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(utf32 const delimiter, preprocessor_token_kind_t const kind, string_encoding_t const enc, char const *const context)
462 const unsigned start_linenr = input.position.lineno;
469 if (resolve_escape_sequences) {
470 utf32 const tc = parse_escape_sequence();
471 if (enc == STRING_ENCODING_CHAR) {
473 warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range");
475 obstack_1grow(&symbol_obstack, tc);
477 obstack_grow_symbol(&symbol_obstack, tc);
480 obstack_1grow(&symbol_obstack, (char)input.c);
482 obstack_1grow(&symbol_obstack, (char)input.c);
489 errorf(&pp_token.base.source_position, "newline while parsing %s", context);
494 source_position_t source_position;
495 source_position.input_name = pp_token.base.source_position.input_name;
496 source_position.lineno = start_linenr;
497 errorf(&source_position, "EOF while parsing %s", context);
502 if (input.c == delimiter) {
506 obstack_grow_symbol(&symbol_obstack, input.c);
514 obstack_1grow(&symbol_obstack, '\0');
515 size_t const size = obstack_object_size(&symbol_obstack) - 1;
516 char *const string = obstack_finish(&symbol_obstack);
518 pp_token.kind = kind;
519 pp_token.string.encoding = enc;
520 pp_token.string.string = make_string(string, size);
523 static void parse_string_literal(string_encoding_t const enc)
525 parse_string('"', TP_STRING_LITERAL, enc, "string literal");
528 static void parse_character_constant(string_encoding_t const enc)
530 parse_string('\'', TP_CHARACTER_CONSTANT, enc, "character constant");
531 if (pp_token.string.string.size == 0) {
532 parse_error("empty character constant");
536 #define SYMBOL_CHARS_WITHOUT_E_P \
587 #define SYMBOL_CHARS \
588 SYMBOL_CHARS_WITHOUT_E_P \
607 * returns next final token from a preprocessor macro expansion
609 static void expand_next(void)
611 assert(current_expansion != NULL);
613 pp_definition_t *definition = current_expansion;
616 if (definition->list_len == 0
617 || definition->expand_pos >= definition->list_len) {
618 /* we're finished with the current macro, move up 1 level in the
620 pp_definition_t *parent = definition->parent_expansion;
621 definition->parent_expansion = NULL;
622 definition->is_expanding = false;
624 /* it was the outermost expansion, parse normal pptoken */
625 if (parent == NULL) {
626 current_expansion = NULL;
627 next_preprocessing_token();
631 current_expansion = definition;
634 pp_token = definition->token_list[definition->expand_pos];
635 pp_token.base.source_position = expansion_pos;
636 ++definition->expand_pos;
638 if (pp_token.kind != TP_IDENTIFIER)
641 /* if it was an identifier then we might need to expand again */
642 pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition;
643 if (symbol_definition != NULL && !symbol_definition->is_expanding) {
644 symbol_definition->parent_expansion = definition;
645 symbol_definition->expand_pos = 0;
646 symbol_definition->is_expanding = true;
647 definition = symbol_definition;
648 current_expansion = definition;
653 static void skip_line_comment(void)
671 static void skip_multiline_comment(void)
673 unsigned start_linenr = input.position.lineno;
678 if (input.c == '*') {
679 /* TODO: nested comment, warn here */
684 if (input.c == '/') {
685 if (input.position.lineno != input.output_line)
686 info.whitespace = input.position.colno;
697 source_position_t source_position;
698 source_position.input_name = pp_token.base.source_position.input_name;
699 source_position.lineno = start_linenr;
700 errorf(&source_position, "at end of file while looking for comment end");
711 static void skip_whitespace(void)
721 info.at_line_begin = true;
727 if (input.c == '/') {
731 } else if (input.c == '*') {
733 skip_multiline_comment();
746 static void eat_pp(preprocessor_token_kind_t const type)
749 assert(pp_token.kind == type);
750 next_preprocessing_token();
753 static void parse_symbol(void)
755 obstack_1grow(&symbol_obstack, (char) input.c);
762 obstack_1grow(&symbol_obstack, (char) input.c);
772 obstack_1grow(&symbol_obstack, '\0');
773 char *string = obstack_finish(&symbol_obstack);
775 /* might be a wide string or character constant ( L"string"/L'c' ) */
776 if (input.c == '"' && string[0] == 'L' && string[1] == '\0') {
777 obstack_free(&symbol_obstack, string);
778 parse_string_literal(STRING_ENCODING_WIDE);
780 } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') {
781 obstack_free(&symbol_obstack, string);
782 parse_character_constant(STRING_ENCODING_WIDE);
786 symbol_t *symbol = symbol_table_insert(string);
788 pp_token.kind = symbol->pp_ID;
789 pp_token.base.symbol = symbol;
791 /* we can free the memory from symbol obstack if we already had an entry in
792 * the symbol table */
793 if (symbol->string != string) {
794 obstack_free(&symbol_obstack, string);
798 static void parse_number(void)
800 obstack_1grow(&symbol_obstack, (char) input.c);
807 SYMBOL_CHARS_WITHOUT_E_P
808 obstack_1grow(&symbol_obstack, (char) input.c);
816 obstack_1grow(&symbol_obstack, (char) input.c);
818 if (input.c == '+' || input.c == '-') {
819 obstack_1grow(&symbol_obstack, (char) input.c);
830 obstack_1grow(&symbol_obstack, '\0');
831 size_t size = obstack_object_size(&symbol_obstack);
832 char *string = obstack_finish(&symbol_obstack);
834 pp_token.kind = TP_NUMBER;
835 pp_token.number.number = make_string(string, size);
839 #define MAYBE_PROLOG \
844 #define MAYBE(ch, set_type) \
847 pp_token.kind = set_type; \
850 #define ELSE_CODE(code) \
857 #define ELSE(set_type) \
859 pp_token.kind = set_type; \
862 static void next_preprocessing_token(void)
864 if (current_expansion != NULL) {
869 info.at_line_begin = false;
870 info.had_whitespace = false;
872 pp_token.base.source_position = input.position;
873 pp_token.base.symbol = NULL;
879 info.had_whitespace = true;
884 info.at_line_begin = true;
885 info.had_whitespace = true;
898 parse_string_literal(STRING_ENCODING_CHAR);
902 parse_character_constant(STRING_ENCODING_CHAR);
924 MAYBE('.', TP_DOTDOTDOT)
933 MAYBE('&', TP_ANDAND)
934 MAYBE('=', TP_ANDEQUAL)
938 MAYBE('=', TP_ASTERISKEQUAL)
942 MAYBE('+', TP_PLUSPLUS)
943 MAYBE('=', TP_PLUSEQUAL)
947 MAYBE('>', TP_MINUSGREATER)
948 MAYBE('-', TP_MINUSMINUS)
949 MAYBE('=', TP_MINUSEQUAL)
953 MAYBE('=', TP_EXCLAMATIONMARKEQUAL)
957 MAYBE('=', TP_SLASHEQUAL)
960 info.had_whitespace = true;
961 skip_multiline_comment();
965 info.had_whitespace = true;
972 MAYBE('=', TP_PERCENTEQUAL)
977 MAYBE(':', TP_HASHHASH)
989 MAYBE('=', TP_LESSEQUAL)
992 MAYBE('=', TP_LESSLESSEQUAL)
997 MAYBE('=', TP_GREATEREQUAL)
1000 MAYBE('=', TP_GREATERGREATEREQUAL)
1001 ELSE(TP_GREATERGREATER)
1005 MAYBE('=', TP_CARETEQUAL)
1009 MAYBE('=', TP_PIPEEQUAL)
1010 MAYBE('|', TP_PIPEPIPE)
1018 MAYBE('=', TP_EQUALEQUAL)
1022 MAYBE('#', TP_HASHHASH)
1024 pp_token.kind = '#';
1038 pp_token.kind = input.c;
1043 if (input_stack != NULL) {
1045 pop_restore_input();
1047 print_line_directive(&input.position, "2");
1050 pp_token.base.source_position.lineno++;
1051 info.at_line_begin = true;
1052 pp_token.kind = TP_EOF;
1058 if (!ignore_unknown_chars) {
1059 errorf(&pp_token.base.source_position,
1060 "unknown character '%c' found\n", input.c);
1063 pp_token.kind = input.c;
1069 static void print_quoted_string(const char *const string)
1072 for (const char *c = string; *c != 0; ++c) {
1074 case '"': fputs("\\\"", out); break;
1075 case '\\': fputs("\\\\", out); break;
1076 case '\a': fputs("\\a", out); break;
1077 case '\b': fputs("\\b", out); break;
1078 case '\f': fputs("\\f", out); break;
1079 case '\n': fputs("\\n", out); break;
1080 case '\r': fputs("\\r", out); break;
1081 case '\t': fputs("\\t", out); break;
1082 case '\v': fputs("\\v", out); break;
1083 case '\?': fputs("\\?", out); break;
1086 fprintf(out, "\\%03o", (unsigned)*c);
1096 static void print_line_directive(const source_position_t *pos, const char *add)
1098 fprintf(out, "# %u ", pos->lineno);
1099 print_quoted_string(pos->input_name);
1105 printed_input_name = pos->input_name;
1106 input.output_line = pos->lineno-1;
1109 static bool emit_newlines(void)
1111 unsigned delta = pp_token.base.source_position.lineno - input.output_line;
1117 print_line_directive(&pp_token.base.source_position, NULL);
1120 for (unsigned i = 0; i < delta; ++i) {
1124 input.output_line = pp_token.base.source_position.lineno;
1126 for (unsigned i = 0; i < info.whitespace; ++i)
1132 static void emit_pp_token(void)
1137 if (!emit_newlines() &&
1138 (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind)))
1141 switch (pp_token.kind) {
1143 fputs(pp_token.base.symbol->string, out);
1146 fputs(pp_token.number.number.begin, out);
1149 case TP_STRING_LITERAL:
1150 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1152 fputs(pp_token.string.string.begin, out);
1156 case TP_CHARACTER_CONSTANT:
1157 fputs(get_string_encoding_prefix(pp_token.string.encoding), out);
1159 fputs(pp_token.string.string.begin, out);
1163 print_pp_token_kind(out, pp_token.kind);
1166 last_token = pp_token.kind;
1169 static void eat_pp_directive(void)
1171 while (!info.at_line_begin) {
1172 next_preprocessing_token();
1176 static bool strings_equal(const string_t *string1, const string_t *string2)
1178 size_t size = string1->size;
1179 if (size != string2->size)
1182 const char *c1 = string1->begin;
1183 const char *c2 = string2->begin;
1184 for (size_t i = 0; i < size; ++i, ++c1, ++c2) {
1191 static bool pp_tokens_equal(const token_t *token1, const token_t *token2)
1193 if (token1->kind != token2->kind)
1196 switch (token1->kind) {
1198 return token1->base.symbol == token2->base.symbol;
1201 case TP_CHARACTER_CONSTANT:
1202 case TP_STRING_LITERAL:
1203 return strings_equal(&token1->string.string, &token2->string.string);
1210 static bool pp_definitions_equal(const pp_definition_t *definition1,
1211 const pp_definition_t *definition2)
1213 if (definition1->list_len != definition2->list_len)
1216 size_t len = definition1->list_len;
1217 const token_t *t1 = definition1->token_list;
1218 const token_t *t2 = definition2->token_list;
1219 for (size_t i = 0; i < len; ++i, ++t1, ++t2) {
1220 if (!pp_tokens_equal(t1, t2))
1226 static void parse_define_directive(void)
1229 assert(obstack_object_size(&pp_obstack) == 0);
1231 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1232 errorf(&pp_token.base.source_position,
1233 "expected identifier after #define, got '%t'", &pp_token);
1236 symbol_t *const symbol = pp_token.base.symbol;
1238 pp_definition_t *new_definition
1239 = obstack_alloc(&pp_obstack, sizeof(new_definition[0]));
1240 memset(new_definition, 0, sizeof(new_definition[0]));
1241 new_definition->source_position = input.position;
1243 /* this is probably the only place where spaces are significant in the
1244 * lexer (except for the fact that they separate tokens). #define b(x)
1245 * is something else than #define b (x) */
1246 if (input.c == '(') {
1248 next_preprocessing_token();
1249 /* get next token after '(' */
1250 next_preprocessing_token();
1253 switch (pp_token.kind) {
1255 new_definition->is_variadic = true;
1256 next_preprocessing_token();
1257 if (pp_token.kind != ')') {
1258 errorf(&input.position,
1259 "'...' not at end of macro argument list");
1264 obstack_ptr_grow(&pp_obstack, pp_token.base.symbol);
1265 next_preprocessing_token();
1267 if (pp_token.kind == ',') {
1268 next_preprocessing_token();
1272 if (pp_token.kind != ')') {
1273 errorf(&pp_token.base.source_position,
1274 "expected ',' or ')' after identifier, got '%t'",
1280 next_preprocessing_token();
1281 goto finish_argument_list;
1283 errorf(&pp_token.base.source_position,
1284 "expected identifier, '...' or ')' in #define argument list, got '%t'",
1290 finish_argument_list:
1291 new_definition->has_parameters = true;
1292 new_definition->n_parameters
1293 = obstack_object_size(&pp_obstack) / sizeof(new_definition->parameters[0]);
1294 new_definition->parameters = obstack_finish(&pp_obstack);
1296 next_preprocessing_token();
1299 /* construct a new pp_definition on the obstack */
1300 assert(obstack_object_size(&pp_obstack) == 0);
1301 size_t list_len = 0;
1302 while (!info.at_line_begin) {
1303 obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token));
1305 next_preprocessing_token();
1308 new_definition->list_len = list_len;
1309 new_definition->token_list = obstack_finish(&pp_obstack);
1311 pp_definition_t *old_definition = symbol->pp_definition;
1312 if (old_definition != NULL) {
1313 if (!pp_definitions_equal(old_definition, new_definition)) {
1314 warningf(WARN_OTHER, &input.position, "multiple definition of macro '%Y' (first defined %P)", symbol, &old_definition->source_position);
1316 /* reuse the old definition */
1317 obstack_free(&pp_obstack, new_definition);
1318 new_definition = old_definition;
1322 symbol->pp_definition = new_definition;
1326 if (obstack_object_size(&pp_obstack) > 0) {
1327 char *ptr = obstack_finish(&pp_obstack);
1328 obstack_free(&pp_obstack, ptr);
1333 static void parse_undef_directive(void)
1337 if (pp_token.kind != TP_IDENTIFIER) {
1338 errorf(&input.position,
1339 "expected identifier after #undef, got '%t'", &pp_token);
1344 pp_token.base.symbol->pp_definition = NULL;
1345 next_preprocessing_token();
1347 if (!info.at_line_begin) {
1348 warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive");
1353 static void parse_headername(void)
1355 const source_position_t start_position = input.position;
1356 string_t string = {NULL, 0};
1357 assert(obstack_object_size(&symbol_obstack) == 0);
1359 /* behind an #include we can have the special headername lexems.
1360 * They're only allowed behind an #include so they're not recognized
1361 * by the normal next_preprocessing_token. We handle them as a special
1363 if (info.at_line_begin) {
1364 parse_error("expected headername after #include");
1368 /* check wether we have a "... or <... headername */
1377 parse_error("header name without closing '>'");
1382 goto finished_headername;
1384 obstack_1grow(&symbol_obstack, (char) input.c);
1387 /* we should never be here */
1396 parse_error("header name without closing '>'");
1401 goto finished_headername;
1403 obstack_1grow(&symbol_obstack, (char) input.c);
1406 /* we should never be here */
1409 /* TODO: do normal pp_token parsing and concatenate results */
1410 panic("pp_token concat include not implemented yet");
1413 finished_headername:
1414 obstack_1grow(&symbol_obstack, '\0');
1415 const size_t size = (size_t)obstack_object_size(&symbol_obstack);
1416 char *const headername = obstack_finish(&symbol_obstack);
1417 string = make_string(headername, size);
1420 pp_token.base.source_position = start_position;
1421 pp_token.kind = TP_HEADERNAME;
1422 pp_token.string.string = string;
1425 static bool do_include(bool system_include, const char *headername)
1427 size_t headername_len = strlen(headername);
1428 if (!system_include) {
1429 /* put dirname of current input on obstack */
1430 const char *filename = input.position.input_name;
1431 const char *last_slash = strrchr(filename, '/');
1432 if (last_slash != NULL) {
1433 size_t len = last_slash - filename;
1434 obstack_grow(&symbol_obstack, filename, len + 1);
1435 obstack_grow0(&symbol_obstack, headername, headername_len);
1436 char *complete_path = obstack_finish(&symbol_obstack);
1437 headername = identify_string(complete_path);
1440 FILE *file = fopen(headername, "r");
1442 switch_input(file, headername);
1447 assert(obstack_object_size(&symbol_obstack) == 0);
1448 /* check searchpath */
1449 for (searchpath_entry_t *entry = searchpath; entry != NULL;
1450 entry = entry->next) {
1451 const char *path = entry->path;
1452 size_t len = strlen(path);
1453 obstack_grow(&symbol_obstack, path, len);
1454 if (path[len-1] != '/')
1455 obstack_1grow(&symbol_obstack, '/');
1456 obstack_grow(&symbol_obstack, headername, headername_len+1);
1458 char *complete_path = obstack_finish(&symbol_obstack);
1459 FILE *file = fopen(complete_path, "r");
1461 const char *filename = identify_string(complete_path);
1462 switch_input(file, filename);
1465 obstack_free(&symbol_obstack, complete_path);
1472 /* read till next newline character, only for parse_include_directive(),
1473 * use eat_pp_directive() in all other cases */
1474 static void skip_till_newline(void)
1476 /* skip till newline */
1489 static bool parse_include_directive(void)
1491 /* don't eat the TP_include here!
1492 * we need an alternative parsing for the next token */
1494 bool system_include = input.c == '<';
1496 string_t headername = pp_token.string.string;
1497 if (headername.begin == NULL) {
1503 if (!info.at_line_begin) {
1504 warningf(WARN_OTHER, &pp_token.base.source_position,
1505 "extra tokens at end of #include directive");
1506 skip_till_newline();
1509 if (n_inputs > INCLUDE_LIMIT) {
1510 errorf(&pp_token.base.source_position, "#include nested too deeply");
1512 next_preprocessing_token();
1519 bool res = do_include(system_include, pp_token.string.string.begin);
1521 errorf(&pp_token.base.source_position,
1522 "failed including '%S': %s", pp_token.string, strerror(errno));
1523 pop_restore_input();
1530 static pp_conditional_t *push_conditional(void)
1532 pp_conditional_t *conditional
1533 = obstack_alloc(&pp_obstack, sizeof(*conditional));
1534 memset(conditional, 0, sizeof(*conditional));
1536 conditional->parent = conditional_stack;
1537 conditional_stack = conditional;
1542 static void pop_conditional(void)
1544 assert(conditional_stack != NULL);
1545 conditional_stack = conditional_stack->parent;
1548 static void check_unclosed_conditionals(void)
1550 while (conditional_stack != NULL) {
1551 pp_conditional_t *conditional = conditional_stack;
1553 if (conditional->in_else) {
1554 errorf(&conditional->source_position, "unterminated #else");
1556 errorf(&conditional->source_position, "unterminated condition");
1562 static void parse_ifdef_ifndef_directive(void)
1564 bool is_ifndef = (pp_token.kind == TP_ifndef);
1566 next_preprocessing_token();
1570 pp_conditional_t *conditional = push_conditional();
1571 conditional->source_position = pp_token.base.source_position;
1572 conditional->skip = true;
1576 if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) {
1577 errorf(&pp_token.base.source_position,
1578 "expected identifier after #%s, got '%t'",
1579 is_ifndef ? "ifndef" : "ifdef", &pp_token);
1582 /* just take the true case in the hope to avoid further errors */
1585 /* evaluate wether we are in true or false case */
1586 condition = !pp_token.base.symbol->pp_definition == is_ifndef;
1588 next_preprocessing_token();
1590 if (!info.at_line_begin) {
1591 errorf(&pp_token.base.source_position,
1592 "extra tokens at end of #%s",
1593 is_ifndef ? "ifndef" : "ifdef");
1598 pp_conditional_t *conditional = push_conditional();
1599 conditional->source_position = pp_token.base.source_position;
1600 conditional->condition = condition;
1607 static void parse_else_directive(void)
1611 if (!info.at_line_begin) {
1613 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else");
1618 pp_conditional_t *conditional = conditional_stack;
1619 if (conditional == NULL) {
1620 errorf(&pp_token.base.source_position, "#else without prior #if");
1624 if (conditional->in_else) {
1625 errorf(&pp_token.base.source_position,
1626 "#else after #else (condition started %P)",
1627 conditional->source_position);
1632 conditional->in_else = true;
1633 if (!conditional->skip) {
1634 skip_mode = conditional->condition;
1636 conditional->source_position = pp_token.base.source_position;
1639 static void parse_endif_directive(void)
1643 if (!info.at_line_begin) {
1645 warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif");
1650 pp_conditional_t *conditional = conditional_stack;
1651 if (conditional == NULL) {
1652 errorf(&pp_token.base.source_position, "#endif without prior #if");
1656 if (!conditional->skip) {
1662 static void parse_preprocessing_directive(void)
1666 if (info.at_line_begin) {
1667 /* empty directive */
1672 switch (pp_token.kind) {
1675 parse_ifdef_ifndef_directive();
1678 parse_else_directive();
1681 parse_endif_directive();
1688 switch (pp_token.kind) {
1690 parse_define_directive();
1693 parse_undef_directive();
1697 parse_ifdef_ifndef_directive();
1700 parse_else_directive();
1703 parse_endif_directive();
1706 parse_include_directive();
1709 if (info.at_line_begin) {
1710 /* the nop directive "#" */
1713 errorf(&pp_token.base.source_position,
1714 "invalid preprocessing directive #%t", &pp_token);
1720 assert(info.at_line_begin);
1723 static void prepend_include_path(const char *path)
1725 searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t);
1727 entry->next = searchpath;
1731 static void setup_include_path(void)
1733 /* built-in paths */
1734 prepend_include_path("/usr/include");
1736 /* parse environment variable */
1737 const char *cpath = getenv("CPATH");
1738 if (cpath != NULL && *cpath != '\0') {
1739 const char *begin = cpath;
1743 while (*c != '\0' && *c != ':')
1746 size_t len = c-begin;
1748 /* for gcc compatibility (Matze: I would expect that
1749 * nothing happens for an empty entry...) */
1750 prepend_include_path(".");
1752 char *string = obstack_alloc(&config_obstack, len+1);
1753 memcpy(string, begin, len);
1756 prepend_include_path(string);
1763 } while(*c != '\0');
1767 int pptest_main(int argc, char **argv);
1768 int pptest_main(int argc, char **argv)
1770 init_symbol_table();
1773 obstack_init(&config_obstack);
1774 obstack_init(&pp_obstack);
1775 obstack_init(&input_obstack);
1776 strset_init(&stringset);
1778 setup_include_path();
1780 /* simplistic commandline parser */
1781 const char *filename = NULL;
1782 const char *output = NULL;
1783 for (int i = 1; i < argc; ++i) {
1784 const char *opt = argv[i];
1785 if (streq(opt, "-I")) {
1786 prepend_include_path(argv[++i]);
1788 } else if (streq(opt, "-E")) {
1790 } else if (streq(opt, "-o")) {
1793 } else if (opt[0] == '-') {
1794 fprintf(stderr, "Unknown option '%s'\n", opt);
1796 if (filename != NULL)
1797 fprintf(stderr, "Multiple inputs not supported\n");
1801 if (filename == NULL) {
1802 fprintf(stderr, "No input specified\n");
1806 if (output == NULL) {
1809 out = fopen(output, "w");
1811 fprintf(stderr, "Couldn't open output '%s'\n", output);
1816 /* just here for gcc compatibility */
1817 fprintf(out, "# 1 \"%s\"\n", filename);
1818 fprintf(out, "# 1 \"<built-in>\"\n");
1819 fprintf(out, "# 1 \"<command-line>\"\n");
1821 FILE *file = fopen(filename, "r");
1823 fprintf(stderr, "Couldn't open input '%s'\n", filename);
1826 switch_input(file, filename);
1829 if (pp_token.kind == '#' && info.at_line_begin) {
1830 parse_preprocessing_directive();
1832 } else if (pp_token.kind == TP_EOF) {
1833 goto end_of_main_loop;
1834 } else if (pp_token.kind == TP_IDENTIFIER) {
1835 symbol_t *const symbol = pp_token.base.symbol;
1836 pp_definition_t *const pp_definition = symbol->pp_definition;
1837 if (pp_definition != NULL && !pp_definition->is_expanding) {
1838 expansion_pos = pp_token.base.source_position;
1839 if (pp_definition->has_parameters) {
1840 source_position_t position = pp_token.base.source_position;
1841 add_token_info_t old_info = info;
1842 next_preprocessing_token();
1843 add_token_info_t new_info = info;
1845 /* no opening brace -> no expansion */
1846 if (pp_token.kind == '(') {
1849 /* parse arguments (TODO) */
1850 while (pp_token.kind != TP_EOF && pp_token.kind != ')')
1851 next_preprocessing_token();
1853 token_t next_token = pp_token;
1854 /* restore identifier token */
1855 pp_token.kind = TP_IDENTIFIER;
1856 pp_token.base.symbol = symbol;
1857 pp_token.base.source_position = position;
1862 pp_token = next_token;
1867 pp_definition->expand_pos = 0;
1868 pp_definition->is_expanding = true;
1869 current_expansion = pp_definition;
1876 next_preprocessing_token();
1881 check_unclosed_conditionals();
1886 obstack_free(&input_obstack, NULL);
1887 obstack_free(&pp_obstack, NULL);
1888 obstack_free(&config_obstack, NULL);
1890 strset_destroy(&stringset);
1893 exit_symbol_table();