/* * TCC - Tiny C Compiler * * Copyright (c) 2001-2004 Fabrice Bellard * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2 of the License, or (at your option) any later version. * * This library is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ #include "tcc.h" /********************************************************/ /* global variables */ ST_DATA int tok_flags; /* additional informations about token */ #define TOK_FLAG_BOL 0x0001 /* beginning of line before */ #define TOK_FLAG_BOF 0x0002 /* beginning of file before */ #define TOK_FLAG_ENDIF 0x0004 /* a endif was found matching starting #ifdef */ #define TOK_FLAG_EOF 0x0008 /* end of file */ ST_DATA int parse_flags; #define PARSE_FLAG_PREPROCESS 0x0001 /* activate preprocessing */ #define PARSE_FLAG_TOK_NUM 0x0002 /* return numbers instead of TOK_PPNUM */ #define PARSE_FLAG_LINEFEED 0x0004 /* line feed is returned as a token. line feed is also returned at eof */ #define PARSE_FLAG_ASM_COMMENTS 0x0008 /* '#' can be used for line comment */ #define PARSE_FLAG_SPACES 0x0010 /* next() returns space tokens (for -E) */ ST_DATA struct BufferedFile *file; ST_DATA int ch, tok; ST_DATA CValue tokc; ST_DATA const int *macro_ptr; ST_DATA CString tokcstr; /* current parsed string, if any */ /* display benchmark infos */ ST_DATA int total_lines; ST_DATA int total_bytes; ST_DATA int tok_ident; ST_DATA TokenSym **table_ident; /* ------------------------------------------------------------------------- */ static int *macro_ptr_allocated; static const int *unget_saved_macro_ptr; static int unget_saved_buffer[TOK_MAX_SIZE + 1]; static int unget_buffer_enabled; static TokenSym *hash_ident[TOK_HASH_SIZE]; static char token_buf[STRING_MAX_SIZE + 1]; /* true if isid(c) || isnum(c) */ static unsigned char isidnum_table[256-CH_EOF]; static const char tcc_keywords[] = #define DEF(id, str) str "\0" #include "tcctok.h" #undef DEF ; /* WARNING: the content of this string encodes token numbers */ static const unsigned char tok_two_chars[] = "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253" "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266"; struct macro_level { struct macro_level *prev; const int *p; }; ST_FUNC void next_nomacro(void); static void next_nomacro_spc(void); static void macro_subst( TokenString *tok_str, Sym **nested_list, const int *macro_str, struct macro_level **can_read_stream ); ST_FUNC void skip(int c) { if (tok != c) error("'%c' expected (got \"%s\"", c, get_tok_str(tok, NULL)); next(); } /* ------------------------------------------------------------------------- */ /* CString handling */ static void cstr_realloc(CString *cstr, int new_size) { int size; void *data; size = cstr->size_allocated; if (size == 0) size = 8; /* no need to allocate a too small first string */ while (size < new_size) size = size * 2; data = tcc_realloc(cstr->data_allocated, size); if (!data) error("memory full"); cstr->data_allocated = data; cstr->size_allocated = size; cstr->data = data; } /* add a byte */ ST_INLN void cstr_ccat(CString *cstr, int ch) { int size; size = cstr->size + 1; if (size > cstr->size_allocated) cstr_realloc(cstr, size); ((unsigned char *)cstr->data)[size - 1] = ch; cstr->size = size; } ST_FUNC void cstr_cat(CString *cstr, const char *str) { int c; for(;;) { c = *str; if (c == '\0') break; cstr_ccat(cstr, c); str++; } } /* add a wide char */ ST_FUNC void cstr_wccat(CString *cstr, int ch) { int size; size = cstr->size + sizeof(nwchar_t); if (size > cstr->size_allocated) cstr_realloc(cstr, size); *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch; cstr->size = size; } ST_FUNC void cstr_new(CString *cstr) { memset(cstr, 0, sizeof(CString)); } /* free string and reset it to NULL */ ST_FUNC void cstr_free(CString *cstr) { tcc_free(cstr->data_allocated); cstr_new(cstr); } /* XXX: unicode ? */ ST_FUNC void add_char(CString *cstr, int c) { if (c == '\'' || c == '\"' || c == '\\') { /* XXX: could be more precise if char or string */ cstr_ccat(cstr, '\\'); } if (c >= 32 && c <= 126) { cstr_ccat(cstr, c); } else { cstr_ccat(cstr, '\\'); if (c == '\n') { cstr_ccat(cstr, 'n'); } else { cstr_ccat(cstr, '0' + ((c >> 6) & 7)); cstr_ccat(cstr, '0' + ((c >> 3) & 7)); cstr_ccat(cstr, '0' + (c & 7)); } } } /* ------------------------------------------------------------------------- */ /* allocate a new token */ static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len) { TokenSym *ts, **ptable; int i; if (tok_ident >= SYM_FIRST_ANOM) error("memory full"); /* expand token table if needed */ i = tok_ident - TOK_IDENT; if ((i % TOK_ALLOC_INCR) == 0) { ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *)); if (!ptable) error("memory full"); table_ident = ptable; } ts = tcc_malloc(sizeof(TokenSym) + len); table_ident[i] = ts; ts->tok = tok_ident++; ts->sym_define = NULL; ts->sym_label = NULL; ts->sym_struct = NULL; ts->sym_identifier = NULL; ts->len = len; ts->hash_next = NULL; memcpy(ts->str, str, len); ts->str[len] = '\0'; *pts = ts; return ts; } #define TOK_HASH_INIT 1 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c)) /* find a token and add it if not found */ ST_FUNC TokenSym *tok_alloc(const char *str, int len) { TokenSym *ts, **pts; int i; unsigned int h; h = TOK_HASH_INIT; for(i=0;ilen == len && !memcmp(ts->str, str, len)) return ts; pts = &(ts->hash_next); } return tok_alloc_new(pts, str, len); } /* XXX: buffer overflow */ /* XXX: float tokens */ ST_FUNC char *get_tok_str(int v, CValue *cv) { static char buf[STRING_MAX_SIZE + 1]; static CString cstr_buf; CString *cstr; char *p; int i, len; /* NOTE: to go faster, we give a fixed buffer for small strings */ cstr_reset(&cstr_buf); cstr_buf.data = buf; cstr_buf.size_allocated = sizeof(buf); p = buf; switch(v) { case TOK_CINT: case TOK_CUINT: /* XXX: not quite exact, but only useful for testing */ sprintf(p, "%u", cv->ui); break; case TOK_CLLONG: case TOK_CULLONG: /* XXX: not quite exact, but only useful for testing */ #ifdef _WIN32 sprintf(p, "%u", (unsigned)cv->ull); #else sprintf(p, "%Lu", cv->ull); #endif break; case TOK_LCHAR: cstr_ccat(&cstr_buf, 'L'); case TOK_CCHAR: cstr_ccat(&cstr_buf, '\''); add_char(&cstr_buf, cv->i); cstr_ccat(&cstr_buf, '\''); cstr_ccat(&cstr_buf, '\0'); break; case TOK_PPNUM: cstr = cv->cstr; len = cstr->size - 1; for(i=0;idata)[i]); cstr_ccat(&cstr_buf, '\0'); break; case TOK_LSTR: cstr_ccat(&cstr_buf, 'L'); case TOK_STR: cstr = cv->cstr; cstr_ccat(&cstr_buf, '\"'); if (v == TOK_STR) { len = cstr->size - 1; for(i=0;idata)[i]); } else { len = (cstr->size / sizeof(nwchar_t)) - 1; for(i=0;idata)[i]); } cstr_ccat(&cstr_buf, '\"'); cstr_ccat(&cstr_buf, '\0'); break; case TOK_LT: v = '<'; goto addv; case TOK_GT: v = '>'; goto addv; case TOK_DOTS: return strcpy(p, "..."); case TOK_A_SHL: return strcpy(p, "<<="); case TOK_A_SAR: return strcpy(p, ">>="); default: if (v < TOK_IDENT) { /* search in two bytes table */ const unsigned char *q = tok_two_chars; while (*q) { if (q[2] == v) { *p++ = q[0]; *p++ = q[1]; *p = '\0'; return buf; } q += 3; } addv: *p++ = v; *p = '\0'; } else if (v < tok_ident) { return table_ident[v - TOK_IDENT]->str; } else if (v >= SYM_FIRST_ANOM) { /* special name for anonymous symbol */ sprintf(p, "L.%u", v - SYM_FIRST_ANOM); } else { /* should never happen */ return NULL; } break; } return cstr_buf.data; } /* fill input buffer and peek next char */ static int tcc_peekc_slow(BufferedFile *bf) { int len; /* only tries to read if really end of buffer */ if (bf->buf_ptr >= bf->buf_end) { if (bf->fd != -1) { #if defined(PARSE_DEBUG) len = 8; #else len = IO_BUF_SIZE; #endif len = read(bf->fd, bf->buffer, len); if (len < 0) len = 0; } else { len = 0; } total_bytes += len; bf->buf_ptr = bf->buffer; bf->buf_end = bf->buffer + len; *bf->buf_end = CH_EOB; } if (bf->buf_ptr < bf->buf_end) { return bf->buf_ptr[0]; } else { bf->buf_ptr = bf->buf_end; return CH_EOF; } } /* return the current character, handling end of block if necessary (but not stray) */ ST_FUNC int handle_eob(void) { return tcc_peekc_slow(file); } /* read next char from current input file and handle end of input buffer */ ST_INLN void inp(void) { ch = *(++(file->buf_ptr)); /* end of buffer/file handling */ if (ch == CH_EOB) ch = handle_eob(); } /* handle '\[\r]\n' */ static int handle_stray_noerror(void) { while (ch == '\\') { inp(); if (ch == '\n') { file->line_num++; inp(); } else if (ch == '\r') { inp(); if (ch != '\n') goto fail; file->line_num++; inp(); } else { fail: return 1; } } return 0; } static void handle_stray(void) { if (handle_stray_noerror()) error("stray '\\' in program"); } /* skip the stray and handle the \\n case. Output an error if incorrect char after the stray */ static int handle_stray1(uint8_t *p) { int c; if (p >= file->buf_end) { file->buf_ptr = p; c = handle_eob(); p = file->buf_ptr; if (c == '\\') goto parse_stray; } else { parse_stray: file->buf_ptr = p; ch = *p; handle_stray(); p = file->buf_ptr; c = *p; } return c; } /* handle just the EOB case, but not stray */ #define PEEKC_EOB(c, p)\ {\ p++;\ c = *p;\ if (c == '\\') {\ file->buf_ptr = p;\ c = handle_eob();\ p = file->buf_ptr;\ }\ } /* handle the complicated stray case */ #define PEEKC(c, p)\ {\ p++;\ c = *p;\ if (c == '\\') {\ c = handle_stray1(p);\ p = file->buf_ptr;\ }\ } /* input with '\[\r]\n' handling. Note that this function cannot handle other characters after '\', so you cannot call it inside strings or comments */ ST_FUNC void minp(void) { inp(); if (ch == '\\') handle_stray(); } /* single line C++ comments */ static uint8_t *parse_line_comment(uint8_t *p) { int c; p++; for(;;) { c = *p; redo: if (c == '\n' || c == CH_EOF) { break; } else if (c == '\\') { file->buf_ptr = p; c = handle_eob(); p = file->buf_ptr; if (c == '\\') { PEEKC_EOB(c, p); if (c == '\n') { file->line_num++; PEEKC_EOB(c, p); } else if (c == '\r') { PEEKC_EOB(c, p); if (c == '\n') { file->line_num++; PEEKC_EOB(c, p); } } } else { goto redo; } } else { p++; } } return p; } /* C comments */ ST_FUNC uint8_t *parse_comment(uint8_t *p) { int c; p++; for(;;) { /* fast skip loop */ for(;;) { c = *p; if (c == '\n' || c == '*' || c == '\\') break; p++; c = *p; if (c == '\n' || c == '*' || c == '\\') break; p++; } /* now we can handle all the cases */ if (c == '\n') { file->line_num++; p++; } else if (c == '*') { p++; for(;;) { c = *p; if (c == '*') { p++; } else if (c == '/') { goto end_of_comment; } else if (c == '\\') { file->buf_ptr = p; c = handle_eob(); p = file->buf_ptr; if (c == '\\') { /* skip '\[\r]\n', otherwise just skip the stray */ while (c == '\\') { PEEKC_EOB(c, p); if (c == '\n') { file->line_num++; PEEKC_EOB(c, p); } else if (c == '\r') { PEEKC_EOB(c, p); if (c == '\n') { file->line_num++; PEEKC_EOB(c, p); } } else { goto after_star; } } } } else { break; } } after_star: ; } else { /* stray, eob or eof */ file->buf_ptr = p; c = handle_eob(); p = file->buf_ptr; if (c == CH_EOF) { error("unexpected end of file in comment"); } else if (c == '\\') { p++; } } } end_of_comment: p++; return p; } #define cinp minp static inline void skip_spaces(void) { while (is_space(ch)) cinp(); } static inline int check_space(int t, int *spc) { if (is_space(t)) { if (*spc) return 1; *spc = 1; } else *spc = 0; return 0; } /* parse a string without interpreting escapes */ static uint8_t *parse_pp_string(uint8_t *p, int sep, CString *str) { int c; p++; for(;;) { c = *p; if (c == sep) { break; } else if (c == '\\') { file->buf_ptr = p; c = handle_eob(); p = file->buf_ptr; if (c == CH_EOF) { unterminated_string: /* XXX: indicate line number of start of string */ error("missing terminating %c character", sep); } else if (c == '\\') { /* escape : just skip \[\r]\n */ PEEKC_EOB(c, p); if (c == '\n') { file->line_num++; p++; } else if (c == '\r') { PEEKC_EOB(c, p); if (c != '\n') expect("'\n' after '\r'"); file->line_num++; p++; } else if (c == CH_EOF) { goto unterminated_string; } else { if (str) { cstr_ccat(str, '\\'); cstr_ccat(str, c); } p++; } } } else if (c == '\n') { file->line_num++; goto add_char; } else if (c == '\r') { PEEKC_EOB(c, p); if (c != '\n') { if (str) cstr_ccat(str, '\r'); } else { file->line_num++; goto add_char; } } else { add_char: if (str) cstr_ccat(str, c); p++; } } p++; return p; } /* skip block of text until #else, #elif or #endif. skip also pairs of #if/#endif */ static void preprocess_skip(void) { int a, start_of_line, c, in_warn_or_error; uint8_t *p; p = file->buf_ptr; a = 0; redo_start: start_of_line = 1; in_warn_or_error = 0; for(;;) { redo_no_start: c = *p; switch(c) { case ' ': case '\t': case '\f': case '\v': case '\r': p++; goto redo_no_start; case '\n': file->line_num++; p++; goto redo_start; case '\\': file->buf_ptr = p; c = handle_eob(); if (c == CH_EOF) { expect("#endif"); } else if (c == '\\') { ch = file->buf_ptr[0]; handle_stray_noerror(); } p = file->buf_ptr; goto redo_no_start; /* skip strings */ case '\"': case '\'': if (in_warn_or_error) goto _default; p = parse_pp_string(p, c, NULL); break; /* skip comments */ case '/': if (in_warn_or_error) goto _default; file->buf_ptr = p; ch = *p; minp(); p = file->buf_ptr; if (ch == '*') { p = parse_comment(p); } else if (ch == '/') { p = parse_line_comment(p); } break; case '#': p++; if (start_of_line) { file->buf_ptr = p; next_nomacro(); p = file->buf_ptr; if (a == 0 && (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF)) goto the_end; if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF) a++; else if (tok == TOK_ENDIF) a--; else if( tok == TOK_ERROR || tok == TOK_WARNING) in_warn_or_error = 1; } break; _default: default: p++; break; } start_of_line = 0; } the_end: ; file->buf_ptr = p; } /* ParseState handling */ /* XXX: currently, no include file info is stored. Thus, we cannot display accurate messages if the function or data definition spans multiple files */ /* save current parse state in 's' */ ST_FUNC void save_parse_state(ParseState *s) { s->line_num = file->line_num; s->macro_ptr = macro_ptr; s->tok = tok; s->tokc = tokc; } /* restore parse state from 's' */ ST_FUNC void restore_parse_state(ParseState *s) { file->line_num = s->line_num; macro_ptr = s->macro_ptr; tok = s->tok; tokc = s->tokc; } /* return the number of additional 'ints' necessary to store the token */ static inline int tok_ext_size(int t) { switch(t) { /* 4 bytes */ case TOK_CINT: case TOK_CUINT: case TOK_CCHAR: case TOK_LCHAR: case TOK_CFLOAT: case TOK_LINENUM: return 1; case TOK_STR: case TOK_LSTR: case TOK_PPNUM: error("unsupported token"); return 1; case TOK_CDOUBLE: case TOK_CLLONG: case TOK_CULLONG: return 2; case TOK_CLDOUBLE: return LDOUBLE_SIZE / 4; default: return 0; } } /* token string handling */ ST_INLN void tok_str_new(TokenString *s) { s->str = NULL; s->len = 0; s->allocated_len = 0; s->last_line_num = -1; } ST_FUNC void tok_str_free(int *str) { tcc_free(str); } static int *tok_str_realloc(TokenString *s) { int *str, len; if (s->allocated_len == 0) { len = 8; } else { len = s->allocated_len * 2; } str = tcc_realloc(s->str, len * sizeof(int)); if (!str) error("memory full"); s->allocated_len = len; s->str = str; return str; } ST_FUNC void tok_str_add(TokenString *s, int t) { int len, *str; len = s->len; str = s->str; if (len >= s->allocated_len) str = tok_str_realloc(s); str[len++] = t; s->len = len; } static void tok_str_add2(TokenString *s, int t, CValue *cv) { int len, *str; len = s->len; str = s->str; /* allocate space for worst case */ if (len + TOK_MAX_SIZE > s->allocated_len) str = tok_str_realloc(s); str[len++] = t; switch(t) { case TOK_CINT: case TOK_CUINT: case TOK_CCHAR: case TOK_LCHAR: case TOK_CFLOAT: case TOK_LINENUM: str[len++] = cv->tab[0]; break; case TOK_PPNUM: case TOK_STR: case TOK_LSTR: { int nb_words; CString *cstr; nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2; while ((len + nb_words) > s->allocated_len) str = tok_str_realloc(s); cstr = (CString *)(str + len); cstr->data = NULL; cstr->size = cv->cstr->size; cstr->data_allocated = NULL; cstr->size_allocated = cstr->size; memcpy((char *)cstr + sizeof(CString), cv->cstr->data, cstr->size); len += nb_words; } break; case TOK_CDOUBLE: case TOK_CLLONG: case TOK_CULLONG: #if LDOUBLE_SIZE == 8 case TOK_CLDOUBLE: #endif str[len++] = cv->tab[0]; str[len++] = cv->tab[1]; break; #if LDOUBLE_SIZE == 12 case TOK_CLDOUBLE: str[len++] = cv->tab[0]; str[len++] = cv->tab[1]; str[len++] = cv->tab[2]; #elif LDOUBLE_SIZE == 16 case TOK_CLDOUBLE: str[len++] = cv->tab[0]; str[len++] = cv->tab[1]; str[len++] = cv->tab[2]; str[len++] = cv->tab[3]; #elif LDOUBLE_SIZE != 8 #error add long double size support #endif break; default: break; } s->len = len; } /* add the current parse token in token string 's' */ ST_FUNC void tok_str_add_tok(TokenString *s) { CValue cval; /* save line number info */ if (file->line_num != s->last_line_num) { s->last_line_num = file->line_num; cval.i = s->last_line_num; tok_str_add2(s, TOK_LINENUM, &cval); } tok_str_add2(s, tok, &tokc); } /* get a token from an integer array and increment pointer accordingly. we code it as a macro to avoid pointer aliasing. */ static inline void TOK_GET(int *t, const int **pp, CValue *cv) { const int *p = *pp; int n, *tab; tab = cv->tab; switch(*t = *p++) { case TOK_CINT: case TOK_CUINT: case TOK_CCHAR: case TOK_LCHAR: case TOK_CFLOAT: case TOK_LINENUM: tab[0] = *p++; break; case TOK_STR: case TOK_LSTR: case TOK_PPNUM: cv->cstr = (CString *)p; cv->cstr->data = (char *)p + sizeof(CString); p += (sizeof(CString) + cv->cstr->size + 3) >> 2; break; case TOK_CDOUBLE: case TOK_CLLONG: case TOK_CULLONG: n = 2; goto copy; case TOK_CLDOUBLE: #if LDOUBLE_SIZE == 16 n = 4; #elif LDOUBLE_SIZE == 12 n = 3; #elif LDOUBLE_SIZE == 8 n = 2; #else # error add long double size support #endif copy: do *tab++ = *p++; while (--n); break; default: break; } *pp = p; } static int macro_is_equal(const int *a, const int *b) { char buf[STRING_MAX_SIZE + 1]; CValue cv; int t; while (*a && *b) { TOK_GET(&t, &a, &cv); pstrcpy(buf, sizeof buf, get_tok_str(t, &cv)); TOK_GET(&t, &b, &cv); if (strcmp(buf, get_tok_str(t, &cv))) return 0; } return !(*a || *b); } /* defines handling */ ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg) { Sym *s; s = define_find(v); if (s && !macro_is_equal(s->d, str)) warning("%s redefined", get_tok_str(v, NULL)); s = sym_push2(&define_stack, v, macro_type, 0); s->d = str; s->next = first_arg; table_ident[v - TOK_IDENT]->sym_define = s; } /* undefined a define symbol. Its name is just set to zero */ ST_FUNC void define_undef(Sym *s) { int v; v = s->v; if (v >= TOK_IDENT && v < tok_ident) table_ident[v - TOK_IDENT]->sym_define = NULL; s->v = 0; } ST_INLN Sym *define_find(int v) { v -= TOK_IDENT; if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT)) return NULL; return table_ident[v]->sym_define; } /* free define stack until top reaches 'b' */ ST_FUNC void free_defines(Sym *b) { Sym *top, *top1; int v; top = define_stack; while (top != b) { top1 = top->prev; /* do not free args or predefined defines */ if (top->d) tok_str_free(top->d); v = top->v; if (v >= TOK_IDENT && v < tok_ident) table_ident[v - TOK_IDENT]->sym_define = NULL; sym_free(top); top = top1; } define_stack = b; } /* label lookup */ ST_FUNC Sym *label_find(int v) { v -= TOK_IDENT; if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT)) return NULL; return table_ident[v]->sym_label; } ST_FUNC Sym *label_push(Sym **ptop, int v, int flags) { Sym *s, **ps; s = sym_push2(ptop, v, 0, 0); s->r = flags; ps = &table_ident[v - TOK_IDENT]->sym_label; if (ptop == &global_label_stack) { /* modify the top most local identifier, so that sym_identifier will point to 's' when popped */ while (*ps != NULL) ps = &(*ps)->prev_tok; } s->prev_tok = *ps; *ps = s; return s; } /* pop labels until element last is reached. Look if any labels are undefined. Define symbols if '&&label' was used. */ ST_FUNC void label_pop(Sym **ptop, Sym *slast) { Sym *s, *s1; for(s = *ptop; s != slast; s = s1) { s1 = s->prev; if (s->r == LABEL_DECLARED) { warning("label '%s' declared but not used", get_tok_str(s->v, NULL)); } else if (s->r == LABEL_FORWARD) { error("label '%s' used but not defined", get_tok_str(s->v, NULL)); } else { if (s->c) { /* define corresponding symbol. A size of 1 is put. */ put_extern_sym(s, cur_text_section, s->jnext, 1); } } /* remove label */ table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok; sym_free(s); } *ptop = slast; } /* eval an expression for #if/#elif */ static int expr_preprocess(void) { int c, t; TokenString str; tok_str_new(&str); while (tok != TOK_LINEFEED && tok != TOK_EOF) { next(); /* do macro subst */ if (tok == TOK_DEFINED) { next_nomacro(); t = tok; if (t == '(') next_nomacro(); c = define_find(tok) != 0; if (t == '(') next_nomacro(); tok = TOK_CINT; tokc.i = c; } else if (tok >= TOK_IDENT) { /* if undefined macro */ tok = TOK_CINT; tokc.i = 0; } tok_str_add_tok(&str); } tok_str_add(&str, -1); /* simulate end of file */ tok_str_add(&str, 0); /* now evaluate C constant expression */ macro_ptr = str.str; next(); c = expr_const(); macro_ptr = NULL; tok_str_free(str.str); return c != 0; } #if defined(PARSE_DEBUG) || defined(PP_DEBUG) static void tok_print(int *str) { int t; CValue cval; printf("<"); while (1) { TOK_GET(&t, &str, &cval); if (!t) break; printf("%s", get_tok_str(t, &cval)); } printf(">\n"); } #endif /* parse after #define */ ST_FUNC void parse_define(void) { Sym *s, *first, **ps; int v, t, varg, is_vaargs, spc; TokenString str; v = tok; if (v < TOK_IDENT) error("invalid macro name '%s'", get_tok_str(tok, &tokc)); /* XXX: should check if same macro (ANSI) */ first = NULL; t = MACRO_OBJ; /* '(' must be just after macro definition for MACRO_FUNC */ next_nomacro_spc(); if (tok == '(') { next_nomacro(); ps = &first; while (tok != ')') { varg = tok; next_nomacro(); is_vaargs = 0; if (varg == TOK_DOTS) { varg = TOK___VA_ARGS__; is_vaargs = 1; } else if (tok == TOK_DOTS && gnu_ext) { is_vaargs = 1; next_nomacro(); } if (varg < TOK_IDENT) error("badly punctuated parameter list"); s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0); *ps = s; ps = &s->next; if (tok != ',') break; next_nomacro(); } if (tok == ')') next_nomacro_spc(); t = MACRO_FUNC; } tok_str_new(&str); spc = 2; /* EOF testing necessary for '-D' handling */ while (tok != TOK_LINEFEED && tok != TOK_EOF) { /* remove spaces around ## and after '#' */ if (TOK_TWOSHARPS == tok) { if (1 == spc) --str.len; spc = 2; } else if ('#' == tok) { spc = 2; } else if (check_space(tok, &spc)) { goto skip; } tok_str_add2(&str, tok, &tokc); skip: next_nomacro_spc(); } if (spc == 1) --str.len; /* remove trailing space */ tok_str_add(&str, 0); #ifdef PP_DEBUG printf("define %s %d: ", get_tok_str(v, NULL), t); tok_print(str.str); #endif define_push(v, t, str.str, first); } static inline int hash_cached_include(int type, const char *filename) { const unsigned char *s; unsigned int h; h = TOK_HASH_INIT; h = TOK_HASH_FUNC(h, type); s = filename; while (*s) { h = TOK_HASH_FUNC(h, *s); s++; } h &= (CACHED_INCLUDES_HASH_SIZE - 1); return h; } /* XXX: use a token or a hash table to accelerate matching ? */ static CachedInclude *search_cached_include(TCCState *s1, int type, const char *filename) { CachedInclude *e; int i, h; h = hash_cached_include(type, filename); i = s1->cached_includes_hash[h]; for(;;) { if (i == 0) break; e = s1->cached_includes[i - 1]; if (e->type == type && !PATHCMP(e->filename, filename)) return e; i = e->hash_next; } return NULL; } static inline void add_cached_include(TCCState *s1, int type, const char *filename, int ifndef_macro) { CachedInclude *e; int h; if (search_cached_include(s1, type, filename)) return; #ifdef INC_DEBUG printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL)); #endif e = tcc_malloc(sizeof(CachedInclude) + strlen(filename)); if (!e) return; e->type = type; strcpy(e->filename, filename); e->ifndef_macro = ifndef_macro; dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e); /* add in hash table */ h = hash_cached_include(type, filename); e->hash_next = s1->cached_includes_hash[h]; s1->cached_includes_hash[h] = s1->nb_cached_includes; } static void pragma_parse(TCCState *s1) { int val; next(); if (tok == TOK_pack) { /* This may be: #pragma pack(1) // set #pragma pack() // reset to default #pragma pack(push,1) // push & set #pragma pack(pop) // restore previous */ next(); skip('('); if (tok == TOK_ASM_pop) { next(); if (s1->pack_stack_ptr <= s1->pack_stack) { stk_error: error("out of pack stack"); } s1->pack_stack_ptr--; } else { val = 0; if (tok != ')') { if (tok == TOK_ASM_push) { next(); if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1) goto stk_error; s1->pack_stack_ptr++; skip(','); } if (tok != TOK_CINT) { pack_error: error("invalid pack pragma"); } val = tokc.i; if (val < 1 || val > 16 || (val & (val - 1)) != 0) goto pack_error; next(); } *s1->pack_stack_ptr = val; skip(')'); } } } /* is_bof is true if first non space token at beginning of file */ ST_FUNC void preprocess(int is_bof) { TCCState *s1 = tcc_state; int i, c, n, saved_parse_flags; char buf[1024], *q; Sym *s; saved_parse_flags = parse_flags; parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM | PARSE_FLAG_LINEFEED; next_nomacro(); redo: switch(tok) { case TOK_DEFINE: next_nomacro(); parse_define(); break; case TOK_UNDEF: next_nomacro(); s = define_find(tok); /* undefine symbol by putting an invalid name */ if (s) define_undef(s); break; case TOK_INCLUDE: case TOK_INCLUDE_NEXT: ch = file->buf_ptr[0]; /* XXX: incorrect if comments : use next_nomacro with a special mode */ skip_spaces(); if (ch == '<') { c = '>'; goto read_name; } else if (ch == '\"') { c = ch; read_name: inp(); q = buf; while (ch != c && ch != '\n' && ch != CH_EOF) { if ((q - buf) < sizeof(buf) - 1) *q++ = ch; if (ch == '\\') { if (handle_stray_noerror() == 0) --q; } else inp(); } *q = '\0'; minp(); #if 0 /* eat all spaces and comments after include */ /* XXX: slightly incorrect */ while (ch1 != '\n' && ch1 != CH_EOF) inp(); #endif } else { /* computed #include : either we have only strings or we have anything enclosed in '<>' */ next(); buf[0] = '\0'; if (tok == TOK_STR) { while (tok != TOK_LINEFEED) { if (tok != TOK_STR) { include_syntax: error("'#include' expects \"FILENAME\" or "); } pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data); next(); } c = '\"'; } else { int len; while (tok != TOK_LINEFEED) { pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc)); next(); } len = strlen(buf); /* check syntax and remove '<>' */ if (len < 2 || buf[0] != '<' || buf[len - 1] != '>') goto include_syntax; memmove(buf, buf + 1, len - 2); buf[len - 2] = '\0'; c = '>'; } } if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE) error("#include recursion too deep"); n = s1->nb_include_paths + s1->nb_sysinclude_paths; for (i = -2; i < n; ++i) { char buf1[sizeof file->filename]; BufferedFile *f; CachedInclude *e; const char *path; int size; if (i == -2) { /* check absolute include path */ if (!IS_ABSPATH(buf)) continue; buf1[0] = 0; } else if (i == -1) { /* search in current dir if "header.h" */ if (c != '\"') continue; size = tcc_basename(file->filename) - file->filename; memcpy(buf1, file->filename, size); buf1[size] = '\0'; } else { /* search in all the include paths */ if (i < s1->nb_include_paths) path = s1->include_paths[i]; else path = s1->sysinclude_paths[i - s1->nb_include_paths]; pstrcpy(buf1, sizeof(buf1), path); pstrcat(buf1, sizeof(buf1), "/"); } pstrcat(buf1, sizeof(buf1), buf); e = search_cached_include(s1, c, buf1); if (e && define_find(e->ifndef_macro)) { /* no need to parse the include because the 'ifndef macro' is defined */ #ifdef INC_DEBUG printf("%s: skipping %s\n", file->filename, buf); #endif f = NULL; } else { f = tcc_open(s1, buf1); if (!f) continue; } if (tok == TOK_INCLUDE_NEXT) { tok = TOK_INCLUDE; if (f) tcc_close(f); continue; } if (!f) goto include_done; #ifdef INC_DEBUG printf("%s: including %s\n", file->filename, buf1); #endif /* XXX: fix current line init */ /* push current file in stack */ *s1->include_stack_ptr++ = file; f->inc_type = c; pstrcpy(f->inc_filename, sizeof(f->inc_filename), buf1); file = f; /* add include file debug info */ if (s1->do_debug) { put_stabs(file->filename, N_BINCL, 0, 0, 0); } tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL; ch = file->buf_ptr[0]; goto the_end; } error("include file '%s' not found", buf); include_done: break; case TOK_IFNDEF: c = 1; goto do_ifdef; case TOK_IF: c = expr_preprocess(); goto do_if; case TOK_IFDEF: c = 0; do_ifdef: next_nomacro(); if (tok < TOK_IDENT) error("invalid argument for '#if%sdef'", c ? "n" : ""); if (is_bof) { if (c) { #ifdef INC_DEBUG printf("#ifndef %s\n", get_tok_str(tok, NULL)); #endif file->ifndef_macro = tok; } } c = (define_find(tok) != 0) ^ c; do_if: if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE) error("memory full"); *s1->ifdef_stack_ptr++ = c; goto test_skip; case TOK_ELSE: if (s1->ifdef_stack_ptr == s1->ifdef_stack) error("#else without matching #if"); if (s1->ifdef_stack_ptr[-1] & 2) error("#else after #else"); c = (s1->ifdef_stack_ptr[-1] ^= 3); goto test_else; case TOK_ELIF: if (s1->ifdef_stack_ptr == s1->ifdef_stack) error("#elif without matching #if"); c = s1->ifdef_stack_ptr[-1]; if (c > 1) error("#elif after #else"); /* last #if/#elif expression was true: we skip */ if (c == 1) goto skip; c = expr_preprocess(); s1->ifdef_stack_ptr[-1] = c; test_else: if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1) file->ifndef_macro = 0; test_skip: if (!(c & 1)) { skip: preprocess_skip(); is_bof = 0; goto redo; } break; case TOK_ENDIF: if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr) error("#endif without matching #if"); s1->ifdef_stack_ptr--; /* '#ifndef macro' was at the start of file. Now we check if an '#endif' is exactly at the end of file */ if (file->ifndef_macro && s1->ifdef_stack_ptr == file->ifdef_stack_ptr) { file->ifndef_macro_saved = file->ifndef_macro; /* need to set to zero to avoid false matches if another #ifndef at middle of file */ file->ifndef_macro = 0; while (tok != TOK_LINEFEED) next_nomacro(); tok_flags |= TOK_FLAG_ENDIF; goto the_end; } break; case TOK_LINE: next(); if (tok != TOK_CINT) error("#line"); file->line_num = tokc.i - 1; /* the line number will be incremented after */ next(); if (tok != TOK_LINEFEED) { if (tok != TOK_STR) error("#line"); pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.cstr->data); } break; case TOK_ERROR: case TOK_WARNING: c = tok; ch = file->buf_ptr[0]; skip_spaces(); q = buf; while (ch != '\n' && ch != CH_EOF) { if ((q - buf) < sizeof(buf) - 1) *q++ = ch; if (ch == '\\') { if (handle_stray_noerror() == 0) --q; } else inp(); } *q = '\0'; if (c == TOK_ERROR) error("#error %s", buf); else warning("#warning %s", buf); break; case TOK_PRAGMA: pragma_parse(s1); break; default: if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_CINT) { /* '!' is ignored to allow C scripts. numbers are ignored to emulate cpp behaviour */ } else { if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS)) warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc)); } break; } /* ignore other preprocess commands or #! for C scripts */ while (tok != TOK_LINEFEED) next_nomacro(); the_end: parse_flags = saved_parse_flags; } /* evaluate escape codes in a string. */ static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long) { int c, n; const uint8_t *p; p = buf; for(;;) { c = *p; if (c == '\0') break; if (c == '\\') { p++; /* escape */ c = *p; switch(c) { case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': /* at most three octal digits */ n = c - '0'; p++; c = *p; if (isoct(c)) { n = n * 8 + c - '0'; p++; c = *p; if (isoct(c)) { n = n * 8 + c - '0'; p++; } } c = n; goto add_char_nonext; case 'x': case 'u': case 'U': p++; n = 0; for(;;) { c = *p; if (c >= 'a' && c <= 'f') c = c - 'a' + 10; else if (c >= 'A' && c <= 'F') c = c - 'A' + 10; else if (isnum(c)) c = c - '0'; else break; n = n * 16 + c; p++; } c = n; goto add_char_nonext; case 'a': c = '\a'; break; case 'b': c = '\b'; break; case 'f': c = '\f'; break; case 'n': c = '\n'; break; case 'r': c = '\r'; break; case 't': c = '\t'; break; case 'v': c = '\v'; break; case 'e': if (!gnu_ext) goto invalid_escape; c = 27; break; case '\'': case '\"': case '\\': case '?': break; default: invalid_escape: if (c >= '!' && c <= '~') warning("unknown escape sequence: \'\\%c\'", c); else warning("unknown escape sequence: \'\\x%x\'", c); break; } } p++; add_char_nonext: if (!is_long) cstr_ccat(outstr, c); else cstr_wccat(outstr, c); } /* add a trailing '\0' */ if (!is_long) cstr_ccat(outstr, '\0'); else cstr_wccat(outstr, '\0'); } /* we use 64 bit numbers */ #define BN_SIZE 2 /* bn = (bn << shift) | or_val */ static void bn_lshift(unsigned int *bn, int shift, int or_val) { int i; unsigned int v; for(i=0;i> (32 - shift); } } static void bn_zero(unsigned int *bn) { int i; for(i=0;i= 'a' && ch <= 'f') t = ch - 'a' + 10; else if (ch >= 'A' && ch <= 'F') t = ch - 'A' + 10; else if (isnum(ch)) t = ch - '0'; else break; if (t >= b) break; if (q >= token_buf + STRING_MAX_SIZE) { num_too_long: error("number too long"); } *q++ = ch; ch = *p++; } if (ch == '.' || ((ch == 'e' || ch == 'E') && b == 10) || ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) { if (b != 10) { /* NOTE: strtox should support that for hexa numbers, but non ISOC99 libcs do not support it, so we prefer to do it by hand */ /* hexadecimal or binary floats */ /* XXX: handle overflows */ *q = '\0'; if (b == 16) shift = 4; else shift = 2; bn_zero(bn); q = token_buf; while (1) { t = *q++; if (t == '\0') { break; } else if (t >= 'a') { t = t - 'a' + 10; } else if (t >= 'A') { t = t - 'A' + 10; } else { t = t - '0'; } bn_lshift(bn, shift, t); } frac_bits = 0; if (ch == '.') { ch = *p++; while (1) { t = ch; if (t >= 'a' && t <= 'f') { t = t - 'a' + 10; } else if (t >= 'A' && t <= 'F') { t = t - 'A' + 10; } else if (t >= '0' && t <= '9') { t = t - '0'; } else { break; } if (t >= b) error("invalid digit"); bn_lshift(bn, shift, t); frac_bits += shift; ch = *p++; } } if (ch != 'p' && ch != 'P') expect("exponent"); ch = *p++; s = 1; exp_val = 0; if (ch == '+') { ch = *p++; } else if (ch == '-') { s = -1; ch = *p++; } if (ch < '0' || ch > '9') expect("exponent digits"); while (ch >= '0' && ch <= '9') { exp_val = exp_val * 10 + ch - '0'; ch = *p++; } exp_val = exp_val * s; /* now we can generate the number */ /* XXX: should patch directly float number */ d = (double)bn[1] * 4294967296.0 + (double)bn[0]; d = ldexp(d, exp_val - frac_bits); t = toup(ch); if (t == 'F') { ch = *p++; tok = TOK_CFLOAT; /* float : should handle overflow */ tokc.f = (float)d; } else if (t == 'L') { ch = *p++; #ifdef TCC_TARGET_PE tok = TOK_CDOUBLE; tokc.d = d; #else tok = TOK_CLDOUBLE; /* XXX: not large enough */ tokc.ld = (long double)d; #endif } else { tok = TOK_CDOUBLE; tokc.d = d; } } else { /* decimal floats */ if (ch == '.') { if (q >= token_buf + STRING_MAX_SIZE) goto num_too_long; *q++ = ch; ch = *p++; float_frac_parse: while (ch >= '0' && ch <= '9') { if (q >= token_buf + STRING_MAX_SIZE) goto num_too_long; *q++ = ch; ch = *p++; } } if (ch == 'e' || ch == 'E') { if (q >= token_buf + STRING_MAX_SIZE) goto num_too_long; *q++ = ch; ch = *p++; if (ch == '-' || ch == '+') { if (q >= token_buf + STRING_MAX_SIZE) goto num_too_long; *q++ = ch; ch = *p++; } if (ch < '0' || ch > '9') expect("exponent digits"); while (ch >= '0' && ch <= '9') { if (q >= token_buf + STRING_MAX_SIZE) goto num_too_long; *q++ = ch; ch = *p++; } } *q = '\0'; t = toup(ch); errno = 0; if (t == 'F') { ch = *p++; tok = TOK_CFLOAT; tokc.f = strtof(token_buf, NULL); } else if (t == 'L') { ch = *p++; #ifdef TCC_TARGET_PE tok = TOK_CDOUBLE; tokc.d = strtod(token_buf, NULL); #else tok = TOK_CLDOUBLE; tokc.ld = strtold(token_buf, NULL); #endif } else { tok = TOK_CDOUBLE; tokc.d = strtod(token_buf, NULL); } } } else { unsigned long long n, n1; int lcount, ucount; /* integer number */ *q = '\0'; q = token_buf; if (b == 10 && *q == '0') { b = 8; q++; } n = 0; while(1) { t = *q++; /* no need for checks except for base 10 / 8 errors */ if (t == '\0') { break; } else if (t >= 'a') { t = t - 'a' + 10; } else if (t >= 'A') { t = t - 'A' + 10; } else { t = t - '0'; if (t >= b) error("invalid digit"); } n1 = n; n = n * b + t; /* detect overflow */ /* XXX: this test is not reliable */ if (n < n1) error("integer constant overflow"); } /* XXX: not exactly ANSI compliant */ if ((n & 0xffffffff00000000LL) != 0) { if ((n >> 63) != 0) tok = TOK_CULLONG; else tok = TOK_CLLONG; } else if (n > 0x7fffffff) { tok = TOK_CUINT; } else { tok = TOK_CINT; } lcount = 0; ucount = 0; for(;;) { t = toup(ch); if (t == 'L') { if (lcount >= 2) error("three 'l's in integer constant"); lcount++; if (lcount == 2) { if (tok == TOK_CINT) tok = TOK_CLLONG; else if (tok == TOK_CUINT) tok = TOK_CULLONG; } ch = *p++; } else if (t == 'U') { if (ucount >= 1) error("two 'u's in integer constant"); ucount++; if (tok == TOK_CINT) tok = TOK_CUINT; else if (tok == TOK_CLLONG) tok = TOK_CULLONG; ch = *p++; } else { break; } } if (tok == TOK_CINT || tok == TOK_CUINT) tokc.ui = n; else tokc.ull = n; } if (ch) error("invalid number\n"); } #define PARSE2(c1, tok1, c2, tok2) \ case c1: \ PEEKC(c, p); \ if (c == c2) { \ p++; \ tok = tok2; \ } else { \ tok = tok1; \ } \ break; /* return next token without macro substitution */ static inline void next_nomacro1(void) { int t, c, is_long; TokenSym *ts; uint8_t *p, *p1; unsigned int h; p = file->buf_ptr; redo_no_start: c = *p; switch(c) { case ' ': case '\t': tok = c; p++; goto keep_tok_flags; case '\f': case '\v': case '\r': p++; goto redo_no_start; case '\\': /* first look if it is in fact an end of buffer */ if (p >= file->buf_end) { file->buf_ptr = p; handle_eob(); p = file->buf_ptr; if (p >= file->buf_end) goto parse_eof; else goto redo_no_start; } else { file->buf_ptr = p; ch = *p; handle_stray(); p = file->buf_ptr; goto redo_no_start; } parse_eof: { TCCState *s1 = tcc_state; if ((parse_flags & PARSE_FLAG_LINEFEED) && !(tok_flags & TOK_FLAG_EOF)) { tok_flags |= TOK_FLAG_EOF; tok = TOK_LINEFEED; goto keep_tok_flags; } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) { tok = TOK_EOF; } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) { error("missing #endif"); } else if (s1->include_stack_ptr == s1->include_stack) { /* no include left : end of file. */ tok = TOK_EOF; } else { tok_flags &= ~TOK_FLAG_EOF; /* pop include file */ /* test if previous '#endif' was after a #ifdef at start of file */ if (tok_flags & TOK_FLAG_ENDIF) { #ifdef INC_DEBUG printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL)); #endif add_cached_include(s1, file->inc_type, file->inc_filename, file->ifndef_macro_saved); } /* add end of include file debug info */ if (tcc_state->do_debug) { put_stabd(N_EINCL, 0, 0); } /* pop include stack */ tcc_close(file); s1->include_stack_ptr--; file = *s1->include_stack_ptr; p = file->buf_ptr; goto redo_no_start; } } break; case '\n': file->line_num++; tok_flags |= TOK_FLAG_BOL; p++; maybe_newline: if (0 == (parse_flags & PARSE_FLAG_LINEFEED)) goto redo_no_start; tok = TOK_LINEFEED; goto keep_tok_flags; case '#': /* XXX: simplify */ PEEKC(c, p); if ((tok_flags & TOK_FLAG_BOL) && (parse_flags & PARSE_FLAG_PREPROCESS)) { file->buf_ptr = p; preprocess(tok_flags & TOK_FLAG_BOF); p = file->buf_ptr; goto maybe_newline; } else { if (c == '#') { p++; tok = TOK_TWOSHARPS; } else { if (parse_flags & PARSE_FLAG_ASM_COMMENTS) { p = parse_line_comment(p - 1); goto redo_no_start; } else { tok = '#'; } } } break; case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g': case 'h': case 'i': case 'j': case 'k': case 'l': case 'm': case 'n': case 'o': case 'p': case 'q': case 'r': case 's': case 't': case 'u': case 'v': case 'w': case 'x': case 'y': case 'z': case 'A': case 'B': case 'C': case 'D': case 'E': case 'F': case 'G': case 'H': case 'I': case 'J': case 'K': case 'M': case 'N': case 'O': case 'P': case 'Q': case 'R': case 'S': case 'T': case 'U': case 'V': case 'W': case 'X': case 'Y': case 'Z': case '_': parse_ident_fast: p1 = p; h = TOK_HASH_INIT; h = TOK_HASH_FUNC(h, c); p++; for(;;) { c = *p; if (!isidnum_table[c-CH_EOF]) break; h = TOK_HASH_FUNC(h, c); p++; } if (c != '\\') { TokenSym **pts; int len; /* fast case : no stray found, so we have the full token and we have already hashed it */ len = p - p1; h &= (TOK_HASH_SIZE - 1); pts = &hash_ident[h]; for(;;) { ts = *pts; if (!ts) break; if (ts->len == len && !memcmp(ts->str, p1, len)) goto token_found; pts = &(ts->hash_next); } ts = tok_alloc_new(pts, p1, len); token_found: ; } else { /* slower case */ cstr_reset(&tokcstr); while (p1 < p) { cstr_ccat(&tokcstr, *p1); p1++; } p--; PEEKC(c, p); parse_ident_slow: while (isidnum_table[c-CH_EOF]) { cstr_ccat(&tokcstr, c); PEEKC(c, p); } ts = tok_alloc(tokcstr.data, tokcstr.size); } tok = ts->tok; break; case 'L': t = p[1]; if (t != '\\' && t != '\'' && t != '\"') { /* fast case */ goto parse_ident_fast; } else { PEEKC(c, p); if (c == '\'' || c == '\"') { is_long = 1; goto str_const; } else { cstr_reset(&tokcstr); cstr_ccat(&tokcstr, 'L'); goto parse_ident_slow; } } break; case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': cstr_reset(&tokcstr); /* after the first digit, accept digits, alpha, '.' or sign if prefixed by 'eEpP' */ parse_num: for(;;) { t = c; cstr_ccat(&tokcstr, c); PEEKC(c, p); if (!(isnum(c) || isid(c) || c == '.' || ((c == '+' || c == '-') && (t == 'e' || t == 'E' || t == 'p' || t == 'P')))) break; } /* We add a trailing '\0' to ease parsing */ cstr_ccat(&tokcstr, '\0'); tokc.cstr = &tokcstr; tok = TOK_PPNUM; break; case '.': /* special dot handling because it can also start a number */ PEEKC(c, p); if (isnum(c)) { cstr_reset(&tokcstr); cstr_ccat(&tokcstr, '.'); goto parse_num; } else if (c == '.') { PEEKC(c, p); if (c != '.') expect("'.'"); PEEKC(c, p); tok = TOK_DOTS; } else { tok = '.'; } break; case '\'': case '\"': is_long = 0; str_const: { CString str; int sep; sep = c; /* parse the string */ cstr_new(&str); p = parse_pp_string(p, sep, &str); cstr_ccat(&str, '\0'); /* eval the escape (should be done as TOK_PPNUM) */ cstr_reset(&tokcstr); parse_escape_string(&tokcstr, str.data, is_long); cstr_free(&str); if (sep == '\'') { int char_size; /* XXX: make it portable */ if (!is_long) char_size = 1; else char_size = sizeof(nwchar_t); if (tokcstr.size <= char_size) error("empty character constant"); if (tokcstr.size > 2 * char_size) warning("multi-character character constant"); if (!is_long) { tokc.i = *(int8_t *)tokcstr.data; tok = TOK_CCHAR; } else { tokc.i = *(nwchar_t *)tokcstr.data; tok = TOK_LCHAR; } } else { tokc.cstr = &tokcstr; if (!is_long) tok = TOK_STR; else tok = TOK_LSTR; } } break; case '<': PEEKC(c, p); if (c == '=') { p++; tok = TOK_LE; } else if (c == '<') { PEEKC(c, p); if (c == '=') { p++; tok = TOK_A_SHL; } else { tok = TOK_SHL; } } else { tok = TOK_LT; } break; case '>': PEEKC(c, p); if (c == '=') { p++; tok = TOK_GE; } else if (c == '>') { PEEKC(c, p); if (c == '=') { p++; tok = TOK_A_SAR; } else { tok = TOK_SAR; } } else { tok = TOK_GT; } break; case '&': PEEKC(c, p); if (c == '&') { p++; tok = TOK_LAND; } else if (c == '=') { p++; tok = TOK_A_AND; } else { tok = '&'; } break; case '|': PEEKC(c, p); if (c == '|') { p++; tok = TOK_LOR; } else if (c == '=') { p++; tok = TOK_A_OR; } else { tok = '|'; } break; case '+': PEEKC(c, p); if (c == '+') { p++; tok = TOK_INC; } else if (c == '=') { p++; tok = TOK_A_ADD; } else { tok = '+'; } break; case '-': PEEKC(c, p); if (c == '-') { p++; tok = TOK_DEC; } else if (c == '=') { p++; tok = TOK_A_SUB; } else if (c == '>') { p++; tok = TOK_ARROW; } else { tok = '-'; } break; PARSE2('!', '!', '=', TOK_NE) PARSE2('=', '=', '=', TOK_EQ) PARSE2('*', '*', '=', TOK_A_MUL) PARSE2('%', '%', '=', TOK_A_MOD) PARSE2('^', '^', '=', TOK_A_XOR) /* comments or operator */ case '/': PEEKC(c, p); if (c == '*') { p = parse_comment(p); goto redo_no_start; } else if (c == '/') { p = parse_line_comment(p); goto redo_no_start; } else if (c == '=') { p++; tok = TOK_A_DIV; } else { tok = '/'; } break; /* simple tokens */ case '(': case ')': case '[': case ']': case '{': case '}': case ',': case ';': case ':': case '?': case '~': case '$': /* only used in assembler */ case '@': /* dito */ tok = c; p++; break; default: error("unrecognized character \\x%02x", c); break; } tok_flags = 0; keep_tok_flags: file->buf_ptr = p; #if defined(PARSE_DEBUG) printf("token = %s\n", get_tok_str(tok, &tokc)); #endif } /* return next token without macro substitution. Can read input from macro_ptr buffer */ static void next_nomacro_spc(void) { if (macro_ptr) { redo: tok = *macro_ptr; if (tok) { TOK_GET(&tok, ¯o_ptr, &tokc); if (tok == TOK_LINENUM) { file->line_num = tokc.i; goto redo; } } } else { next_nomacro1(); } } ST_FUNC void next_nomacro(void) { do { next_nomacro_spc(); } while (is_space(tok)); } /* substitute args in macro_str and return allocated string */ static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args) { int last_tok, t, spc; const int *st; Sym *s; CValue cval; TokenString str; CString cstr; tok_str_new(&str); last_tok = 0; while(1) { TOK_GET(&t, ¯o_str, &cval); if (!t) break; if (t == '#') { /* stringize */ TOK_GET(&t, ¯o_str, &cval); if (!t) break; s = sym_find2(args, t); if (s) { cstr_new(&cstr); st = s->d; spc = 0; while (*st) { TOK_GET(&t, &st, &cval); if (!check_space(t, &spc)) cstr_cat(&cstr, get_tok_str(t, &cval)); } cstr.size -= spc; cstr_ccat(&cstr, '\0'); #ifdef PP_DEBUG printf("stringize: %s\n", (char *)cstr.data); #endif /* add string */ cval.cstr = &cstr; tok_str_add2(&str, TOK_STR, &cval); cstr_free(&cstr); } else { tok_str_add2(&str, t, &cval); } } else if (t >= TOK_IDENT) { s = sym_find2(args, t); if (s) { st = s->d; /* if '##' is present before or after, no arg substitution */ if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) { /* special case for var arg macros : ## eats the ',' if empty VA_ARGS variable. */ /* XXX: test of the ',' is not 100% reliable. should fix it to avoid security problems */ if (gnu_ext && s->type.t && last_tok == TOK_TWOSHARPS && str.len >= 2 && str.str[str.len - 2] == ',') { if (*st == 0) { /* suppress ',' '##' */ str.len -= 2; } else { /* suppress '##' and add variable */ str.len--; goto add_var; } } else { int t1; add_var: for(;;) { TOK_GET(&t1, &st, &cval); if (!t1) break; tok_str_add2(&str, t1, &cval); } } } else { /* NOTE: the stream cannot be read when macro substituing an argument */ macro_subst(&str, nested_list, st, NULL); } } else { tok_str_add(&str, t); } } else { tok_str_add2(&str, t, &cval); } last_tok = t; } tok_str_add(&str, 0); return str.str; } static char const ab_month_name[12][4] = { "Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" }; /* do macro substitution of current token with macro 's' and add result to (tok_str,tok_len). 'nested_list' is the list of all macros we got inside to avoid recursing. Return non zero if no substitution needs to be done */ static int macro_subst_tok(TokenString *tok_str, Sym **nested_list, Sym *s, struct macro_level **can_read_stream) { Sym *args, *sa, *sa1; int mstr_allocated, parlevel, *mstr, t, t1, spc; const int *p; TokenString str; char *cstrval; CValue cval; CString cstr; char buf[32]; /* if symbol is a macro, prepare substitution */ /* special macros */ if (tok == TOK___LINE__) { snprintf(buf, sizeof(buf), "%d", file->line_num); cstrval = buf; t1 = TOK_PPNUM; goto add_cstr1; } else if (tok == TOK___FILE__) { cstrval = file->filename; goto add_cstr; } else if (tok == TOK___DATE__ || tok == TOK___TIME__) { time_t ti; struct tm *tm; time(&ti); tm = localtime(&ti); if (tok == TOK___DATE__) { snprintf(buf, sizeof(buf), "%s %2d %d", ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900); } else { snprintf(buf, sizeof(buf), "%02d:%02d:%02d", tm->tm_hour, tm->tm_min, tm->tm_sec); } cstrval = buf; add_cstr: t1 = TOK_STR; add_cstr1: cstr_new(&cstr); cstr_cat(&cstr, cstrval); cstr_ccat(&cstr, '\0'); cval.cstr = &cstr; tok_str_add2(tok_str, t1, &cval); cstr_free(&cstr); } else { mstr = s->d; mstr_allocated = 0; if (s->type.t == MACRO_FUNC) { /* NOTE: we do not use next_nomacro to avoid eating the next token. XXX: find better solution */ redo: if (macro_ptr) { p = macro_ptr; while (is_space(t = *p) || TOK_LINEFEED == t) ++p; if (t == 0 && can_read_stream) { /* end of macro stream: we must look at the token after in the file */ struct macro_level *ml = *can_read_stream; macro_ptr = NULL; if (ml) { macro_ptr = ml->p; ml->p = NULL; *can_read_stream = ml -> prev; } goto redo; } } else { /* XXX: incorrect with comments */ ch = file->buf_ptr[0]; while (is_space(ch) || ch == '\n') cinp(); t = ch; } if (t != '(') /* no macro subst */ return -1; /* argument macro */ next_nomacro(); next_nomacro(); args = NULL; sa = s->next; /* NOTE: empty args are allowed, except if no args */ for(;;) { /* handle '()' case */ if (!args && !sa && tok == ')') break; if (!sa) error("macro '%s' used with too many args", get_tok_str(s->v, 0)); tok_str_new(&str); parlevel = spc = 0; /* NOTE: non zero sa->t indicates VA_ARGS */ while ((parlevel > 0 || (tok != ')' && (tok != ',' || sa->type.t))) && tok != -1) { if (tok == '(') parlevel++; else if (tok == ')') parlevel--; if (tok == TOK_LINEFEED) tok = ' '; if (!check_space(tok, &spc)) tok_str_add2(&str, tok, &tokc); next_nomacro_spc(); } str.len -= spc; tok_str_add(&str, 0); sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0); sa1->d = str.str; sa = sa->next; if (tok == ')') { /* special case for gcc var args: add an empty var arg argument if it is omitted */ if (sa && sa->type.t && gnu_ext) continue; else break; } if (tok != ',') expect(","); next_nomacro(); } if (sa) { error("macro '%s' used with too few args", get_tok_str(s->v, 0)); } /* now subst each arg */ mstr = macro_arg_subst(nested_list, mstr, args); /* free memory */ sa = args; while (sa) { sa1 = sa->prev; tok_str_free(sa->d); sym_free(sa); sa = sa1; } mstr_allocated = 1; } sym_push2(nested_list, s->v, 0, 0); macro_subst(tok_str, nested_list, mstr, can_read_stream); /* pop nested defined symbol */ sa1 = *nested_list; *nested_list = sa1->prev; sym_free(sa1); if (mstr_allocated) tok_str_free(mstr); } return 0; } /* handle the '##' operator. Return NULL if no '##' seen. Otherwise return the resulting string (which must be freed). */ static inline int *macro_twosharps(const int *macro_str) { const int *ptr; int t; CValue cval; TokenString macro_str1; CString cstr; char *p; int n; /* we search the first '##' */ for(ptr = macro_str;;) { TOK_GET(&t, &ptr, &cval); if (t == TOK_TWOSHARPS) break; /* nothing more to do if end of string */ if (t == 0) return NULL; } /* we saw '##', so we need more processing to handle it */ tok_str_new(¯o_str1); for(ptr = macro_str;;) { TOK_GET(&tok, &ptr, &tokc); if (tok == 0) break; if (tok == TOK_TWOSHARPS) continue; while (*ptr == TOK_TWOSHARPS) { t = *++ptr; if (t && t != TOK_TWOSHARPS) { TOK_GET(&t, &ptr, &cval); /* We concatenate the two tokens */ cstr_new(&cstr); cstr_cat(&cstr, get_tok_str(tok, &tokc)); n = cstr.size; cstr_cat(&cstr, get_tok_str(t, &cval)); cstr_ccat(&cstr, '\0'); p = file->buf_ptr; file->buf_ptr = cstr.data; for (;;) { next_nomacro1(); if (0 == *file->buf_ptr) break; tok_str_add2(¯o_str1, tok, &tokc); warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token", n, cstr.data, (char*)cstr.data + n); } file->buf_ptr = p; cstr_reset(&cstr); } } tok_str_add2(¯o_str1, tok, &tokc); } tok_str_add(¯o_str1, 0); return macro_str1.str; } /* do macro substitution of macro_str and add result to (tok_str,tok_len). 'nested_list' is the list of all macros we got inside to avoid recursing. */ static void macro_subst(TokenString *tok_str, Sym **nested_list, const int *macro_str, struct macro_level ** can_read_stream) { Sym *s; int *macro_str1; const int *ptr; int t, ret, spc; CValue cval; struct macro_level ml; /* first scan for '##' operator handling */ ptr = macro_str; macro_str1 = macro_twosharps(ptr); if (macro_str1) ptr = macro_str1; spc = 0; while (1) { /* NOTE: ptr == NULL can only happen if tokens are read from file stream due to a macro function call */ if (ptr == NULL) break; TOK_GET(&t, &ptr, &cval); if (t == 0) break; s = define_find(t); if (s != NULL) { /* if nested substitution, do nothing */ if (sym_find2(*nested_list, t)) goto no_subst; ml.p = macro_ptr; if (can_read_stream) ml.prev = *can_read_stream, *can_read_stream = &ml; macro_ptr = (int *)ptr; tok = t; ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream); ptr = (int *)macro_ptr; macro_ptr = ml.p; if (can_read_stream && *can_read_stream == &ml) *can_read_stream = ml.prev; if (ret != 0) goto no_subst; } else { no_subst: if (!check_space(t, &spc)) tok_str_add2(tok_str, t, &cval); } } if (macro_str1) tok_str_free(macro_str1); } /* return next token with macro substitution */ ST_FUNC void next(void) { Sym *nested_list, *s; TokenString str; struct macro_level *ml; redo: if (parse_flags & PARSE_FLAG_SPACES) next_nomacro_spc(); else next_nomacro(); if (!macro_ptr) { /* if not reading from macro substituted string, then try to substitute macros */ if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) { s = define_find(tok); if (s) { /* we have a macro: we try to substitute */ tok_str_new(&str); nested_list = NULL; ml = NULL; if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) { /* substitution done, NOTE: maybe empty */ tok_str_add(&str, 0); macro_ptr = str.str; macro_ptr_allocated = str.str; goto redo; } } } } else { if (tok == 0) { /* end of macro or end of unget buffer */ if (unget_buffer_enabled) { macro_ptr = unget_saved_macro_ptr; unget_buffer_enabled = 0; } else { /* end of macro string: free it */ tok_str_free(macro_ptr_allocated); macro_ptr_allocated = NULL; macro_ptr = NULL; } goto redo; } } /* convert preprocessor tokens into C tokens */ if (tok == TOK_PPNUM && (parse_flags & PARSE_FLAG_TOK_NUM)) { parse_number((char *)tokc.cstr->data); } } /* push back current token and set current token to 'last_tok'. Only identifier case handled for labels. */ ST_INLN void unget_tok(int last_tok) { int i, n; int *q; unget_saved_macro_ptr = macro_ptr; unget_buffer_enabled = 1; q = unget_saved_buffer; macro_ptr = q; *q++ = tok; n = tok_ext_size(tok) - 1; for(i=0;iinclude_stack_ptr = s1->include_stack; /* XXX: move that before to avoid having to initialize file->ifdef_stack_ptr ? */ s1->ifdef_stack_ptr = s1->ifdef_stack; file->ifdef_stack_ptr = s1->ifdef_stack_ptr; /* XXX: not ANSI compliant: bound checking says error */ vtop = vstack - 1; s1->pack_stack[0] = 0; s1->pack_stack_ptr = s1->pack_stack; } ST_FUNC void preprocess_new() { int i, c; const char *p, *r; TokenSym *ts; /* init isid table */ for(i=CH_EOF;i<256;i++) isidnum_table[i-CH_EOF] = isid(i) || isnum(i); /* add all tokens */ table_ident = NULL; memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *)); tok_ident = TOK_IDENT; p = tcc_keywords; while (*p) { r = p; for(;;) { c = *r++; if (c == '\0') break; } ts = tok_alloc(p, r - p - 1); p = r; } } /* Preprocess the current file */ ST_FUNC int tcc_preprocess(TCCState *s1) { Sym *define_start; BufferedFile *file_ref, **iptr, **iptr_new; int token_seen, line_ref, d; const char *s; preprocess_init(s1); define_start = define_stack; ch = file->buf_ptr[0]; tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF; parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS | PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES; token_seen = 0; line_ref = 0; file_ref = NULL; iptr = s1->include_stack_ptr; for (;;) { next(); if (tok == TOK_EOF) { break; } else if (file != file_ref) { goto print_line; } else if (tok == TOK_LINEFEED) { if (!token_seen) continue; ++line_ref; token_seen = 0; } else if (!token_seen) { d = file->line_num - line_ref; if (file != file_ref || d < 0 || d >= 8) { print_line: iptr_new = s1->include_stack_ptr; s = iptr_new > iptr ? " 1" : iptr_new < iptr ? " 2" : iptr_new > s1->include_stack ? " 3" : "" ; iptr = iptr_new; fprintf(s1->outfile, "# %d \"%s\"%s\n", file->line_num, file->filename, s); } else { while (d) fputs("\n", s1->outfile), --d; } line_ref = (file_ref = file)->line_num; token_seen = tok != TOK_LINEFEED; if (!token_seen) continue; } fputs(get_tok_str(tok, &tokc), s1->outfile); } free_defines(define_start); return 0; }