12 #include "adt/error.h"
13 #include "adt/strutil.h"
14 #include "adt/strset.h"
15 #include "lang_features.h"
16 #include "diagnostic.h"
17 #include "string_rep.h"
21 #define INCLUDE_LIMIT 199 /* 199 is for gcc "compatibility" */
23 struct pp_argument_t {
28 struct pp_definition_t {
30 source_position_t source_position;
31 pp_definition_t *parent_expansion;
34 bool is_expanding : 1;
35 bool has_parameters : 1;
45 typedef struct pp_conditional_t pp_conditional_t;
46 struct pp_conditional_t {
47 source_position_t source_position;
50 bool skip; /**< conditional in skip mode (then+else gets skipped) */
51 pp_conditional_t *parent;
54 typedef struct pp_input_t pp_input_t;
59 utf32 buf[1024+MAX_PUTBACK];
62 source_position_t position;
67 /** additional info about the current token */
68 typedef struct add_token_info_t {
69 /** whitespace from beginning of line to the token */
71 /** there has been any whitespace before the token */
73 /** the token is at the beginning of the line */
77 typedef struct searchpath_entry_t searchpath_entry_t;
78 struct searchpath_entry_t {
80 searchpath_entry_t *next;
83 static pp_input_t input;
85 static pp_input_t *input_stack;
86 static unsigned n_inputs;
87 static struct obstack input_obstack;
89 static pp_conditional_t *conditional_stack;
91 static token_t pp_token;
92 static bool resolve_escape_sequences = false;
93 static bool ignore_unknown_chars = true;
94 static bool in_pp_directive;
95 static bool skip_mode;
97 static struct obstack pp_obstack;
98 static struct obstack config_obstack;
99 static const char *printed_input_name = NULL;
100 static source_position_t expansion_pos;
101 static pp_definition_t *current_expansion = NULL;
102 static strset_t stringset;
103 static preprocessor_token_kind_t last_token = TP_ERROR;
105 static searchpath_entry_t *searchpath;
107 static add_token_info_t info;
109 static inline void next_char(void);
110 static void next_preprocessing_token(void);
111 static void print_line_directive(const source_position_t *pos, const char *add);
113 static void switch_input(FILE *file, const char *filename)
116 input.input = input_from_stream(file, NULL);
119 input.output_line = 0;
120 input.position.input_name = filename;
121 input.position.lineno = 1;
123 /* indicate that we're at a new input */
124 print_line_directive(&input.position, input_stack != NULL ? "1" : NULL);
126 /* place a virtual '\n' so we realize we're at line begin */
127 input.position.lineno = 0;
129 next_preprocessing_token();
132 static void close_input(void)
134 input_free(input.input);
135 assert(input.file != NULL);
145 static void push_input(void)
147 pp_input_t *saved_input
148 = obstack_alloc(&input_obstack, sizeof(*saved_input));
150 memcpy(saved_input, &input, sizeof(*saved_input));
152 /* adjust buffer positions */
153 if (input.bufpos != NULL)
154 saved_input->bufpos = saved_input->buf + (input.bufpos - input.buf);
155 if (input.bufend != NULL)
156 saved_input->bufend = saved_input->buf + (input.bufend - input.buf);
158 saved_input->parent = input_stack;
159 input_stack = saved_input;
163 static void pop_restore_input(void)
165 assert(n_inputs > 0);
166 assert(input_stack != NULL);
168 pp_input_t *saved_input = input_stack;
170 memcpy(&input, saved_input, sizeof(input));
173 /* adjust buffer positions */
174 if (saved_input->bufpos != NULL)
175 input.bufpos = input.buf + (saved_input->bufpos - saved_input->buf);
176 if (saved_input->bufend != NULL)
177 input.bufend = input.buf + (saved_input->bufend - saved_input->buf);
179 input_stack = saved_input->parent;
180 obstack_free(&input_obstack, saved_input);
185 * Prints a parse error message at the current token.
187 * @param msg the error message
189 static void parse_error(const char *msg)
191 errorf(&pp_token.base.source_position, "%s", msg);
194 static inline void next_real_char(void)
196 assert(input.bufpos <= input.bufend);
197 if (input.bufpos >= input.bufend) {
198 size_t n = decode(input.input, input.buf + MAX_PUTBACK,
199 sizeof(input.buf)/sizeof(input.buf[0]) - MAX_PUTBACK);
204 input.bufpos = input.buf + MAX_PUTBACK;
205 input.bufend = input.bufpos + n;
207 input.c = *input.bufpos++;
208 ++input.position.colno;
212 * Put a character back into the buffer.
214 * @param pc the character to put back
216 static inline void put_back(utf32 const pc)
218 assert(input.bufpos > input.buf);
219 *(--input.bufpos - input.buf + input.buf) = (char) pc;
220 --input.position.colno;
223 #define MATCH_NEWLINE(code) \
226 if (input.c == '\n') { \
230 info.whitespace = 0; \
231 ++input.position.lineno; \
232 input.position.colno = 1; \
235 #define eat(c_type) (assert(input.c == c_type), next_char())
237 static void maybe_concat_lines(void)
255 * Set c to the next input character, ie.
256 * after expanding trigraphs.
258 static inline void next_char(void)
262 /* filter trigraphs and concatenated lines */
263 if (UNLIKELY(input.c == '\\')) {
264 maybe_concat_lines();
265 goto end_of_next_char;
268 if (LIKELY(input.c != '?'))
269 goto end_of_next_char;
272 if (LIKELY(input.c != '?')) {
275 goto end_of_next_char;
280 case '=': input.c = '#'; break;
281 case '(': input.c = '['; break;
282 case '/': input.c = '\\'; maybe_concat_lines(); break;
283 case ')': input.c = ']'; break;
284 case '\'': input.c = '^'; break;
285 case '<': input.c = '{'; break;
286 case '!': input.c = '|'; break;
287 case '>': input.c = '}'; break;
288 case '-': input.c = '~'; break;
298 printf("nchar '%c'\n", input.c);
305 * Returns true if the given char is a octal digit.
307 * @param char the character to check
309 static inline bool is_octal_digit(int chr)
327 * Returns the value of a digit.
328 * The only portable way to do it ...
330 static int digit_value(int digit)
356 panic("wrong character given");
361 * Parses an octal character sequence.
363 * @param first_digit the already read first digit
365 static int parse_octal_sequence(const int first_digit)
367 assert(is_octal_digit(first_digit));
368 int value = digit_value(first_digit);
369 if (!is_octal_digit(input.c)) return value;
370 value = 8 * value + digit_value(input.c);
372 if (!is_octal_digit(input.c)) return value;
373 value = 8 * value + digit_value(input.c);
376 if (char_is_signed) {
377 return (signed char) value;
379 return (unsigned char) value;
384 * Parses a hex character sequence.
386 static int parse_hex_sequence(void)
389 while (isxdigit(input.c)) {
390 value = 16 * value + digit_value(input.c);
394 if (char_is_signed) {
395 return (signed char) value;
397 return (unsigned char) value;
402 * Parse an escape sequence.
404 static int parse_escape_sequence(void)
412 case '"': return '"';
413 case '\'': return '\'';
414 case '\\': return '\\';
415 case '?': return '\?';
416 case 'a': return '\a';
417 case 'b': return '\b';
418 case 'f': return '\f';
419 case 'n': return '\n';
420 case 'r': return '\r';
421 case 't': return '\t';
422 case 'v': return '\v';
424 return parse_hex_sequence();
433 return parse_octal_sequence(ec);
435 parse_error("reached end of file while parsing escape sequence");
438 parse_error("unknown escape sequence");
443 static void grow_symbol(utf32 const tc)
445 struct obstack *const o = &symbol_obstack;
447 obstack_1grow(o, tc);
448 } else if (tc < 0x800) {
449 obstack_1grow(o, 0xC0 | (tc >> 6));
450 obstack_1grow(o, 0x80 | (tc & 0x3F));
451 } else if (tc < 0x10000) {
452 obstack_1grow(o, 0xE0 | ( tc >> 12));
453 obstack_1grow(o, 0x80 | ((tc >> 6) & 0x3F));
454 obstack_1grow(o, 0x80 | ( tc & 0x3F));
456 obstack_1grow(o, 0xF0 | ( tc >> 18));
457 obstack_1grow(o, 0x80 | ((tc >> 12) & 0x3F));
458 obstack_1grow(o, 0x80 | ((tc >> 6) & 0x3F));
459 obstack_1grow(o, 0x80 | ( tc & 0x3F));
463 static const char *identify_string(char *string)
465 const char *result = strset_insert(&stringset, string);
466 if (result != string) {
467 obstack_free(&symbol_obstack, string);
472 static string_t make_string(char *string, size_t len)
474 const char *result = identify_string(string);
475 return (string_t) {result, len};
478 static void parse_string_literal(void)
480 const unsigned start_linenr = input.position.lineno;
488 if (resolve_escape_sequences) {
489 tc = parse_escape_sequence();
490 obstack_1grow(&symbol_obstack, (char) tc);
492 obstack_1grow(&symbol_obstack, (char) input.c);
494 obstack_1grow(&symbol_obstack, (char) input.c);
501 source_position_t source_position;
502 source_position.input_name = pp_token.base.source_position.input_name;
503 source_position.lineno = start_linenr;
504 errorf(&source_position, "string has no end");
505 pp_token.kind = TP_ERROR;
514 grow_symbol(input.c);
521 /* add finishing 0 to the string */
522 obstack_1grow(&symbol_obstack, '\0');
523 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
524 char *const string = obstack_finish(&symbol_obstack);
526 pp_token.kind = TP_STRING_LITERAL;
527 pp_token.string.string = make_string(string, size);
531 * Parse a wide string literal and set lexer_token.
533 static void parse_wide_string_literal(void)
535 parse_string_literal();
536 if (pp_token.kind == TP_STRING_LITERAL)
537 pp_token.kind = TP_WIDE_STRING_LITERAL;
540 static void parse_wide_character_constant(void)
547 const utf32 tc = parse_escape_sequence();
553 parse_error("newline while parsing character constant");
559 goto end_of_wide_char_constant;
562 parse_error("EOF while parsing character constant");
563 pp_token.kind = TP_ERROR;
567 grow_symbol(input.c);
573 end_of_wide_char_constant:
574 obstack_1grow(&symbol_obstack, '\0');
575 size_t size = (size_t) obstack_object_size(&symbol_obstack)-1;
576 char *string = obstack_finish(&symbol_obstack);
577 pp_token.kind = TP_WIDE_CHARACTER_CONSTANT;
578 pp_token.string.string = make_string(string, size);
581 parse_error("empty character constant");
585 static void parse_character_constant(void)
587 const unsigned start_linenr = input.position.lineno;
595 tc = parse_escape_sequence();
596 obstack_1grow(&symbol_obstack, (char) tc);
600 parse_error("newline while parsing character constant");
605 source_position_t source_position;
606 source_position.input_name = pp_token.base.source_position.input_name;
607 source_position.lineno = start_linenr;
608 errorf(&source_position, "EOF while parsing character constant");
609 pp_token.kind = TP_ERROR;
615 goto end_of_char_constant;
618 obstack_1grow(&symbol_obstack, (char) input.c);
625 end_of_char_constant:;
626 obstack_1grow(&symbol_obstack, '\0');
627 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
628 char *const string = obstack_finish(&symbol_obstack);
630 pp_token.kind = TP_CHARACTER_CONSTANT;
631 pp_token.string.string = make_string(string, size);
634 parse_error("empty character constant");
638 #define SYMBOL_CHARS_WITHOUT_E_P \
689 #define SYMBOL_CHARS \
690 SYMBOL_CHARS_WITHOUT_E_P \
709 * returns next final token from a preprocessor macro expansion
711 static void expand_next(void)
713 assert(current_expansion != NULL);
715 pp_definition_t *definition = current_expansion;
718 if (definition->list_len == 0
719 || definition->expand_pos >= definition->list_len) {
720 /* we're finished with the current macro, move up 1 level in the
722 pp_definition_t *parent = definition->parent_expansion;
723 definition->parent_expansion = NULL;
724 definition->is_expanding = false;
726 /* it was the outermost expansion, parse normal pptoken */
727 if (parent == NULL) {
728 current_expansion = NULL;
729 next_preprocessing_token();
733 current_expansion = definition;
736 pp_token = definition->token_list[definition->expand_pos];
737 pp_token.base.source_position = expansion_pos;
738 ++definition->expand_pos;
740 if (pp_token.kind != TP_IDENTIFIER)
743 /* if it was an identifier then we might need to expand again */
744 pp_definition_t *symbol_definition = pp_token.identifier.symbol->pp_definition;
745 if (symbol_definition != NULL && !symbol_definition->is_expanding) {
746 symbol_definition->parent_expansion = definition;
747 symbol_definition->expand_pos = 0;
748 symbol_definition->is_expanding = true;
749 definition = symbol_definition;
750 current_expansion = definition;
755 static void skip_line_comment(void)
773 static void skip_multiline_comment(void)
775 unsigned start_linenr = input.position.lineno;
780 if (input.c == '*') {
781 /* TODO: nested comment, warn here */
786 if (input.c == '/') {
788 info.whitespace += input.position.colno-1;
794 info.at_line_begin |= !in_pp_directive;
799 source_position_t source_position;
800 source_position.input_name = pp_token.base.source_position.input_name;
801 source_position.lineno = start_linenr;
802 errorf(&source_position, "at end of file while looking for comment end");
813 static void skip_whitespace(void)
823 info.at_line_begin = true;
829 if (input.c == '/') {
833 } else if (input.c == '*') {
835 skip_multiline_comment();
848 static void eat_pp(int type)
851 assert(pp_token.kind == type);
852 next_preprocessing_token();
855 static void parse_symbol(void)
857 obstack_1grow(&symbol_obstack, (char) input.c);
864 obstack_1grow(&symbol_obstack, (char) input.c);
874 obstack_1grow(&symbol_obstack, '\0');
875 char *string = obstack_finish(&symbol_obstack);
877 /* might be a wide string or character constant ( L"string"/L'c' ) */
878 if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
879 obstack_free(&symbol_obstack, string);
880 parse_wide_string_literal();
882 } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
883 obstack_free(&symbol_obstack, string);
884 parse_wide_character_constant();
888 symbol_t *symbol = symbol_table_insert(string);
890 pp_token.kind = symbol->pp_ID;
891 pp_token.identifier.symbol = symbol;
893 /* we can free the memory from symbol obstack if we already had an entry in
894 * the symbol table */
895 if (symbol->string != string) {
896 obstack_free(&symbol_obstack, string);
900 static void parse_number(void)
902 obstack_1grow(&symbol_obstack, (char) input.c);
909 SYMBOL_CHARS_WITHOUT_E_P
910 obstack_1grow(&symbol_obstack, (char) input.c);
918 obstack_1grow(&symbol_obstack, (char) input.c);
920 if (input.c == '+' || input.c == '-') {
921 obstack_1grow(&symbol_obstack, (char) input.c);
932 obstack_1grow(&symbol_obstack, '\0');
933 size_t size = obstack_object_size(&symbol_obstack);
934 char *string = obstack_finish(&symbol_obstack);
936 pp_token.kind = TP_NUMBER;
937 pp_token.number.number = make_string(string, size);
941 #define MAYBE_PROLOG \
946 #define MAYBE(ch, set_type) \
949 pp_token.kind = set_type; \
952 #define ELSE_CODE(code) \
959 #define ELSE(set_type) \
961 pp_token.kind = set_type; \
964 static void next_preprocessing_token(void)
966 if (current_expansion != NULL) {
971 info.at_line_begin = false;
972 info.had_whitespace = false;
974 pp_token.base.source_position = input.position;
979 info.had_whitespace = true;
984 info.at_line_begin = true;
985 info.had_whitespace = true;
998 parse_string_literal();
1002 parse_character_constant();
1024 MAYBE('.', TP_DOTDOTDOT)
1028 pp_token.kind = '.';
1033 MAYBE('&', TP_ANDAND)
1034 MAYBE('=', TP_ANDEQUAL)
1038 MAYBE('=', TP_ASTERISKEQUAL)
1042 MAYBE('+', TP_PLUSPLUS)
1043 MAYBE('=', TP_PLUSEQUAL)
1047 MAYBE('>', TP_MINUSGREATER)
1048 MAYBE('-', TP_MINUSMINUS)
1049 MAYBE('=', TP_MINUSEQUAL)
1053 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
1057 MAYBE('=', TP_SLASHEQUAL)
1060 info.had_whitespace = true;
1061 skip_multiline_comment();
1065 info.had_whitespace = true;
1066 skip_line_comment();
1072 MAYBE('=', TP_PERCENTEQUAL)
1077 MAYBE(':', TP_HASHHASH)
1081 pp_token.kind = '#';
1089 MAYBE('=', TP_LESSEQUAL)
1092 MAYBE('=', TP_LESSLESSEQUAL)
1097 MAYBE('=', TP_GREATEREQUAL)
1100 MAYBE('=', TP_GREATERGREATEREQUAL)
1101 ELSE(TP_GREATERGREATER)
1105 MAYBE('=', TP_CARETEQUAL)
1109 MAYBE('=', TP_PIPEEQUAL)
1110 MAYBE('|', TP_PIPEPIPE)
1118 MAYBE('=', TP_EQUALEQUAL)
1122 MAYBE('#', TP_HASHHASH)
1124 pp_token.kind = '#';
1138 pp_token.kind = input.c;
1143 if (input_stack != NULL) {
1145 pop_restore_input();
1147 print_line_directive(&input.position, "2");
1150 pp_token.base.source_position.lineno++;
1151 info.at_line_begin = true;
1152 pp_token.kind = TP_EOF;
1158 if (!ignore_unknown_chars) {
1159 errorf(&pp_token.base.source_position,
1160 "unknown character '%c' found\n", input.c);
1161 pp_token.kind = TP_ERROR;
1163 pp_token.kind = input.c;
1169 static void print_quoted_string(const char *const string)
1172 for (const char *c = string; *c != 0; ++c) {
1174 case '"': fputs("\\\"", out); break;
1175 case '\\': fputs("\\\\", out); break;
1176 case '\a': fputs("\\a", out); break;
1177 case '\b': fputs("\\b", out); break;
1178 case '\f': fputs("\\f", out); break;
1179 case '\n': fputs("\\n", out); break;
1180 case '\r': fputs("\\r", out); break;
1181 case '\t': fputs("\\t", out); break;
1182 case '\v': fputs("\\v", out); break;
1183 case '\?': fputs("\\?", out); break;
1186 fprintf(out, "\\%03o", (unsigned)*c);
1196 static void print_line_directive(const source_position_t *pos, const char *add)
1198 fprintf(out, "# %u ", pos->lineno);
1199 print_quoted_string(pos->input_name);
1205 printed_input_name = pos->input_name;
1206 input.output_line = pos->lineno-1;
1209 static void emit_newlines(void)
1211 unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1215 print_line_directive(&pp_token.base.source_position, NULL);
1218 for (unsigned i = 0; i < delta; ++i) {
1222 input.output_line = pp_token.base.source_position.lineno;
1225 static void emit_pp_token(void)
1230 if (info.at_line_begin) {
1233 for (unsigned i = 0; i < info.whitespace; ++i)
1236 } else if (info.had_whitespace ||
1237 tokens_would_paste(last_token, pp_token.kind)) {
1241 switch (pp_token.kind) {
1243 fputs(pp_token.identifier.symbol->string, out);
1246 fputs(pp_token.number.number.begin, out);
1248 case TP_WIDE_STRING_LITERAL:
1250 case TP_STRING_LITERAL:
1252 fputs(pp_token.string.string.begin, out);
1255 case TP_WIDE_CHARACTER_CONSTANT:
1257 case TP_CHARACTER_CONSTANT:
1259 fputs(pp_token.string.string.begin, out);
1263 print_pp_token_kind(out, pp_token.kind);
1266 last_token = pp_token.kind;
1269 static void eat_pp_directive(void)
1271 while (!info.at_line_begin) {
1272 next_preprocessing_token();
1276 static bool strings_equal(const string_t *string1, const string_t *string2)
1278 size_t size = string1->size;
1279 if (size != string2->size)
1282 const char *c1 = string1->begin;
1283 const char *c2 = string2->begin;
1284 for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1291 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1293 if (token1->kind != token2->kind)
1296 switch (token1->kind) {
1298 return token1->identifier.symbol == token2->identifier.symbol;
1300 case TP_CHARACTER_CONSTANT:
1301 case TP_STRING_LITERAL:
1302 return strings_equal(&token1->string.string, &token2->string.string);
1309 static bool pp_definitions_equal(const pp_definition_t *definition1,
1310 const pp_definition_t *definition2)
1312 if (definition1->list_len != definition2->list_len)
1315 size_t len = definition1->list_len;
1316 const token_t *t1 = definition1->token_list;
1317 const token_t *t2 = definition2->token_list;
1318 for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1319 if (!pp_tokens_equal(t1, t2))
1325 static void parse_define_directive(void)
1328 assert(obstack_object_size(&pp_obstack) == 0);
1330 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1331 errorf(&pp_token.base.source_position,
1332 "expected identifier after #define, got '%t'", &pp_token);
1335 symbol_t *symbol = pp_token.identifier.symbol;
1337 pp_definition_t *new_definition
1338 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1339 memset(new_definition, 0, sizeof(new_definition[0]));
1340 new_definition->source_position = input.position;
1342 /* this is probably the only place where spaces are significant in the
1343 * lexer (except for the fact that they separate tokens). #define b(x)
1344 * is something else than #define b (x) */
1345 if (input.c == '(') {
1347 next_preprocessing_token();
1348 /* get next token after '(' */
1349 next_preprocessing_token();
1352 switch (pp_token.kind) {
1354 new_definition->is_variadic = true;
1355 next_preprocessing_token();
1356 if (pp_token.kind != ')') {
1357 errorf(&input.position,
1358 "'...' not at end of macro argument list");
1363 obstack_ptr_grow(&pp_obstack, pp_token.identifier.symbol);
1364 next_preprocessing_token();
1366 if (pp_token.kind == ',') {
1367 next_preprocessing_token();
1371 if (pp_token.kind != ')') {
1372 errorf(&pp_token.base.source_position,
1373 "expected ',' or ')' after identifier, got '%t'",
1379 next_preprocessing_token();
1380 goto finish_argument_list;
1382 errorf(&pp_token.base.source_position,
1383 "expected identifier, '...' or ')' in #define argument list, got '%t'",
1389 finish_argument_list:
1390 new_definition->has_parameters = true;
1391 new_definition->n_parameters
1392 = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1393 new_definition->parameters = obstack_finish(&pp_obstack);
1395 next_preprocessing_token();
1398 /* construct a new pp_definition on the obstack */
1399 assert(obstack_object_size(&pp_obstack) == 0);
1400 size_t list_len = 0;
1401 while (!info.at_line_begin) {
1402 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1404 next_preprocessing_token();
1407 new_definition->list_len = list_len;
1408 new_definition->token_list = obstack_finish(&pp_obstack);
1410 pp_definition_t *old_definition = symbol->pp_definition;
1411 if (old_definition != NULL) {
1412 if (!pp_definitions_equal(old_definition, new_definition)) {
1413 warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1415 /* reuse the old definition */
1416 obstack_free(&pp_obstack, new_definition);
1417 new_definition = old_definition;
1421 symbol->pp_definition = new_definition;
1425 if (obstack_object_size(&pp_obstack) > 0) {
1426 char *ptr = obstack_finish(&pp_obstack);
1427 obstack_free(&pp_obstack, ptr);
1432 static void parse_undef_directive(void)
1436 if (pp_token.kind != TP_IDENTIFIER) {
1437 errorf(&input.position,
1438 "expected identifier after #undef, got '%t'", &pp_token);
1443 symbol_t *symbol = pp_token.identifier.symbol;
1444 symbol->pp_definition = NULL;
1445 next_preprocessing_token();
1447 if (!info.at_line_begin) {
1448 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1453 static const char *parse_headername(void)
1455 /* behind an #include we can have the special headername lexems.
1456 * They're only allowed behind an #include so they're not recognized
1457 * by the normal next_preprocessing_token. We handle them as a special
1459 if (info.at_line_begin) {
1460 parse_error("expected headername after #include");
1464 assert(obstack_object_size(&symbol_obstack) == 0);
1466 /* check wether we have a "... or <... headername */
1475 parse_error("header name without closing '>'");
1480 goto finished_headername;
1482 obstack_1grow(&symbol_obstack, (char) input.c);
1485 /* we should never be here */
1494 parse_error("header name without closing '>'");
1499 goto finished_headername;
1501 obstack_1grow(&symbol_obstack, (char) input.c);
1504 /* we should never be here */
1507 /* TODO: do normal pp_token parsing and concatenate results */
1508 panic("pp_token concat include not implemented yet");
1511 finished_headername:
1512 obstack_1grow(&symbol_obstack, '\0');
1513 char *headername = obstack_finish(&symbol_obstack);
1515 /* TODO: iterate search-path to find the file */
1519 return identify_string(headername);
1522 static bool do_include(bool system_include, const char *headername)
1524 if (!system_include) {
1525 /* for "bla" includes first try current dir
1526 * TODO: this isn't correct, should be the directory of the source file
1528 FILE *file = fopen(headername, "r");
1530 switch_input(file, headername);
1535 size_t headername_len = strlen(headername);
1536 assert(obstack_object_size(&pp_obstack) == 0);
1537 /* check searchpath */
1538 for (searchpath_entry_t *entry = searchpath; entry != NULL;
1539 entry = entry->next) {
1540 const char *path = entry->path;
1541 size_t len = strlen(path);
1542 obstack_grow(&pp_obstack, path, len);
1543 if (path[len-1] != '/')
1544 obstack_1grow(&pp_obstack, '/');
1545 obstack_grow(&pp_obstack, headername, headername_len+1);
1547 char *complete_path = obstack_finish(&pp_obstack);
1548 FILE *file = fopen(complete_path, "r");
1550 const char *filename = identify_string(complete_path);
1551 switch_input(file, filename);
1554 obstack_free(&pp_obstack, complete_path);
1560 static bool parse_include_directive(void)
1562 /* don't eat the TP_include here!
1563 * we need an alternative parsing for the next token */
1565 bool system_include = input.c == '<';
1566 const char *headername = parse_headername();
1567 if (headername == NULL) {
1572 if (!info.at_line_begin) {
1573 warningf(WARN_OTHER, &pp_token.base.source_position,
1574 "extra tokens at end of #include directive");
1578 if (n_inputs > INCLUDE_LIMIT) {
1579 errorf(&pp_token.base.source_position, "#include nested too deeply");
1581 next_preprocessing_token();
1585 /* we have to reenable space counting and macro expansion here,
1586 * because it is still disabled in directive parsing,
1587 * but we will trigger a preprocessing token reading of the new file
1588 * now and need expansions/space counting */
1589 in_pp_directive = false;
1594 bool res = do_include(system_include, headername);
1596 errorf(&pp_token.base.source_position,
1597 "failed including '%s': %s", headername, strerror(errno));
1598 pop_restore_input();
1605 static pp_conditional_t *push_conditional(void)
1607 pp_conditional_t *conditional
1608 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1609 memset(conditional, 0, sizeof(*conditional));
1611 conditional->parent = conditional_stack;
1612 conditional_stack = conditional;
1617 static void pop_conditional(void)
1619 assert(conditional_stack != NULL);
1620 conditional_stack = conditional_stack->parent;
1623 static void check_unclosed_conditionals(void)
1625 while (conditional_stack != NULL) {
1626 pp_conditional_t *conditional = conditional_stack;
1628 if (conditional->in_else) {
1629 errorf(&conditional->source_position, "unterminated #else");
1631 errorf(&conditional->source_position, "unterminated condition");
1637 static void parse_ifdef_ifndef_directive(void)
1639 bool is_ifndef = (pp_token.kind == TP_ifndef);
1641 next_preprocessing_token();
1645 pp_conditional_t *conditional = push_conditional();
1646 conditional->source_position = pp_token.base.source_position;
1647 conditional->skip = true;
1651 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1652 errorf(&pp_token.base.source_position,
1653 "expected identifier after #%s, got '%t'",
1654 is_ifndef ? "ifndef" : "ifdef", &pp_token);
1657 /* just take the true case in the hope to avoid further errors */
1660 symbol_t *symbol = pp_token.identifier.symbol;
1661 pp_definition_t *pp_definition = symbol->pp_definition;
1662 next_preprocessing_token();
1664 if (!info.at_line_begin) {
1665 errorf(&pp_token.base.source_position,
1666 "extra tokens at end of #%s",
1667 is_ifndef ? "ifndef" : "ifdef");
1671 /* evaluate wether we are in true or false case */
1672 condition = is_ifndef ? pp_definition == NULL : pp_definition != NULL;
1675 pp_conditional_t *conditional = push_conditional();
1676 conditional->source_position = pp_token.base.source_position;
1677 conditional->condition = condition;
1684 static void parse_else_directive(void)
1688 if (!info.at_line_begin) {
1690 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1695 pp_conditional_t *conditional = conditional_stack;
1696 if (conditional == NULL) {
1697 errorf(&pp_token.base.source_position, "#else without prior #if");
1701 if (conditional->in_else) {
1702 errorf(&pp_token.base.source_position,
1703 "#else after #else (condition started %P)",
1704 conditional->source_position);
1709 conditional->in_else = true;
1710 if (!conditional->skip) {
1711 skip_mode = conditional->condition;
1713 conditional->source_position = pp_token.base.source_position;
1716 static void parse_endif_directive(void)
1720 if (!info.at_line_begin) {
1722 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1727 pp_conditional_t *conditional = conditional_stack;
1728 if (conditional == NULL) {
1729 errorf(&pp_token.base.source_position, "#endif without prior #if");
1733 if (!conditional->skip) {
1739 static void parse_preprocessing_directive(void)
1741 in_pp_directive = true;
1745 switch (pp_token.kind) {
1748 parse_ifdef_ifndef_directive();
1751 parse_else_directive();
1754 parse_endif_directive();
1761 switch (pp_token.kind) {
1763 parse_define_directive();
1766 parse_undef_directive();
1770 parse_ifdef_ifndef_directive();
1773 parse_else_directive();
1776 parse_endif_directive();
1779 parse_include_directive();
1782 if (info.at_line_begin) {
1783 /* the nop directive "#" */
1786 errorf(&pp_token.base.source_position,
1787 "invalid preprocessing directive #%t", &pp_token);
1793 in_pp_directive = false;
1794 assert(info.at_line_begin);
1797 static void prepend_include_path(const char *path)
1799 searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1801 entry->next = searchpath;
1805 static void setup_include_path(void)
1807 /* built-in paths */
1808 prepend_include_path("/usr/include");
1810 /* parse environment variable */
1811 const char *cpath = getenv("CPATH");
1812 if (cpath != NULL && *cpath != '\0') {
1813 const char *begin = cpath;
1817 while (*c != '\0' && *c != ':')
1820 size_t len = c-begin;
1822 /* for gcc compatibility (Matze: I would expect that
1823 * nothing happens for an empty entry...) */
1824 prepend_include_path(".");
1826 char *string = obstack_alloc(&config_obstack, len+1);
1827 memcpy(string, begin, len);
1830 prepend_include_path(string);
1837 } while(*c != '\0');
1841 int pptest_main(int argc, char **argv);
1842 int pptest_main(int argc, char **argv)
1844 init_symbol_table();
1847 obstack_init(&config_obstack);
1848 obstack_init(&pp_obstack);
1849 obstack_init(&input_obstack);
1850 strset_init(&stringset);
1852 setup_include_path();
1854 /* simplistic commandline parser */
1855 const char *filename = NULL;
1856 for (int i = 1; i < argc; ++i) {
1857 const char *opt = argv[i];
1858 if (streq(opt, "-I")) {
1859 prepend_include_path(argv[++i]);
1861 } else if (streq(opt, "-E")) {
1863 } else if (opt[0] == '-') {
1864 fprintf(stderr, "Unknown option '%s'\n", opt);
1866 if (filename != NULL)
1867 fprintf(stderr, "Multiple inputs not supported\n");
1871 if (filename == NULL) {
1872 fprintf(stderr, "No input specified\n");
1878 /* just here for gcc compatibility */
1879 fprintf(out, "# 1 \"%s\"\n", filename);
1880 fprintf(out, "# 1 \"<built-in>\"\n");
1881 fprintf(out, "# 1 \"<command-line>\"\n");
1883 FILE *file = fopen(filename, "r");
1885 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1888 switch_input(file, filename);
1891 if (pp_token.kind == '#' && info.at_line_begin) {
1892 parse_preprocessing_directive();
1894 } else if (pp_token.kind == TP_EOF) {
1895 goto end_of_main_loop;
1896 } else if (pp_token.kind == TP_IDENTIFIER && !in_pp_directive) {
1897 symbol_t *symbol = pp_token.identifier.symbol;
1898 pp_definition_t *pp_definition = symbol->pp_definition;
1899 if (pp_definition != NULL && !pp_definition->is_expanding) {
1900 expansion_pos = pp_token.base.source_position;
1901 if (pp_definition->has_parameters) {
1902 source_position_t position = pp_token.base.source_position;
1903 add_token_info_t old_info = info;
1904 next_preprocessing_token();
1905 add_token_info_t new_info = info;
1907 /* no opening brace -> no expansion */
1908 if (pp_token.kind == '(') {
1911 /* parse arguments (TODO) */
1912 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1913 next_preprocessing_token();
1915 token_t next_token = pp_token;
1916 /* restore identifier token */
1917 pp_token.kind = TP_IDENTIFIER;
1918 pp_token.identifier.symbol = symbol;
1919 pp_token.base.source_position = position;
1924 pp_token = next_token;
1929 pp_definition->expand_pos = 0;
1930 pp_definition->is_expanding = true;
1931 current_expansion = pp_definition;
1938 next_preprocessing_token();
1943 check_unclosed_conditionals();
1946 obstack_free(&input_obstack, NULL);
1947 obstack_free(&pp_obstack, NULL);
1948 obstack_free(&config_obstack, NULL);
1950 strset_destroy(&stringset);
1953 exit_symbol_table();