal '%S'", &lexer_token.number.number);
395 lexer_token.number.number.begin = "0";
396 lexer_token.number.number.size = 1;
399 parse_number_suffix();
403 * Returns true if the given char is a octal digit.
405 * @param char the character to check
407 static bool is_octal_digit(utf32 chr)
409 return '0' <= chr && chr <= '7';
413 * Parses a number and sets the lexer_token.
415 static void parse_number(void)
417 bool is_float = false;
418 bool has_digits = false;
420 assert(obstack_object_size(&symbol_obstack) == 0);
422 obstack_1grow(&symbol_obstack, (char)c);
424 if (c == 'x' || c == 'X') {
425 obstack_1grow(&symbol_obstack, (char)c);
436 obstack_1grow(&symbol_obstack, (char) c);
442 obstack_1grow(&symbol_obstack, '.');
447 obstack_1grow(&symbol_obstack, (char) c);
451 if (c == 'e' || c == 'E') {
453 obstack_1grow(&symbol_obstack, 'e');
458 obstack_1grow(&symbol_obstack, '\0');
459 size_t size = obstack_object_size(&symbol_obstack) - 1;
460 char *string = obstack_finish(&symbol_obstack);
461 lexer_token.number.number = identify_string(string, size);
463 /* is it an octal number? */
465 lexer_token.kind = T_FLOATINGPOINT;
466 } else if (string[0] == '0') {
467 lexer_token.kind = T_INTEGER_OCTAL;
469 /* check for invalid octal digits */
470 for (size_t i= 0; i < size; ++i) {
473 errorf(&lexer_token.base.source_position,
474 "invalid digit '%c' in octal number", t);
477 lexer_token.kind = T_INTEGER;
481 errorf(&lexer_token.base.source_position, "invalid number literal '%S'",
482 &lexer_token.number.number);
485 parse_number_suffix();
489 * Returns the value of a digit.
490 * The only portable way to do it ...
492 static int digit_value(utf32 const digit)
518 internal_error("wrong character given");
523 * Parses an octal character sequence.
525 * @param first_digit the already read first digit
527 static utf32 parse_octal_sequence(utf32 const first_digit)
529 assert(is_octal_digit(first_digit));
530 utf32 value = digit_value(first_digit);
531 if (!is_octal_digit(c)) return value;
532 value = 8 * value + digit_value(c);
534 if (!is_octal_digit(c)) return value;
535 value = 8 * value + digit_value(c);
541 * Parses a hex character sequence.
543 static utf32 parse_hex_sequence(void)
546 while (isxdigit(c)) {
547 value = 16 * value + digit_value(c);
554 * Parse an escape sequence.
556 static utf32 parse_escape_sequence(void)
564 case '"': return '"';
565 case '\'': return '\'';
566 case '\\': return '\\';
567 case '?': return '\?';
568 case 'a': return '\a';
569 case 'b': return '\b';
570 case 'f': return '\f';
571 case 'n': return '\n';
572 case 'r': return '\r';
573 case 't': return '\t';
574 case 'v': return '\v';
576 return parse_hex_sequence();
585 return parse_octal_sequence(ec);
587 parse_error("reached end of file while parsing escape sequence");
589 /* \E is not documented, but handled, by GCC. It is acceptable according
590 * to §6.11.4, whereas \e is not. */
594 return 27; /* hopefully 27 is ALWAYS the code for ESCAPE */
598 parse_error("universal character parsing not implemented yet");
603 /* §6.4.4.4:8 footnote 64 */
604 parse_error("unknown escape sequence");
609 * Concatenate two strings.
611 string_t concat_strings(const string_t *const s1, const string_t *const s2)
613 const size_t len1 = s1->size - 1;
614 const size_t len2 = s2->size - 1;
616 char *const concat = obstack_alloc(&symbol_obstack, len1 + len2 + 1);
617 memcpy(concat, s1->begin, len1);
618 memcpy(concat + len1, s2->begin, len2 + 1);
620 return identify_string(concat, len1 + len2 + 1);
623 string_t make_string(const char *string)
625 size_t len = strlen(string) + 1;
626 char *const space = obstack_alloc(&symbol_obstack, len);
627 memcpy(space, string, len);
629 return identify_string(space, len);
633 * Parse a string literal and set lexer_token.
635 static void parse_string_literal(void)
642 utf32 const tc = parse_escape_sequence();
644 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
646 obstack_1grow(&symbol_obstack, tc);
651 errorf(&lexer_token.base.source_position, "string has no end");
659 obstack_grow_symbol(&symbol_obstack, c);
667 /* TODO: concatenate multiple strings separated by whitespace... */
669 /* add finishing 0 to the string */
670 obstack_1grow(&symbol_obstack, '\0');
671 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
672 char *string = obstack_finish(&symbol_obstack);
674 lexer_token.kind = T_STRING_LITERAL;
675 lexer_token.string.string = identify_string(string, size);
679 * Parse a wide character constant and set lexer_token.
681 static void parse_wide_character_constant(void)
688 const utf32 tc = parse_escape_sequence();
689 obstack_grow_symbol(&symbol_obstack, tc);
694 parse_error("newline while parsing character constant");
700 goto end_of_wide_char_constant;
703 errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
704 goto end_of_wide_char_constant;
707 obstack_grow_symbol(&symbol_obstack, c);
713 end_of_wide_char_constant:;
714 obstack_1grow(&symbol_obstack, '\0');
715 size_t size = (size_t) obstack_object_size(&symbol_obstack) - 1;
716 char *string = obstack_finish(&symbol_obstack);
718 lexer_token.kind = T_WIDE_CHARACTER_CONSTANT;
719 lexer_token.string.string = identify_string(string, size);
722 errorf(&lexer_token.base.source_position, "empty character constant");
727 * Parse a wide string literal and set lexer_token.
729 static void parse_wide_string_literal(void)
731 parse_string_literal();
732 if (lexer_token.kind == T_STRING_LITERAL)
733 lexer_token.kind = T_WIDE_STRING_LITERAL;
737 * Parse a character constant and set lexer_token.
739 static void parse_character_constant(void)
746 utf32 const tc = parse_escape_sequence();
748 warningf(WARN_OTHER, &lexer_pos, "escape sequence out of range");
750 obstack_1grow(&symbol_obstack, tc);
755 parse_error("newline while parsing character constant");
761 goto end_of_char_constant;
764 errorf(&lexer_token.base.source_position, "EOF while parsing character constant");
765 goto end_of_char_constant;
768 obstack_grow_symbol(&symbol_obstack, c);
775 end_of_char_constant:;
776 obstack_1grow(&symbol_obstack, '\0');
777 const size_t size = (size_t)obstack_object_size(&symbol_obstack)-1;
778 char *const string = obstack_finish(&symbol_obstack);
780 lexer_token.kind = T_CHARACTER_CONSTANT;
781 lexer_token.string.string = identify_string(string, size);
784 errorf(&lexer_token.base.source_position, "empty character constant");
789 * Skip a multiline comment.
791 static void skip_multiline_comment(void)
798 /* nested comment, warn here */
799 warningf(WARN_COMMENT, &lexer_pos, "'/*' within comment");
810 MATCH_NEWLINE(break;)
813 errorf(&lexer_token.base.source_position,
814 "at end of file while looking for comment end");
826 * Skip a single line comment.
828 static void skip_line_comment(void)
841 if (c == '\n' || c == '\r') {
842 warningf(WARN_COMMENT, &lexer_pos, "multi-line comment");
854 /** The current preprocessor token. */
855 static token_t pp_token;
858 * Read the next preprocessor token.
860 static inline void next_pp_token(void)
862 lexer_next_preprocessing_token();
863 pp_token = lexer_token;
867 * Eat all preprocessor tokens until newline.
869 static void eat_until_newline(void)
871 while (pp_token.kind != '\n' && pp_token.kind != T_EOF) {
877 * Parse the line directive.
879 static void parse_line_directive(void)
881 if (pp_token.kind != T_INTEGER) {
882 parse_error("expected integer");
884 /* use offset -1 as this is about the next line */
885 lexer_pos.lineno = atoi(pp_token.number.number.begin) - 1;
888 if (pp_token.kind == T_STRING_LITERAL) {
889 lexer_pos.input_name = pp_token.string.string.begin;
890 lexer_pos.is_system_header = false;
893 /* attempt to parse numeric flags as outputted by gcc preprocessor */
894 while (pp_token.kind == T_INTEGER) {
896 * 1 - indicates start of a new file
897 * 2 - indicates return from a file
898 * 3 - indicates system header
899 * 4 - indicates implicit extern "C" in C++ mode
901 * currently we're only interested in "3"
903 if (streq(pp_token.number.number.begin, "3")) {
904 lexer_pos.is_system_header = true;
916 typedef enum stdc_pragma_kind_t {
920 STDC_CX_LIMITED_RANGE
921 } stdc_pragma_kind_t;
924 * STDC pragma values.
926 typedef enum stdc_pragma_value_kind_t {
931 } stdc_pragma_value_kind_t;
934 * Parse a pragma directive.
936 static void parse_pragma(void)
938 bool unknown_pragma = true;
941 if (pp_token.kind != T_IDENTIFIER) {
942 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
943 "expected identifier after #pragma");
948 symbol_t *symbol = pp_token.identifier.symbol;
949 if (symbol->pp_ID == TP_STDC) {
950 stdc_pragma_kind_t kind = STDC_UNKNOWN;
955 switch (pp_token.identifier.symbol->pp_ID) {
957 kind = STDC_FP_CONTRACT;
960 kind = STDC_FENV_ACCESS;
962 case TP_CX_LIMITED_RANGE:
963 kind = STDC_CX_LIMITED_RANGE;
968 if (kind != STDC_UNKNOWN) {
969 stdc_pragma_value_kind_t value = STDC_VALUE_UNKNOWN;
971 switch (pp_token.identifier.symbol->pp_ID) {
973 value = STDC_VALUE_ON;
976 value = STDC_VALUE_OFF;
979 value = STDC_VALUE_DEFAULT;
984 if (value != STDC_VALUE_UNKNOWN) {
985 unknown_pragma = false;
987 errorf(&pp_token.base.source_position,
988 "bad STDC pragma argument");
993 unknown_pragma = true;
996 if (unknown_pragma) {
997 warningf(WARN_UNKNOWN_PRAGMAS, &pp_token.base.source_position,
998 "encountered unknown #pragma");
1003 * Parse a preprocessor non-null directive.
1005 static void parse_preprocessor_identifier(void)
1007 assert(pp_token.kind == T_IDENTIFIER);
1008 symbol_t *symbol = pp_token.identifier.symbol;
1010 switch (symbol->pp_ID) {
1013 parse_line_directive();
1019 /* TODO; output the rest of the line */
1020 parse_error("#error directive");
1026 * Parse a preprocessor directive.
1028 static void parse_preprocessor_directive(void)
1032 switch (pp_token.kind) {
1034 parse_preprocessor_identifier();
1037 parse_line_directive();
1040 /* NULL directive, see §6.10.7 */
1043 parse_error("invalid preprocessor directive");
1044 eat_until_newline();
1049 #define MAYBE_PROLOG \
1054 #define MAYBE(ch, set_type) \
1057 lexer_token.kind = set_type; \
1060 /* must use this as last thing */
1061 #define MAYBE_MODE(ch, set_type, mode) \
1063 if (c_mode & mode) { \
1065 lexer_token.kind = set_type; \
1070 #define ELSE_CODE(code) \
1075 } /* end of while (true) */ \
1077 #define ELSE(set_type) \
1079 lexer_token.kind = set_type; \
1082 void lexer_next_preprocessing_token(void)
1085 lexer_token.base.source_position = lexer_pos;
1094 lexer_token.kind = '\n';
1100 /* might be a wide string ( L"string" ) */
1101 if (lexer_token.identifier.symbol == symbol_L) {
1103 case '"': parse_wide_string_literal(); break;
1104 case '\'': parse_wide_character_constant(); break;
1114 parse_string_literal();
1118 parse_character_constant();
1131 MAYBE('.', T_DOTDOTDOT)
1135 lexer_token.kind = '.';
1140 MAYBE('&', T_ANDAND)
1141 MAYBE('=', T_ANDEQUAL)
1145 MAYBE('=', T_ASTERISKEQUAL)
1149 MAYBE('+', T_PLUSPLUS)
1150 MAYBE('=', T_PLUSEQUAL)
1154 MAYBE('>', T_MINUSGREATER)
1155 MAYBE('-', T_MINUSMINUS)
1156 MAYBE('=', T_MINUSEQUAL)
1160 MAYBE('=', T_EXCLAMATIONMARKEQUAL)
1164 MAYBE('=', T_SLASHEQUAL)
1167 skip_multiline_comment();
1168 lexer_next_preprocessing_token();
1172 skip_line_comment();
1173 lexer_next_preprocessing_token();
1179 MAYBE('=', T_PERCENTEQUAL)
1184 MAYBE(':', T_HASHHASH)
1188 lexer_token.kind = '#';
1196 MAYBE('=', T_LESSEQUAL)
1199 MAYBE('=', T_LESSLESSEQUAL)
1204 MAYBE('=', T_GREATEREQUAL)
1207 MAYBE('=', T_GREATERGREATEREQUAL)
1208 ELSE(T_GREATERGREATER)
1212 MAYBE('=', T_CARETEQUAL)
1216 MAYBE('=', T_PIPEEQUAL)
1217 MAYBE('|', T_PIPEPIPE)
1222 MAYBE_MODE(':', T_COLONCOLON, _CXX)
1226 MAYBE('=', T_EQUALEQUAL)
1230 MAYBE('#', T_HASHHASH)
1244 lexer_token.kind = c;
1249 lexer_token.kind = T_EOF;
1254 errorf(&lexer_pos, "unknown character '%c' found", c);
1261 void lexer_next_token(void)
1263 lexer_next_preprocessing_token();
1265 while (lexer_token.kind == '\n') {
1267 lexer_next_preprocessing_token();
1270 if (lexer_token.kind == '#') {
1271 parse_preprocessor_directive();
1276 void init_lexer(void)
1278 strset_init(&stringset);
1279 symbol_L = symbol_table_insert("L");
1282 static void input_error(unsigned delta_lines, unsigned delta_cols,
1283 const char *message)
1285 lexer_pos.lineno += delta_lines;
1286 lexer_pos.colno += delta_cols;
1287 errorf(&lexer_pos, "%s", message);
1290 void lexer_switch_input(input_t *new_input, const char *input_name)
1292 lexer_pos.lineno = 0;
1293 lexer_pos.colno = 0;
1294 lexer_pos.input_name = input_name;
1296 set_input_error_callback(input_error);
1301 /* place a virtual \n at the beginning so the lexer knows that we're
1302 * at the beginning of a line */
1306 void exit_lexer(void)
1308 strset_destroy(&stringset);
1311 static __attribute__((unused))
1312 void dbg_pos(const source_position_t source_position)
1314 fprintf(stdout, "%s:%u:%u\n", source_position.input_name,
1315 source_position.lineno, (unsigned)source_position.colno);