X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=preprocessor.c;h=a9c10d429ba8fcf2e2a48c68ef20401e283478e0;hb=47b73b80e0b34b1e716253384d0a7c56f2a5e268;hp=0c7204313b1861f1f8252bbd14d85ead0f42500f;hpb=1e652037c790eb8e55a3d79b6031628d3df5e9d8;p=cparser diff --git a/preprocessor.c b/preprocessor.c index 0c72043..a9c10d4 100644 --- a/preprocessor.c +++ b/preprocessor.c @@ -10,6 +10,8 @@ #include "symbol_t.h" #include "adt/util.h" #include "adt/error.h" +#include "adt/strutil.h" +#include "adt/strset.h" #include "lang_features.h" #include "diagnostic.h" #include "string_rep.h" @@ -58,11 +60,27 @@ struct pp_input_t { const utf32 *bufend; const utf32 *bufpos; source_position_t position; - bool had_non_space; pp_input_t *parent; + unsigned output_line; }; -static pp_input_t input; +/** additional info about the current token */ +typedef struct add_token_info_t { + /** whitespace from beginning of line to the token */ + unsigned whitespace; + /** there has been any whitespace before the token */ + bool had_whitespace; + /** the token is at the beginning of the line */ + bool at_line_begin; +} add_token_info_t; + +typedef struct searchpath_entry_t searchpath_entry_t; +struct searchpath_entry_t { + const char *path; + searchpath_entry_t *next; +}; + +static pp_input_t input; static pp_input_t *input_stack; static unsigned n_inputs; @@ -70,56 +88,48 @@ static struct obstack input_obstack; static pp_conditional_t *conditional_stack; -static token_t pp_token; -static bool resolve_escape_sequences = false; -static bool do_print_spaces = true; -static bool do_expansions; -static bool skip_mode; -static FILE *out; -static struct obstack pp_obstack; -static unsigned counted_newlines; -static unsigned counted_spaces; -static const char *printed_input_name = NULL; -static pp_definition_t *current_expansion = NULL; +static token_t pp_token; +static bool resolve_escape_sequences = false; +static bool ignore_unknown_chars = true; +static bool skip_mode; +static FILE *out; +static struct obstack pp_obstack; +static struct obstack config_obstack; +static const char *printed_input_name = NULL; +static source_position_t expansion_pos; +static pp_definition_t *current_expansion = NULL; +static strset_t stringset; +static preprocessor_token_kind_t last_token; + +static searchpath_entry_t *searchpath; + +static add_token_info_t info; static inline void next_char(void); static void next_preprocessing_token(void); static void print_line_directive(const source_position_t *pos, const char *add); -static bool open_input(const char *filename) +static void switch_input(FILE *file, const char *filename) { - FILE *file = fopen(filename, "r"); - if (file == NULL) - return false; - input.file = file; input.input = input_from_stream(file, NULL); input.bufend = NULL; input.bufpos = NULL; - input.had_non_space = false; + input.output_line = 0; input.position.input_name = filename; input.position.lineno = 1; /* indicate that we're at a new input */ print_line_directive(&input.position, input_stack != NULL ? "1" : NULL); - counted_newlines = 0; - counted_spaces = 0; - - /* read first char and first token */ - next_char(); + /* place a virtual '\n' so we realize we're at line begin */ + input.position.lineno = 0; + input.c = '\n'; next_preprocessing_token(); - - return true; } static void close_input(void) { - /* ensure we have a newline at EOF */ - if (input.had_non_space) { - fputc('\n', out); - } - input_free(input.input); assert(input.file != NULL); @@ -177,15 +187,14 @@ static void pop_restore_input(void) */ static void parse_error(const char *msg) { - errorf(&pp_token.source_position, "%s", msg); + errorf(&pp_token.base.source_position, "%s", msg); } static inline void next_real_char(void) { assert(input.bufpos <= input.bufend); if (input.bufpos >= input.bufend) { - size_t n = decode(input.input, input.buf + MAX_PUTBACK, - sizeof(input.buf)/sizeof(input.buf[0]) - MAX_PUTBACK); + size_t const n = decode(input.input, input.buf + MAX_PUTBACK, lengthof(input.buf) - MAX_PUTBACK); if (n == 0) { input.c = EOF; return; @@ -209,18 +218,18 @@ static inline void put_back(utf32 const pc) --input.position.colno; } -#define MATCH_NEWLINE(code) \ - case '\r': \ - next_char(); \ - if(input.c == '\n') { \ - next_char(); \ - } \ - ++input.position.lineno; \ - code \ - case '\n': \ - next_char(); \ - ++input.position.lineno; \ - code +#define NEWLINE \ + '\r': \ + next_char(); \ + if (input.c == '\n') { \ + case '\n': \ + next_char(); \ + } \ + info.whitespace = 0; \ + ++input.position.lineno; \ + input.position.colno = 1; \ + goto newline; \ + newline // Let it look like an ordinary case label. #define eat(c_type) (assert(input.c == c_type), next_char()) @@ -228,8 +237,9 @@ static void maybe_concat_lines(void) { eat('\\'); - switch(input.c) { - MATCH_NEWLINE(return;) + switch (input.c) { + case NEWLINE: + return; default: break; @@ -248,23 +258,23 @@ static inline void next_char(void) next_real_char(); /* filter trigraphs and concatenated lines */ - if(UNLIKELY(input.c == '\\')) { + if (UNLIKELY(input.c == '\\')) { maybe_concat_lines(); goto end_of_next_char; } - if(LIKELY(input.c != '?')) + if (LIKELY(input.c != '?')) goto end_of_next_char; next_real_char(); - if(LIKELY(input.c != '?')) { + if (LIKELY(input.c != '?')) { put_back(input.c); input.c = '?'; goto end_of_next_char; } next_real_char(); - switch(input.c) { + switch (input.c) { case '=': input.c = '#'; break; case '(': input.c = '['; break; case '/': input.c = '\\'; maybe_concat_lines(); break; @@ -296,7 +306,7 @@ end_of_next_char:; */ static inline bool is_octal_digit(int chr) { - switch(chr) { + switch (chr) { case '0': case '1': case '2': @@ -350,53 +360,44 @@ static int digit_value(int digit) * * @param first_digit the already read first digit */ -static int parse_octal_sequence(const int first_digit) +static utf32 parse_octal_sequence(const utf32 first_digit) { assert(is_octal_digit(first_digit)); - int value = digit_value(first_digit); + utf32 value = digit_value(first_digit); if (!is_octal_digit(input.c)) return value; value = 8 * value + digit_value(input.c); next_char(); if (!is_octal_digit(input.c)) return value; value = 8 * value + digit_value(input.c); next_char(); + return value; - if(char_is_signed) { - return (signed char) value; - } else { - return (unsigned char) value; - } } /** * Parses a hex character sequence. */ -static int parse_hex_sequence(void) +static utf32 parse_hex_sequence(void) { - int value = 0; - while(isxdigit(input.c)) { + utf32 value = 0; + while (isxdigit(input.c)) { value = 16 * value + digit_value(input.c); next_char(); } - - if(char_is_signed) { - return (signed char) value; - } else { - return (unsigned char) value; - } + return value; } /** * Parse an escape sequence. */ -static int parse_escape_sequence(void) +static utf32 parse_escape_sequence(void) { eat('\\'); - int ec = input.c; + utf32 const ec = input.c; next_char(); - switch(ec) { + switch (ec) { case '"': return '"'; case '\'': return '\''; case '\\': return '\\'; @@ -422,164 +423,113 @@ static int parse_escape_sequence(void) case EOF: parse_error("reached end of file while parsing escape sequence"); return EOF; - default: - parse_error("unknown escape sequence"); + /* \E is not documented, but handled, by GCC. It is acceptable according + * to §6.11.4, whereas \e is not. */ + case 'E': + case 'e': + if (c_mode & _GNUC) + return 27; /* hopefully 27 is ALWAYS the code for ESCAPE */ + break; + case 'u': + case 'U': + parse_error("universal character parsing not implemented yet"); return EOF; + default: + break; } + /* §6.4.4.4:8 footnote 64 */ + parse_error("unknown escape sequence"); + return EOF; +} + +static const char *identify_string(char *string) +{ + const char *result = strset_insert(&stringset, string); + if (result != string) { + obstack_free(&symbol_obstack, string); + } + return result; +} + +static string_t make_string(char *string, size_t len) +{ + const char *result = identify_string(string); + return (string_t) {result, len}; } -static void parse_string_literal(void) +static void parse_string(utf32 const delimiter, preprocessor_token_kind_t const kind, string_encoding_t const enc, char const *const context) { const unsigned start_linenr = input.position.lineno; - eat('"'); + eat(delimiter); - int tc; - while(1) { - switch(input.c) { - case '\\': - if(resolve_escape_sequences) { - tc = parse_escape_sequence(); - obstack_1grow(&symbol_obstack, (char) tc); + while (true) { + switch (input.c) { + case '\\': { + if (resolve_escape_sequences) { + utf32 const tc = parse_escape_sequence(); + if (enc == STRING_ENCODING_CHAR) { + if (tc >= 0x100) { + warningf(WARN_OTHER, &pp_token.base.source_position, "escape sequence out of range"); + } + obstack_1grow(&symbol_obstack, tc); + } else { + obstack_grow_symbol(&symbol_obstack, tc); + } } else { - obstack_1grow(&symbol_obstack, (char) input.c); + obstack_1grow(&symbol_obstack, (char)input.c); next_char(); - obstack_1grow(&symbol_obstack, (char) input.c); + obstack_1grow(&symbol_obstack, (char)input.c); next_char(); } break; + } + + case NEWLINE: + errorf(&pp_token.base.source_position, "newline while parsing %s", context); + break; case EOF: { source_position_t source_position; - source_position.input_name = pp_token.source_position.input_name; + source_position.input_name = pp_token.base.source_position.input_name; source_position.lineno = start_linenr; - errorf(&source_position, "string has no end"); - pp_token.type = TP_ERROR; - return; - } - - case '"': - next_char(); + errorf(&source_position, "EOF while parsing %s", context); goto end_of_string; - - default: - obstack_1grow(&symbol_obstack, (char) input.c); - next_char(); - break; } - } - -end_of_string: - /* add finishing 0 to the string */ - obstack_1grow(&symbol_obstack, '\0'); - const size_t size = (size_t)obstack_object_size(&symbol_obstack); - const char *const string = obstack_finish(&symbol_obstack); - -#if 0 /* TODO hash */ - /* check if there is already a copy of the string */ - result = strset_insert(&stringset, string); - if(result != string) { - obstack_free(&symbol_obstack, string); - } -#else - const char *const result = string; -#endif - - pp_token.type = TP_STRING_LITERAL; - pp_token.literal.begin = result; - pp_token.literal.size = size; -} - -static void parse_wide_character_constant(void) -{ - eat('\''); - - int found_char = 0; - while(1) { - switch(input.c) { - case '\\': - found_char = parse_escape_sequence(); - break; - - MATCH_NEWLINE( - parse_error("newline while parsing character constant"); - break; - ) - - case '\'': - next_char(); - goto end_of_wide_char_constant; - - case EOF: - parse_error("EOF while parsing character constant"); - pp_token.type = TP_ERROR; - return; default: - if(found_char != 0) { - parse_error("more than 1 characters in character " - "constant"); - goto end_of_wide_char_constant; + if (input.c == delimiter) { + next_char(); + goto end_of_string; } else { - found_char = input.c; + obstack_grow_symbol(&symbol_obstack, input.c); next_char(); + break; } - break; } } -end_of_wide_char_constant: - pp_token.type = TP_WIDE_CHARACTER_CONSTANT; - /* TODO... */ +end_of_string:; + obstack_1grow(&symbol_obstack, '\0'); + size_t const size = obstack_object_size(&symbol_obstack) - 1; + char *const string = obstack_finish(&symbol_obstack); + + pp_token.kind = kind; + pp_token.string.encoding = enc; + pp_token.string.string = make_string(string, size); } -static void parse_character_constant(void) +static void parse_string_literal(string_encoding_t const enc) { - const unsigned start_linenr = input.position.lineno; - - eat('\''); - - int tc; - while(1) { - switch(input.c) { - case '\\': - tc = parse_escape_sequence(); - obstack_1grow(&symbol_obstack, (char) tc); - break; - - MATCH_NEWLINE( - parse_error("newline while parsing character constant"); - break; - ) - - case EOF: { - source_position_t source_position; - source_position.input_name = pp_token.source_position.input_name; - source_position.lineno = start_linenr; - errorf(&source_position, "EOF while parsing character constant"); - pp_token.type = TP_ERROR; - return; - } - - case '\'': - next_char(); - goto end_of_char_constant; - - default: - obstack_1grow(&symbol_obstack, (char) input.c); - next_char(); - break; + parse_string('"', TP_STRING_LITERAL, enc, "string literal"); +} - } +static void parse_character_constant(string_encoding_t const enc) +{ + parse_string('\'', TP_CHARACTER_CONSTANT, enc, "character constant"); + if (pp_token.string.string.size == 0) { + parse_error("empty character constant"); } - -end_of_char_constant:; - const size_t size = (size_t)obstack_object_size(&symbol_obstack); - const char *const string = obstack_finish(&symbol_obstack); - - pp_token.type = TP_CHARACTER_CONSTANT; - pp_token.literal.begin = string; - pp_token.literal.size = size; } #define SYMBOL_CHARS_WITHOUT_E_P \ @@ -662,7 +612,7 @@ static void expand_next(void) pp_definition_t *definition = current_expansion; restart: - if(definition->list_len == 0 + if (definition->list_len == 0 || definition->expand_pos >= definition->list_len) { /* we're finished with the current macro, move up 1 level in the * expansion stack */ @@ -671,7 +621,7 @@ restart: definition->is_expanding = false; /* it was the outermost expansion, parse normal pptoken */ - if(parent == NULL) { + if (parent == NULL) { current_expansion = NULL; next_preprocessing_token(); return; @@ -681,14 +631,15 @@ restart: goto restart; } pp_token = definition->token_list[definition->expand_pos]; + pp_token.base.source_position = expansion_pos; ++definition->expand_pos; - if(pp_token.type != TP_IDENTIFIER) + if (pp_token.kind != TP_IDENTIFIER) return; /* if it was an identifier then we might need to expand again */ - pp_definition_t *symbol_definition = pp_token.symbol->pp_definition; - if(symbol_definition != NULL && !symbol_definition->is_expanding) { + pp_definition_t *const symbol_definition = pp_token.base.symbol->pp_definition; + if (symbol_definition != NULL && !symbol_definition->is_expanding) { symbol_definition->parent_expansion = definition; symbol_definition->expand_pos = 0; symbol_definition->is_expanding = true; @@ -700,16 +651,13 @@ restart: static void skip_line_comment(void) { - if(do_print_spaces) - counted_spaces++; - - while(1) { - switch(input.c) { + while (true) { + switch (input.c) { case EOF: return; - case '\n': case '\r': + case '\n': return; default: @@ -721,12 +669,9 @@ static void skip_line_comment(void) static void skip_multiline_comment(void) { - if(do_print_spaces) - counted_spaces++; - unsigned start_linenr = input.position.lineno; - while(1) { - switch(input.c) { + while (true) { + switch (input.c) { case '/': next_char(); if (input.c == '*') { @@ -735,23 +680,20 @@ static void skip_multiline_comment(void) break; case '*': next_char(); - if(input.c == '/') { + if (input.c == '/') { + if (input.position.lineno != input.output_line) + info.whitespace = input.position.colno; next_char(); return; } break; - MATCH_NEWLINE( - if(do_print_spaces) { - counted_newlines++; - counted_spaces = 0; - } + case NEWLINE: break; - ) case EOF: { source_position_t source_position; - source_position.input_name = pp_token.source_position.input_name; + source_position.input_name = pp_token.base.source_position.input_name; source_position.lineno = start_linenr; errorf(&source_position, "at end of file while looking for comment end"); return; @@ -764,17 +706,19 @@ static void skip_multiline_comment(void) } } -/* skip spaces advancing at the start of the next preprocessing token */ -static void skip_spaces(bool skip_newline) +static void skip_whitespace(void) { while (true) { switch (input.c) { case ' ': case '\t': - if (do_print_spaces) - counted_spaces++; next_char(); continue; + + case NEWLINE: + info.at_line_begin = true; + return; + case '/': next_char(); if (input.c == '/') { @@ -790,40 +734,16 @@ static void skip_spaces(bool skip_newline) input.c = '/'; } return; - - case '\r': - if (!skip_newline) - return; - - next_char(); - if(input.c == '\n') { - next_char(); - } - ++input.position.lineno; - if (do_print_spaces) - ++counted_newlines; - continue; - - case '\n': - if (!skip_newline) - return; - - next_char(); - ++input.position.lineno; - if (do_print_spaces) - ++counted_newlines; - continue; - default: return; } } } -static void eat_pp(int type) +static void eat_pp(preprocessor_token_kind_t const type) { (void) type; - assert(pp_token.type == type); + assert(pp_token.kind == type); next_preprocessing_token(); } @@ -832,8 +752,8 @@ static void parse_symbol(void) obstack_1grow(&symbol_obstack, (char) input.c); next_char(); - while(1) { - switch(input.c) { + while (true) { + switch (input.c) { DIGITS SYMBOL_CHARS obstack_1grow(&symbol_obstack, (char) input.c); @@ -852,49 +772,24 @@ end_symbol: /* might be a wide string or character constant ( L"string"/L'c' ) */ if (input.c == '"' && string[0] == 'L' && string[1] == '\0') { obstack_free(&symbol_obstack, string); - /* TODO */ + parse_string_literal(STRING_ENCODING_WIDE); return; } else if (input.c == '\'' && string[0] == 'L' && string[1] == '\0') { obstack_free(&symbol_obstack, string); - parse_wide_character_constant(); + parse_character_constant(STRING_ENCODING_WIDE); return; } symbol_t *symbol = symbol_table_insert(string); - pp_token.type = symbol->pp_ID; - pp_token.symbol = symbol; + pp_token.kind = symbol->pp_ID; + pp_token.base.symbol = symbol; /* we can free the memory from symbol obstack if we already had an entry in * the symbol table */ if (symbol->string != string) { obstack_free(&symbol_obstack, string); } - if (!do_expansions) - return; - - pp_definition_t *pp_definition = symbol->pp_definition; - if (pp_definition == NULL) - return; - - if (pp_definition->has_parameters) { - skip_spaces(true); - /* no opening brace -> no expansion */ - if (input.c != '(') - return; - next_preprocessing_token(); - eat_pp('('); - - /* parse arguments (TODO) */ - while (pp_token.type != TP_EOF && pp_token.type != ')') - next_preprocessing_token(); - next_preprocessing_token(); - } - - pp_definition->expand_pos = 0; - pp_definition->is_expanding = true, - current_expansion = pp_definition; - expand_next(); } static void parse_number(void) @@ -902,8 +797,8 @@ static void parse_number(void) obstack_1grow(&symbol_obstack, (char) input.c); next_char(); - while(1) { - switch(input.c) { + while (true) { + switch (input.c) { case '.': DIGITS SYMBOL_CHARS_WITHOUT_E_P @@ -917,7 +812,7 @@ static void parse_number(void) case 'P': obstack_1grow(&symbol_obstack, (char) input.c); next_char(); - if(input.c == '+' || input.c == '-') { + if (input.c == '+' || input.c == '-') { obstack_1grow(&symbol_obstack, (char) input.c); next_char(); } @@ -933,21 +828,20 @@ end_number: size_t size = obstack_object_size(&symbol_obstack); char *string = obstack_finish(&symbol_obstack); - pp_token.type = TP_NUMBER; - pp_token.literal.begin = string; - pp_token.literal.size = size; + pp_token.kind = TP_NUMBER; + pp_token.number.number = make_string(string, size); } #define MAYBE_PROLOG \ next_char(); \ - while(1) { \ - switch(input.c) { + while (true) { \ + switch (input.c) { #define MAYBE(ch, set_type) \ case ch: \ next_char(); \ - pp_token.type = set_type; \ + pp_token.kind = set_type; \ return; #define ELSE_CODE(code) \ @@ -955,37 +849,38 @@ end_number: code \ return; \ } \ - } /* end of while(1) */ \ + } #define ELSE(set_type) \ ELSE_CODE( \ - pp_token.type = set_type; \ + pp_token.kind = set_type; \ ) static void next_preprocessing_token(void) { - if(current_expansion != NULL) { + if (current_expansion != NULL) { expand_next(); return; } - pp_token.source_position = input.position; - + info.at_line_begin = false; + info.had_whitespace = false; restart: - switch(input.c) { + pp_token.base.source_position = input.position; + pp_token.base.symbol = NULL; + + switch (input.c) { case ' ': case '\t': - if (do_print_spaces) - counted_spaces++; + ++info.whitespace; + info.had_whitespace = true; next_char(); goto restart; - MATCH_NEWLINE( - counted_newlines++; - counted_spaces = 0; - pp_token.type = '\n'; - return; - ) + case NEWLINE: + info.at_line_begin = true; + info.had_whitespace = true; + goto restart; SYMBOL_CHARS parse_symbol(); @@ -996,11 +891,11 @@ restart: return; case '"': - parse_string_literal(); + parse_string_literal(STRING_ENCODING_CHAR); return; case '\'': - parse_character_constant(); + parse_character_constant(STRING_ENCODING_CHAR); return; case '.': @@ -1026,7 +921,7 @@ restart: ELSE_CODE( put_back(input.c); input.c = '.'; - pp_token.type = '.'; + pp_token.kind = '.'; ) ELSE('.') case '&': @@ -1058,10 +953,12 @@ restart: MAYBE('=', TP_SLASHEQUAL) case '*': next_char(); + info.had_whitespace = true; skip_multiline_comment(); goto restart; case '/': next_char(); + info.had_whitespace = true; skip_line_comment(); goto restart; ELSE('/') @@ -1077,7 +974,7 @@ restart: ELSE_CODE( put_back(input.c); input.c = '%'; - pp_token.type = '#'; + pp_token.kind = '#'; ) ELSE('#') ELSE('%') @@ -1119,7 +1016,9 @@ restart: case '#': MAYBE_PROLOG MAYBE('#', TP_HASHHASH) - ELSE('#') + ELSE_CODE( + pp_token.kind = '#'; + ) case '?': case '[': @@ -1132,7 +1031,7 @@ restart: case ';': case ',': case '\\': - pp_token.type = input.c; + pp_token.kind = input.c; next_char(); return; @@ -1140,22 +1039,26 @@ restart: if (input_stack != NULL) { close_input(); pop_restore_input(); - counted_newlines = 0; - counted_spaces = 0; - /* hack to output correct line number */ + fputc('\n', out); print_line_directive(&input.position, "2"); - next_preprocessing_token(); + goto restart; } else { - pp_token.type = TP_EOF; + pp_token.base.source_position.lineno++; + info.at_line_begin = true; + pp_token.kind = TP_EOF; } return; default: next_char(); - errorf(&pp_token.source_position, "unknown character '%c' found\n", - input.c); - pp_token.type = TP_ERROR; - return; + if (!ignore_unknown_chars) { + errorf(&pp_token.base.source_position, + "unknown character '%c' found\n", input.c); + goto restart; + } else { + pp_token.kind = input.c; + return; + } } } @@ -1163,7 +1066,7 @@ static void print_quoted_string(const char *const string) { fputc('"', out); for (const char *c = string; *c != 0; ++c) { - switch(*c) { + switch (*c) { case '"': fputs("\\\"", out); break; case '\\': fputs("\\\\", out); break; case '\a': fputs("\\a", out); break; @@ -1175,7 +1078,7 @@ static void print_quoted_string(const char *const string) case '\v': fputs("\\v", out); break; case '\?': fputs("\\?", out); break; default: - if(!isprint(*c)) { + if (!isprint(*c)) { fprintf(out, "\\%03o", (unsigned)*c); break; } @@ -1194,27 +1097,32 @@ static void print_line_directive(const source_position_t *pos, const char *add) fputc(' ', out); fputs(add, out); } - fputc('\n', out); printed_input_name = pos->input_name; + input.output_line = pos->lineno-1; } -static void print_spaces(void) +static bool emit_newlines(void) { - if (counted_newlines >= 9) { - if (input.had_non_space) { - fputc('\n', out); - } - print_line_directive(&pp_token.source_position, NULL); - counted_newlines = 0; + unsigned delta = pp_token.base.source_position.lineno - input.output_line; + if (delta == 0) + return false; + + if (delta >= 9) { + fputc('\n', out); + print_line_directive(&pp_token.base.source_position, NULL); + fputc('\n', out); } else { - for (unsigned i = 0; i < counted_newlines; ++i) + for (unsigned i = 0; i < delta; ++i) { fputc('\n', out); - counted_newlines = 0; + } } - for (unsigned i = 0; i < counted_spaces; ++i) + input.output_line = pp_token.base.source_position.lineno; + + for (unsigned i = 0; i < info.whitespace; ++i) fputc(' ', out); - counted_spaces = 0; + + return true; } static void emit_pp_token(void) @@ -1222,34 +1130,41 @@ static void emit_pp_token(void) if (skip_mode) return; - if (pp_token.type != '\n') { - print_spaces(); - input.had_non_space = true; - } + if (!emit_newlines() && + (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind))) + fputc(' ', out); - switch(pp_token.type) { + switch (pp_token.kind) { case TP_IDENTIFIER: - fputs(pp_token.symbol->string, out); + fputs(pp_token.base.symbol->string, out); break; case TP_NUMBER: - fputs(pp_token.literal.begin, out); + fputs(pp_token.number.number.begin, out); break; + case TP_STRING_LITERAL: + fputs(get_string_encoding_prefix(pp_token.string.encoding), out); fputc('"', out); - fputs(pp_token.literal.begin, out); + fputs(pp_token.string.string.begin, out); fputc('"', out); break; - case '\n': + + case TP_CHARACTER_CONSTANT: + fputs(get_string_encoding_prefix(pp_token.string.encoding), out); + fputc('\'', out); + fputs(pp_token.string.string.begin, out); + fputc('\'', out); break; default: - print_pp_token_type(out, pp_token.type); + print_pp_token_kind(out, pp_token.kind); break; } + last_token = pp_token.kind; } static void eat_pp_directive(void) { - while(pp_token.type != '\n' && pp_token.type != TP_EOF) { + while (!info.at_line_begin) { next_preprocessing_token(); } } @@ -1257,13 +1172,13 @@ static void eat_pp_directive(void) static bool strings_equal(const string_t *string1, const string_t *string2) { size_t size = string1->size; - if(size != string2->size) + if (size != string2->size) return false; const char *c1 = string1->begin; const char *c2 = string2->begin; - for(size_t i = 0; i < size; ++i, ++c1, ++c2) { - if(*c1 != *c2) + for (size_t i = 0; i < size; ++i, ++c1, ++c2) { + if (*c1 != *c2) return false; } return true; @@ -1271,19 +1186,17 @@ static bool strings_equal(const string_t *string1, const string_t *string2) static bool pp_tokens_equal(const token_t *token1, const token_t *token2) { - if(token1->type != token2->type) + if (token1->kind != token2->kind) return false; - switch(token1->type) { - case TP_HEADERNAME: - /* TODO */ - return false; + switch (token1->kind) { case TP_IDENTIFIER: - return token1->symbol == token2->symbol; + return token1->base.symbol == token2->base.symbol; + case TP_NUMBER: case TP_CHARACTER_CONSTANT: case TP_STRING_LITERAL: - return strings_equal(&token1->literal, &token2->literal); + return strings_equal(&token1->string.string, &token2->string.string); default: return true; @@ -1293,14 +1206,14 @@ static bool pp_tokens_equal(const token_t *token1, const token_t *token2) static bool pp_definitions_equal(const pp_definition_t *definition1, const pp_definition_t *definition2) { - if(definition1->list_len != definition2->list_len) + if (definition1->list_len != definition2->list_len) return false; size_t len = definition1->list_len; const token_t *t1 = definition1->token_list; const token_t *t2 = definition2->token_list; - for(size_t i = 0; i < len; ++i, ++t1, ++t2) { - if(!pp_tokens_equal(t1, t2)) + for (size_t i = 0; i < len; ++i, ++t1, ++t2) { + if (!pp_tokens_equal(t1, t2)) return false; } return true; @@ -1311,12 +1224,12 @@ static void parse_define_directive(void) eat_pp(TP_define); assert(obstack_object_size(&pp_obstack) == 0); - if (pp_token.type != TP_IDENTIFIER) { - errorf(&pp_token.source_position, + if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) { + errorf(&pp_token.base.source_position, "expected identifier after #define, got '%t'", &pp_token); goto error_out; } - symbol_t *symbol = pp_token.symbol; + symbol_t *const symbol = pp_token.base.symbol; pp_definition_t *new_definition = obstack_alloc(&pp_obstack, sizeof(new_definition[0])); @@ -1333,27 +1246,27 @@ static void parse_define_directive(void) next_preprocessing_token(); while (true) { - switch (pp_token.type) { + switch (pp_token.kind) { case TP_DOTDOTDOT: new_definition->is_variadic = true; next_preprocessing_token(); - if (pp_token.type != ')') { + if (pp_token.kind != ')') { errorf(&input.position, "'...' not at end of macro argument list"); goto error_out; } break; case TP_IDENTIFIER: - obstack_ptr_grow(&pp_obstack, pp_token.symbol); + obstack_ptr_grow(&pp_obstack, pp_token.base.symbol); next_preprocessing_token(); - if (pp_token.type == ',') { + if (pp_token.kind == ',') { next_preprocessing_token(); break; } - if (pp_token.type != ')') { - errorf(&pp_token.source_position, + if (pp_token.kind != ')') { + errorf(&pp_token.base.source_position, "expected ',' or ')' after identifier, got '%t'", &pp_token); goto error_out; @@ -1363,7 +1276,7 @@ static void parse_define_directive(void) next_preprocessing_token(); goto finish_argument_list; default: - errorf(&pp_token.source_position, + errorf(&pp_token.base.source_position, "expected identifier, '...' or ')' in #define argument list, got '%t'", &pp_token); goto error_out; @@ -1382,7 +1295,7 @@ static void parse_define_directive(void) /* construct a new pp_definition on the obstack */ assert(obstack_object_size(&pp_obstack) == 0); size_t list_len = 0; - while (pp_token.type != '\n' && pp_token.type != TP_EOF) { + while (!info.at_line_begin) { obstack_grow(&pp_obstack, &pp_token, sizeof(pp_token)); ++list_len; next_preprocessing_token(); @@ -1417,138 +1330,179 @@ static void parse_undef_directive(void) { eat_pp(TP_undef); - if(pp_token.type != TP_IDENTIFIER) { + if (pp_token.kind != TP_IDENTIFIER) { errorf(&input.position, "expected identifier after #undef, got '%t'", &pp_token); eat_pp_directive(); return; } - symbol_t *symbol = pp_token.symbol; - symbol->pp_definition = NULL; + pp_token.base.symbol->pp_definition = NULL; next_preprocessing_token(); - if(pp_token.type != '\n') { + if (!info.at_line_begin) { warningf(WARN_OTHER, &input.position, "extra tokens at end of #undef directive"); } - /* eat until '\n' */ eat_pp_directive(); } -static const char *parse_headername(void) +static void parse_headername(void) { + const source_position_t start_position = input.position; + string_t string = {NULL, 0}; + assert(obstack_object_size(&symbol_obstack) == 0); + /* behind an #include we can have the special headername lexems. * They're only allowed behind an #include so they're not recognized * by the normal next_preprocessing_token. We handle them as a special * exception here */ - - /* skip spaces so we reach start of next preprocessing token */ - skip_spaces(false); - - assert(obstack_object_size(&input_obstack) == 0); + if (info.at_line_begin) { + parse_error("expected headername after #include"); + goto finish_error; + } /* check wether we have a "... or <... headername */ switch (input.c) { - case '<': - /* for now until we have proper searchpath handling */ - obstack_1grow(&input_obstack, '.'); - obstack_1grow(&input_obstack, '/'); - + { + utf32 delimiter; + case '<': delimiter = '>'; goto parse_name; + case '"': delimiter = '"'; goto parse_name; +parse_name: next_char(); while (true) { switch (input.c) { + case NEWLINE: case EOF: - /* fallthrough */ - MATCH_NEWLINE( - parse_error("header name without closing '>'"); - return NULL; - ) - case '>': - next_char(); - goto finished_headername; - } - obstack_1grow(&input_obstack, (char) input.c); - next_char(); - } - /* we should never be here */ + errorf(&pp_token.base.source_position, "header name without closing '%c'", (char)delimiter); + goto finish_error; - case '"': - /* for now until we have proper searchpath handling */ - obstack_1grow(&input_obstack, '.'); - obstack_1grow(&input_obstack, '/'); - - next_char(); - while (true) { - switch (input.c) { - case EOF: - /* fallthrough */ - MATCH_NEWLINE( - parse_error("header name without closing '>'"); - return NULL; - ) - case '"': - next_char(); - goto finished_headername; + default: + if (input.c == delimiter) { + next_char(); + goto finished_headername; + } else { + obstack_1grow(&symbol_obstack, (char)input.c); + next_char(); + } + break; } - obstack_1grow(&input_obstack, (char) input.c); - next_char(); } /* we should never be here */ + } default: - /* TODO: do normale pp_token parsing and concatenate results */ + /* TODO: do normal pp_token parsing and concatenate results */ panic("pp_token concat include not implemented yet"); } finished_headername: - obstack_1grow(&input_obstack, '\0'); - char *headername = obstack_finish(&input_obstack); + obstack_1grow(&symbol_obstack, '\0'); + const size_t size = (size_t)obstack_object_size(&symbol_obstack); + char *const headername = obstack_finish(&symbol_obstack); + string = make_string(headername, size); + +finish_error: + pp_token.base.source_position = start_position; + pp_token.kind = TP_HEADERNAME; + pp_token.string.string = string; +} - /* TODO: iterate search-path to find the file */ +static bool do_include(bool system_include, const char *headername) +{ + size_t headername_len = strlen(headername); + if (!system_include) { + /* put dirname of current input on obstack */ + const char *filename = input.position.input_name; + const char *last_slash = strrchr(filename, '/'); + if (last_slash != NULL) { + size_t len = last_slash - filename; + obstack_grow(&symbol_obstack, filename, len + 1); + obstack_grow0(&symbol_obstack, headername, headername_len); + char *complete_path = obstack_finish(&symbol_obstack); + headername = identify_string(complete_path); + } - next_preprocessing_token(); + FILE *file = fopen(headername, "r"); + if (file != NULL) { + switch_input(file, headername); + return true; + } + } - return headername; + assert(obstack_object_size(&symbol_obstack) == 0); + /* check searchpath */ + for (searchpath_entry_t *entry = searchpath; entry != NULL; + entry = entry->next) { + const char *path = entry->path; + size_t len = strlen(path); + obstack_grow(&symbol_obstack, path, len); + if (path[len-1] != '/') + obstack_1grow(&symbol_obstack, '/'); + obstack_grow(&symbol_obstack, headername, headername_len+1); + + char *complete_path = obstack_finish(&symbol_obstack); + FILE *file = fopen(complete_path, "r"); + if (file != NULL) { + const char *filename = identify_string(complete_path); + switch_input(file, filename); + return true; + } else { + obstack_free(&symbol_obstack, complete_path); + } + } + + return false; +} + +/* read till next newline character, only for parse_include_directive(), + * use eat_pp_directive() in all other cases */ +static void skip_till_newline(void) +{ + /* skip till newline */ + while (true) { + switch (input.c) { + case NEWLINE: + case EOF: + return; + } + next_char(); + } } static bool parse_include_directive(void) { /* don't eat the TP_include here! * we need an alternative parsing for the next token */ - - print_spaces(); - - const char *headername = parse_headername(); - if (headername == NULL) { + skip_whitespace(); + bool system_include = input.c == '<'; + parse_headername(); + string_t headername = pp_token.string.string; + if (headername.begin == NULL) { eat_pp_directive(); return false; } - if (pp_token.type != '\n' && pp_token.type != TP_EOF) { - warningf(WARN_OTHER, &pp_token.source_position, "extra tokens at end of #include directive"); - eat_pp_directive(); + skip_whitespace(); + if (!info.at_line_begin) { + warningf(WARN_OTHER, &pp_token.base.source_position, + "extra tokens at end of #include directive"); + skip_till_newline(); } if (n_inputs > INCLUDE_LIMIT) { - errorf(&pp_token.source_position, "#include nested too deeply"); + errorf(&pp_token.base.source_position, "#include nested too deeply"); /* eat \n or EOF */ next_preprocessing_token(); return false; } - /* we have to reenable space counting and macro expansion here, - * because it is still disabled in directive parsing, - * but we will trigger a preprocessing token reading of the new file - * now and need expansions/space counting */ - do_print_spaces = true; - do_expansions = true; - /* switch inputs */ + emit_newlines(); push_input(); - bool res = open_input(headername); + bool res = do_include(system_include, pp_token.string.string.begin); if (!res) { - errorf(&pp_token.source_position, - "failed including '%s': %s", headername, strerror(errno)); + errorf(&pp_token.base.source_position, + "failed including '%S': %s", pp_token.string, strerror(errno)); pop_restore_input(); return false; } @@ -1590,20 +1544,20 @@ static void check_unclosed_conditionals(void) static void parse_ifdef_ifndef_directive(void) { - bool is_ifndef = (pp_token.type == TP_ifndef); + bool is_ifndef = (pp_token.kind == TP_ifndef); bool condition; next_preprocessing_token(); if (skip_mode) { eat_pp_directive(); pp_conditional_t *conditional = push_conditional(); - conditional->source_position = pp_token.source_position; + conditional->source_position = pp_token.base.source_position; conditional->skip = true; return; } - if (pp_token.type != TP_IDENTIFIER) { - errorf(&pp_token.source_position, + if (pp_token.kind != TP_IDENTIFIER || info.at_line_begin) { + errorf(&pp_token.base.source_position, "expected identifier after #%s, got '%t'", is_ifndef ? "ifndef" : "ifdef", &pp_token); eat_pp_directive(); @@ -1611,23 +1565,21 @@ static void parse_ifdef_ifndef_directive(void) /* just take the true case in the hope to avoid further errors */ condition = true; } else { - symbol_t *symbol = pp_token.symbol; - pp_definition_t *pp_definition = symbol->pp_definition; + /* evaluate wether we are in true or false case */ + condition = !pp_token.base.symbol->pp_definition == is_ifndef; + next_preprocessing_token(); - if (pp_token.type != '\n') { - errorf(&pp_token.source_position, + if (!info.at_line_begin) { + errorf(&pp_token.base.source_position, "extra tokens at end of #%s", is_ifndef ? "ifndef" : "ifdef"); eat_pp_directive(); } - - /* evaluate wether we are in true or false case */ - condition = is_ifndef ? pp_definition == NULL : pp_definition != NULL; } pp_conditional_t *conditional = push_conditional(); - conditional->source_position = pp_token.source_position; + conditional->source_position = pp_token.base.source_position; conditional->condition = condition; if (!condition) { @@ -1639,21 +1591,21 @@ static void parse_else_directive(void) { eat_pp(TP_else); - if (pp_token.type != '\n') { + if (!info.at_line_begin) { if (!skip_mode) { - warningf(WARN_OTHER, &pp_token.source_position, "extra tokens at end of #else"); + warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #else"); } eat_pp_directive(); } pp_conditional_t *conditional = conditional_stack; if (conditional == NULL) { - errorf(&pp_token.source_position, "#else without prior #if"); + errorf(&pp_token.base.source_position, "#else without prior #if"); return; } if (conditional->in_else) { - errorf(&pp_token.source_position, + errorf(&pp_token.base.source_position, "#else after #else (condition started %P)", conditional->source_position); skip_mode = true; @@ -1664,23 +1616,23 @@ static void parse_else_directive(void) if (!conditional->skip) { skip_mode = conditional->condition; } - conditional->source_position = pp_token.source_position; + conditional->source_position = pp_token.base.source_position; } static void parse_endif_directive(void) { eat_pp(TP_endif); - if (pp_token.type != '\n') { + if (!info.at_line_begin) { if (!skip_mode) { - warningf(WARN_OTHER, &pp_token.source_position, "extra tokens at end of #endif"); + warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #endif"); } eat_pp_directive(); } pp_conditional_t *conditional = conditional_stack; if (conditional == NULL) { - errorf(&pp_token.source_position, "#endif without prior #if"); + errorf(&pp_token.base.source_position, "#endif without prior #if"); return; } @@ -1692,12 +1644,15 @@ static void parse_endif_directive(void) static void parse_preprocessing_directive(void) { - do_print_spaces = false; - do_expansions = false; eat_pp('#'); + if (info.at_line_begin) { + /* empty directive */ + return; + } + if (skip_mode) { - switch(pp_token.type) { + switch (pp_token.kind) { case TP_ifdef: case TP_ifndef: parse_ifdef_ifndef_directive(); @@ -1713,7 +1668,7 @@ static void parse_preprocessing_directive(void) break; } } else { - switch(pp_token.type) { + switch (pp_token.kind) { case TP_define: parse_define_directive(); break; @@ -1730,33 +1685,67 @@ static void parse_preprocessing_directive(void) case TP_endif: parse_endif_directive(); break; - case TP_include: { - bool in_new_source = parse_include_directive(); - /* no need to do anything if source file switched */ - if (in_new_source) - return; - break; - } - case '\n': - /* the nop directive */ + case TP_include: + parse_include_directive(); break; default: - errorf(&pp_token.source_position, + if (info.at_line_begin) { + /* the nop directive "#" */ + break; + } + errorf(&pp_token.base.source_position, "invalid preprocessing directive #%t", &pp_token); eat_pp_directive(); break; } } - do_print_spaces = true; - do_expansions = true; + assert(info.at_line_begin); +} - /* eat '\n' */ - assert(pp_token.type == '\n' || pp_token.type == TP_EOF); - next_preprocessing_token(); +static void prepend_include_path(const char *path) +{ + searchpath_entry_t *entry = OALLOCZ(&config_obstack, searchpath_entry_t); + entry->path = path; + entry->next = searchpath; + searchpath = entry; } -#define GCC_COMPAT_MODE +static void setup_include_path(void) +{ + /* built-in paths */ + prepend_include_path("/usr/include"); + + /* parse environment variable */ + const char *cpath = getenv("CPATH"); + if (cpath != NULL && *cpath != '\0') { + const char *begin = cpath; + const char *c; + do { + c = begin; + while (*c != '\0' && *c != ':') + ++c; + + size_t len = c-begin; + if (len == 0) { + /* for gcc compatibility (Matze: I would expect that + * nothing happens for an empty entry...) */ + prepend_include_path("."); + } else { + char *string = obstack_alloc(&config_obstack, len+1); + memcpy(string, begin, len); + string[len] = '\0'; + + prepend_include_path(string); + } + + begin = c+1; + /* skip : */ + if (*begin == ':') + ++begin; + } while(*c != '\0'); + } +} int pptest_main(int argc, char **argv); int pptest_main(int argc, char **argv) @@ -1764,48 +1753,124 @@ int pptest_main(int argc, char **argv) init_symbol_table(); init_tokens(); + obstack_init(&config_obstack); obstack_init(&pp_obstack); obstack_init(&input_obstack); + strset_init(&stringset); - const char *filename = "t.c"; - if (argc > 1) - filename = argv[1]; + setup_include_path(); - out = stdout; + /* simplistic commandline parser */ + const char *filename = NULL; + const char *output = NULL; + for (int i = 1; i < argc; ++i) { + const char *opt = argv[i]; + if (streq(opt, "-I")) { + prepend_include_path(argv[++i]); + continue; + } else if (streq(opt, "-E")) { + /* ignore */ + } else if (streq(opt, "-o")) { + output = argv[++i]; + continue; + } else if (opt[0] == '-') { + fprintf(stderr, "Unknown option '%s'\n", opt); + } else { + if (filename != NULL) + fprintf(stderr, "Multiple inputs not supported\n"); + filename = argv[i]; + } + } + if (filename == NULL) { + fprintf(stderr, "No input specified\n"); + return 1; + } + + if (output == NULL) { + out = stdout; + } else { + out = fopen(output, "w"); + if (out == NULL) { + fprintf(stderr, "Couldn't open output '%s'\n", output); + return 1; + } + } -#ifdef GCC_COMPAT_MODE - /* this is here so we can directly compare "gcc -E" output and our output */ + /* just here for gcc compatibility */ fprintf(out, "# 1 \"%s\"\n", filename); - fputs("# 1 \"\"\n", out); - fputs("# 1 \"\"\n", out); -#endif + fprintf(out, "# 1 \"\"\n"); + fprintf(out, "# 1 \"\"\n"); - bool ok = open_input(filename); - assert(ok); + FILE *file = fopen(filename, "r"); + if (file == NULL) { + fprintf(stderr, "Couldn't open input '%s'\n", filename); + return 1; + } + switch_input(file, filename); - while(true) { - /* we're at a line begin */ - if(pp_token.type == '#') { + while (true) { + if (pp_token.kind == '#' && info.at_line_begin) { parse_preprocessing_directive(); - } else { - /* parse+emit a line */ - while(pp_token.type != '\n') { - if(pp_token.type == TP_EOF) - goto end_of_main_loop; - emit_pp_token(); - next_preprocessing_token(); + continue; + } else if (pp_token.kind == TP_EOF) { + goto end_of_main_loop; + } else if (pp_token.kind == TP_IDENTIFIER) { + symbol_t *const symbol = pp_token.base.symbol; + pp_definition_t *const pp_definition = symbol->pp_definition; + if (pp_definition != NULL && !pp_definition->is_expanding) { + expansion_pos = pp_token.base.source_position; + if (pp_definition->has_parameters) { + source_position_t position = pp_token.base.source_position; + add_token_info_t old_info = info; + next_preprocessing_token(); + add_token_info_t new_info = info; + + /* no opening brace -> no expansion */ + if (pp_token.kind == '(') { + eat_pp('('); + + /* parse arguments (TODO) */ + while (pp_token.kind != TP_EOF && pp_token.kind != ')') + next_preprocessing_token(); + } else { + token_t next_token = pp_token; + /* restore identifier token */ + pp_token.kind = TP_IDENTIFIER; + pp_token.base.symbol = symbol; + pp_token.base.source_position = position; + info = old_info; + emit_pp_token(); + + info = new_info; + pp_token = next_token; + continue; + } + info = old_info; + } + pp_definition->expand_pos = 0; + pp_definition->is_expanding = true; + current_expansion = pp_definition; + expand_next(); + continue; } - emit_pp_token(); - next_preprocessing_token(); } + + emit_pp_token(); + next_preprocessing_token(); } end_of_main_loop: + fputc('\n', out); check_unclosed_conditionals(); close_input(); + if (out != stdout) + fclose(out); obstack_free(&input_obstack, NULL); obstack_free(&pp_obstack, NULL); + obstack_free(&config_obstack, NULL); + + strset_destroy(&stringset); exit_tokens(); exit_symbol_table();