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 skip_mode;
96 static struct obstack pp_obstack;
97 static struct obstack config_obstack;
98 static const char *printed_input_name = NULL;
99 static source_position_t expansion_pos;
100 static pp_definition_t *current_expansion = NULL;
101 static strset_t stringset;
102 static preprocessor_token_kind_t last_token;
104 static searchpath_entry_t *searchpath;
106 static add_token_info_t info;
108 static inline void next_char(void);
109 static void next_preprocessing_token(void);
110 static void print_line_directive(const source_position_t *pos, const char *add);
112 static void switch_input(FILE *file, const char *filename)
115 input.input = input_from_stream(file, NULL);
118 input.output_line = 0;
119 input.position.input_name = filename;
120 input.position.lineno = 1;
122 /* indicate that we're at a new input */
123 print_line_directive(&input.position, input_stack != NULL ? "1" : NULL);
125 /* place a virtual '\n' so we realize we're at line begin */
126 input.position.lineno = 0;
128 next_preprocessing_token();
131 static void close_input(void)
133 input_free(input.input);
134 assert(input.file != NULL);
144 static void push_input(void)
146 pp_input_t *saved_input
147 = obstack_alloc(&input_obstack, sizeof(*saved_input));
149 memcpy(saved_input, &input, sizeof(*saved_input));
151 /* adjust buffer positions */
152 if (input.bufpos != NULL)
153 saved_input->bufpos = saved_input->buf + (input.bufpos - input.buf);
154 if (input.bufend != NULL)
155 saved_input->bufend = saved_input->buf + (input.bufend - input.buf);
157 saved_input->parent = input_stack;
158 input_stack = saved_input;
162 static void pop_restore_input(void)
164 assert(n_inputs > 0);
165 assert(input_stack != NULL);
167 pp_input_t *saved_input = input_stack;
169 memcpy(&input, saved_input, sizeof(input));
172 /* adjust buffer positions */
173 if (saved_input->bufpos != NULL)
174 input.bufpos = input.buf + (saved_input->bufpos - saved_input->buf);
175 if (saved_input->bufend != NULL)
176 input.bufend = input.buf + (saved_input->bufend - saved_input->buf);
178 input_stack = saved_input->parent;
179 obstack_free(&input_obstack, saved_input);
184 * Prints a parse error message at the current token.
186 * @param msg the error message
188 static void parse_error(const char *msg)
190 errorf(&pp_token.base.source_position, "%s", msg);
193 static inline void next_real_char(void)
195 assert(input.bufpos <= input.bufend);
196 if (input.bufpos >= input.bufend) {
197 size_t const n = decode(input.input, input.buf + MAX_PUTBACK, lengthof(input.buf) - MAX_PUTBACK);
202 input.bufpos = input.buf + MAX_PUTBACK;
203 input.bufend = input.bufpos + n;
205 input.c = *input.bufpos++;
206 ++input.position.colno;
210 * Put a character back into the buffer.
212 * @param pc the character to put back
214 static inline void put_back(utf32 const pc)
216 assert(input.bufpos > input.buf);
217 *(--input.bufpos - input.buf + input.buf) = (char) pc;
218 --input.position.colno;
221 #define MATCH_NEWLINE(code) \
224 if (input.c == '\n') { \
228 info.whitespace = 0; \
229 ++input.position.lineno; \
230 input.position.colno = 1; \
233 #define eat(c_type) (assert(input.c == c_type), next_char())
235 static void maybe_concat_lines(void)
253 * Set c to the next input character, ie.
254 * after expanding trigraphs.
256 static inline void next_char(void)
260 /* filter trigraphs and concatenated lines */
261 if (UNLIKELY(input.c == '\\')) {
262 maybe_concat_lines();
263 goto end_of_next_char;
266 if (LIKELY(input.c != '?'))
267 goto end_of_next_char;
270 if (LIKELY(input.c != '?')) {
273 goto end_of_next_char;
278 case '=': input.c = '#'; break;
279 case '(': input.c = '['; break;
280 case '/': input.c = '\\'; maybe_concat_lines(); break;
281 case ')': input.c = ']'; break;
282 case '\'': input.c = '^'; break;
283 case '<': input.c = '{'; break;
284 case '!': input.c = '|'; break;
285 case '>': input.c = '}'; break;
286 case '-': input.c = '~'; break;
296 printf("nchar '%c'\n", input.c);
303 * Returns true if the given char is a octal digit.
305 * @param char the character to check
307 static inline bool is_octal_digit(int chr)
325 * Returns the value of a digit.
326 * The only portable way to do it ...
328 static int digit_value(int digit)
354 panic("wrong character given");
359 * Parses an octal character sequence.
361 * @param first_digit the already read first digit
363 static utf32 parse_octal_sequence(const utf32 first_digit)
365 assert(is_octal_digit(first_digit));
366 utf32 value = digit_value(first_digit);
367 if (!is_octal_digit(input.c)) return value;
368 value = 8 * value + digit_value(input.c);
370 if (!is_octal_digit(input.c)) return value;
371 value = 8 * value + digit_value(input.c);
378 * Parses a hex character sequence.
380 static utf32 parse_hex_sequence(void)
383 while (isxdigit(input.c)) {
384 value = 16 * value + digit_value(input.c);
391 * Parse an escape sequence.
393 static utf32 parse_escape_sequence(void)
397 utf32 const ec = input.c;
401 case '"': return '"';
402 case '\'': return '\'';
403 case '\\': return '\\';
404 case '?': return '\?';
405 case 'a': return '\a';
406 case 'b': return '\b';
407 case 'f': return '\f';
408 case 'n': return '\n';
409 case 'r': return '\r';
410 case 't': return '\t';
411 case 'v': return '\v';
413 return parse_hex_sequence();
422 return parse_octal_sequence(ec);
424 parse_error("reached end of file while parsing escape sequence");
426 /* \E is not documented, but handled, by GCC. It is acceptable according
427 * to §6.11.4, whereas \e is not. */
431 return 27; /* hopefully 27 is ALWAYS the code for ESCAPE */
435 parse_error("universal character parsing not implemented yet");
440 /* §6.4.4.4:8 footnote 64 */
441 parse_error("unknown escape sequence");
445 static const char *identify_string(char *string)
447 const char *result = strset_insert(&stringset, string);
448 if (result != string) {
449 obstack_free(&symbol_obstack, string);
454 static string_t make_string(char *string, size_t len)
456 const char *result = identify_string(string);
457 return (string_t) {result, len};
460 static void parse_string_literal(string_encoding_t const enc)
462 const unsigned start_linenr = input.position.lineno;
470 if (resolve_escape_sequences) {
471 tc = parse_escape_sequence();
472 obstack_1grow(&symbol_obstack, (char) tc);
474 obstack_1grow(&symbol_obstack, (char) input.c);
476 obstack_1grow(&symbol_obstack, (char) input.c);
483 source_position_t source_position;
484 source_position.input_name = pp_token.base.source_position.input_name;
485 source_position.lineno = start_linenr;
486 errorf(&source_position, "string has no end");
495 obstack_grow_symbol(&symbol_obstack, input.c);
502 /* add finishing 0 to the string */
503 obstack_1grow(&symbol_obstack, '\0');
504 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
505 char *const string = obstack_finish(&symbol_obstack);
507 pp_token.kind = TP_STRING_LITERAL;
508 pp_token.string.encoding = enc;
509 pp_token.string.string = make_string(string, size);
512 static void parse_character_constant(string_encoding_t const enc)
514 const unsigned start_linenr = input.position.lineno;
521 utf32 const tc = parse_escape_sequence();
522 if (enc == STRING_ENCODING_CHAR) {
524 warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
526 obstack_1grow(&symbol_obstack, tc);
528 obstack_grow_symbol(&symbol_obstack, tc);
534 parse_error("newline while parsing character constant");
539 source_position_t source_position;
540 source_position.input_name = pp_token.base.source_position.input_name;
541 source_position.lineno = start_linenr;
542 errorf(&source_position, "EOF while parsing character constant");
543 goto end_of_char_constant;
548 goto end_of_char_constant;
551 obstack_grow_symbol(&symbol_obstack, input.c);
558 end_of_char_constant:;
559 obstack_1grow(&symbol_obstack, '\0');
560 size_t const size = obstack_object_size(&symbol_obstack) - 1;
561 char *const string = obstack_finish(&symbol_obstack);
563 pp_token.kind = TP_CHARACTER_CONSTANT;
564 pp_token.string.encoding = enc;
565 pp_token.string.string = make_string(string, size);
568 parse_error("empty character constant");
572 #define SYMBOL_CHARS_WITHOUT_E_P \
623 #define SYMBOL_CHARS \
624 SYMBOL_CHARS_WITHOUT_E_P \
643 * returns next final token from a preprocessor macro expansion
645 static void expand_next(void)
647 assert(current_expansion != NULL);
649 pp_definition_t *definition = current_expansion;
652 if (definition->list_len == 0
653 || definition->expand_pos >= definition->list_len) {
654 /* we're finished with the current macro, move up 1 level in the
656 pp_definition_t *parent = definition->parent_expansion;
657 definition->parent_expansion = NULL;
658 definition->is_expanding = false;
660 /* it was the outermost expansion, parse normal pptoken */
661 if (parent == NULL) {
662 current_expansion = NULL;
663 next_preprocessing_token();
667 current_expansion = definition;
670 pp_token = definition->token_list[definition->expand_pos];
671 pp_token.base.source_position = expansion_pos;
672 ++definition->expand_pos;
674 if (pp_token.kind != TP_IDENTIFIER)
677 /* if it was an identifier then we might need to expand again */
678 pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
679 if (symbol_definition != NULL && !symbol_definition->is_expanding) {
680 symbol_definition->parent_expansion = definition;
681 symbol_definition->expand_pos = 0;
682 symbol_definition->is_expanding = true;
683 definition = symbol_definition;
684 current_expansion = definition;
689 static void skip_line_comment(void)
707 static void skip_multiline_comment(void)
709 unsigned start_linenr = input.position.lineno;
714 if (input.c == '*') {
715 /* TODO: nested comment, warn here */
720 if (input.c == '/') {
721 if (input.position.lineno != input.output_line)
722 info.whitespace = input.position.colno;
733 source_position_t source_position;
734 source_position.input_name = pp_token.base.source_position.input_name;
735 source_position.lineno = start_linenr;
736 errorf(&source_position, "at end of file while looking for comment end");
747 static void skip_whitespace(void)
757 info.at_line_begin = true;
763 if (input.c == '/') {
767 } else if (input.c == '*') {
769 skip_multiline_comment();
782 static void eat_pp(preprocessor_token_kind_t const type)
785 assert(pp_token.kind == type);
786 next_preprocessing_token();
789 static void parse_symbol(void)
791 obstack_1grow(&symbol_obstack, (char) input.c);
798 obstack_1grow(&symbol_obstack, (char) input.c);
808 obstack_1grow(&symbol_obstack, '\0');
809 char *string = obstack_finish(&symbol_obstack);
811 /* might be a wide string or character constant ( L"string"/L'c' ) */
812 if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
813 obstack_free(&symbol_obstack, string);
814 parse_string_literal(STRING_ENCODING_WIDE);
816 } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
817 obstack_free(&symbol_obstack, string);
818 parse_character_constant(STRING_ENCODING_WIDE);
822 symbol_t *symbol = symbol_table_insert(string);
824 pp_token.kind = symbol->pp_ID;
825 pp_token.base.symbol = symbol;
827 /* we can free the memory from symbol obstack if we already had an entry in
828 * the symbol table */
829 if (symbol->string != string) {
830 obstack_free(&symbol_obstack, string);
834 static void parse_number(void)
836 obstack_1grow(&symbol_obstack, (char) input.c);
843 SYMBOL_CHARS_WITHOUT_E_P
844 obstack_1grow(&symbol_obstack, (char) input.c);
852 obstack_1grow(&symbol_obstack, (char) input.c);
854 if (input.c == '+' || input.c == '-') {
855 obstack_1grow(&symbol_obstack, (char) input.c);
866 obstack_1grow(&symbol_obstack, '\0');
867 size_t size = obstack_object_size(&symbol_obstack);
868 char *string = obstack_finish(&symbol_obstack);
870 pp_token.kind = TP_NUMBER;
871 pp_token.number.number = make_string(string, size);
875 #define MAYBE_PROLOG \
880 #define MAYBE(ch, set_type) \
883 pp_token.kind = set_type; \
886 #define ELSE_CODE(code) \
893 #define ELSE(set_type) \
895 pp_token.kind = set_type; \
898 static void next_preprocessing_token(void)
900 if (current_expansion != NULL) {
905 info.at_line_begin = false;
906 info.had_whitespace = false;
908 pp_token.base.source_position = input.position;
909 pp_token.base.symbol = NULL;
915 info.had_whitespace = true;
920 info.at_line_begin = true;
921 info.had_whitespace = true;
934 parse_string_literal(STRING_ENCODING_CHAR);
938 parse_character_constant(STRING_ENCODING_CHAR);
960 MAYBE('.', TP_DOTDOTDOT)
969 MAYBE('&', TP_ANDAND)
970 MAYBE('=', TP_ANDEQUAL)
974 MAYBE('=', TP_ASTERISKEQUAL)
978 MAYBE('+', TP_PLUSPLUS)
979 MAYBE('=', TP_PLUSEQUAL)
983 MAYBE('>', TP_MINUSGREATER)
984 MAYBE('-', TP_MINUSMINUS)
985 MAYBE('=', TP_MINUSEQUAL)
989 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
993 MAYBE('=', TP_SLASHEQUAL)
996 info.had_whitespace = true;
997 skip_multiline_comment();
1001 info.had_whitespace = true;
1002 skip_line_comment();
1008 MAYBE('=', TP_PERCENTEQUAL)
1013 MAYBE(':', TP_HASHHASH)
1017 pp_token.kind = '#';
1025 MAYBE('=', TP_LESSEQUAL)
1028 MAYBE('=', TP_LESSLESSEQUAL)
1033 MAYBE('=', TP_GREATEREQUAL)
1036 MAYBE('=', TP_GREATERGREATEREQUAL)
1037 ELSE(TP_GREATERGREATER)
1041 MAYBE('=', TP_CARETEQUAL)
1045 MAYBE('=', TP_PIPEEQUAL)
1046 MAYBE('|', TP_PIPEPIPE)
1054 MAYBE('=', TP_EQUALEQUAL)
1058 MAYBE('#', TP_HASHHASH)
1060 pp_token.kind = '#';
1074 pp_token.kind = input.c;
1079 if (input_stack != NULL) {
1081 pop_restore_input();
1083 print_line_directive(&input.position, "2");
1086 pp_token.base.source_position.lineno++;
1087 info.at_line_begin = true;
1088 pp_token.kind = TP_EOF;
1094 if (!ignore_unknown_chars) {
1095 errorf(&pp_token.base.source_position,
1096 "unknown character '%c' found\n", input.c);
1099 pp_token.kind = input.c;
1105 static void print_quoted_string(const char *const string)
1108 for (const char *c = string; *c != 0; ++c) {
1110 case '"': fputs("\\\"", out); break;
1111 case '\\': fputs("\\\\", out); break;
1112 case '\a': fputs("\\a", out); break;
1113 case '\b': fputs("\\b", out); break;
1114 case '\f': fputs("\\f", out); break;
1115 case '\n': fputs("\\n", out); break;
1116 case '\r': fputs("\\r", out); break;
1117 case '\t': fputs("\\t", out); break;
1118 case '\v': fputs("\\v", out); break;
1119 case '\?': fputs("\\?", out); break;
1122 fprintf(out, "\\%03o", (unsigned)*c);
1132 static void print_line_directive(const source_position_t *pos, const char *add)
1134 fprintf(out, "# %u ", pos->lineno);
1135 print_quoted_string(pos->input_name);
1141 printed_input_name = pos->input_name;
1142 input.output_line = pos->lineno-1;
1145 static bool emit_newlines(void)
1147 unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1153 print_line_directive(&pp_token.base.source_position, NULL);
1156 for (unsigned i = 0; i < delta; ++i) {
1160 input.output_line = pp_token.base.source_position.lineno;
1162 for (unsigned i = 0; i < info.whitespace; ++i)
1168 static void emit_pp_token(void)
1173 if (!emit_newlines() &&
1174 (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1177 switch (pp_token.kind) {
1179 fputs(pp_token.base.symbol->string, out);
1182 fputs(pp_token.number.number.begin, out);
1185 case TP_STRING_LITERAL:
1186 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1188 fputs(pp_token.string.string.begin, out);
1192 case TP_CHARACTER_CONSTANT:
1193 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1195 fputs(pp_token.string.string.begin, out);
1199 print_pp_token_kind(out, pp_token.kind);
1202 last_token = pp_token.kind;
1205 static void eat_pp_directive(void)
1207 while (!info.at_line_begin) {
1208 next_preprocessing_token();
1212 static bool strings_equal(const string_t *string1, const string_t *string2)
1214 size_t size = string1->size;
1215 if (size != string2->size)
1218 const char *c1 = string1->begin;
1219 const char *c2 = string2->begin;
1220 for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1227 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1229 if (token1->kind != token2->kind)
1232 switch (token1->kind) {
1234 return token1->base.symbol == token2->base.symbol;
1237 case TP_CHARACTER_CONSTANT:
1238 case TP_STRING_LITERAL:
1239 return strings_equal(&token1->string.string, &token2->string.string);
1246 static bool pp_definitions_equal(const pp_definition_t *definition1,
1247 const pp_definition_t *definition2)
1249 if (definition1->list_len != definition2->list_len)
1252 size_t len = definition1->list_len;
1253 const token_t *t1 = definition1->token_list;
1254 const token_t *t2 = definition2->token_list;
1255 for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1256 if (!pp_tokens_equal(t1, t2))
1262 static void parse_define_directive(void)
1265 assert(obstack_object_size(&pp_obstack) == 0);
1267 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1268 errorf(&pp_token.base.source_position,
1269 "expected identifier after #define, got '%t'", &pp_token);
1272 symbol_t *const symbol = pp_token.base.symbol;
1274 pp_definition_t *new_definition
1275 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1276 memset(new_definition, 0, sizeof(new_definition[0]));
1277 new_definition->source_position = input.position;
1279 /* this is probably the only place where spaces are significant in the
1280 * lexer (except for the fact that they separate tokens). #define b(x)
1281 * is something else than #define b (x) */
1282 if (input.c == '(') {
1284 next_preprocessing_token();
1285 /* get next token after '(' */
1286 next_preprocessing_token();
1289 switch (pp_token.kind) {
1291 new_definition->is_variadic = true;
1292 next_preprocessing_token();
1293 if (pp_token.kind != ')') {
1294 errorf(&input.position,
1295 "'...' not at end of macro argument list");
1300 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1301 next_preprocessing_token();
1303 if (pp_token.kind == ',') {
1304 next_preprocessing_token();
1308 if (pp_token.kind != ')') {
1309 errorf(&pp_token.base.source_position,
1310 "expected ',' or ')' after identifier, got '%t'",
1316 next_preprocessing_token();
1317 goto finish_argument_list;
1319 errorf(&pp_token.base.source_position,
1320 "expected identifier, '...' or ')' in #define argument list, got '%t'",
1326 finish_argument_list:
1327 new_definition->has_parameters = true;
1328 new_definition->n_parameters
1329 = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1330 new_definition->parameters = obstack_finish(&pp_obstack);
1332 next_preprocessing_token();
1335 /* construct a new pp_definition on the obstack */
1336 assert(obstack_object_size(&pp_obstack) == 0);
1337 size_t list_len = 0;
1338 while (!info.at_line_begin) {
1339 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1341 next_preprocessing_token();
1344 new_definition->list_len = list_len;
1345 new_definition->token_list = obstack_finish(&pp_obstack);
1347 pp_definition_t *old_definition = symbol->pp_definition;
1348 if (old_definition != NULL) {
1349 if (!pp_definitions_equal(old_definition, new_definition)) {
1350 warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1352 /* reuse the old definition */
1353 obstack_free(&pp_obstack, new_definition);
1354 new_definition = old_definition;
1358 symbol->pp_definition = new_definition;
1362 if (obstack_object_size(&pp_obstack) > 0) {
1363 char *ptr = obstack_finish(&pp_obstack);
1364 obstack_free(&pp_obstack, ptr);
1369 static void parse_undef_directive(void)
1373 if (pp_token.kind != TP_IDENTIFIER) {
1374 errorf(&input.position,
1375 "expected identifier after #undef, got '%t'", &pp_token);
1380 pp_token.base.symbol->pp_definition = NULL;
1381 next_preprocessing_token();
1383 if (!info.at_line_begin) {
1384 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1389 static void parse_headername(void)
1391 const source_position_t start_position = input.position;
1392 string_t string = {NULL, 0};
1393 assert(obstack_object_size(&symbol_obstack) == 0);
1395 /* behind an #include we can have the special headername lexems.
1396 * They're only allowed behind an #include so they're not recognized
1397 * by the normal next_preprocessing_token. We handle them as a special
1399 if (info.at_line_begin) {
1400 parse_error("expected headername after #include");
1404 /* check wether we have a "... or <... headername */
1413 parse_error("header name without closing '>'");
1418 goto finished_headername;
1420 obstack_1grow(&symbol_obstack, (char) input.c);
1423 /* we should never be here */
1432 parse_error("header name without closing '>'");
1437 goto finished_headername;
1439 obstack_1grow(&symbol_obstack, (char) input.c);
1442 /* we should never be here */
1445 /* TODO: do normal pp_token parsing and concatenate results */
1446 panic("pp_token concat include not implemented yet");
1449 finished_headername:
1450 obstack_1grow(&symbol_obstack, '\0');
1451 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
1452 char *const headername = obstack_finish(&symbol_obstack);
1453 string = make_string(headername, size);
1456 pp_token.base.source_position = start_position;
1457 pp_token.kind = TP_HEADERNAME;
1458 pp_token.string.string = string;
1461 static bool do_include(bool system_include, const char *headername)
1463 size_t headername_len = strlen(headername);
1464 if (!system_include) {
1465 /* put dirname of current input on obstack */
1466 const char *filename = input.position.input_name;
1467 const char *last_slash = strrchr(filename, '/');
1468 if (last_slash != NULL) {
1469 size_t len = last_slash - filename;
1470 obstack_grow(&symbol_obstack, filename, len + 1);
1471 obstack_grow0(&symbol_obstack, headername, headername_len);
1472 char *complete_path = obstack_finish(&symbol_obstack);
1473 headername = identify_string(complete_path);
1476 FILE *file = fopen(headername, "r");
1478 switch_input(file, headername);
1483 assert(obstack_object_size(&symbol_obstack) == 0);
1484 /* check searchpath */
1485 for (searchpath_entry_t *entry = searchpath; entry != NULL;
1486 entry = entry->next) {
1487 const char *path = entry->path;
1488 size_t len = strlen(path);
1489 obstack_grow(&symbol_obstack, path, len);
1490 if (path[len-1] != '/')
1491 obstack_1grow(&symbol_obstack, '/');
1492 obstack_grow(&symbol_obstack, headername, headername_len+1);
1494 char *complete_path = obstack_finish(&symbol_obstack);
1495 FILE *file = fopen(complete_path, "r");
1497 const char *filename = identify_string(complete_path);
1498 switch_input(file, filename);
1501 obstack_free(&symbol_obstack, complete_path);
1508 /* read till next newline character, only for parse_include_directive(),
1509 * use eat_pp_directive() in all other cases */
1510 static void skip_till_newline(void)
1512 /* skip till newline */
1525 static bool parse_include_directive(void)
1527 /* don't eat the TP_include here!
1528 * we need an alternative parsing for the next token */
1530 bool system_include = input.c == '<';
1532 string_t headername = pp_token.string.string;
1533 if (headername.begin == NULL) {
1539 if (!info.at_line_begin) {
1540 warningf(WARN_OTHER, &pp_token.base.source_position,
1541 "extra tokens at end of #include directive");
1542 skip_till_newline();
1545 if (n_inputs > INCLUDE_LIMIT) {
1546 errorf(&pp_token.base.source_position, "#include nested too deeply");
1548 next_preprocessing_token();
1555 bool res = do_include(system_include, pp_token.string.string.begin);
1557 errorf(&pp_token.base.source_position,
1558 "failed including '%S': %s", pp_token.string, strerror(errno));
1559 pop_restore_input();
1566 static pp_conditional_t *push_conditional(void)
1568 pp_conditional_t *conditional
1569 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1570 memset(conditional, 0, sizeof(*conditional));
1572 conditional->parent = conditional_stack;
1573 conditional_stack = conditional;
1578 static void pop_conditional(void)
1580 assert(conditional_stack != NULL);
1581 conditional_stack = conditional_stack->parent;
1584 static void check_unclosed_conditionals(void)
1586 while (conditional_stack != NULL) {
1587 pp_conditional_t *conditional = conditional_stack;
1589 if (conditional->in_else) {
1590 errorf(&conditional->source_position, "unterminated #else");
1592 errorf(&conditional->source_position, "unterminated condition");
1598 static void parse_ifdef_ifndef_directive(void)
1600 bool is_ifndef = (pp_token.kind == TP_ifndef);
1602 next_preprocessing_token();
1606 pp_conditional_t *conditional = push_conditional();
1607 conditional->source_position = pp_token.base.source_position;
1608 conditional->skip = true;
1612 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1613 errorf(&pp_token.base.source_position,
1614 "expected identifier after #%s, got '%t'",
1615 is_ifndef ? "ifndef" : "ifdef", &pp_token);
1618 /* just take the true case in the hope to avoid further errors */
1621 /* evaluate wether we are in true or false case */
1622 condition = !pp_token.base.symbol->pp_definition == is_ifndef;
1624 next_preprocessing_token();
1626 if (!info.at_line_begin) {
1627 errorf(&pp_token.base.source_position,
1628 "extra tokens at end of #%s",
1629 is_ifndef ? "ifndef" : "ifdef");
1634 pp_conditional_t *conditional = push_conditional();
1635 conditional->source_position = pp_token.base.source_position;
1636 conditional->condition = condition;
1643 static void parse_else_directive(void)
1647 if (!info.at_line_begin) {
1649 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1654 pp_conditional_t *conditional = conditional_stack;
1655 if (conditional == NULL) {
1656 errorf(&pp_token.base.source_position, "#else without prior #if");
1660 if (conditional->in_else) {
1661 errorf(&pp_token.base.source_position,
1662 "#else after #else (condition started %P)",
1663 conditional->source_position);
1668 conditional->in_else = true;
1669 if (!conditional->skip) {
1670 skip_mode = conditional->condition;
1672 conditional->source_position = pp_token.base.source_position;
1675 static void parse_endif_directive(void)
1679 if (!info.at_line_begin) {
1681 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1686 pp_conditional_t *conditional = conditional_stack;
1687 if (conditional == NULL) {
1688 errorf(&pp_token.base.source_position, "#endif without prior #if");
1692 if (!conditional->skip) {
1698 static void parse_preprocessing_directive(void)
1702 if (info.at_line_begin) {
1703 /* empty directive */
1708 switch (pp_token.kind) {
1711 parse_ifdef_ifndef_directive();
1714 parse_else_directive();
1717 parse_endif_directive();
1724 switch (pp_token.kind) {
1726 parse_define_directive();
1729 parse_undef_directive();
1733 parse_ifdef_ifndef_directive();
1736 parse_else_directive();
1739 parse_endif_directive();
1742 parse_include_directive();
1745 if (info.at_line_begin) {
1746 /* the nop directive "#" */
1749 errorf(&pp_token.base.source_position,
1750 "invalid preprocessing directive #%t", &pp_token);
1756 assert(info.at_line_begin);
1759 static void prepend_include_path(const char *path)
1761 searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1763 entry->next = searchpath;
1767 static void setup_include_path(void)
1769 /* built-in paths */
1770 prepend_include_path("/usr/include");
1772 /* parse environment variable */
1773 const char *cpath = getenv("CPATH");
1774 if (cpath != NULL && *cpath != '\0') {
1775 const char *begin = cpath;
1779 while (*c != '\0' && *c != ':')
1782 size_t len = c-begin;
1784 /* for gcc compatibility (Matze: I would expect that
1785 * nothing happens for an empty entry...) */
1786 prepend_include_path(".");
1788 char *string = obstack_alloc(&config_obstack, len+1);
1789 memcpy(string, begin, len);
1792 prepend_include_path(string);
1799 } while(*c != '\0');
1803 int pptest_main(int argc, char **argv);
1804 int pptest_main(int argc, char **argv)
1806 init_symbol_table();
1809 obstack_init(&config_obstack);
1810 obstack_init(&pp_obstack);
1811 obstack_init(&input_obstack);
1812 strset_init(&stringset);
1814 setup_include_path();
1816 /* simplistic commandline parser */
1817 const char *filename = NULL;
1818 const char *output = NULL;
1819 for (int i = 1; i < argc; ++i) {
1820 const char *opt = argv[i];
1821 if (streq(opt, "-I")) {
1822 prepend_include_path(argv[++i]);
1824 } else if (streq(opt, "-E")) {
1826 } else if (streq(opt, "-o")) {
1829 } else if (opt[0] == '-') {
1830 fprintf(stderr, "Unknown option '%s'\n", opt);
1832 if (filename != NULL)
1833 fprintf(stderr, "Multiple inputs not supported\n");
1837 if (filename == NULL) {
1838 fprintf(stderr, "No input specified\n");
1842 if (output == NULL) {
1845 out = fopen(output, "w");
1847 fprintf(stderr, "Couldn't open output '%s'\n", output);
1852 /* just here for gcc compatibility */
1853 fprintf(out, "# 1 \"%s\"\n", filename);
1854 fprintf(out, "# 1 \"<built-in>\"\n");
1855 fprintf(out, "# 1 \"<command-line>\"\n");
1857 FILE *file = fopen(filename, "r");
1859 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1862 switch_input(file, filename);
1865 if (pp_token.kind == '#' && info.at_line_begin) {
1866 parse_preprocessing_directive();
1868 } else if (pp_token.kind == TP_EOF) {
1869 goto end_of_main_loop;
1870 } else if (pp_token.kind == TP_IDENTIFIER) {
1871 symbol_t *const symbol = pp_token.base.symbol;
1872 pp_definition_t *const pp_definition = symbol->pp_definition;
1873 if (pp_definition != NULL && !pp_definition->is_expanding) {
1874 expansion_pos = pp_token.base.source_position;
1875 if (pp_definition->has_parameters) {
1876 source_position_t position = pp_token.base.source_position;
1877 add_token_info_t old_info = info;
1878 next_preprocessing_token();
1879 add_token_info_t new_info = info;
1881 /* no opening brace -> no expansion */
1882 if (pp_token.kind == '(') {
1885 /* parse arguments (TODO) */
1886 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1887 next_preprocessing_token();
1889 token_t next_token = pp_token;
1890 /* restore identifier token */
1891 pp_token.kind = TP_IDENTIFIER;
1892 pp_token.base.symbol = symbol;
1893 pp_token.base.source_position = position;
1898 pp_token = next_token;
1903 pp_definition->expand_pos = 0;
1904 pp_definition->is_expanding = true;
1905 current_expansion = pp_definition;
1912 next_preprocessing_token();
1917 check_unclosed_conditionals();
1922 obstack_free(&input_obstack, NULL);
1923 obstack_free(&pp_obstack, NULL);
1924 obstack_free(&config_obstack, NULL);
1926 strset_destroy(&stringset);
1929 exit_symbol_table();