diff options
author | Dmitry Atamanov <data-man@users.noreply.github.com> | 2017-10-28 10:25:56 +0300 |
---|---|---|
committer | Andreas Rumpf <rumpf_a@web.de> | 2017-10-28 09:25:56 +0200 |
commit | d2c7d391c8b69a6a590a2f702ed58bea033f6325 (patch) | |
tree | c74a1b46e1166ddb87453ddc49cea84e1baaa5ab /tinyc/tccpp.c | |
parent | 9c00f6decd4453a4233450a60ccef05b20e9f24a (diff) | |
download | Nim-d2c7d391c8b69a6a590a2f702ed58bea033f6325.tar.gz |
TinyC upgrade (#6593)
Diffstat (limited to 'tinyc/tccpp.c')
-rw-r--r-- | tinyc/tccpp.c | 2844 |
1 files changed, 1906 insertions, 938 deletions
diff --git a/tinyc/tccpp.c b/tinyc/tccpp.c index ff17d8bed..76f9e428e 100644 --- a/tinyc/tccpp.c +++ b/tinyc/tccpp.c @@ -18,6 +18,45 @@ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ +#include "tcc.h" + +/********************************************************/ +/* global variables */ + +ST_DATA int tok_flags; +ST_DATA int parse_flags; + +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 TokenSym *hash_ident[TOK_HASH_SIZE]; +static char token_buf[STRING_MAX_SIZE + 1]; +static CString cstr_buf; +static CString macro_equal_buf; +static TokenString tokstr_buf; +static unsigned char isidnum_table[256 - CH_EOF]; +static int pp_debug_tok, pp_debug_symv; +static int pp_once; +static int pp_expr; +static int pp_counter; +static void tok_print(const char *msg, const int *str); + +static struct TinyAlloc *toksym_alloc; +static struct TinyAlloc *tokstr_alloc; +static struct TinyAlloc *cstr_alloc; + +static TokenString *macro_stack; static const char tcc_keywords[] = #define DEF(id, str) str "\0" @@ -26,23 +65,350 @@ static const char tcc_keywords[] = ; /* WARNING: the content of this string encodes token numbers */ -static 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\313..\250##\266"; +static const unsigned char tok_two_chars[] = +/* outdated -- gr + "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253" + "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266"; +*/{ + '<','=', TOK_LE, + '>','=', TOK_GE, + '!','=', TOK_NE, + '&','&', TOK_LAND, + '|','|', TOK_LOR, + '+','+', TOK_INC, + '-','-', TOK_DEC, + '=','=', TOK_EQ, + '<','<', TOK_SHL, + '>','>', TOK_SAR, + '+','=', TOK_A_ADD, + '-','=', TOK_A_SUB, + '*','=', TOK_A_MUL, + '/','=', TOK_A_DIV, + '%','=', TOK_A_MOD, + '&','=', TOK_A_AND, + '^','=', TOK_A_XOR, + '|','=', TOK_A_OR, + '-','>', TOK_ARROW, + '.','.', TOK_TWODOTS, + '#','#', TOK_TWOSHARPS, + 0 +}; -/* true if isid(c) || isnum(c) */ -static unsigned char isidnum_table[256-CH_EOF]; +static void next_nomacro_spc(void); +ST_FUNC void skip(int c) +{ + if (tok != c) + tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc)); + next(); +} -struct macro_level { - struct macro_level *prev; - int *p; -}; +ST_FUNC void expect(const char *msg) +{ + tcc_error("%s expected", msg); +} -static 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); +/* ------------------------------------------------------------------------- */ +/* Custom allocator for tiny objects */ + +#define USE_TAL + +#ifndef USE_TAL +#define tal_free(al, p) tcc_free(p) +#define tal_realloc(al, p, size) tcc_realloc(p, size) +#define tal_new(a,b,c) +#define tal_delete(a) +#else +#if !defined(MEM_DEBUG) +#define tal_free(al, p) tal_free_impl(al, p) +#define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size) +#define TAL_DEBUG_PARAMS +#else +#define TAL_DEBUG 1 +//#define TAL_INFO 1 /* collect and dump allocators stats */ +#define tal_free(al, p) tal_free_impl(al, p, __FILE__, __LINE__) +#define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size, __FILE__, __LINE__) +#define TAL_DEBUG_PARAMS , const char *file, int line +#define TAL_DEBUG_FILE_LEN 40 +#endif + +#define TOKSYM_TAL_SIZE (768 * 1024) /* allocator for tiny TokenSym in table_ident */ +#define TOKSTR_TAL_SIZE (768 * 1024) /* allocator for tiny TokenString instances */ +#define CSTR_TAL_SIZE (256 * 1024) /* allocator for tiny CString instances */ +#define TOKSYM_TAL_LIMIT 256 /* prefer unique limits to distinguish allocators debug msgs */ +#define TOKSTR_TAL_LIMIT 128 /* 32 * sizeof(int) */ +#define CSTR_TAL_LIMIT 1024 + +typedef struct TinyAlloc { + unsigned limit; + unsigned size; + uint8_t *buffer; + uint8_t *p; + unsigned nb_allocs; + struct TinyAlloc *next, *top; +#ifdef TAL_INFO + unsigned nb_peak; + unsigned nb_total; + unsigned nb_missed; + uint8_t *peak_p; +#endif +} TinyAlloc; +typedef struct tal_header_t { + unsigned size; +#ifdef TAL_DEBUG + int line_num; /* negative line_num used for double free check */ + char file_name[TAL_DEBUG_FILE_LEN + 1]; +#endif +} tal_header_t; + +/* ------------------------------------------------------------------------- */ + +static TinyAlloc *tal_new(TinyAlloc **pal, unsigned limit, unsigned size) +{ + TinyAlloc *al = tcc_mallocz(sizeof(TinyAlloc)); + al->p = al->buffer = tcc_malloc(size); + al->limit = limit; + al->size = size; + if (pal) *pal = al; + return al; +} +static void tal_delete(TinyAlloc *al) +{ + TinyAlloc *next; + +tail_call: + if (!al) + return; +#ifdef TAL_INFO + fprintf(stderr, "limit=%5d, size=%5g MB, nb_peak=%6d, nb_total=%8d, nb_missed=%6d, usage=%5.1f%%\n", + al->limit, al->size / 1024.0 / 1024.0, al->nb_peak, al->nb_total, al->nb_missed, + (al->peak_p - al->buffer) * 100.0 / al->size); +#endif +#ifdef TAL_DEBUG + if (al->nb_allocs > 0) { + uint8_t *p; + fprintf(stderr, "TAL_DEBUG: memory leak %d chunk(s) (limit= %d)\n", + al->nb_allocs, al->limit); + p = al->buffer; + while (p < al->p) { + tal_header_t *header = (tal_header_t *)p; + if (header->line_num > 0) { + fprintf(stderr, "%s:%d: chunk of %d bytes leaked\n", + header->file_name, header->line_num, header->size); + } + p += header->size + sizeof(tal_header_t); + } +#if MEM_DEBUG-0 == 2 + exit(2); +#endif + } +#endif + next = al->next; + tcc_free(al->buffer); + tcc_free(al); + al = next; + goto tail_call; +} + +static void tal_free_impl(TinyAlloc *al, void *p TAL_DEBUG_PARAMS) +{ + if (!p) + return; +tail_call: + if (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size) { +#ifdef TAL_DEBUG + tal_header_t *header = (((tal_header_t *)p) - 1); + if (header->line_num < 0) { + fprintf(stderr, "%s:%d: TAL_DEBUG: double frees chunk from\n", + file, line); + fprintf(stderr, "%s:%d: %d bytes\n", + header->file_name, (int)-header->line_num, (int)header->size); + } else + header->line_num = -header->line_num; +#endif + al->nb_allocs--; + if (!al->nb_allocs) + al->p = al->buffer; + } else if (al->next) { + al = al->next; + goto tail_call; + } + else + tcc_free(p); +} + +static void *tal_realloc_impl(TinyAlloc **pal, void *p, unsigned size TAL_DEBUG_PARAMS) +{ + tal_header_t *header; + void *ret; + int is_own; + unsigned adj_size = (size + 3) & -4; + TinyAlloc *al = *pal; + +tail_call: + is_own = (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size); + if ((!p || is_own) && size <= al->limit) { + if (al->p + adj_size + sizeof(tal_header_t) < al->buffer + al->size) { + header = (tal_header_t *)al->p; + header->size = adj_size; +#ifdef TAL_DEBUG + { int ofs = strlen(file) - TAL_DEBUG_FILE_LEN; + strncpy(header->file_name, file + (ofs > 0 ? ofs : 0), TAL_DEBUG_FILE_LEN); + header->file_name[TAL_DEBUG_FILE_LEN] = 0; + header->line_num = line; } +#endif + ret = al->p + sizeof(tal_header_t); + al->p += adj_size + sizeof(tal_header_t); + if (is_own) { + header = (((tal_header_t *)p) - 1); + memcpy(ret, p, header->size); +#ifdef TAL_DEBUG + header->line_num = -header->line_num; +#endif + } else { + al->nb_allocs++; + } +#ifdef TAL_INFO + if (al->nb_peak < al->nb_allocs) + al->nb_peak = al->nb_allocs; + if (al->peak_p < al->p) + al->peak_p = al->p; + al->nb_total++; +#endif + return ret; + } else if (is_own) { + al->nb_allocs--; + ret = tal_realloc(*pal, 0, size); + header = (((tal_header_t *)p) - 1); + memcpy(ret, p, header->size); +#ifdef TAL_DEBUG + header->line_num = -header->line_num; +#endif + return ret; + } + if (al->next) { + al = al->next; + } else { + TinyAlloc *bottom = al, *next = al->top ? al->top : al; + + al = tal_new(pal, next->limit, next->size * 2); + al->next = next; + bottom->top = al; + } + goto tail_call; + } + if (is_own) { + al->nb_allocs--; + ret = tcc_malloc(size); + header = (((tal_header_t *)p) - 1); + memcpy(ret, p, header->size); +#ifdef TAL_DEBUG + header->line_num = -header->line_num; +#endif + } else if (al->next) { + al = al->next; + goto tail_call; + } else + ret = tcc_realloc(p, size); +#ifdef TAL_INFO + al->nb_missed++; +#endif + return ret; +} + +#endif /* USE_TAL */ + +/* ------------------------------------------------------------------------- */ +/* CString handling */ +static void cstr_realloc(CString *cstr, int new_size) +{ + int size; + + size = cstr->size_allocated; + if (size < 8) + size = 8; /* no need to allocate a too small first string */ + while (size < new_size) + size = size * 2; + cstr->data = tal_realloc(cstr_alloc, cstr->data, size); + cstr->size_allocated = size; +} + +/* 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 len) +{ + int size; + if (len <= 0) + len = strlen(str) + 1 + len; + size = cstr->size + len; + if (size > cstr->size_allocated) + cstr_realloc(cstr, size); + memmove(((unsigned char *)cstr->data) + cstr->size, str, len); + cstr->size = size; +} + +/* 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) +{ + tal_free(cstr_alloc, cstr->data); + cstr_new(cstr); +} + +/* reset string to empty */ +ST_FUNC void cstr_reset(CString *cstr) +{ + cstr->size = 0; +} + +/* XXX: unicode ? */ +static 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) { @@ -50,18 +416,16 @@ static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len) int i; if (tok_ident >= SYM_FIRST_ANOM) - error("memory full"); + tcc_error("memory full (symbols)"); /* 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); + ts = tal_realloc(toksym_alloc, 0, sizeof(TokenSym) + len); table_ident[i] = ts; ts->tok = tok_ident++; ts->sym_define = NULL; @@ -77,10 +441,11 @@ static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len) } #define TOK_HASH_INIT 1 -#define TOK_HASH_FUNC(h, c) ((h) * 263 + (c)) +#define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c)) + /* find a token and add it if not found */ -static TokenSym *tok_alloc(const char *str, int len) +ST_FUNC TokenSym *tok_alloc(const char *str, int len) { TokenSym *ts, **pts; int i; @@ -105,31 +470,27 @@ static TokenSym *tok_alloc(const char *str, int len) /* XXX: buffer overflow */ /* XXX: float tokens */ -char *get_tok_str(int v, CValue *cv) +ST_FUNC const char *get_tok_str(int v, CValue *cv) { - static char buf[STRING_MAX_SIZE + 1]; - static CString cstr_buf; - CString *cstr; - unsigned char *q; 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; + p = cstr_buf.data; 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_CLONG: + case TOK_CULONG: case TOK_CLLONG: case TOK_CULLONG: /* XXX: not quite exact, but only useful for testing */ - sprintf(p, "%Lu", cv->ull); +#ifdef _WIN32 + sprintf(p, "%u", (unsigned)cv->i); +#else + sprintf(p, "%llu", (unsigned long long)cv->i); +#endif break; case TOK_LCHAR: cstr_ccat(&cstr_buf, 'L'); @@ -140,29 +501,39 @@ char *get_tok_str(int v, CValue *cv) cstr_ccat(&cstr_buf, '\0'); break; case TOK_PPNUM: - cstr = cv->cstr; - len = cstr->size - 1; - for(i=0;i<len;i++) - add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]); - cstr_ccat(&cstr_buf, '\0'); - break; + case TOK_PPSTR: + return (char*)cv->str.data; 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; + len = cv->str.size - 1; for(i=0;i<len;i++) - add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]); + add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]); } else { - len = (cstr->size / sizeof(nwchar_t)) - 1; + len = (cv->str.size / sizeof(nwchar_t)) - 1; for(i=0;i<len;i++) - add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]); + add_char(&cstr_buf, ((nwchar_t *)cv->str.data)[i]); } cstr_ccat(&cstr_buf, '\"'); cstr_ccat(&cstr_buf, '\0'); break; + + case TOK_CFLOAT: + cstr_cat(&cstr_buf, "<float>", 0); + break; + case TOK_CDOUBLE: + cstr_cat(&cstr_buf, "<double>", 0); + break; + case TOK_CLDOUBLE: + cstr_cat(&cstr_buf, "<long double>", 0); + break; + case TOK_LINENUM: + cstr_cat(&cstr_buf, "<linenumber>", 0); + break; + + /* above tokens have value, the ones below don't */ case TOK_LT: v = '<'; goto addv; @@ -175,19 +546,25 @@ char *get_tok_str(int v, CValue *cv) return strcpy(p, "<<="); case TOK_A_SAR: return strcpy(p, ">>="); + case TOK_EOF: + return strcpy(p, "<eof>"); default: if (v < TOK_IDENT) { /* search in two bytes table */ - q = tok_two_chars; + const unsigned char *q = tok_two_chars; while (*q) { if (q[2] == v) { *p++ = q[0]; *p++ = q[1]; *p = '\0'; - return buf; + return cstr_buf.data; } q += 3; } + if (v >= 127) { + sprintf(cstr_buf.data, "<%02x>", v); + return cstr_buf.data; + } addv: *p++ = v; *p = '\0'; @@ -205,15 +582,18 @@ char *get_tok_str(int v, CValue *cv) return cstr_buf.data; } -/* fill input buffer and peek next char */ -static int tcc_peekc_slow(BufferedFile *bf) +/* return the current character, handling end of block if necessary + (but not stray) */ +ST_FUNC int handle_eob(void) { + BufferedFile *bf = file; int len; + /* only tries to read if really end of buffer */ if (bf->buf_ptr >= bf->buf_end) { - if (bf->fd != -1) { + if (bf->fd >= 0) { #if defined(PARSE_DEBUG) - len = 8; + len = 1; #else len = IO_BUF_SIZE; #endif @@ -236,15 +616,8 @@ static int tcc_peekc_slow(BufferedFile *bf) } } -/* return the current character, handling end of block if necessary - (but not stray) */ -static int handle_eob(void) -{ - return tcc_peekc_slow(file); -} - /* read next char from current input file and handle end of input buffer */ -static inline void inp(void) +ST_INLN void inp(void) { ch = *(++(file->buf_ptr)); /* end of buffer/file handling */ @@ -277,7 +650,7 @@ static int handle_stray_noerror(void) static void handle_stray(void) { if (handle_stray_noerror()) - error("stray '\\' in program"); + tcc_error("stray '\\' in program"); } /* skip the stray and handle the \\n case. Output an error if @@ -286,20 +659,21 @@ static int handle_stray1(uint8_t *p) { int c; + file->buf_ptr = p; if (p >= file->buf_end) { - file->buf_ptr = p; c = handle_eob(); + if (c != '\\') + return c; 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; } + ch = *p; + if (handle_stray_noerror()) { + if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS)) + tcc_error("stray '\\' in program"); + *--file->buf_ptr = '\\'; + } + p = file->buf_ptr; + c = *p; return c; } @@ -329,14 +703,13 @@ static int handle_stray1(uint8_t *p) /* input with '\[\r]\n' handling. Note that this function cannot handle other characters after '\', so you cannot call it inside strings or comments */ -static void minp(void) +ST_FUNC void minp(void) { inp(); if (ch == '\\') handle_stray(); } - /* single line C++ comments */ static uint8_t *parse_line_comment(uint8_t *p) { @@ -375,10 +748,10 @@ static uint8_t *parse_line_comment(uint8_t *p) } /* C comments */ -static uint8_t *parse_comment(uint8_t *p) +ST_FUNC uint8_t *parse_comment(uint8_t *p) { int c; - + p++; for(;;) { /* fast skip loop */ @@ -408,6 +781,8 @@ static uint8_t *parse_comment(uint8_t *p) file->buf_ptr = p; c = handle_eob(); p = file->buf_ptr; + if (c == CH_EOF) + tcc_error("unexpected end of file in comment"); if (c == '\\') { /* skip '\[\r]\n', otherwise just skip the stray */ while (c == '\\') { @@ -437,7 +812,7 @@ static uint8_t *parse_comment(uint8_t *p) c = handle_eob(); p = file->buf_ptr; if (c == CH_EOF) { - error("unexpected end of file in comment"); + tcc_error("unexpected end of file in comment"); } else if (c == '\\') { p++; } @@ -448,17 +823,24 @@ static uint8_t *parse_comment(uint8_t *p) return p; } +ST_FUNC int set_idnum(int c, int val) +{ + int prev = isidnum_table[c - CH_EOF]; + isidnum_table[c - CH_EOF] = val; + return prev; +} + #define cinp minp static inline void skip_spaces(void) { - while (is_space(ch)) + while (isidnum_table[ch - CH_EOF] & IS_SPC) cinp(); } static inline int check_space(int t, int *spc) { - if (is_space(t)) { + if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) { if (*spc) return 1; *spc = 1; @@ -484,7 +866,7 @@ static uint8_t *parse_pp_string(uint8_t *p, if (c == CH_EOF) { unterminated_string: /* XXX: indicate line number of start of string */ - error("missing terminating %c character", sep); + tcc_error("missing terminating %c character", sep); } else if (c == '\\') { /* escape : just skip \[\r]\n */ PEEKC_EOB(c, p); @@ -532,7 +914,7 @@ static uint8_t *parse_pp_string(uint8_t *p, /* skip block of text until #else, #elif or #endif. skip also pairs of #if/#endif */ -void preprocess_skip(void) +static void preprocess_skip(void) { int a, start_of_line, c, in_warn_or_error; uint8_t *p; @@ -604,7 +986,12 @@ redo_start: a--; else if( tok == TOK_ERROR || tok == TOK_WARNING) in_warn_or_error = 1; - } + else if (tok == TOK_LINEFEED) + goto redo_start; + else if (parse_flags & PARSE_FLAG_ASM_FILE) + p = parse_line_comment(p - 1); + } else if (parse_flags & PARSE_FLAG_ASM_FILE) + p = parse_line_comment(p - 1); break; _default: default: @@ -617,35 +1004,12 @@ _default: 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' */ -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' */ -void restore_parse_state(ParseState *s) -{ - file->line_num = s->line_num; - macro_ptr = s->macro_ptr; - tok = s->tok; - tokc = s->tokc; -} - +#if 0 /* return the number of additional 'ints' necessary to store the token */ -static inline int tok_ext_size(int t) +static inline int tok_size(const int *p) { - switch(t) { + switch(*p) { /* 4 bytes */ case TOK_CINT: case TOK_CUINT: @@ -653,77 +1017,125 @@ static inline int tok_ext_size(int t) case TOK_LCHAR: case TOK_CFLOAT: case TOK_LINENUM: - return 1; + return 1 + 1; case TOK_STR: case TOK_LSTR: case TOK_PPNUM: - error("unsupported token"); - return 1; + case TOK_PPSTR: + return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2); + case TOK_CLONG: + case TOK_CULONG: + return 1 + LONG_SIZE / 4; case TOK_CDOUBLE: case TOK_CLLONG: case TOK_CULLONG: - return 2; + return 1 + 2; case TOK_CLDOUBLE: - return LDOUBLE_SIZE / 4; + return 1 + LDOUBLE_SIZE / 4; default: - return 0; + return 1 + 0; } } +#endif /* token string handling */ - -static inline void tok_str_new(TokenString *s) +ST_INLN void tok_str_new(TokenString *s) { s->str = NULL; - s->len = 0; + s->len = s->lastlen = 0; s->allocated_len = 0; s->last_line_num = -1; } -static void tok_str_free(int *str) +ST_FUNC TokenString *tok_str_alloc(void) { - tcc_free(str); + TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str); + tok_str_new(str); + return str; } -static int *tok_str_realloc(TokenString *s) +ST_FUNC int *tok_str_dup(TokenString *s) { - int *str, len; + int *str; - 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; + str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int)); + memcpy(str, s->str, s->len * sizeof(int)); return str; } -static void tok_str_add(TokenString *s, int t) +ST_FUNC void tok_str_free_str(int *str) +{ + tal_free(tokstr_alloc, str); +} + +ST_FUNC void tok_str_free(TokenString *str) +{ + tok_str_free_str(str->str); + tal_free(tokstr_alloc, str); +} + +ST_FUNC int *tok_str_realloc(TokenString *s, int new_size) +{ + int *str, size; + + size = s->allocated_len; + if (size < 16) + size = 16; + while (size < new_size) + size = size * 2; + if (size > s->allocated_len) { + str = tal_realloc(tokstr_alloc, s->str, size * sizeof(int)); + s->allocated_len = size; + s->str = str; + } + return s->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 = tok_str_realloc(s, len + 1); str[len++] = t; s->len = len; } +ST_FUNC void begin_macro(TokenString *str, int alloc) +{ + str->alloc = alloc; + str->prev = macro_stack; + str->prev_ptr = macro_ptr; + str->save_line_num = file->line_num; + macro_ptr = str->str; + macro_stack = str; +} + +ST_FUNC void end_macro(void) +{ + TokenString *str = macro_stack; + macro_stack = str->prev; + macro_ptr = str->prev_ptr; + file->line_num = str->save_line_num; + if (str->alloc == 2) { + str->alloc = 3; /* just mark as finished */ + } else { + tok_str_free(str); + } +} + static void tok_str_add2(TokenString *s, int t, CValue *cv) { int len, *str; - len = s->len; + len = s->lastlen = s->len; str = s->str; /* allocate space for worst case */ - if (len + TOK_MAX_SIZE > s->allocated_len) - str = tok_str_realloc(s); + if (len + TOK_MAX_SIZE >= s->allocated_len) + str = tok_str_realloc(s, len + TOK_MAX_SIZE + 1); str[len++] = t; switch(t) { case TOK_CINT: @@ -732,31 +1144,34 @@ static void tok_str_add2(TokenString *s, int t, CValue *cv) case TOK_LCHAR: case TOK_CFLOAT: case TOK_LINENUM: +#if LONG_SIZE == 4 + case TOK_CLONG: + case TOK_CULONG: +#endif str[len++] = cv->tab[0]; break; case TOK_PPNUM: + case TOK_PPSTR: 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); + /* Insert the string into the int array. */ + size_t nb_words = + 1 + (cv->str.size + sizeof(int) - 1) / sizeof(int); + if (len + nb_words >= s->allocated_len) + str = tok_str_realloc(s, len + nb_words + 1); + str[len] = cv->str.size; + memcpy(&str[len + 1], cv->str.data, cv->str.size); len += nb_words; } break; case TOK_CDOUBLE: case TOK_CLLONG: case TOK_CULLONG: +#if LONG_SIZE == 8 + case TOK_CLONG: + case TOK_CULONG: +#endif #if LDOUBLE_SIZE == 8 case TOK_CLDOUBLE: #endif @@ -785,7 +1200,7 @@ static void tok_str_add2(TokenString *s, int t, CValue *cv) } /* add the current parse token in token string 's' */ -static void tok_str_add_tok(TokenString *s) +ST_FUNC void tok_str_add_tok(TokenString *s) { CValue cval; @@ -798,84 +1213,115 @@ static void tok_str_add_tok(TokenString *s) 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++) { +#if LONG_SIZE == 4 + case TOK_CLONG: +#endif + case TOK_CINT: + case TOK_CCHAR: + case TOK_LCHAR: + case TOK_LINENUM: + cv->i = *p++; + break; +#if LONG_SIZE == 4 + case TOK_CULONG: +#endif + case TOK_CUINT: + cv->i = (unsigned)*p++; + break; + case TOK_CFLOAT: + tab[0] = *p++; + break; + case TOK_STR: + case TOK_LSTR: + case TOK_PPNUM: + case TOK_PPSTR: + cv->str.size = *p++; + cv->str.data = p; + p += (cv->str.size + sizeof(int) - 1) / sizeof(int); + break; + case TOK_CDOUBLE: + case TOK_CLLONG: + case TOK_CULLONG: +#if LONG_SIZE == 8 + case TOK_CLONG: + case TOK_CULONG: +#endif + n = 2; + goto copy; + case TOK_CLDOUBLE: #if LDOUBLE_SIZE == 16 -#define LDOUBLE_GET(p, cv) \ - cv.tab[0] = p[0]; \ - cv.tab[1] = p[1]; \ - cv.tab[2] = p[2]; \ - cv.tab[3] = p[3]; + n = 4; #elif LDOUBLE_SIZE == 12 -#define LDOUBLE_GET(p, cv) \ - cv.tab[0] = p[0]; \ - cv.tab[1] = p[1]; \ - cv.tab[2] = p[2]; + n = 3; #elif LDOUBLE_SIZE == 8 -#define LDOUBLE_GET(p, cv) \ - cv.tab[0] = p[0]; \ - cv.tab[1] = p[1]; + n = 2; #else -#error add long double size support +# 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) +{ + CValue cv; + int t; -/* get a token from an integer array and increment pointer - accordingly. we code it as a macro to avoid pointer aliasing. */ -#define TOK_GET(t, p, cv) \ -{ \ - t = *p++; \ - switch(t) { \ - case TOK_CINT: \ - case TOK_CUINT: \ - case TOK_CCHAR: \ - case TOK_LCHAR: \ - case TOK_CFLOAT: \ - case TOK_LINENUM: \ - cv.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: \ - cv.tab[0] = p[0]; \ - cv.tab[1] = p[1]; \ - p += 2; \ - break; \ - case TOK_CLDOUBLE: \ - LDOUBLE_GET(p, cv); \ - p += LDOUBLE_SIZE / 4; \ - break; \ - default: \ - break; \ - } \ + if (!a || !b) + return 1; + + while (*a && *b) { + /* first time preallocate macro_equal_buf, next time only reset position to start */ + cstr_reset(¯o_equal_buf); + TOK_GET(&t, &a, &cv); + cstr_cat(¯o_equal_buf, get_tok_str(t, &cv), 0); + TOK_GET(&t, &b, &cv); + if (strcmp(macro_equal_buf.data, get_tok_str(t, &cv))) + return 0; + } + return !(*a || *b); } /* defines handling */ -static inline void define_push(int v, int macro_type, int *str, Sym *first_arg) +ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg) { - Sym *s; + Sym *s, *o; - s = sym_push2(&define_stack, v, macro_type, (long)str); + o = define_find(v); + s = sym_push2(&define_stack, v, macro_type, 0); + s->d = str; s->next = first_arg; table_ident[v - TOK_IDENT]->sym_define = s; + + if (o && !macro_is_equal(o->d, s->d)) + tcc_warning("%s redefined", get_tok_str(v, NULL)); } /* undefined a define symbol. Its name is just set to zero */ -static void define_undef(Sym *s) +ST_FUNC void define_undef(Sym *s) { - int v; - v = s->v; + int v = s->v; if (v >= TOK_IDENT && v < tok_ident) table_ident[v - TOK_IDENT]->sym_define = NULL; - s->v = 0; } -static inline Sym *define_find(int v) +ST_INLN Sym *define_find(int v) { v -= TOK_IDENT; if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT)) @@ -884,28 +1330,31 @@ static inline Sym *define_find(int v) } /* free define stack until top reaches 'b' */ -static void free_defines(Sym *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->c) - tok_str_free((int *)top->c); - v = top->v; - if (v >= TOK_IDENT && v < tok_ident) - table_ident[v - TOK_IDENT]->sym_define = NULL; + while (define_stack != b) { + Sym *top = define_stack; + define_stack = top->prev; + tok_str_free_str(top->d); + define_undef(top); sym_free(top); - top = top1; } - define_stack = b; + + /* restore remaining (-D or predefined) symbols if they were + #undef'd in the file */ + while (b) { + int v = b->v; + if (v >= TOK_IDENT && v < tok_ident) { + Sym **d = &table_ident[v - TOK_IDENT]->sym_define; + if (!*d) + *d = b; + } + b = b->prev; + } } /* label lookup */ -static Sym *label_find(int v) +ST_FUNC Sym *label_find(int v) { v -= TOK_IDENT; if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT)) @@ -913,7 +1362,7 @@ static Sym *label_find(int v) return table_ident[v]->sym_label; } -static Sym *label_push(Sym **ptop, int v, int flags) +ST_FUNC Sym *label_push(Sym **ptop, int v, int flags) { Sym *s, **ps; s = sym_push2(ptop, v, 0, 0); @@ -932,37 +1381,55 @@ static Sym *label_push(Sym **ptop, int v, int flags) /* pop labels until element last is reached. Look if any labels are undefined. Define symbols if '&&label' was used. */ -static void label_pop(Sym **ptop, Sym *slast) +ST_FUNC void label_pop(Sym **ptop, Sym *slast, int keep) { 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)); + tcc_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", + tcc_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, (long)s->next, 1); + 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); + if (!keep) + sym_free(s); } - *ptop = slast; + if (!keep) + *ptop = slast; +} + +/* fake the nth "#if defined test_..." for tcc -dt -run */ +static void maybe_run_test(TCCState *s) +{ + const char *p; + if (s->include_stack_ptr != s->include_stack) + return; + p = get_tok_str(tok, NULL); + if (0 != memcmp(p, "test_", 5)) + return; + if (0 != --s->run_test) + return; + fprintf(s->ppfp, "\n[%s]\n" + !(s->dflag & 32), p), fflush(s->ppfp); + define_push(tok, MACRO_OBJ, NULL, NULL); } /* eval an expression for #if/#elif */ static int expr_preprocess(void) { int c, t; - TokenString str; + TokenString *str; - tok_str_new(&str); + str = tok_str_alloc(); + pp_expr = 1; while (tok != TOK_LINEFEED && tok != TOK_EOF) { next(); /* do macro subst */ if (tok == TOK_DEFINED) { @@ -970,9 +1437,16 @@ static int expr_preprocess(void) t = tok; if (t == '(') next_nomacro(); + if (tok < TOK_IDENT) + expect("identifier"); + if (tcc_state->run_test) + maybe_run_test(tcc_state); c = define_find(tok) != 0; - if (t == '(') + if (t == '(') { next_nomacro(); + if (tok != ')') + expect("')'"); + } tok = TOK_CINT; tokc.i = c; } else if (tok >= TOK_IDENT) { @@ -980,55 +1454,45 @@ static int expr_preprocess(void) tok = TOK_CINT; tokc.i = 0; } - tok_str_add_tok(&str); + tok_str_add_tok(str); } - tok_str_add(&str, -1); /* simulate end of file */ - tok_str_add(&str, 0); + pp_expr = 0; + tok_str_add(str, -1); /* simulate end of file */ + tok_str_add(str, 0); /* now evaluate C constant expression */ - macro_ptr = str.str; + begin_macro(str, 1); next(); c = expr_const(); - macro_ptr = NULL; - tok_str_free(str.str); + end_macro(); 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 */ -static void parse_define(void) +ST_FUNC void parse_define(void) { Sym *s, *first, **ps; int v, t, varg, is_vaargs, spc; - TokenString str; - + int saved_parse_flags = parse_flags; + v = tok; - if (v < TOK_IDENT) - error("invalid macro name '%s'", get_tok_str(tok, &tokc)); + if (v < TOK_IDENT || v == TOK_DEFINED) + tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc)); /* XXX: should check if same macro (ANSI) */ first = NULL; t = MACRO_OBJ; + /* We have to parse the whole define as if not in asm mode, in particular + no line comment with '#' must be ignored. Also for function + macros the argument list must be parsed without '.' being an ID + character. */ + parse_flags = ((parse_flags & ~PARSE_FLAG_ASM_FILE) | PARSE_FLAG_SPACES); /* '(' must be just after macro definition for MACRO_FUNC */ next_nomacro_spc(); if (tok == '(') { + int dotid = set_idnum('.', 0); next_nomacro(); ps = &first; - while (tok != ')') { + if (tok != ')') for (;;) { varg = tok; next_nomacro(); is_vaargs = 0; @@ -1040,129 +1504,161 @@ static void parse_define(void) next_nomacro(); } if (varg < TOK_IDENT) - error("badly punctuated parameter list"); + bad_list: + tcc_error("bad macro parameter list"); s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0); *ps = s; ps = &s->next; - if (tok != ',') + if (tok == ')') break; + if (tok != ',' || is_vaargs) + goto bad_list; next_nomacro(); } - if (tok == ')') - next_nomacro_spc(); + next_nomacro_spc(); t = MACRO_FUNC; + set_idnum('.', dotid); } - tok_str_new(&str); + + tokstr_buf.len = 0; spc = 2; - /* EOF testing necessary for '-D' handling */ + parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED; + /* The body of a macro definition should be parsed such that identifiers + are parsed like the file mode determines (i.e. with '.' being an + ID character in asm mode). But '#' should be retained instead of + regarded as line comment leader, so still don't set ASM_FILE + in parse_flags. */ while (tok != TOK_LINEFEED && tok != TOK_EOF) { - /* remove spaces around ## and after '#' */ + /* remove spaces around ## and after '#' */ if (TOK_TWOSHARPS == tok) { + if (2 == spc) + goto bad_twosharp; if (1 == spc) - --str.len; - spc = 2; + --tokstr_buf.len; + spc = 3; + tok = TOK_PPJOIN; } else if ('#' == tok) { - spc = 2; + spc = 4; } else if (check_space(tok, &spc)) { goto skip; } - tok_str_add2(&str, tok, &tokc); + tok_str_add2(&tokstr_buf, tok, &tokc); skip: next_nomacro_spc(); } + + parse_flags = saved_parse_flags; 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); + --tokstr_buf.len; /* remove trailing space */ + tok_str_add(&tokstr_buf, 0); + if (3 == spc) +bad_twosharp: + tcc_error("'##' cannot appear at either end of macro"); + define_push(v, t, tok_str_dup(&tokstr_buf), first); } -static inline int hash_cached_include(int type, const char *filename) +static CachedInclude *search_cached_include(TCCState *s1, const char *filename, int add) { const unsigned char *s; unsigned int h; + CachedInclude *e; + int i; h = TOK_HASH_INIT; - h = TOK_HASH_FUNC(h, type); - s = filename; + s = (unsigned char *) filename; while (*s) { +#ifdef _WIN32 + h = TOK_HASH_FUNC(h, toup(*s)); +#else h = TOK_HASH_FUNC(h, *s); +#endif 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)) + if (0 == 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 (!add) + return NULL; - 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); + e->ifndef_macro = e->once = 0; + dynarray_add(&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; +#ifdef INC_DEBUG + printf("adding cached '%s'\n", filename); +#endif + return e; } 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_nomacro(); + if (tok == TOK_push_macro || tok == TOK_pop_macro) { + int t = tok, v; + Sym *s; + + if (next(), tok != '(') + goto pragma_err; + if (next(), tok != TOK_STR) + goto pragma_err; + v = tok_alloc(tokc.str.data, tokc.str.size - 1)->tok; + if (next(), tok != ')') + goto pragma_err; + if (t == TOK_push_macro) { + while (NULL == (s = define_find(v))) + define_push(v, 0, NULL, NULL); + s->type.ref = s; /* set push boundary */ + } else { + for (s = define_stack; s; s = s->prev) + if (s->v == v && s->type.ref == s) { + s->type.ref = NULL; + break; + } + } + if (s) + table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL; + else + tcc_warning("unbalanced #pragma pop_macro"); + pp_debug_tok = t, pp_debug_symv = v; + + } else if (tok == TOK_once) { + search_cached_include(s1, file->filename, 1)->once = pp_once; + + } else if (s1->output_type == TCC_OUTPUT_PREPROCESS) { + /* tcc -E: keep pragmas below unchanged */ + unget_tok(' '); + unget_tok(TOK_PRAGMA); + unget_tok('#'); + unget_tok(TOK_LINEFEED); + + } else 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"); + tcc_error("out of pack stack"); } s1->pack_stack_ptr--; } else { - val = 0; + int val = 0; if (tok != ')') { if (tok == TOK_ASM_push) { next(); @@ -1171,23 +1667,51 @@ static void pragma_parse(TCCState *s1) s1->pack_stack_ptr++; skip(','); } - if (tok != TOK_CINT) { - pack_error: - error("invalid pack pragma"); - } + if (tok != TOK_CINT) + goto pragma_err; val = tokc.i; if (val < 1 || val > 16 || (val & (val - 1)) != 0) - goto pack_error; + goto pragma_err; next(); } *s1->pack_stack_ptr = val; - skip(')'); } + if (tok != ')') + goto pragma_err; + + } else if (tok == TOK_comment) { + char *p; int t; + next(); + skip('('); + t = tok; + next(); + skip(','); + if (tok != TOK_STR) + goto pragma_err; + p = tcc_strdup((char *)tokc.str.data); + next(); + if (tok != ')') + goto pragma_err; + if (t == TOK_lib) { + dynarray_add(&s1->pragma_libs, &s1->nb_pragma_libs, p); + } else { + if (t == TOK_option) + tcc_set_options(s1, p); + tcc_free(p); + } + + } else if (s1->warn_unsupported) { + tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc)); } + return; + +pragma_err: + tcc_error("malformed #pragma directive"); + return; } /* is_bof is true if first non space token at beginning of file */ -static void preprocess(int is_bof) +ST_FUNC void preprocess(int is_bof) { TCCState *s1 = tcc_state; int i, c, n, saved_parse_flags; @@ -1195,17 +1719,26 @@ static void preprocess(int is_bof) Sym *s; saved_parse_flags = parse_flags; - parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM | - PARSE_FLAG_LINEFEED; + parse_flags = PARSE_FLAG_PREPROCESS + | PARSE_FLAG_TOK_NUM + | PARSE_FLAG_TOK_STR + | PARSE_FLAG_LINEFEED + | (parse_flags & PARSE_FLAG_ASM_FILE) + ; + next_nomacro(); redo: switch(tok) { case TOK_DEFINE: + pp_debug_tok = tok; next_nomacro(); + pp_debug_symv = tok; parse_define(); break; case TOK_UNDEF: + pp_debug_tok = tok; next_nomacro(); + pp_debug_symv = tok; s = define_find(tok); /* undefine symbol by putting an invalid name */ if (s) @@ -1242,116 +1775,93 @@ static void preprocess(int is_bof) inp(); #endif } else { - /* computed #include : either we have only strings or - we have anything enclosed in '<>' */ + int len; + /* computed #include : concatenate everything up to linefeed, + the result must be one of the two accepted forms. + Don't convert pp-tokens to tokens here. */ + parse_flags = (PARSE_FLAG_PREPROCESS + | PARSE_FLAG_LINEFEED + | (parse_flags & PARSE_FLAG_ASM_FILE)); next(); buf[0] = '\0'; - if (tok == TOK_STR) { - while (tok != TOK_LINEFEED) { - if (tok != TOK_STR) { - include_syntax: - error("'#include' expects \"FILENAME\" or <FILENAME>"); - } - 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 = '>'; - } + 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] != '"') && + (buf[0] != '<' || buf[len-1] != '>')))) + tcc_error("'#include' expects \"FILENAME\" or <FILENAME>"); + c = buf[len-1]; + memmove(buf, buf + 1, len - 2); + buf[len - 2] = '\0'; } 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) { + tcc_error("#include recursion too deep"); + /* store current file in stack, but increment stack later below */ + *s1->include_stack_ptr = file; + i = tok == TOK_INCLUDE_NEXT ? file->include_next_index : 0; + n = 2 + s1->nb_include_paths + s1->nb_sysinclude_paths; + for (; i < n; ++i) { char buf1[sizeof file->filename]; - BufferedFile *f; CachedInclude *e; const char *path; - int size; - if (i == -2) { + if (i == 0) { /* check absolute include path */ if (!IS_ABSPATH(buf)) continue; buf1[0] = 0; - } else if (i == -1) { - /* search in current dir if "header.h" */ + } else if (i == 1) { + /* search in file's dir if "header.h" */ if (c != '\"') continue; - size = tcc_basename(file->filename) - file->filename; - memcpy(buf1, file->filename, size); - buf1[size] = '\0'; + /* https://savannah.nongnu.org/bugs/index.php?50847 */ + path = file->true_filename; + pstrncpy(buf1, path, tcc_basename(path) - path); } 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]; + int j = i - 2, k = j - s1->nb_include_paths; + path = k < 0 ? s1->include_paths[j] : s1->sysinclude_paths[k]; 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)) { + e = search_cached_include(s1, buf1, 0); + if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) { /* no need to parse the include because the 'ifndef macro' - is defined */ + is defined (or had #pragma once) */ #ifdef INC_DEBUG - printf("%s: skipping %s\n", file->filename, buf); + printf("%s: skipping cached %s\n", file->filename, buf1); #endif - f = NULL; - } else { - f = tcc_open(s1, buf1); - if (!f) - continue; + goto include_done; } - if (tok == TOK_INCLUDE_NEXT) { - tok = TOK_INCLUDE; - if (f) - tcc_close(f); + if (tcc_open(s1, buf1) < 0) continue; - } - - if (!f) - goto include_done; + file->include_next_index = i + 1; #ifdef INC_DEBUG - printf("%s: including %s\n", file->filename, buf1); + printf("%s: including %s\n", file->prev->filename, file->filename); #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; + /* update target deps */ + dynarray_add(&s1->target_deps, &s1->nb_target_deps, + tcc_strdup(buf1)); + /* push current file in stack */ + ++s1->include_stack_ptr; /* add include file debug info */ - if (tcc_state->do_debug) { + 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); + tcc_error("include file '%s' not found", buf); include_done: break; case TOK_IFNDEF: @@ -1365,7 +1875,7 @@ include_done: do_ifdef: next_nomacro(); if (tok < TOK_IDENT) - error("invalid argument for '#if%sdef'", c ? "n" : ""); + tcc_error("invalid argument for '#if%sdef'", c ? "n" : ""); if (is_bof) { if (c) { #ifdef INC_DEBUG @@ -1377,30 +1887,34 @@ include_done: c = (define_find(tok) != 0) ^ c; do_if: if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE) - error("memory full"); + tcc_error("memory full (ifdef)"); *s1->ifdef_stack_ptr++ = c; goto test_skip; case TOK_ELSE: if (s1->ifdef_stack_ptr == s1->ifdef_stack) - error("#else without matching #if"); + tcc_error("#else without matching #if"); if (s1->ifdef_stack_ptr[-1] & 2) - error("#else after #else"); + tcc_error("#else after #else"); c = (s1->ifdef_stack_ptr[-1] ^= 3); - goto test_skip; + goto test_else; case TOK_ELIF: if (s1->ifdef_stack_ptr == s1->ifdef_stack) - error("#elif without matching #if"); + tcc_error("#elif without matching #if"); c = s1->ifdef_stack_ptr[-1]; if (c > 1) - error("#elif after #else"); + tcc_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; + if (c == 1) { + c = 0; + } else { + 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; @@ -1408,7 +1922,7 @@ include_done: break; case TOK_ENDIF: if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr) - error("#endif without matching #if"); + tcc_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 */ @@ -1424,18 +1938,33 @@ include_done: goto the_end; } break; + case TOK_PPNUM: + n = strtoul((char*)tokc.str.data, &q, 10); + goto _line_num; case TOK_LINE: next(); if (tok != TOK_CINT) - error("#line"); - file->line_num = tokc.i - 1; /* the line number will be incremented after */ + _line_err: + tcc_error("wrong #line format"); + n = tokc.i; + _line_num: next(); if (tok != TOK_LINEFEED) { - if (tok != TOK_STR) - error("#line"); - pstrcpy(file->filename, sizeof(file->filename), - (char *)tokc.cstr->data); + if (tok == TOK_STR) { + if (file->true_filename == file->filename) + file->true_filename = tcc_strdup(file->filename); + pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.str.data); + } else if (parse_flags & PARSE_FLAG_ASM_FILE) + break; + else + goto _line_err; + --n; } + if (file->fd > 0) + total_lines += file->line_num - n; + file->line_num = n; + if (s1->do_debug) + put_stabs(file->filename, N_BINCL, 0, 0, 0); break; case TOK_ERROR: case TOK_WARNING: @@ -1454,22 +1983,26 @@ include_done: } *q = '\0'; if (c == TOK_ERROR) - error("#error %s", buf); + tcc_error("#error %s", buf); else - warning("#warning %s", buf); + tcc_warning("#warning %s", buf); break; case TOK_PRAGMA: pragma_parse(s1); break; + case TOK_LINEFEED: + goto the_end; 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 gas line comment in an 'S' file. */ + if (saved_parse_flags & PARSE_FLAG_ASM_FILE) + goto ignore; + if (tok == '!' && is_bof) + /* '!' is ignored at beginning to allow C scripts. */ + goto ignore; + tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc)); + ignore: + file->buf_ptr = parse_line_comment(file->buf_ptr - 1); + goto the_end; } /* ignore other preprocess commands or #! for C scripts */ while (tok != TOK_LINEFEED) @@ -1565,18 +2098,84 @@ static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long default: invalid_escape: if (c >= '!' && c <= '~') - warning("unknown escape sequence: \'\\%c\'", c); + tcc_warning("unknown escape sequence: \'\\%c\'", c); else - warning("unknown escape sequence: \'\\x%x\'", c); + tcc_warning("unknown escape sequence: \'\\x%x\'", c); break; } + } else if (is_long && c >= 0x80) { + /* assume we are processing UTF-8 sequence */ + /* reference: The Unicode Standard, Version 10.0, ch3.9 */ + + int cont; /* count of continuation bytes */ + int skip; /* how many bytes should skip when error occurred */ + int i; + + /* decode leading byte */ + if (c < 0xC2) { + skip = 1; goto invalid_utf8_sequence; + } else if (c <= 0xDF) { + cont = 1; n = c & 0x1f; + } else if (c <= 0xEF) { + cont = 2; n = c & 0xf; + } else if (c <= 0xF4) { + cont = 3; n = c & 0x7; + } else { + skip = 1; goto invalid_utf8_sequence; + } + + /* decode continuation bytes */ + for (i = 1; i <= cont; i++) { + int l = 0x80, h = 0xBF; + + /* adjust limit for second byte */ + if (i == 1) { + switch (c) { + case 0xE0: l = 0xA0; break; + case 0xED: h = 0x9F; break; + case 0xF0: l = 0x90; break; + case 0xF4: h = 0x8F; break; + } + } + + if (p[i] < l || p[i] > h) { + skip = i; goto invalid_utf8_sequence; + } + + n = (n << 6) | (p[i] & 0x3f); + } + + /* advance pointer */ + p += 1 + cont; + c = n; + goto add_char_nonext; + + /* error handling */ + invalid_utf8_sequence: + tcc_warning("ill-formed UTF-8 subsequence starting with: \'\\x%x\'", c); + c = 0xFFFD; + p += skip; + goto add_char_nonext; + } p++; add_char_nonext: if (!is_long) cstr_ccat(outstr, c); - else + else { +#ifdef TCC_TARGET_PE + /* store as UTF-16 */ + if (c < 0x10000) { + cstr_wccat(outstr, c); + } else { + c -= 0x10000; + cstr_wccat(outstr, (c >> 10) + 0xD800); + cstr_wccat(outstr, (c & 0x3FF) + 0xDC00); + } +#else cstr_wccat(outstr, c); +#endif + } } /* add a trailing '\0' */ if (!is_long) @@ -1585,11 +2184,59 @@ static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long cstr_wccat(outstr, '\0'); } +static void parse_string(const char *s, int len) +{ + uint8_t buf[1000], *p = buf; + int is_long, sep; + + if ((is_long = *s == 'L')) + ++s, --len; + sep = *s++; + len -= 2; + if (len >= sizeof buf) + p = tcc_malloc(len + 1); + memcpy(p, s, len); + p[len] = 0; + + cstr_reset(&tokcstr); + parse_escape_string(&tokcstr, p, is_long); + if (p != buf) + tcc_free(p); + + if (sep == '\'') { + int char_size, i, n, c; + /* XXX: make it portable */ + if (!is_long) + tok = TOK_CCHAR, char_size = 1; + else + tok = TOK_LCHAR, char_size = sizeof(nwchar_t); + n = tokcstr.size / char_size - 1; + if (n < 1) + tcc_error("empty character constant"); + if (n > 1) + tcc_warning("multi-character character constant"); + for (c = i = 0; i < n; ++i) { + if (is_long) + c = ((nwchar_t *)tokcstr.data)[i]; + else + c = (c << 8) | ((char *)tokcstr.data)[i]; + } + tokc.i = c; + } else { + tokc.str.size = tokcstr.size; + tokc.str.data = tokcstr.data; + if (!is_long) + tok = TOK_STR; + else + tok = TOK_LSTR; + } +} + /* we use 64 bit numbers */ #define BN_SIZE 2 /* bn = (bn << shift) | or_val */ -void bn_lshift(unsigned int *bn, int shift, int or_val) +static void bn_lshift(unsigned int *bn, int shift, int or_val) { int i; unsigned int v; @@ -1600,7 +2247,7 @@ void bn_lshift(unsigned int *bn, int shift, int or_val) } } -void bn_zero(unsigned int *bn) +static void bn_zero(unsigned int *bn) { int i; for(i=0;i<BN_SIZE;i++) { @@ -1610,7 +2257,7 @@ void bn_zero(unsigned int *bn) /* parse number in null terminated string 'p' and return it in the current token */ -void parse_number(const char *p) +static void parse_number(const char *p) { int b, t, shift, frac_bits, s, exp_val, ch; char *q; @@ -1652,7 +2299,7 @@ void parse_number(const char *p) break; if (q >= token_buf + STRING_MAX_SIZE) { num_too_long: - error("number too long"); + tcc_error("number too long"); } *q++ = ch; ch = *p++; @@ -1670,7 +2317,7 @@ void parse_number(const char *p) if (b == 16) shift = 4; else - shift = 2; + shift = 1; bn_zero(bn); q = token_buf; while (1) { @@ -1701,7 +2348,7 @@ void parse_number(const char *p) break; } if (t >= b) - error("invalid digit"); + tcc_error("invalid digit"); bn_lshift(bn, shift, t); frac_bits += shift; ch = *p++; @@ -1738,9 +2385,14 @@ void parse_number(const char *p) 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; @@ -1789,8 +2441,13 @@ void parse_number(const char *p) 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); @@ -1798,7 +2455,8 @@ void parse_number(const char *p) } } else { unsigned long long n, n1; - int lcount, ucount; + int lcount, ucount, ov = 0; + const char *p1; /* integer number */ *q = '\0'; @@ -1811,71 +2469,80 @@ void parse_number(const char *p) while(1) { t = *q++; /* no need for checks except for base 10 / 8 errors */ - if (t == '\0') { + if (t == '\0') break; - } else if (t >= 'a') { + else if (t >= 'a') t = t - 'a' + 10; - } else if (t >= 'A') { + else if (t >= 'A') t = t - 'A' + 10; - } else { + else t = t - '0'; - if (t >= b) - error("invalid digit"); - } + if (t >= b) + tcc_error("invalid digit"); n1 = n; n = n * b + t; /* detect overflow */ - /* XXX: this test is not reliable */ - if (n < n1) - error("integer constant overflow"); + if (n1 >= 0x1000000000000000ULL && n / b != n1) + ov = 1; } - - /* 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; + + /* Determine the characteristics (unsigned and/or 64bit) the type of + the constant must have according to the constant suffix(es) */ + lcount = ucount = 0; + p1 = p; for(;;) { t = toup(ch); if (t == 'L') { if (lcount >= 2) - error("three 'l's in integer constant"); + tcc_error("three 'l's in integer constant"); + if (lcount && *(p - 1) != ch) + tcc_error("incorrect integer suffix: %s", p1); 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"); + tcc_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; + + /* Determine if it needs 64 bits and/or unsigned in order to fit */ + if (ucount == 0 && b == 10) { + if (lcount <= (LONG_SIZE == 4)) { + if (n >= 0x80000000U) + lcount = (LONG_SIZE == 4) + 1; + } + if (n >= 0x8000000000000000ULL) + ov = 1, ucount = 1; + } else { + if (lcount <= (LONG_SIZE == 4)) { + if (n >= 0x100000000ULL) + lcount = (LONG_SIZE == 4) + 1; + else if (n >= 0x80000000U) + ucount = 1; + } + if (n >= 0x8000000000000000ULL) + ucount = 1; + } + + if (ov) + tcc_warning("integer constant overflow"); + + tok = TOK_CINT; + if (lcount) { + tok = TOK_CLONG; + if (lcount == 2) + tok = TOK_CLLONG; + } + if (ucount) + ++tok; /* TOK_CU... */ + tokc.i = n; } if (ch) - error("invalid number\n"); + tcc_error("invalid number\n"); } @@ -1893,7 +2560,7 @@ void parse_number(const char *p) /* return next token without macro substitution */ static inline void next_nomacro1(void) { - int t, c, is_long; + int t, c, is_long, len; TokenSym *ts; uint8_t *p, *p1; unsigned int h; @@ -1906,7 +2573,11 @@ static inline void next_nomacro1(void) case '\t': tok = c; p++; - goto keep_tok_flags; + if (parse_flags & PARSE_FLAG_SPACES) + goto keep_tok_flags; + while (isidnum_table[*p - CH_EOF] & IS_SPC) + ++p; + goto redo_no_start; case '\f': case '\v': case '\r': @@ -1914,22 +2585,12 @@ static inline void next_nomacro1(void) 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; + c = handle_stray1(p); + p = file->buf_ptr; + if (c == '\\') + goto parse_simple; + if (c != CH_EOF) goto redo_no_start; - } - parse_eof: { TCCState *s1 = tcc_state; if ((parse_flags & PARSE_FLAG_LINEFEED) @@ -1937,8 +2598,11 @@ static inline void next_nomacro1(void) tok_flags |= TOK_FLAG_EOF; tok = TOK_LINEFEED; goto keep_tok_flags; - } else if (s1->include_stack_ptr == s1->include_stack || - !(parse_flags & PARSE_FLAG_PREPROCESS)) { + } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) { + tok = TOK_EOF; + } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) { + tcc_error("missing #endif"); + } else if (s1->include_stack_ptr == s1->include_stack) { /* no include left : end of file. */ tok = TOK_EOF; } else { @@ -1951,8 +2615,9 @@ static inline void next_nomacro1(void) #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); + search_cached_include(s1, file->filename, 1) + ->ifndef_macro = file->ifndef_macro_saved; + tok_flags &= ~TOK_FLAG_ENDIF; } /* add end of include file debug info */ @@ -1960,10 +2625,11 @@ static inline void next_nomacro1(void) put_stabd(N_EINCL, 0, 0); } /* pop include stack */ - tcc_close(file); + tcc_close(); s1->include_stack_ptr--; - file = *s1->include_stack_ptr; p = file->buf_ptr; + if (p == file->buffer) + tok_flags = TOK_FLAG_BOF|TOK_FLAG_BOL; goto redo_no_start; } } @@ -1973,6 +2639,7 @@ static inline void next_nomacro1(void) 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; @@ -1986,13 +2653,13 @@ static inline void next_nomacro1(void) file->buf_ptr = p; preprocess(tok_flags & TOK_FLAG_BOF); p = file->buf_ptr; - goto redo_no_start; + goto maybe_newline; } else { if (c == '#') { p++; tok = TOK_TWOSHARPS; } else { - if (parse_flags & PARSE_FLAG_ASM_COMMENTS) { + if (parse_flags & PARSE_FLAG_ASM_FILE) { p = parse_line_comment(p - 1); goto redo_no_start; } else { @@ -2001,6 +2668,12 @@ static inline void next_nomacro1(void) } } break; + + /* dollar is allowed to start identifiers when not parsing asm */ + case '$': + if (!(isidnum_table[c - CH_EOF] & IS_ID) + || (parse_flags & PARSE_FLAG_ASM_FILE)) + goto parse_simple; case 'a': case 'b': case 'c': case 'd': case 'e': case 'f': case 'g': case 'h': @@ -2021,21 +2694,14 @@ static inline void next_nomacro1(void) p1 = p; h = TOK_HASH_INIT; h = TOK_HASH_FUNC(h, c); - p++; - for(;;) { - c = *p; - if (!isidnum_table[c-CH_EOF]) - break; + while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM)) h = TOK_HASH_FUNC(h, c); - p++; - } + len = p - p1; 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(;;) { @@ -2046,20 +2712,17 @@ static inline void next_nomacro1(void) goto token_found; pts = &(ts->hash_next); } - ts = tok_alloc_new(pts, p1, len); + ts = tok_alloc_new(pts, (char *) p1, len); token_found: ; } else { /* slower case */ cstr_reset(&tokcstr); - - while (p1 < p) { - cstr_ccat(&tokcstr, *p1); - p1++; - } + cstr_cat(&tokcstr, (char *) p1, len); p--; PEEKC(c, p); parse_ident_slow: - while (isidnum_table[c-CH_EOF]) { + while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM)) + { cstr_ccat(&tokcstr, c); PEEKC(c, p); } @@ -2084,41 +2747,57 @@ static inline void next_nomacro1(void) } } break; + case '0': case '1': case '2': case '3': case '4': case '5': case '6': case '7': case '8': case '9': - - cstr_reset(&tokcstr); + t = c; + PEEKC(c, p); /* after the first digit, accept digits, alpha, '.' or sign if prefixed by 'eEpP' */ parse_num: + cstr_reset(&tokcstr); for(;;) { + cstr_ccat(&tokcstr, t); + if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM)) + || c == '.' + || ((c == '+' || c == '-') + && (((t == 'e' || t == 'E') + && !(parse_flags & PARSE_FLAG_ASM_FILE + /* 0xe+1 is 3 tokens in asm */ + && ((char*)tokcstr.data)[0] == '0' + && toup(((char*)tokcstr.data)[1]) == 'X')) + || t == 'p' || t == 'P')))) + break; 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; + tokc.str.size = tokcstr.size; + tokc.str.data = tokcstr.data; 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, '.'); + t = '.'; goto parse_num; + } else if ((isidnum_table['.' - CH_EOF] & IS_ID) + && (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))) { + *--p = c = '.'; + goto parse_ident_fast; } else if (c == '.') { PEEKC(c, p); - if (c != '.') - expect("'.'"); - PEEKC(c, p); - tok = TOK_DOTS; + if (c == '.') { + p++; + tok = TOK_DOTS; + } else { + *--p = '.'; /* may underflow into file->unget[] */ + tok = '.'; + } } else { tok = '.'; } @@ -2127,48 +2806,16 @@ static inline void next_nomacro1(void) 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; - } - } + cstr_reset(&tokcstr); + if (is_long) + cstr_ccat(&tokcstr, 'L'); + cstr_ccat(&tokcstr, c); + p = parse_pp_string(p, c, &tokcstr); + cstr_ccat(&tokcstr, c); + cstr_ccat(&tokcstr, '\0'); + tokc.str.size = tokcstr.size; + tokc.str.data = tokcstr.data; + tok = TOK_PPSTR; break; case '<': @@ -2188,7 +2835,6 @@ static inline void next_nomacro1(void) tok = TOK_LT; } break; - case '>': PEEKC(c, p); if (c == '=') { @@ -2273,10 +2919,13 @@ static inline void next_nomacro1(void) PEEKC(c, p); if (c == '*') { p = parse_comment(p); - goto redo_no_start; + /* comments replaced by a blank */ + tok = ' '; + goto keep_tok_flags; } else if (c == '/') { p = parse_line_comment(p); - goto redo_no_start; + tok = ' '; + goto keep_tok_flags; } else if (c == '=') { p++; tok = TOK_A_DIV; @@ -2297,20 +2946,24 @@ static inline void next_nomacro1(void) case ':': case '?': case '~': - case '$': /* only used in assembler */ - case '@': /* dito */ + case '@': /* only used in assembler */ + parse_simple: tok = c; p++; break; default: - error("unrecognized character \\x%02x", c); + if (c >= 0x80 && c <= 0xFF) /* utf8 identifiers */ + goto parse_ident_fast; + if (parse_flags & PARSE_FLAG_ASM_FILE) + goto parse_simple; + tcc_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)); + printf("token = %d %s\n", tok, get_tok_str(tok, &tokc)); #endif } @@ -2322,7 +2975,7 @@ static void next_nomacro_spc(void) redo: tok = *macro_ptr; if (tok) { - TOK_GET(tok, macro_ptr, tokc); + TOK_GET(&tok, ¯o_ptr, &tokc); if (tok == TOK_LINENUM) { file->line_num = tokc.i; goto redo; @@ -2331,72 +2984,92 @@ static void next_nomacro_spc(void) } else { next_nomacro1(); } + //printf("token = %s\n", get_tok_str(tok, &tokc)); } -static void next_nomacro(void) +ST_FUNC void next_nomacro(void) { do { next_nomacro_spc(); - } while (is_space(tok)); + } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC)); } -/* substitute args in macro_str and return allocated string */ -static int *macro_arg_subst(Sym **nested_list, int *macro_str, Sym *args) + +static void macro_subst( + TokenString *tok_str, + Sym **nested_list, + const int *macro_str + ); + +/* substitute arguments in replacement lists in macro_str by the values in + args (field d) and return allocated string */ +static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args) { - int *st, last_tok, t, spc; + int t, t0, t1, spc; + const int *st; Sym *s; CValue cval; TokenString str; CString cstr; tok_str_new(&str); - last_tok = 0; + t0 = t1 = 0; while(1) { - TOK_GET(t, macro_str, cval); + TOK_GET(&t, ¯o_str, &cval); if (!t) break; if (t == '#') { /* stringize */ - TOK_GET(t, macro_str, cval); + TOK_GET(&t, ¯o_str, &cval); if (!t) - break; + goto bad_stringy; s = sym_find2(args, t); if (s) { cstr_new(&cstr); - st = (int *)s->c; + cstr_ccat(&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)); + while (*st >= 0) { + TOK_GET(&t, &st, &cval); + if (t != TOK_PLCHLDR + && t != TOK_NOSUBST + && 0 == check_space(t, &spc)) { + const char *s = get_tok_str(t, &cval); + while (*s) { + if (t == TOK_PPSTR && *s != '\'') + add_char(&cstr, *s); + else + cstr_ccat(&cstr, *s); + ++s; + } + } } cstr.size -= spc; + cstr_ccat(&cstr, '\"'); cstr_ccat(&cstr, '\0'); #ifdef PP_DEBUG - printf("stringize: %s\n", (char *)cstr.data); + printf("\nstringize: <%s>\n", (char *)cstr.data); #endif /* add string */ - cval.cstr = &cstr; - tok_str_add2(&str, TOK_STR, &cval); + cval.str.size = cstr.size; + cval.str.data = cstr.data; + tok_str_add2(&str, TOK_PPSTR, &cval); cstr_free(&cstr); } else { - tok_str_add2(&str, t, &cval); + bad_stringy: + expect("macro parameter after '#'"); } } else if (t >= TOK_IDENT) { s = sym_find2(args, t); if (s) { - st = (int *)s->c; + int l0 = str.len; + 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) { + if (*macro_str == TOK_PPJOIN || t1 == TOK_PPJOIN) { + /* special case for var arg macros : ## eats the ',' + if empty VA_ARGS variable. */ + if (t1 == TOK_PPJOIN && t0 == ',' && gnu_ext && s->type.t) { + if (*st <= 0) { /* suppress ',' '##' */ str.len -= 2; } else { @@ -2404,28 +3077,39 @@ static int *macro_arg_subst(Sym **nested_list, int *macro_str, Sym *args) 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); + add_var: + if (!s->next) { + /* Expand arguments tokens and store them. In most + cases we could also re-expand each argument if + used multiple times, but not if the argument + contains the __COUNTER__ macro. */ + TokenString str2; + sym_push2(&s->next, s->v, s->type.t, 0); + tok_str_new(&str2); + macro_subst(&str2, nested_list, st); + tok_str_add(&str2, 0); + s->next->d = str2.str; + } + st = s->next->d; } + for(;;) { + int t2; + TOK_GET(&t2, &st, &cval); + if (t2 <= 0) + break; + tok_str_add2(&str, t2, &cval); + } + if (str.len == l0) /* expanded to empty string */ + tok_str_add(&str, TOK_PLCHLDR); } else { tok_str_add(&str, t); } } else { tok_str_add2(&str, t, &cval); } - last_tok = t; + t0 = t1, t1 = t; } tok_str_add(&str, 0); return str.str; @@ -2437,25 +3121,182 @@ static char const ab_month_name[12][4] = "Jul", "Aug", "Sep", "Oct", "Nov", "Dec" }; +static int paste_tokens(int t1, CValue *v1, int t2, CValue *v2) +{ + CString cstr; + int n, ret = 1; + + cstr_new(&cstr); + if (t1 != TOK_PLCHLDR) + cstr_cat(&cstr, get_tok_str(t1, v1), -1); + n = cstr.size; + if (t2 != TOK_PLCHLDR) + cstr_cat(&cstr, get_tok_str(t2, v2), -1); + cstr_ccat(&cstr, '\0'); + + tcc_open_bf(tcc_state, ":paste:", cstr.size); + memcpy(file->buffer, cstr.data, cstr.size); + tok_flags = 0; + for (;;) { + next_nomacro1(); + if (0 == *file->buf_ptr) + break; + if (is_space(tok)) + continue; + tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid" + " preprocessing token", n, cstr.data, (char*)cstr.data + n); + ret = 0; + break; + } + tcc_close(); + //printf("paste <%s>\n", (char*)cstr.data); + cstr_free(&cstr); + return ret; +} + +/* handle the '##' operator. Return NULL if no '##' seen. Otherwise + return the resulting string (which must be freed). */ +static inline int *macro_twosharps(const int *ptr0) +{ + int t; + CValue cval; + TokenString macro_str1; + int start_of_nosubsts = -1; + const int *ptr; + + /* we search the first '##' */ + for (ptr = ptr0;;) { + TOK_GET(&t, &ptr, &cval); + if (t == TOK_PPJOIN) + break; + if (t == 0) + return NULL; + } + + tok_str_new(¯o_str1); + + //tok_print(" $$$", ptr0); + for (ptr = ptr0;;) { + TOK_GET(&t, &ptr, &cval); + if (t == 0) + break; + if (t == TOK_PPJOIN) + continue; + while (*ptr == TOK_PPJOIN) { + int t1; CValue cv1; + /* given 'a##b', remove nosubsts preceding 'a' */ + if (start_of_nosubsts >= 0) + macro_str1.len = start_of_nosubsts; + /* given 'a##b', remove nosubsts preceding 'b' */ + while ((t1 = *++ptr) == TOK_NOSUBST) + ; + if (t1 && t1 != TOK_PPJOIN) { + TOK_GET(&t1, &ptr, &cv1); + if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) { + if (paste_tokens(t, &cval, t1, &cv1)) { + t = tok, cval = tokc; + } else { + tok_str_add2(¯o_str1, t, &cval); + t = t1, cval = cv1; + } + } + } + } + if (t == TOK_NOSUBST) { + if (start_of_nosubsts < 0) + start_of_nosubsts = macro_str1.len; + } else { + start_of_nosubsts = -1; + } + tok_str_add2(¯o_str1, t, &cval); + } + tok_str_add(¯o_str1, 0); + //tok_print(" ###", macro_str1.str); + return macro_str1.str; +} + +/* peek or read [ws_str == NULL] next token from function macro call, + walking up macro levels up to the file if necessary */ +static int next_argstream(Sym **nested_list, TokenString *ws_str) +{ + int t; + const int *p; + Sym *sa; + + for (;;) { + if (macro_ptr) { + p = macro_ptr, t = *p; + if (ws_str) { + while (is_space(t) || TOK_LINEFEED == t || TOK_PLCHLDR == t) + tok_str_add(ws_str, t), t = *++p; + } + if (t == 0) { + end_macro(); + /* also, end of scope for nested defined symbol */ + sa = *nested_list; + while (sa && sa->v == 0) + sa = sa->prev; + if (sa) + sa->v = 0; + continue; + } + } else { + ch = handle_eob(); + if (ws_str) { + while (is_space(ch) || ch == '\n' || ch == '/') { + if (ch == '/') { + int c; + uint8_t *p = file->buf_ptr; + PEEKC(c, p); + if (c == '*') { + p = parse_comment(p); + file->buf_ptr = p - 1; + } else if (c == '/') { + p = parse_line_comment(p); + file->buf_ptr = p - 1; + } else + break; + ch = ' '; + } + if (ch == '\n') + file->line_num++; + if (!(ch == '\f' || ch == '\v' || ch == '\r')) + tok_str_add(ws_str, ch); + cinp(); + } + } + t = ch; + } + + if (ws_str) + return t; + next_nomacro_spc(); + return tok; + } +} + /* 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) +static int macro_subst_tok( + TokenString *tok_str, + Sym **nested_list, + Sym *s) { Sym *args, *sa, *sa1; - int mstr_allocated, parlevel, *mstr, t, t1, *p, spc; + int parlevel, t, t1, spc; 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); + if (tok == TOK___LINE__ || tok == TOK___COUNTER__) { + t = tok == TOK___LINE__ ? file->line_num : pp_counter++; + snprintf(buf, sizeof(buf), "%d", t); cstrval = buf; t1 = TOK_PPNUM; goto add_cstr1; @@ -2480,65 +3321,71 @@ static int macro_subst_tok(TokenString *tok_str, t1 = TOK_STR; add_cstr1: cstr_new(&cstr); - cstr_cat(&cstr, cstrval); - cstr_ccat(&cstr, '\0'); - cval.cstr = &cstr; + cstr_cat(&cstr, cstrval, 0); + cval.str.size = cstr.size; + cval.str.data = cstr.data; tok_str_add2(tok_str, t1, &cval); cstr_free(&cstr); - } else { - mstr = (int *)s->c; - mstr_allocated = 0; + } else if (s->d) { + int saved_parse_flags = parse_flags; + int *joined_str = NULL; + int *mstr = s->d; + 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; + /* whitespace between macro name and argument list */ + TokenString ws_str; + tok_str_new(&ws_str); + + spc = 0; + parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED + | PARSE_FLAG_ACCEPT_STRAYS; + + /* get next token from argument stream */ + t = next_argstream(nested_list, &ws_str); + if (t != '(') { + /* not a macro substitution after all, restore the + * macro token plus all whitespace we've read. + * whitespace is intentionally not merged to preserve + * newlines. */ + parse_flags = saved_parse_flags; + tok_str_add(tok_str, tok); + if (parse_flags & PARSE_FLAG_SPACES) { + int i; + for (i = 0; i < ws_str.len; i++) + tok_str_add(tok_str, ws_str.str[i]); } + tok_str_free_str(ws_str.str); + return 0; } else { - /* XXX: incorrect with comments */ - ch = file->buf_ptr[0]; - while (is_space(ch) || ch == '\n') - cinp(); - t = ch; + tok_str_free_str(ws_str.str); } - if (t != '(') /* no macro subst */ - return -1; - + do { + next_nomacro(); /* eat '(' */ + } while (tok == TOK_PLCHLDR); + /* argument macro */ - next_nomacro(); - next_nomacro(); args = NULL; sa = s->next; /* NOTE: empty args are allowed, except if no args */ for(;;) { + do { + next_argstream(nested_list, NULL); + } while (is_space(tok) || TOK_LINEFEED == tok); + empty_arg: /* handle '()' case */ if (!args && !sa && tok == ')') break; if (!sa) - error("macro '%s' used with too many args", + tcc_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) { + (tok != ',' || sa->type.t)))) { + if (tok == TOK_EOF || tok == 0) + break; if (tok == '(') parlevel++; else if (tok == ')') @@ -2547,335 +3394,273 @@ static int macro_subst_tok(TokenString *tok_str, tok = ' '; if (!check_space(tok, &spc)) tok_str_add2(&str, tok, &tokc); - next_nomacro_spc(); + next_argstream(nested_list, NULL); } + if (parlevel) + expect(")"); str.len -= spc; + tok_str_add(&str, -1); tok_str_add(&str, 0); - sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, (long)str.str); + 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; + goto empty_arg; + break; } if (tok != ',') expect(","); - next_nomacro(); } if (sa) { - error("macro '%s' used with too few args", + tcc_error("macro '%s' used with too few args", get_tok_str(s->v, 0)); } + parse_flags = saved_parse_flags; + /* now subst each arg */ mstr = macro_arg_subst(nested_list, mstr, args); /* free memory */ sa = args; while (sa) { sa1 = sa->prev; - tok_str_free((int *)sa->c); + tok_str_free_str(sa->d); + if (sa->next) { + tok_str_free_str(sa->next->d); + sym_free(sa->next); + } 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); + parse_flags = saved_parse_flags; + joined_str = macro_twosharps(mstr); + macro_subst(tok_str, nested_list, joined_str ? joined_str : mstr); + /* pop nested defined symbol */ sa1 = *nested_list; *nested_list = sa1->prev; sym_free(sa1); - if (mstr_allocated) - tok_str_free(mstr); + if (joined_str) + tok_str_free_str(joined_str); + if (mstr != s->d) + tok_str_free_str(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) -{ - TokenSym *ts; - const int *ptr, *saved_macro_ptr; - int t; - const char *p1, *p2; - CValue cval; - TokenString macro_str1; - CString cstr; - - /* 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 */ - cstr_new(&cstr); - tok_str_new(¯o_str1); - saved_macro_ptr = macro_ptr; - /* XXX: get rid of the use of macro_ptr here */ - macro_ptr = (int *)macro_str; - for(;;) { - next_nomacro_spc(); - if (tok == 0) - break; - if (tok == TOK_TWOSHARPS) - continue; - while (*macro_ptr == TOK_TWOSHARPS) { - t = *++macro_ptr; - if (t && t != TOK_TWOSHARPS) { - TOK_GET(t, macro_ptr, cval); - /* We concatenate the two tokens if we have an - identifier or a preprocessing number */ - cstr_reset(&cstr); - p1 = get_tok_str(tok, &tokc); - cstr_cat(&cstr, p1); - p2 = get_tok_str(t, &cval); - cstr_cat(&cstr, p2); - cstr_ccat(&cstr, '\0'); - - if ((tok >= TOK_IDENT || tok == TOK_PPNUM) && - (t >= TOK_IDENT || t == TOK_PPNUM)) { - if (tok == TOK_PPNUM) { - /* if number, then create a number token */ - /* NOTE: no need to allocate because - tok_str_add2() does it */ - cstr_reset(&tokcstr); - tokcstr = cstr; - cstr_new(&cstr); - tokc.cstr = &tokcstr; - } else { - /* if identifier, we must do a test to - validate we have a correct identifier */ - if (t == TOK_PPNUM) { - const char *p; - int c; - - p = p2; - for(;;) { - c = *p; - if (c == '\0') - break; - p++; - if (!isnum(c) && !isid(c)) - goto error_pasting; - } - } - ts = tok_alloc(cstr.data, strlen(cstr.data)); - tok = ts->tok; /* modify current token */ - } - } else { - const char *str = cstr.data; - const unsigned char *q; - - /* we look for a valid token */ - /* XXX: do more extensive checks */ - if (!strcmp(str, ">>=")) { - tok = TOK_A_SAR; - } else if (!strcmp(str, "<<=")) { - tok = TOK_A_SHL; - } else if (strlen(str) == 2) { - /* search in two bytes table */ - q = tok_two_chars; - for(;;) { - if (!*q) - goto error_pasting; - if (q[0] == str[0] && q[1] == str[1]) - break; - q += 3; - } - tok = q[2]; - } else { - error_pasting: - /* NOTE: because get_tok_str use a static buffer, - we must save it */ - cstr_reset(&cstr); - p1 = get_tok_str(tok, &tokc); - cstr_cat(&cstr, p1); - cstr_ccat(&cstr, '\0'); - p2 = get_tok_str(t, &cval); - warning("pasting \"%s\" and \"%s\" does not give a valid preprocessing token", cstr.data, p2); - /* cannot merge tokens: just add them separately */ - tok_str_add2(¯o_str1, tok, &tokc); - /* XXX: free associated memory ? */ - tok = t; - tokc = cval; - } - } - } - } - tok_str_add2(¯o_str1, tok, &tokc); - } - macro_ptr = (int *)saved_macro_ptr; - cstr_free(&cstr); - 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) +static void macro_subst( + TokenString *tok_str, + Sym **nested_list, + const int *macro_str + ) { Sym *s; - int *macro_str1; - const int *ptr; - int t, ret, spc; + int t, spc, nosubst; 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; + spc = nosubst = 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) + TOK_GET(&t, ¯o_str, &cval); + if (t <= 0) break; - s = define_find(t); - if (s != NULL) { - /* if nested substitution, do nothing */ - if (sym_find2(*nested_list, t)) + + if (t >= TOK_IDENT && 0 == nosubst) { + s = define_find(t); + if (s == NULL) 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) + + /* if nested substitution, do nothing */ + if (sym_find2(*nested_list, t)) { + /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */ + tok_str_add2(tok_str, TOK_NOSUBST, NULL); goto no_subst; + } + + { + TokenString str; + str.str = (int*)macro_str; + begin_macro(&str, 2); + + tok = t; + macro_subst_tok(tok_str, nested_list, s); + + if (str.alloc == 3) { + /* already finished by reading function macro arguments */ + break; + } + + macro_str = macro_ptr; + end_macro (); + } + if (tok_str->len) + spc = is_space(t = tok_str->str[tok_str->lastlen]); } else { - no_subst: - if (!check_space(t, &spc)) + if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS)) + tcc_error("stray '\\' in program"); +no_subst: + if (!check_space(t, &spc)) tok_str_add2(tok_str, t, &cval); + + if (nosubst) { + if (nosubst > 1 && (spc || (++nosubst == 3 && t == '('))) + continue; + nosubst = 0; + } + if (t == TOK_NOSUBST) + nosubst = 1; } + /* GCC supports 'defined' as result of a macro substitution */ + if (t == TOK_DEFINED && pp_expr) + nosubst = 2; } - if (macro_str1) - tok_str_free(macro_str1); } /* return next token with macro substitution */ -static void next(void) +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; - } - } + + if (macro_ptr) { + if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) { + /* discard preprocessor markers */ + goto redo; + } else if (tok == 0) { + /* end of macro or unget token string */ + end_macro(); + 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 = NULL; - } + } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) { + Sym *s; + /* if reading from file, try to substitute macros */ + s = define_find(tok); + if (s) { + Sym *nested_list = NULL; + tokstr_buf.len = 0; + macro_subst_tok(&tokstr_buf, &nested_list, s); + tok_str_add(&tokstr_buf, 0); + begin_macro(&tokstr_buf, 2); goto redo; } } - /* convert preprocessor tokens into C tokens */ - if (tok == TOK_PPNUM && - (parse_flags & PARSE_FLAG_TOK_NUM)) { - parse_number((char *)tokc.cstr->data); + if (tok == TOK_PPNUM) { + if (parse_flags & PARSE_FLAG_TOK_NUM) + parse_number((char *)tokc.str.data); + } else if (tok == TOK_PPSTR) { + if (parse_flags & PARSE_FLAG_TOK_STR) + parse_string((char *)tokc.str.data, tokc.str.size - 1); } } /* push back current token and set current token to 'last_tok'. Only identifier case handled for labels. */ -static inline void unget_tok(int last_tok) +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;i<n;i++) - *q++ = tokc.tab[i]; - *q = 0; /* end of token string */ + + TokenString *str = tok_str_alloc(); + tok_str_add2(str, tok, &tokc); + tok_str_add(str, 0); + begin_macro(str, 1); tok = last_tok; } - -/* better than nothing, but needs extension to handle '-E' option - correctly too */ -static void preprocess_init(TCCState *s1) +ST_FUNC void preprocess_start(TCCState *s1, int is_asm) { + CString cstr; + int i; + s1->include_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; + pp_expr = 0; + pp_counter = 0; + pp_debug_tok = pp_debug_symv = 0; + pp_once++; + pvtop = vtop = vstack - 1; s1->pack_stack[0] = 0; s1->pack_stack_ptr = s1->pack_stack; + + set_idnum('$', s1->dollars_in_identifiers ? IS_ID : 0); + set_idnum('.', is_asm ? IS_ID : 0); + + cstr_new(&cstr); + cstr_cat(&cstr, "\"", -1); + cstr_cat(&cstr, file->filename, -1); + cstr_cat(&cstr, "\"", 0); + tcc_define_symbol(s1, "__BASE_FILE__", cstr.data); + + cstr_reset(&cstr); + for (i = 0; i < s1->nb_cmd_include_files; i++) { + cstr_cat(&cstr, "#include \"", -1); + cstr_cat(&cstr, s1->cmd_include_files[i], -1); + cstr_cat(&cstr, "\"\n", -1); + } + if (cstr.size) { + *s1->include_stack_ptr++ = file; + tcc_open_bf(s1, "<command line>", cstr.size); + memcpy(file->buffer, cstr.data, cstr.size); + } + cstr_free(&cstr); + + if (is_asm) + tcc_define_symbol(s1, "__ASSEMBLER__", NULL); + + parse_flags = is_asm ? PARSE_FLAG_ASM_FILE : 0; + tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF; } -void preprocess_new() +/* cleanup from error/setjmp */ +ST_FUNC void preprocess_end(TCCState *s1) +{ + while (macro_stack) + end_macro(); + macro_ptr = NULL; +} + +ST_FUNC void tccpp_new(TCCState *s) { int i, c; const char *p, *r; - TokenSym *ts; + + /* might be used in error() before preprocess_start() */ + s->include_stack_ptr = s->include_stack; + s->ppfp = stdout; /* init isid table */ - for(i=CH_EOF;i<256;i++) - isidnum_table[i-CH_EOF] = isid(i) || isnum(i); + for(i = CH_EOF; i<128; i++) + set_idnum(i, + is_space(i) ? IS_SPC + : isid(i) ? IS_ID + : isnum(i) ? IS_NUM + : 0); + + for(i = 128; i<256; i++) + set_idnum(i, IS_ID); + + /* init allocators */ + tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE); + tal_new(&tokstr_alloc, TOKSTR_TAL_LIMIT, TOKSTR_TAL_SIZE); + tal_new(&cstr_alloc, CSTR_TAL_LIMIT, CSTR_TAL_SIZE); - /* add all tokens */ - table_ident = NULL; memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *)); + cstr_new(&cstr_buf); + cstr_realloc(&cstr_buf, STRING_MAX_SIZE); + tok_str_new(&tokstr_buf); + tok_str_realloc(&tokstr_buf, TOKSTR_MAX_SIZE); tok_ident = TOK_IDENT; p = tcc_keywords; @@ -2886,50 +3671,233 @@ void preprocess_new() if (c == '\0') break; } - ts = tok_alloc(p, r - p - 1); + tok_alloc(p, r - p - 1); p = r; } } +ST_FUNC void tccpp_delete(TCCState *s) +{ + int i, n; + + /* free -D and compiler defines */ + free_defines(NULL); + + /* free tokens */ + n = tok_ident - TOK_IDENT; + for(i = 0; i < n; i++) + tal_free(toksym_alloc, table_ident[i]); + tcc_free(table_ident); + table_ident = NULL; + + /* free static buffers */ + cstr_free(&tokcstr); + cstr_free(&cstr_buf); + cstr_free(¯o_equal_buf); + tok_str_free_str(tokstr_buf.str); + + /* free allocators */ + tal_delete(toksym_alloc); + toksym_alloc = NULL; + tal_delete(tokstr_alloc); + tokstr_alloc = NULL; + tal_delete(cstr_alloc); + cstr_alloc = NULL; +} + +/* ------------------------------------------------------------------------- */ +/* tcc -E [-P[1]] [-dD} support */ + +static void tok_print(const char *msg, const int *str) +{ + FILE *fp; + int t, s = 0; + CValue cval; + + fp = tcc_state->ppfp; + fprintf(fp, "%s", msg); + while (str) { + TOK_GET(&t, &str, &cval); + if (!t) + break; + fprintf(fp, " %s" + s, get_tok_str(t, &cval)), s = 1; + } + fprintf(fp, "\n"); +} + +static void pp_line(TCCState *s1, BufferedFile *f, int level) +{ + int d = f->line_num - f->line_ref; + + if (s1->dflag & 4) + return; + + if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE) { + ; + } else if (level == 0 && f->line_ref && d < 8) { + while (d > 0) + fputs("\n", s1->ppfp), --d; + } else if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) { + fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename); + } else { + fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename, + level > 0 ? " 1" : level < 0 ? " 2" : ""); + } + f->line_ref = f->line_num; +} + +static void define_print(TCCState *s1, int v) +{ + FILE *fp; + Sym *s; + + s = define_find(v); + if (NULL == s || NULL == s->d) + return; + + fp = s1->ppfp; + fprintf(fp, "#define %s", get_tok_str(v, NULL)); + if (s->type.t == MACRO_FUNC) { + Sym *a = s->next; + fprintf(fp,"("); + if (a) + for (;;) { + fprintf(fp,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL)); + if (!(a = a->next)) + break; + fprintf(fp,","); + } + fprintf(fp,")"); + } + tok_print("", s->d); +} + +static void pp_debug_defines(TCCState *s1) +{ + int v, t; + const char *vs; + FILE *fp; + + t = pp_debug_tok; + if (t == 0) + return; + + file->line_num--; + pp_line(s1, file, 0); + file->line_ref = ++file->line_num; + + fp = s1->ppfp; + v = pp_debug_symv; + vs = get_tok_str(v, NULL); + if (t == TOK_DEFINE) { + define_print(s1, v); + } else if (t == TOK_UNDEF) { + fprintf(fp, "#undef %s\n", vs); + } else if (t == TOK_push_macro) { + fprintf(fp, "#pragma push_macro(\"%s\")\n", vs); + } else if (t == TOK_pop_macro) { + fprintf(fp, "#pragma pop_macro(\"%s\")\n", vs); + } + pp_debug_tok = 0; +} + +static void pp_debug_builtins(TCCState *s1) +{ + int v; + for (v = TOK_IDENT; v < tok_ident; ++v) + define_print(s1, v); +} + +/* Add a space between tokens a and b to avoid unwanted textual pasting */ +static int pp_need_space(int a, int b) +{ + return 'E' == a ? '+' == b || '-' == b + : '+' == a ? TOK_INC == b || '+' == b + : '-' == a ? TOK_DEC == b || '-' == b + : a >= TOK_IDENT ? b >= TOK_IDENT + : a == TOK_PPNUM ? b >= TOK_IDENT + : 0; +} + +/* maybe hex like 0x1e */ +static int pp_check_he0xE(int t, const char *p) +{ + if (t == TOK_PPNUM && toup(strchr(p, 0)[-1]) == 'E') + return 'E'; + return t; +} + /* Preprocess the current file */ -static int tcc_preprocess(TCCState *s1) +ST_FUNC int tcc_preprocess(TCCState *s1) { - Sym *define_start; - BufferedFile *file_ref; - int token_seen, line_ref; + BufferedFile **iptr; + int token_seen, spcs, level; + const char *p; + char white[400]; + + parse_flags = PARSE_FLAG_PREPROCESS + | (parse_flags & PARSE_FLAG_ASM_FILE) + | PARSE_FLAG_LINEFEED + | PARSE_FLAG_SPACES + | PARSE_FLAG_ACCEPT_STRAYS + ; + /* Credits to Fabrice Bellard's initial revision to demonstrate its + capability to compile and run itself, provided all numbers are + given as decimals. tcc -E -P10 will do. */ + if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_P10) + parse_flags |= PARSE_FLAG_TOK_NUM, s1->Pflag = 1; + +#ifdef PP_BENCH + /* for PP benchmarks */ + do next(); while (tok != TOK_EOF); + return 0; +#endif - 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; + if (s1->dflag & 1) { + pp_debug_builtins(s1); + s1->dflag &= ~1; + } + token_seen = TOK_LINEFEED, spcs = 0; + pp_line(s1, file, 0); for (;;) { + iptr = s1->include_stack_ptr; next(); - if (tok == TOK_EOF) { + if (tok == TOK_EOF) break; + + level = s1->include_stack_ptr - iptr; + if (level) { + if (level > 0) + pp_line(s1, *iptr, 0); + pp_line(s1, file, level); + } + if (s1->dflag & 7) { + pp_debug_defines(s1); + if (s1->dflag & 4) + continue; + } + + if (is_space(tok)) { + if (spcs < sizeof white - 1) + white[spcs++] = tok; + continue; } else if (tok == TOK_LINEFEED) { - if (!token_seen) + spcs = 0; + if (token_seen == TOK_LINEFEED) continue; - ++line_ref; - token_seen = 0; - } else if (!token_seen) { - int d = file->line_num - line_ref; - if (file != file_ref || d < 0 || d >= 8) - fprintf(s1->outfile, "# %d \"%s\"\n", file->line_num, file->filename); - else - while (d) - fputs("\n", s1->outfile), --d; - line_ref = (file_ref = file)->line_num; - token_seen = 1; + ++file->line_ref; + } else if (token_seen == TOK_LINEFEED) { + pp_line(s1, file, 0); + } else if (spcs == 0 && pp_need_space(token_seen, tok)) { + white[spcs++] = ' '; } - fputs(get_tok_str(tok, &tokc), s1->outfile); + + white[spcs] = 0, fputs(white, s1->ppfp), spcs = 0; + fputs(p = get_tok_str(tok, &tokc), s1->ppfp); + token_seen = pp_check_he0xE(tok, p); } - free_defines(define_start); return 0; } +/* ------------------------------------------------------------------------- */ |