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 utf32 parse_octal_sequence(const utf32 first_digit)
367 assert(is_octal_digit(first_digit));
368 utf32 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);
380 * Parses a hex character sequence.
382 static utf32 parse_hex_sequence(void)
385 while (isxdigit(input.c)) {
386 value = 16 * value + digit_value(input.c);
393 * Parse an escape sequence.
395 static utf32 parse_escape_sequence(void)
399 utf32 const ec = input.c;
403 case '"': return '"';
404 case '\'': return '\'';
405 case '\\': return '\\';
406 case '?': return '\?';
407 case 'a': return '\a';
408 case 'b': return '\b';
409 case 'f': return '\f';
410 case 'n': return '\n';
411 case 'r': return '\r';
412 case 't': return '\t';
413 case 'v': return '\v';
415 return parse_hex_sequence();
424 return parse_octal_sequence(ec);
426 parse_error("reached end of file while parsing escape sequence");
428 /* \E is not documented, but handled, by GCC. It is acceptable according
429 * to §6.11.4, whereas \e is not. */
433 return 27; /* hopefully 27 is ALWAYS the code for ESCAPE */
437 parse_error("universal character parsing not implemented yet");
442 /* §6.4.4.4:8 footnote 64 */
443 parse_error("unknown escape sequence");
447 static const char *identify_string(char *string)
449 const char *result = strset_insert(&stringset, string);
450 if (result != string) {
451 obstack_free(&symbol_obstack, string);
456 static string_t make_string(char *string, size_t len)
458 const char *result = identify_string(string);
459 return (string_t) {result, len};
462 static void parse_string_literal(void)
464 const unsigned start_linenr = input.position.lineno;
472 if (resolve_escape_sequences) {
473 tc = parse_escape_sequence();
474 obstack_1grow(&symbol_obstack, (char) tc);
476 obstack_1grow(&symbol_obstack, (char) input.c);
478 obstack_1grow(&symbol_obstack, (char) input.c);
485 source_position_t source_position;
486 source_position.input_name = pp_token.base.source_position.input_name;
487 source_position.lineno = start_linenr;
488 errorf(&source_position, "string has no end");
489 pp_token.kind = TP_ERROR;
498 obstack_grow_symbol(&symbol_obstack, input.c);
505 /* add finishing 0 to the string */
506 obstack_1grow(&symbol_obstack, '\0');
507 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
508 char *const string = obstack_finish(&symbol_obstack);
510 pp_token.kind = TP_STRING_LITERAL;
511 pp_token.string.string = make_string(string, size);
515 * Parse a wide string literal and set lexer_token.
517 static void parse_wide_string_literal(void)
519 parse_string_literal();
520 if (pp_token.kind == TP_STRING_LITERAL)
521 pp_token.kind = TP_WIDE_STRING_LITERAL;
524 static void parse_wide_character_constant(void)
531 const utf32 tc = parse_escape_sequence();
532 obstack_grow_symbol(&symbol_obstack, tc);
537 parse_error("newline while parsing character constant");
543 goto end_of_wide_char_constant;
546 parse_error("EOF while parsing character constant");
547 pp_token.kind = TP_ERROR;
551 obstack_grow_symbol(&symbol_obstack, input.c);
557 end_of_wide_char_constant:
558 obstack_1grow(&symbol_obstack, '\0');
559 size_t size = (size_t) obstack_object_size(&symbol_obstack)-1;
560 char *string = obstack_finish(&symbol_obstack);
561 pp_token.kind = TP_WIDE_CHARACTER_CONSTANT;
562 pp_token.string.string = make_string(string, size);
565 parse_error("empty character constant");
569 static void parse_character_constant(void)
571 const unsigned start_linenr = input.position.lineno;
579 tc = parse_escape_sequence();
580 obstack_1grow(&symbol_obstack, (char) tc);
584 parse_error("newline while parsing character constant");
589 source_position_t source_position;
590 source_position.input_name = pp_token.base.source_position.input_name;
591 source_position.lineno = start_linenr;
592 errorf(&source_position, "EOF while parsing character constant");
593 pp_token.kind = TP_ERROR;
599 goto end_of_char_constant;
602 obstack_1grow(&symbol_obstack, (char) input.c);
609 end_of_char_constant:;
610 obstack_1grow(&symbol_obstack, '\0');
611 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
612 char *const string = obstack_finish(&symbol_obstack);
614 pp_token.kind = TP_CHARACTER_CONSTANT;
615 pp_token.string.string = make_string(string, size);
618 parse_error("empty character constant");
622 #define SYMBOL_CHARS_WITHOUT_E_P \
673 #define SYMBOL_CHARS \
674 SYMBOL_CHARS_WITHOUT_E_P \
693 * returns next final token from a preprocessor macro expansion
695 static void expand_next(void)
697 assert(current_expansion != NULL);
699 pp_definition_t *definition = current_expansion;
702 if (definition->list_len == 0
703 || definition->expand_pos >= definition->list_len) {
704 /* we're finished with the current macro, move up 1 level in the
706 pp_definition_t *parent = definition->parent_expansion;
707 definition->parent_expansion = NULL;
708 definition->is_expanding = false;
710 /* it was the outermost expansion, parse normal pptoken */
711 if (parent == NULL) {
712 current_expansion = NULL;
713 next_preprocessing_token();
717 current_expansion = definition;
720 pp_token = definition->token_list[definition->expand_pos];
721 pp_token.base.source_position = expansion_pos;
722 ++definition->expand_pos;
724 if (pp_token.kind != TP_IDENTIFIER)
727 /* if it was an identifier then we might need to expand again */
728 pp_definition_t *symbol_definition = pp_token.identifier.symbol->pp_definition;
729 if (symbol_definition != NULL && !symbol_definition->is_expanding) {
730 symbol_definition->parent_expansion = definition;
731 symbol_definition->expand_pos = 0;
732 symbol_definition->is_expanding = true;
733 definition = symbol_definition;
734 current_expansion = definition;
739 static void skip_line_comment(void)
757 static void skip_multiline_comment(void)
759 unsigned start_linenr = input.position.lineno;
764 if (input.c == '*') {
765 /* TODO: nested comment, warn here */
770 if (input.c == '/') {
772 info.whitespace += input.position.colno-1;
778 info.at_line_begin |= !in_pp_directive;
783 source_position_t source_position;
784 source_position.input_name = pp_token.base.source_position.input_name;
785 source_position.lineno = start_linenr;
786 errorf(&source_position, "at end of file while looking for comment end");
797 static void skip_whitespace(void)
807 info.at_line_begin = true;
813 if (input.c == '/') {
817 } else if (input.c == '*') {
819 skip_multiline_comment();
832 static void eat_pp(int type)
835 assert(pp_token.kind == type);
836 next_preprocessing_token();
839 static void parse_symbol(void)
841 obstack_1grow(&symbol_obstack, (char) input.c);
848 obstack_1grow(&symbol_obstack, (char) input.c);
858 obstack_1grow(&symbol_obstack, '\0');
859 char *string = obstack_finish(&symbol_obstack);
861 /* might be a wide string or character constant ( L"string"/L'c' ) */
862 if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
863 obstack_free(&symbol_obstack, string);
864 parse_wide_string_literal();
866 } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
867 obstack_free(&symbol_obstack, string);
868 parse_wide_character_constant();
872 symbol_t *symbol = symbol_table_insert(string);
874 pp_token.kind = symbol->pp_ID;
875 pp_token.identifier.symbol = symbol;
877 /* we can free the memory from symbol obstack if we already had an entry in
878 * the symbol table */
879 if (symbol->string != string) {
880 obstack_free(&symbol_obstack, string);
884 static void parse_number(void)
886 obstack_1grow(&symbol_obstack, (char) input.c);
893 SYMBOL_CHARS_WITHOUT_E_P
894 obstack_1grow(&symbol_obstack, (char) input.c);
902 obstack_1grow(&symbol_obstack, (char) input.c);
904 if (input.c == '+' || input.c == '-') {
905 obstack_1grow(&symbol_obstack, (char) input.c);
916 obstack_1grow(&symbol_obstack, '\0');
917 size_t size = obstack_object_size(&symbol_obstack);
918 char *string = obstack_finish(&symbol_obstack);
920 pp_token.kind = TP_NUMBER;
921 pp_token.number.number = make_string(string, size);
925 #define MAYBE_PROLOG \
930 #define MAYBE(ch, set_type) \
933 pp_token.kind = set_type; \
936 #define ELSE_CODE(code) \
943 #define ELSE(set_type) \
945 pp_token.kind = set_type; \
948 static void next_preprocessing_token(void)
950 if (current_expansion != NULL) {
955 info.at_line_begin = false;
956 info.had_whitespace = false;
958 pp_token.base.source_position = input.position;
963 info.had_whitespace = true;
968 info.at_line_begin = true;
969 info.had_whitespace = true;
982 parse_string_literal();
986 parse_character_constant();
1008 MAYBE('.', TP_DOTDOTDOT)
1012 pp_token.kind = '.';
1017 MAYBE('&', TP_ANDAND)
1018 MAYBE('=', TP_ANDEQUAL)
1022 MAYBE('=', TP_ASTERISKEQUAL)
1026 MAYBE('+', TP_PLUSPLUS)
1027 MAYBE('=', TP_PLUSEQUAL)
1031 MAYBE('>', TP_MINUSGREATER)
1032 MAYBE('-', TP_MINUSMINUS)
1033 MAYBE('=', TP_MINUSEQUAL)
1037 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
1041 MAYBE('=', TP_SLASHEQUAL)
1044 info.had_whitespace = true;
1045 skip_multiline_comment();
1049 info.had_whitespace = true;
1050 skip_line_comment();
1056 MAYBE('=', TP_PERCENTEQUAL)
1061 MAYBE(':', TP_HASHHASH)
1065 pp_token.kind = '#';
1073 MAYBE('=', TP_LESSEQUAL)
1076 MAYBE('=', TP_LESSLESSEQUAL)
1081 MAYBE('=', TP_GREATEREQUAL)
1084 MAYBE('=', TP_GREATERGREATEREQUAL)
1085 ELSE(TP_GREATERGREATER)
1089 MAYBE('=', TP_CARETEQUAL)
1093 MAYBE('=', TP_PIPEEQUAL)
1094 MAYBE('|', TP_PIPEPIPE)
1102 MAYBE('=', TP_EQUALEQUAL)
1106 MAYBE('#', TP_HASHHASH)
1108 pp_token.kind = '#';
1122 pp_token.kind = input.c;
1127 if (input_stack != NULL) {
1129 pop_restore_input();
1131 print_line_directive(&input.position, "2");
1134 pp_token.base.source_position.lineno++;
1135 info.at_line_begin = true;
1136 pp_token.kind = TP_EOF;
1142 if (!ignore_unknown_chars) {
1143 errorf(&pp_token.base.source_position,
1144 "unknown character '%c' found\n", input.c);
1145 pp_token.kind = TP_ERROR;
1147 pp_token.kind = input.c;
1153 static void print_quoted_string(const char *const string)
1156 for (const char *c = string; *c != 0; ++c) {
1158 case '"': fputs("\\\"", out); break;
1159 case '\\': fputs("\\\\", out); break;
1160 case '\a': fputs("\\a", out); break;
1161 case '\b': fputs("\\b", out); break;
1162 case '\f': fputs("\\f", out); break;
1163 case '\n': fputs("\\n", out); break;
1164 case '\r': fputs("\\r", out); break;
1165 case '\t': fputs("\\t", out); break;
1166 case '\v': fputs("\\v", out); break;
1167 case '\?': fputs("\\?", out); break;
1170 fprintf(out, "\\%03o", (unsigned)*c);
1180 static void print_line_directive(const source_position_t *pos, const char *add)
1182 fprintf(out, "# %u ", pos->lineno);
1183 print_quoted_string(pos->input_name);
1189 printed_input_name = pos->input_name;
1190 input.output_line = pos->lineno-1;
1193 static void emit_newlines(void)
1195 unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1199 print_line_directive(&pp_token.base.source_position, NULL);
1202 for (unsigned i = 0; i < delta; ++i) {
1206 input.output_line = pp_token.base.source_position.lineno;
1209 static void emit_pp_token(void)
1214 if (info.at_line_begin) {
1217 for (unsigned i = 0; i < info.whitespace; ++i)
1220 } else if (info.had_whitespace ||
1221 tokens_would_paste(last_token, pp_token.kind)) {
1225 switch (pp_token.kind) {
1227 fputs(pp_token.identifier.symbol->string, out);
1230 fputs(pp_token.number.number.begin, out);
1232 case TP_WIDE_STRING_LITERAL:
1234 case TP_STRING_LITERAL:
1236 fputs(pp_token.string.string.begin, out);
1239 case TP_WIDE_CHARACTER_CONSTANT:
1241 case TP_CHARACTER_CONSTANT:
1243 fputs(pp_token.string.string.begin, out);
1247 print_pp_token_kind(out, pp_token.kind);
1250 last_token = pp_token.kind;
1253 static void eat_pp_directive(void)
1255 while (!info.at_line_begin) {
1256 next_preprocessing_token();
1260 static bool strings_equal(const string_t *string1, const string_t *string2)
1262 size_t size = string1->size;
1263 if (size != string2->size)
1266 const char *c1 = string1->begin;
1267 const char *c2 = string2->begin;
1268 for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1275 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1277 if (token1->kind != token2->kind)
1280 switch (token1->kind) {
1282 return token1->identifier.symbol == token2->identifier.symbol;
1284 case TP_CHARACTER_CONSTANT:
1285 case TP_STRING_LITERAL:
1286 return strings_equal(&token1->string.string, &token2->string.string);
1293 static bool pp_definitions_equal(const pp_definition_t *definition1,
1294 const pp_definition_t *definition2)
1296 if (definition1->list_len != definition2->list_len)
1299 size_t len = definition1->list_len;
1300 const token_t *t1 = definition1->token_list;
1301 const token_t *t2 = definition2->token_list;
1302 for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1303 if (!pp_tokens_equal(t1, t2))
1309 static void parse_define_directive(void)
1312 assert(obstack_object_size(&pp_obstack) == 0);
1314 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1315 errorf(&pp_token.base.source_position,
1316 "expected identifier after #define, got '%t'", &pp_token);
1319 symbol_t *symbol = pp_token.identifier.symbol;
1321 pp_definition_t *new_definition
1322 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1323 memset(new_definition, 0, sizeof(new_definition[0]));
1324 new_definition->source_position = input.position;
1326 /* this is probably the only place where spaces are significant in the
1327 * lexer (except for the fact that they separate tokens). #define b(x)
1328 * is something else than #define b (x) */
1329 if (input.c == '(') {
1331 next_preprocessing_token();
1332 /* get next token after '(' */
1333 next_preprocessing_token();
1336 switch (pp_token.kind) {
1338 new_definition->is_variadic = true;
1339 next_preprocessing_token();
1340 if (pp_token.kind != ')') {
1341 errorf(&input.position,
1342 "'...' not at end of macro argument list");
1347 obstack_ptr_grow(&pp_obstack, pp_token.identifier.symbol);
1348 next_preprocessing_token();
1350 if (pp_token.kind == ',') {
1351 next_preprocessing_token();
1355 if (pp_token.kind != ')') {
1356 errorf(&pp_token.base.source_position,
1357 "expected ',' or ')' after identifier, got '%t'",
1363 next_preprocessing_token();
1364 goto finish_argument_list;
1366 errorf(&pp_token.base.source_position,
1367 "expected identifier, '...' or ')' in #define argument list, got '%t'",
1373 finish_argument_list:
1374 new_definition->has_parameters = true;
1375 new_definition->n_parameters
1376 = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1377 new_definition->parameters = obstack_finish(&pp_obstack);
1379 next_preprocessing_token();
1382 /* construct a new pp_definition on the obstack */
1383 assert(obstack_object_size(&pp_obstack) == 0);
1384 size_t list_len = 0;
1385 while (!info.at_line_begin) {
1386 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1388 next_preprocessing_token();
1391 new_definition->list_len = list_len;
1392 new_definition->token_list = obstack_finish(&pp_obstack);
1394 pp_definition_t *old_definition = symbol->pp_definition;
1395 if (old_definition != NULL) {
1396 if (!pp_definitions_equal(old_definition, new_definition)) {
1397 warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1399 /* reuse the old definition */
1400 obstack_free(&pp_obstack, new_definition);
1401 new_definition = old_definition;
1405 symbol->pp_definition = new_definition;
1409 if (obstack_object_size(&pp_obstack) > 0) {
1410 char *ptr = obstack_finish(&pp_obstack);
1411 obstack_free(&pp_obstack, ptr);
1416 static void parse_undef_directive(void)
1420 if (pp_token.kind != TP_IDENTIFIER) {
1421 errorf(&input.position,
1422 "expected identifier after #undef, got '%t'", &pp_token);
1427 symbol_t *symbol = pp_token.identifier.symbol;
1428 symbol->pp_definition = NULL;
1429 next_preprocessing_token();
1431 if (!info.at_line_begin) {
1432 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1437 static const char *parse_headername(void)
1439 /* behind an #include we can have the special headername lexems.
1440 * They're only allowed behind an #include so they're not recognized
1441 * by the normal next_preprocessing_token. We handle them as a special
1443 if (info.at_line_begin) {
1444 parse_error("expected headername after #include");
1448 assert(obstack_object_size(&symbol_obstack) == 0);
1450 /* check wether we have a "... or <... headername */
1459 parse_error("header name without closing '>'");
1464 goto finished_headername;
1466 obstack_1grow(&symbol_obstack, (char) input.c);
1469 /* we should never be here */
1478 parse_error("header name without closing '>'");
1483 goto finished_headername;
1485 obstack_1grow(&symbol_obstack, (char) input.c);
1488 /* we should never be here */
1491 /* TODO: do normal pp_token parsing and concatenate results */
1492 panic("pp_token concat include not implemented yet");
1495 finished_headername:
1496 obstack_1grow(&symbol_obstack, '\0');
1497 char *headername = obstack_finish(&symbol_obstack);
1499 /* TODO: iterate search-path to find the file */
1503 return identify_string(headername);
1506 static bool do_include(bool system_include, const char *headername)
1508 if (!system_include) {
1509 /* for "bla" includes first try current dir
1510 * TODO: this isn't correct, should be the directory of the source file
1512 FILE *file = fopen(headername, "r");
1514 switch_input(file, headername);
1519 size_t headername_len = strlen(headername);
1520 assert(obstack_object_size(&pp_obstack) == 0);
1521 /* check searchpath */
1522 for (searchpath_entry_t *entry = searchpath; entry != NULL;
1523 entry = entry->next) {
1524 const char *path = entry->path;
1525 size_t len = strlen(path);
1526 obstack_grow(&pp_obstack, path, len);
1527 if (path[len-1] != '/')
1528 obstack_1grow(&pp_obstack, '/');
1529 obstack_grow(&pp_obstack, headername, headername_len+1);
1531 char *complete_path = obstack_finish(&pp_obstack);
1532 FILE *file = fopen(complete_path, "r");
1534 const char *filename = identify_string(complete_path);
1535 switch_input(file, filename);
1538 obstack_free(&pp_obstack, complete_path);
1544 static bool parse_include_directive(void)
1546 /* don't eat the TP_include here!
1547 * we need an alternative parsing for the next token */
1549 bool system_include = input.c == '<';
1550 const char *headername = parse_headername();
1551 if (headername == NULL) {
1556 if (!info.at_line_begin) {
1557 warningf(WARN_OTHER, &pp_token.base.source_position,
1558 "extra tokens at end of #include directive");
1562 if (n_inputs > INCLUDE_LIMIT) {
1563 errorf(&pp_token.base.source_position, "#include nested too deeply");
1565 next_preprocessing_token();
1569 /* we have to reenable space counting and macro expansion here,
1570 * because it is still disabled in directive parsing,
1571 * but we will trigger a preprocessing token reading of the new file
1572 * now and need expansions/space counting */
1573 in_pp_directive = false;
1578 bool res = do_include(system_include, headername);
1580 errorf(&pp_token.base.source_position,
1581 "failed including '%s': %s", headername, strerror(errno));
1582 pop_restore_input();
1589 static pp_conditional_t *push_conditional(void)
1591 pp_conditional_t *conditional
1592 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1593 memset(conditional, 0, sizeof(*conditional));
1595 conditional->parent = conditional_stack;
1596 conditional_stack = conditional;
1601 static void pop_conditional(void)
1603 assert(conditional_stack != NULL);
1604 conditional_stack = conditional_stack->parent;
1607 static void check_unclosed_conditionals(void)
1609 while (conditional_stack != NULL) {
1610 pp_conditional_t *conditional = conditional_stack;
1612 if (conditional->in_else) {
1613 errorf(&conditional->source_position, "unterminated #else");
1615 errorf(&conditional->source_position, "unterminated condition");
1621 static void parse_ifdef_ifndef_directive(void)
1623 bool is_ifndef = (pp_token.kind == TP_ifndef);
1625 next_preprocessing_token();
1629 pp_conditional_t *conditional = push_conditional();
1630 conditional->source_position = pp_token.base.source_position;
1631 conditional->skip = true;
1635 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1636 errorf(&pp_token.base.source_position,
1637 "expected identifier after #%s, got '%t'",
1638 is_ifndef ? "ifndef" : "ifdef", &pp_token);
1641 /* just take the true case in the hope to avoid further errors */
1644 symbol_t *symbol = pp_token.identifier.symbol;
1645 pp_definition_t *pp_definition = symbol->pp_definition;
1646 next_preprocessing_token();
1648 if (!info.at_line_begin) {
1649 errorf(&pp_token.base.source_position,
1650 "extra tokens at end of #%s",
1651 is_ifndef ? "ifndef" : "ifdef");
1655 /* evaluate wether we are in true or false case */
1656 condition = is_ifndef ? pp_definition == NULL : pp_definition != NULL;
1659 pp_conditional_t *conditional = push_conditional();
1660 conditional->source_position = pp_token.base.source_position;
1661 conditional->condition = condition;
1668 static void parse_else_directive(void)
1672 if (!info.at_line_begin) {
1674 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1679 pp_conditional_t *conditional = conditional_stack;
1680 if (conditional == NULL) {
1681 errorf(&pp_token.base.source_position, "#else without prior #if");
1685 if (conditional->in_else) {
1686 errorf(&pp_token.base.source_position,
1687 "#else after #else (condition started %P)",
1688 conditional->source_position);
1693 conditional->in_else = true;
1694 if (!conditional->skip) {
1695 skip_mode = conditional->condition;
1697 conditional->source_position = pp_token.base.source_position;
1700 static void parse_endif_directive(void)
1704 if (!info.at_line_begin) {
1706 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1711 pp_conditional_t *conditional = conditional_stack;
1712 if (conditional == NULL) {
1713 errorf(&pp_token.base.source_position, "#endif without prior #if");
1717 if (!conditional->skip) {
1723 static void parse_preprocessing_directive(void)
1725 in_pp_directive = true;
1729 switch (pp_token.kind) {
1732 parse_ifdef_ifndef_directive();
1735 parse_else_directive();
1738 parse_endif_directive();
1745 switch (pp_token.kind) {
1747 parse_define_directive();
1750 parse_undef_directive();
1754 parse_ifdef_ifndef_directive();
1757 parse_else_directive();
1760 parse_endif_directive();
1763 parse_include_directive();
1766 if (info.at_line_begin) {
1767 /* the nop directive "#" */
1770 errorf(&pp_token.base.source_position,
1771 "invalid preprocessing directive #%t", &pp_token);
1777 in_pp_directive = false;
1778 assert(info.at_line_begin);
1781 static void prepend_include_path(const char *path)
1783 searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1785 entry->next = searchpath;
1789 static void setup_include_path(void)
1791 /* built-in paths */
1792 prepend_include_path("/usr/include");
1794 /* parse environment variable */
1795 const char *cpath = getenv("CPATH");
1796 if (cpath != NULL && *cpath != '\0') {
1797 const char *begin = cpath;
1801 while (*c != '\0' && *c != ':')
1804 size_t len = c-begin;
1806 /* for gcc compatibility (Matze: I would expect that
1807 * nothing happens for an empty entry...) */
1808 prepend_include_path(".");
1810 char *string = obstack_alloc(&config_obstack, len+1);
1811 memcpy(string, begin, len);
1814 prepend_include_path(string);
1821 } while(*c != '\0');
1825 int pptest_main(int argc, char **argv);
1826 int pptest_main(int argc, char **argv)
1828 init_symbol_table();
1831 obstack_init(&config_obstack);
1832 obstack_init(&pp_obstack);
1833 obstack_init(&input_obstack);
1834 strset_init(&stringset);
1836 setup_include_path();
1838 /* simplistic commandline parser */
1839 const char *filename = NULL;
1840 for (int i = 1; i < argc; ++i) {
1841 const char *opt = argv[i];
1842 if (streq(opt, "-I")) {
1843 prepend_include_path(argv[++i]);
1845 } else if (streq(opt, "-E")) {
1847 } else if (opt[0] == '-') {
1848 fprintf(stderr, "Unknown option '%s'\n", opt);
1850 if (filename != NULL)
1851 fprintf(stderr, "Multiple inputs not supported\n");
1855 if (filename == NULL) {
1856 fprintf(stderr, "No input specified\n");
1862 /* just here for gcc compatibility */
1863 fprintf(out, "# 1 \"%s\"\n", filename);
1864 fprintf(out, "# 1 \"<built-in>\"\n");
1865 fprintf(out, "# 1 \"<command-line>\"\n");
1867 FILE *file = fopen(filename, "r");
1869 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1872 switch_input(file, filename);
1875 if (pp_token.kind == '#' && info.at_line_begin) {
1876 parse_preprocessing_directive();
1878 } else if (pp_token.kind == TP_EOF) {
1879 goto end_of_main_loop;
1880 } else if (pp_token.kind == TP_IDENTIFIER && !in_pp_directive) {
1881 symbol_t *symbol = pp_token.identifier.symbol;
1882 pp_definition_t *pp_definition = symbol->pp_definition;
1883 if (pp_definition != NULL && !pp_definition->is_expanding) {
1884 expansion_pos = pp_token.base.source_position;
1885 if (pp_definition->has_parameters) {
1886 source_position_t position = pp_token.base.source_position;
1887 add_token_info_t old_info = info;
1888 next_preprocessing_token();
1889 add_token_info_t new_info = info;
1891 /* no opening brace -> no expansion */
1892 if (pp_token.kind == '(') {
1895 /* parse arguments (TODO) */
1896 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1897 next_preprocessing_token();
1899 token_t next_token = pp_token;
1900 /* restore identifier token */
1901 pp_token.kind = TP_IDENTIFIER;
1902 pp_token.identifier.symbol = symbol;
1903 pp_token.base.source_position = position;
1908 pp_token = next_token;
1913 pp_definition->expand_pos = 0;
1914 pp_definition->is_expanding = true;
1915 current_expansion = pp_definition;
1922 next_preprocessing_token();
1927 check_unclosed_conditionals();
1930 obstack_free(&input_obstack, NULL);
1931 obstack_free(&pp_obstack, NULL);
1932 obstack_free(&config_obstack, NULL);
1934 strset_destroy(&stringset);
1937 exit_symbol_table();