X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;ds=sidebyside;f=preprocessor.c;h=a9c10d429ba8fcf2e2a48c68ef20401e283478e0;hb=0a880c90fd6d1ec23c4ec1956f3b92694ec0a139;hp=a8cf16db43ea462ab582b2caf809f60780dae773;hpb=f845d7fde121d7548f2ccb49eff3bee6cb842a05;p=cparser diff --git a/preprocessor.c b/preprocessor.c index a8cf16d..a9c10d4 100644 --- a/preprocessor.c +++ b/preprocessor.c @@ -91,7 +91,6 @@ static pp_conditional_t *conditional_stack; static token_t pp_token; static bool resolve_escape_sequences = false; static bool ignore_unknown_chars = true; -static bool in_pp_directive; static bool skip_mode; static FILE *out; static struct obstack pp_obstack; @@ -100,7 +99,7 @@ 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 = TP_ERROR; +static preprocessor_token_kind_t last_token; static searchpath_entry_t *searchpath; @@ -195,8 +194,7 @@ 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; @@ -220,17 +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') { \ - case '\n': \ - next_char(); \ - } \ - info.whitespace = 0; \ - ++input.position.lineno; \ - input.position.colno = 1; \ - 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()) @@ -239,9 +238,8 @@ static void maybe_concat_lines(void) eat('\\'); switch (input.c) { - MATCH_NEWLINE( + case NEWLINE: return; - ) default: break; @@ -459,162 +457,77 @@ static string_t make_string(char *string, size_t len) 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); while (true) { switch (input.c) { case '\\': { - utf32 tc; if (resolve_escape_sequences) { - tc = parse_escape_sequence(); - obstack_1grow(&symbol_obstack, (char) tc); + 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.base.source_position.input_name; source_position.lineno = start_linenr; - errorf(&source_position, "string has no end"); - pp_token.kind = TP_ERROR; - return; - } - - case '"': - next_char(); + errorf(&source_position, "EOF while parsing %s", context); goto end_of_string; + } default: - obstack_grow_symbol(&symbol_obstack, input.c); - next_char(); - break; + if (input.c == delimiter) { + next_char(); + goto end_of_string; + } else { + obstack_grow_symbol(&symbol_obstack, input.c); + next_char(); + break; + } } } -end_of_string: - /* add finishing 0 to the string */ +end_of_string:; obstack_1grow(&symbol_obstack, '\0'); - const size_t size = (size_t)obstack_object_size(&symbol_obstack); - char *const string = obstack_finish(&symbol_obstack); + size_t const size = obstack_object_size(&symbol_obstack) - 1; + char *const string = obstack_finish(&symbol_obstack); - pp_token.kind = TP_STRING_LITERAL; - pp_token.string.string = make_string(string, size); + pp_token.kind = kind; + pp_token.string.encoding = enc; + pp_token.string.string = make_string(string, size); } -/** - * Parse a wide string literal and set lexer_token. - */ -static void parse_wide_string_literal(void) +static void parse_string_literal(string_encoding_t const enc) { - parse_string_literal(); - if (pp_token.kind == TP_STRING_LITERAL) - pp_token.kind = TP_WIDE_STRING_LITERAL; + parse_string('"', TP_STRING_LITERAL, enc, "string literal"); } -static void parse_wide_character_constant(void) +static void parse_character_constant(string_encoding_t const enc) { - eat('\''); - - while (true) { - switch (input.c) { - case '\\': { - const utf32 tc = parse_escape_sequence(); - obstack_grow_symbol(&symbol_obstack, tc); - 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.kind = TP_ERROR; - return; - - default: - obstack_grow_symbol(&symbol_obstack, input.c); - next_char(); - break; - } - } - -end_of_wide_char_constant: - obstack_1grow(&symbol_obstack, '\0'); - size_t size = (size_t) obstack_object_size(&symbol_obstack)-1; - char *string = obstack_finish(&symbol_obstack); - pp_token.kind = TP_WIDE_CHARACTER_CONSTANT; - pp_token.string.string = make_string(string, size); - - if (size == 0) { - parse_error("empty character constant"); - } -} - -static void parse_character_constant(void) -{ - const unsigned start_linenr = input.position.lineno; - - eat('\''); - - int tc; - while (true) { - 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.base.source_position.input_name; - source_position.lineno = start_linenr; - errorf(&source_position, "EOF while parsing character constant"); - pp_token.kind = TP_ERROR; - return; - } - - case '\'': - next_char(); - goto end_of_char_constant; - - default: - obstack_1grow(&symbol_obstack, (char) input.c); - next_char(); - break; - - } - } - -end_of_char_constant:; - obstack_1grow(&symbol_obstack, '\0'); - const size_t size = (size_t)obstack_object_size(&symbol_obstack); - char *const string = obstack_finish(&symbol_obstack); - - pp_token.kind = TP_CHARACTER_CONSTANT; - pp_token.string.string = make_string(string, size); - - if (size == 0) { + parse_string('\'', TP_CHARACTER_CONSTANT, enc, "character constant"); + if (pp_token.string.string.size == 0) { parse_error("empty character constant"); } } @@ -725,7 +638,7 @@ restart: return; /* if it was an identifier then we might need to expand again */ - pp_definition_t *symbol_definition = pp_token.identifier.symbol->pp_definition; + 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; @@ -768,16 +681,15 @@ static void skip_multiline_comment(void) case '*': next_char(); if (input.c == '/') { + if (input.position.lineno != input.output_line) + info.whitespace = input.position.colno; next_char(); - info.whitespace += input.position.colno-1; return; } break; - MATCH_NEWLINE( - info.at_line_begin |= !in_pp_directive; + case NEWLINE: break; - ) case EOF: { source_position_t source_position; @@ -803,10 +715,9 @@ static void skip_whitespace(void) next_char(); continue; - MATCH_NEWLINE( + case NEWLINE: info.at_line_begin = true; return; - ) case '/': next_char(); @@ -829,7 +740,7 @@ static void skip_whitespace(void) } } -static void eat_pp(int type) +static void eat_pp(preprocessor_token_kind_t const type) { (void) type; assert(pp_token.kind == type); @@ -861,18 +772,18 @@ 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); - parse_wide_string_literal(); + 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.kind = symbol->pp_ID; - pp_token.identifier.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 */ @@ -956,6 +867,8 @@ static void next_preprocessing_token(void) info.had_whitespace = false; restart: pp_token.base.source_position = input.position; + pp_token.base.symbol = NULL; + switch (input.c) { case ' ': case '\t': @@ -964,11 +877,10 @@ restart: next_char(); goto restart; - MATCH_NEWLINE( + case NEWLINE: info.at_line_begin = true; info.had_whitespace = true; goto restart; - ) SYMBOL_CHARS parse_symbol(); @@ -979,11 +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 '.': @@ -1142,11 +1054,11 @@ restart: if (!ignore_unknown_chars) { errorf(&pp_token.base.source_position, "unknown character '%c' found\n", input.c); - pp_token.kind = TP_ERROR; + goto restart; } else { pp_token.kind = input.c; + return; } - return; } } @@ -1190,9 +1102,11 @@ static void print_line_directive(const source_position_t *pos, const char *add) input.output_line = pos->lineno-1; } -static void emit_newlines(void) +static bool emit_newlines(void) { unsigned delta = pp_token.base.source_position.lineno - input.output_line; + if (delta == 0) + return false; if (delta >= 9) { fputc('\n', out); @@ -1204,6 +1118,11 @@ static void emit_newlines(void) } } input.output_line = pp_token.base.source_position.lineno; + + for (unsigned i = 0; i < info.whitespace; ++i) + fputc(' ', out); + + return true; } static void emit_pp_token(void) @@ -1211,34 +1130,27 @@ static void emit_pp_token(void) if (skip_mode) return; - if (info.at_line_begin) { - emit_newlines(); - - for (unsigned i = 0; i < info.whitespace; ++i) - fputc(' ', out); - - } else if (info.had_whitespace || - tokens_would_paste(last_token, pp_token.kind)) { + if (!emit_newlines() && + (info.had_whitespace || tokens_would_paste(last_token, pp_token.kind))) fputc(' ', out); - } switch (pp_token.kind) { case TP_IDENTIFIER: - fputs(pp_token.identifier.symbol->string, out); + fputs(pp_token.base.symbol->string, out); break; case TP_NUMBER: fputs(pp_token.number.number.begin, out); break; - case TP_WIDE_STRING_LITERAL: - fputc('L', out); + case TP_STRING_LITERAL: + fputs(get_string_encoding_prefix(pp_token.string.encoding), out); fputc('"', out); fputs(pp_token.string.string.begin, out); fputc('"', out); break; - case TP_WIDE_CHARACTER_CONSTANT: - fputc('L', out); + 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); @@ -1279,7 +1191,8 @@ static bool pp_tokens_equal(const token_t *token1, const token_t *token2) switch (token1->kind) { case TP_IDENTIFIER: - return token1->identifier.symbol == token2->identifier.symbol; + return token1->base.symbol == token2->base.symbol; + case TP_NUMBER: case TP_CHARACTER_CONSTANT: case TP_STRING_LITERAL: @@ -1316,7 +1229,7 @@ static void parse_define_directive(void) "expected identifier after #define, got '%t'", &pp_token); goto error_out; } - symbol_t *symbol = pp_token.identifier.symbol; + symbol_t *const symbol = pp_token.base.symbol; pp_definition_t *new_definition = obstack_alloc(&pp_obstack, sizeof(new_definition[0])); @@ -1344,7 +1257,7 @@ static void parse_define_directive(void) } break; case TP_IDENTIFIER: - obstack_ptr_grow(&pp_obstack, pp_token.identifier.symbol); + obstack_ptr_grow(&pp_obstack, pp_token.base.symbol); next_preprocessing_token(); if (pp_token.kind == ',') { @@ -1424,8 +1337,7 @@ static void parse_undef_directive(void) return; } - symbol_t *symbol = pp_token.identifier.symbol; - symbol->pp_definition = NULL; + pp_token.base.symbol->pp_definition = NULL; next_preprocessing_token(); if (!info.at_line_begin) { @@ -1434,58 +1346,49 @@ static void parse_undef_directive(void) 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 */ if (info.at_line_begin) { parse_error("expected headername after #include"); - return NULL; + goto finish_error; } - assert(obstack_object_size(&symbol_obstack) == 0); - /* check wether we have a "... or <... headername */ switch (input.c) { - case '<': + { + 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(&symbol_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 '"': - 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(&symbol_obstack, (char) input.c); - next_char(); } /* we should never be here */ + } default: /* TODO: do normal pp_token parsing and concatenate results */ @@ -1494,21 +1397,31 @@ static const char *parse_headername(void) finished_headername: obstack_1grow(&symbol_obstack, '\0'); - char *headername = obstack_finish(&symbol_obstack); - - /* TODO: iterate search-path to find the file */ - - skip_whitespace(); - - return identify_string(headername); + 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; } static bool do_include(bool system_include, const char *headername) { + size_t headername_len = strlen(headername); if (!system_include) { - /* for "bla" includes first try current dir - * TODO: this isn't correct, should be the directory of the source file - */ + /* 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); + } + FILE *file = fopen(headername, "r"); if (file != NULL) { switch_input(file, headername); @@ -1516,47 +1429,64 @@ static bool do_include(bool system_include, const char *headername) } } - size_t headername_len = strlen(headername); - assert(obstack_object_size(&pp_obstack) == 0); + 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(&pp_obstack, path, len); + obstack_grow(&symbol_obstack, path, len); if (path[len-1] != '/') - obstack_1grow(&pp_obstack, '/'); - obstack_grow(&pp_obstack, headername, headername_len+1); + obstack_1grow(&symbol_obstack, '/'); + obstack_grow(&symbol_obstack, headername, headername_len+1); - char *complete_path = obstack_finish(&pp_obstack); + 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); } - obstack_free(&pp_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 */ skip_whitespace(); bool system_include = input.c == '<'; - const char *headername = parse_headername(); - if (headername == NULL) { + parse_headername(); + string_t headername = pp_token.string.string; + if (headername.begin == NULL) { eat_pp_directive(); return false; } + skip_whitespace(); if (!info.at_line_begin) { warningf(WARN_OTHER, &pp_token.base.source_position, "extra tokens at end of #include directive"); - eat_pp_directive(); + skip_till_newline(); } if (n_inputs > INCLUDE_LIMIT) { @@ -1566,19 +1496,13 @@ static bool parse_include_directive(void) 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 */ - in_pp_directive = false; - /* switch inputs */ emit_newlines(); push_input(); - bool res = do_include(system_include, headername); + bool res = do_include(system_include, pp_token.string.string.begin); if (!res) { errorf(&pp_token.base.source_position, - "failed including '%s': %s", headername, strerror(errno)); + "failed including '%S': %s", pp_token.string, strerror(errno)); pop_restore_input(); return false; } @@ -1641,8 +1565,9 @@ 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.identifier.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 (!info.at_line_begin) { @@ -1651,9 +1576,6 @@ static void parse_ifdef_ifndef_directive(void) 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(); @@ -1722,9 +1644,13 @@ static void parse_endif_directive(void) static void parse_preprocessing_directive(void) { - in_pp_directive = true; eat_pp('#'); + if (info.at_line_begin) { + /* empty directive */ + return; + } + if (skip_mode) { switch (pp_token.kind) { case TP_ifdef: @@ -1774,7 +1700,6 @@ static void parse_preprocessing_directive(void) } } - in_pp_directive = false; assert(info.at_line_begin); } @@ -1837,6 +1762,7 @@ int pptest_main(int argc, char **argv) /* 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")) { @@ -1844,6 +1770,9 @@ int pptest_main(int argc, char **argv) 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 { @@ -1857,7 +1786,15 @@ int pptest_main(int argc, char **argv) return 1; } - out = stdout; + if (output == NULL) { + out = stdout; + } else { + out = fopen(output, "w"); + if (out == NULL) { + fprintf(stderr, "Couldn't open output '%s'\n", output); + return 1; + } + } /* just here for gcc compatibility */ fprintf(out, "# 1 \"%s\"\n", filename); @@ -1877,9 +1814,9 @@ int pptest_main(int argc, char **argv) continue; } else if (pp_token.kind == TP_EOF) { goto end_of_main_loop; - } else if (pp_token.kind == TP_IDENTIFIER && !in_pp_directive) { - symbol_t *symbol = pp_token.identifier.symbol; - pp_definition_t *pp_definition = symbol->pp_definition; + } 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) { @@ -1899,7 +1836,7 @@ int pptest_main(int argc, char **argv) token_t next_token = pp_token; /* restore identifier token */ pp_token.kind = TP_IDENTIFIER; - pp_token.identifier.symbol = symbol; + pp_token.base.symbol = symbol; pp_token.base.source_position = position; info = old_info; emit_pp_token(); @@ -1926,6 +1863,8 @@ 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);