macoy
/
tinycc
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
My fork of the Tiny C Compiler, designed to support self-modifying and hot-reloading applications
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
3240 Commits
4 Branches
10 Tags
5.6 MiB
 
 
 
 
 
 
Branch: mob
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from 'mob'
${ noResults }
tinycc/tccpp.c

4022 lines
111 KiB
Raw Permalink Blame History

/*
* TCC - Tiny C Compiler
*
* Copyright (c) 2001-2004 Fabrice Bellard
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#define USING_GLOBALS
#include "tcc.h"
/* #define to 1 to enable (see parse_pp_string()) */
#define ACCEPT_LF_IN_STRINGS 0
/********************************************************/
/* global variables */
ST_DATA int tok_flags;
ST_DATA int parse_flags;
ST_DATA struct BufferedFile *file;
ST_DATA int 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 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 TokenString *macro_stack;
static const char tcc_keywords[] =
#define DEF(id, str) str "\0"
#include "tcctok.h"
#undef DEF
;
/* WARNING: the content of this string encodes token numbers */
static const unsigned char tok_two_chars[] =
/* 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,
'#','#', TOK_PPJOIN,
0
};
static void next_nomacro(void);
ST_FUNC void skip(int c)
{
if (tok != c)
tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
next();
}
ST_FUNC void expect(const char *msg)
{
tcc_error("%s expected", msg);
}
/* ------------------------------------------------------------------------- */
/* 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 - al->buffer + adj_size + sizeof(tal_header_t) < 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);
if (p) 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);
if (p) 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);
if (p) 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 = tcc_realloc(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_INLN char *unicode_to_utf8 (char *b, uint32_t Uc)
{
if (Uc<0x80) *b++=Uc;
else if (Uc<0x800) *b++=192+Uc/64, *b++=128+Uc%64;
else if (Uc-0xd800u<0x800) goto error;
else if (Uc<0x10000) *b++=224+Uc/4096, *b++=128+Uc/64%64, *b++=128+Uc%64;
else if (Uc<0x110000) *b++=240+Uc/262144, *b++=128+Uc/4096%64, *b++=128+Uc/64%64, *b++=128+Uc%64;
else error: tcc_error("0x%x is not a valid universal character", Uc);
return b;
}
/* add a unicode character expanded into utf8 */
ST_INLN void cstr_u8cat(CString *cstr, int ch)
{
char buf[4], *e;
e = unicode_to_utf8(buf, (uint32_t)ch);
cstr_cat(cstr, buf, e - buf);
}
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)
{
tcc_free(cstr->data);
cstr_new(cstr);
}
/* reset string to empty */
ST_FUNC void cstr_reset(CString *cstr)
{
cstr->size = 0;
}
ST_FUNC int cstr_vprintf(CString *cstr, const char *fmt, va_list ap)
{
va_list v;
int len, size = 80;
for (;;) {
size += cstr->size;
if (size > cstr->size_allocated)
cstr_realloc(cstr, size);
size = cstr->size_allocated - cstr->size;
va_copy(v, ap);
len = vsnprintf((char*)cstr->data + cstr->size, size, fmt, v);
va_end(v);
if (len > 0 && len < size)
break;
size *= 2;
}
cstr->size += len;
return len;
}
ST_FUNC int cstr_printf(CString *cstr, const char *fmt, ...)
{
va_list ap; int len;
va_start(ap, fmt);
len = cstr_vprintf(cstr, fmt, ap);
va_end(ap);
return len;
}
/* 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)
{
TokenSym *ts, **ptable;
int i;
if (tok_ident >= SYM_FIRST_ANOM)
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 *));
table_ident = ptable;
}
ts = tal_realloc(toksym_alloc, 0, sizeof(TokenSym) + len);
table_ident[i] = ts;
ts->tok = tok_ident++;
ts->sym_define = NULL;
ts->sym_label = NULL;
ts->sym_struct = NULL;
ts->sym_identifier = NULL;
ts->len = len;
ts->hash_next = NULL;
memcpy(ts->str, str, len);
ts->str[len] = '\0';
*pts = ts;
return ts;
}
#define TOK_HASH_INIT 1
#define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c))
/* find a token and add it if not found */
ST_FUNC TokenSym *tok_alloc(const char *str, int len)
{
TokenSym *ts, **pts;
int i;
unsigned int h;
h = TOK_HASH_INIT;
for(i=0;i<len;i++)
h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
h &= (TOK_HASH_SIZE - 1);
pts = &hash_ident[h];
for(;;) {
ts = *pts;
if (!ts)
break;
if (ts->len == len && !memcmp(ts->str, str, len))
return ts;
pts = &(ts->hash_next);
}
return tok_alloc_new(pts, str, len);
}
ST_FUNC int tok_alloc_const(const char *str)
{
return tok_alloc(str, strlen(str))->tok;
}
/* XXX: buffer overflow */
/* XXX: float tokens */
ST_FUNC const char *get_tok_str(int v, CValue *cv)
{
char *p;
int i, len;
cstr_reset(&cstr_buf);
p = cstr_buf.data;
switch(v) {
case TOK_CINT:
case TOK_CUINT:
case TOK_CLONG:
case TOK_CULONG:
case TOK_CLLONG:
case TOK_CULLONG:
/* XXX: not quite exact, but only useful for testing */
#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');
case TOK_CCHAR:
cstr_ccat(&cstr_buf, '\'');
add_char(&cstr_buf, cv->i);
cstr_ccat(&cstr_buf, '\'');
cstr_ccat(&cstr_buf, '\0');
break;
case TOK_PPNUM:
case TOK_PPSTR:
return (char*)cv->str.data;
case TOK_LSTR:
cstr_ccat(&cstr_buf, 'L');
case TOK_STR:
cstr_ccat(&cstr_buf, '\"');
if (v == TOK_STR) {
len = cv->str.size - 1;
for(i=0;i<len;i++)
add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]);
} else {
len = (cv->str.size / sizeof(nwchar_t)) - 1;
for(i=0;i<len;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;
case TOK_GT:
v = '>';
goto addv;
case TOK_DOTS:
return strcpy(p, "...");
case TOK_A_SHL:
return strcpy(p, "<<=");
case TOK_A_SAR:
return strcpy(p, ">>=");
case TOK_EOF:
return strcpy(p, "<eof>");
default:
if (v < TOK_IDENT) {
/* search in two bytes table */
const unsigned char *q = tok_two_chars;
while (*q) {
if (q[2] == v) {
*p++ = q[0];
*p++ = q[1];
*p = '\0';
return cstr_buf.data;
}
q += 3;
}
if (v >= 127) {
sprintf(cstr_buf.data, "<%02x>", v);
return cstr_buf.data;
}
addv:
*p++ = v;
case 0: /* nameless anonymous symbol */
*p = '\0';
} else if (v < tok_ident) {
return table_ident[v - TOK_IDENT]->str;
} else if (v >= SYM_FIRST_ANOM) {
/* special name for anonymous symbol */
sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
} else {
/* should never happen */
return NULL;
}
break;
}
return cstr_buf.data;
}
static inline int check_space(int t, int *spc)
{
if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
if (*spc)
return 1;
*spc = 1;
} else
*spc = 0;
return 0;
}
/* return the current character, handling end of block if necessary
(but not stray) */
static 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 >= 0) {
#if defined(PARSE_DEBUG)
len = 1;
#else
len = IO_BUF_SIZE;
#endif
len = read(bf->fd, bf->buffer, len);
if (len < 0)
len = 0;
} else {
len = 0;
}
total_bytes += len;
bf->buf_ptr = bf->buffer;
bf->buf_end = bf->buffer + len;
*bf->buf_end = CH_EOB;
}
if (bf->buf_ptr < bf->buf_end) {
return bf->buf_ptr[0];
} else {
bf->buf_ptr = bf->buf_end;
return CH_EOF;
}
}
/* read next char from current input file and handle end of input buffer */
static int next_c(void)
{
int ch = *++file->buf_ptr;
/* end of buffer/file handling */
if (ch == CH_EOB && file->buf_ptr >= file->buf_end)
ch = handle_eob();
return ch;
}
/* input with '\[\r]\n' handling. */
static int handle_stray_noerror(int err)
{
int ch;
while ((ch = next_c()) == '\\') {
ch = next_c();
if (ch == '\n') {
newl:
file->line_num++;
} else {
if (ch == '\r') {
ch = next_c();
if (ch == '\n')
goto newl;
*--file->buf_ptr = '\r';
}
if (err)
tcc_error("stray '\\' in program");
/* may take advantage of 'BufferedFile.unget[4}' */
return *--file->buf_ptr = '\\';
}
}
return ch;
}
#define ninp() handle_stray_noerror(0)
/* handle '\\' in strings, comments and skipped regions */
static int handle_bs(uint8_t **p)
{
int c;
file->buf_ptr = *p - 1;
c = ninp();
*p = file->buf_ptr;
return c;
}
/* skip the stray and handle the \\n case. Output an error if
incorrect char after the stray */
static int handle_stray(uint8_t **p)
{
int c;
file->buf_ptr = *p - 1;
c = handle_stray_noerror(!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS));
*p = file->buf_ptr;
return c;
}
/* handle the complicated stray case */
#define PEEKC(c, p)\
{\
c = *++p;\
if (c == '\\')\
c = handle_stray(&p); \
}
static int skip_spaces(void)
{
int ch;
--file->buf_ptr;
do {
ch = ninp();
} while (isidnum_table[ch - CH_EOF] & IS_SPC);
return ch;
}
/* single line C++ comments */
static uint8_t *parse_line_comment(uint8_t *p)
{
int c;
for(;;) {
for (;;) {
c = *++p;
redo:
if (c == '\n' || c == '\\')
break;
c = *++p;
if (c == '\n' || c == '\\')
break;
}
if (c == '\n')
break;
c = handle_bs(&p);
if (c == CH_EOF)
break;
if (c != '\\')
goto redo;
}
return p;
}
/* C comments */
static uint8_t *parse_comment(uint8_t *p)
{
int c;
for(;;) {
/* fast skip loop */
for(;;) {
c = *++p;
redo:
if (c == '\n' || c == '*' || c == '\\')
break;
c = *++p;
if (c == '\n' || c == '*' || c == '\\')
break;
}
/* now we can handle all the cases */
if (c == '\n') {
file->line_num++;
} else if (c == '*') {
do {
c = *++p;
} while (c == '*');
if (c == '\\')
c = handle_bs(&p);
if (c == '/')
break;
goto check_eof;
} else {
c = handle_bs(&p);
check_eof:
if (c == CH_EOF)
tcc_error("unexpected end of file in comment");
if (c != '\\')
goto redo;
}
}
return p + 1;
}
/* parse a string without interpreting escapes */
static uint8_t *parse_pp_string(uint8_t *p, int sep, CString *str)
{
int c;
for(;;) {
c = *++p;
redo:
if (c == sep) {
break;
} else if (c == '\\') {
c = handle_bs(&p);
if (c == CH_EOF) {
unterminated_string:
/* XXX: indicate line number of start of string */
tok_flags &= ~TOK_FLAG_BOL;
tcc_error("missing terminating %c character", sep);
} else if (c == '\\') {
if (str)
cstr_ccat(str, c);
c = *++p;
/* add char after '\\' unconditionally */
if (c == '\\') {
c = handle_bs(&p);
if (c == CH_EOF)
goto unterminated_string;
}
goto add_char;
} else {
goto redo;
}
} else if (c == '\n') {
add_lf:
if (ACCEPT_LF_IN_STRINGS) {
file->line_num++;
goto add_char;
} else if (str) { /* not skipping */
goto unterminated_string;
} else {
//tcc_warning("missing terminating %c character", sep);
return p;
}
} else if (c == '\r') {
c = *++p;
if (c == '\\')
c = handle_bs(&p);
if (c == '\n')
goto add_lf;
if (c == CH_EOF)
goto unterminated_string;
if (str)
cstr_ccat(str, '\r');
goto redo;
} else {
add_char:
if (str)
cstr_ccat(str, c);
}
}
p++;
return p;
}
/* skip block of text until #else, #elif or #endif. skip also pairs of
#if/#endif */
static void preprocess_skip(void)
{
int a, start_of_line, c, in_warn_or_error;
uint8_t *p;
p = file->buf_ptr;
a = 0;
redo_start:
start_of_line = 1;
in_warn_or_error = 0;
for(;;) {
redo_no_start:
c = *p;
switch(c) {
case ' ':
case '\t':
case '\f':
case '\v':
case '\r':
p++;
goto redo_no_start;
case '\n':
file->line_num++;
p++;
goto redo_start;
case '\\':
c = handle_bs(&p);
if (c == CH_EOF)
expect("#endif");
if (c == '\\')
++p;
goto redo_no_start;
/* skip strings */
case '\"':
case '\'':
if (in_warn_or_error)
goto _default;
tok_flags &= ~TOK_FLAG_BOL;
p = parse_pp_string(p, c, NULL);
break;
/* skip comments */
case '/':
if (in_warn_or_error)
goto _default;
++p;
c = handle_bs(&p);
if (c == '*') {
p = parse_comment(p);
} else if (c == '/') {
p = parse_line_comment(p);
}
break;
case '#':
p++;
if (start_of_line) {
file->buf_ptr = p;
next_nomacro();
p = file->buf_ptr;
if (a == 0 &&
(tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
goto the_end;
if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
a++;
else if (tok == TOK_ENDIF)
a--;
else if( tok == TOK_ERROR || tok == TOK_WARNING)
in_warn_or_error = 1;
else if (tok == TOK_LINEFEED)
goto redo_start;
else if (parse_flags & PARSE_FLAG_ASM_FILE)
p = parse_line_comment(p - 1);
}
#if !defined(TCC_TARGET_ARM)
else if (parse_flags & PARSE_FLAG_ASM_FILE)
p = parse_line_comment(p - 1);
#else
/* ARM assembly uses '#' for constants */
#endif
break;
_default:
default:
p++;
break;
}
start_of_line = 0;
}
the_end: ;
file->buf_ptr = p;
}
#if 0
/* return the number of additional 'ints' necessary to store the
token */
static inline int tok_size(const int *p)
{
switch(*p) {
/* 4 bytes */
case TOK_CINT:
case TOK_CUINT:
case TOK_CCHAR:
case TOK_LCHAR:
case TOK_CFLOAT:
case TOK_LINENUM:
return 1 + 1;
case TOK_STR:
case TOK_LSTR:
case TOK_PPNUM:
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 1 + 2;
case TOK_CLDOUBLE:
return 1 + LDOUBLE_SIZE / 4;
default:
return 1 + 0;
}
}
#endif
/* token string handling */
ST_INLN void tok_str_new(TokenString *s)
{
s->str = NULL;
s->len = s->lastlen = 0;
s->allocated_len = 0;
s->last_line_num = -1;
}
ST_FUNC TokenString *tok_str_alloc(void)
{
TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str);
tok_str_new(str);
return str;
}
ST_FUNC int *tok_str_dup(TokenString *s)
{
int *str;
str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int));
memcpy(str, s->str, s->len * sizeof(int));
return str;
}
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, 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;
str->len = 0; /* matters if str not alloced, may be tokstr_buf */
if (str->alloc != 0) {
if (str->alloc == 2)
str->str = NULL; /* don't free */
tok_str_free(str);
}
}
static void tok_str_add2(TokenString *s, int t, CValue *cv)
{
int len, *str;
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, len + TOK_MAX_SIZE + 1);
str[len++] = t;
switch(t) {
case TOK_CINT:
case TOK_CUINT:
case TOK_CCHAR:
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:
{
/* 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
str[len++] = cv->tab[0];
str[len++] = cv->tab[1];
break;
#if LDOUBLE_SIZE == 12
case TOK_CLDOUBLE:
str[len++] = cv->tab[0];
str[len++] = cv->tab[1];
str[len++] = cv->tab[2];
#elif LDOUBLE_SIZE == 16
case TOK_CLDOUBLE:
str[len++] = cv->tab[0];
str[len++] = cv->tab[1];
str[len++] = cv->tab[2];
str[len++] = cv->tab[3];
#elif LDOUBLE_SIZE != 8
#error add long double size support
#endif
break;
default:
break;
}
s->len = len;
}
/* add the current parse token in token string 's' */
ST_FUNC void tok_str_add_tok(TokenString *s)
{
CValue cval;
/* save line number info */
if (file->line_num != s->last_line_num) {
s->last_line_num = file->line_num;
cval.i = s->last_line_num;
tok_str_add2(s, TOK_LINENUM, &cval);
}
tok_str_add2(s, tok, &tokc);
}
/* get a token from an integer array and increment pointer. */
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
n = 4;
#elif LDOUBLE_SIZE == 12
n = 3;
#elif LDOUBLE_SIZE == 8
n = 2;
#else
# error add long double size support
#endif
copy:
do
*tab++ = *p++;
while (--n);
break;
default:
break;
}
*pp = p;
}
#if 0
# define TOK_GET(t,p,c) tok_get(t,p,c)
#else
# define TOK_GET(t,p,c) do { \
int _t = **(p); \
if (TOK_HAS_VALUE(_t)) \
tok_get(t, p, c); \
else \
*(t) = _t, ++*(p); \
} while (0)
#endif
static int macro_is_equal(const int *a, const int *b)
{
CValue cv;
int t;
if (!a || !b)
return 1;
while (*a && *b) {
/* first time preallocate macro_equal_buf, next time only reset position to start */
cstr_reset(&macro_equal_buf);
TOK_GET(&t, &a, &cv);
cstr_cat(&macro_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 */
ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
{
Sym *s, *o;
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 */
ST_FUNC void define_undef(Sym *s)
{
int v = s->v;
if (v >= TOK_IDENT && v < tok_ident)
table_ident[v - TOK_IDENT]->sym_define = NULL;
}
ST_INLN Sym *define_find(int v)
{
v -= TOK_IDENT;
if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
return NULL;
return table_ident[v]->sym_define;
}
/* free define stack until top reaches 'b' */
ST_FUNC void free_defines(Sym *b)
{
while (define_stack != b) {
Sym *top = define_stack;
define_stack = top->prev;
tok_str_free_str(top->d);
define_undef(top);
sym_free(top);
}
}
/* label lookup */
ST_FUNC Sym *label_find(int v)
{
v -= TOK_IDENT;
if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
return NULL;
return table_ident[v]->sym_label;
}
ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
{
Sym *s, **ps;
s = sym_push2(ptop, v, VT_STATIC, 0);
s->r = flags;
ps = &table_ident[v - TOK_IDENT]->sym_label;
if (ptop == &global_label_stack) {
/* modify the top most local identifier, so that
sym_identifier will point to 's' when popped */
while (*ps != NULL)
ps = &(*ps)->prev_tok;
}
s->prev_tok = *ps;
*ps = s;
return s;
}
/* pop labels until element last is reached. Look if any labels are
undefined. Define symbols if '&&label' was used. */
ST_FUNC void label_pop(Sym **ptop, Sym *slast, int keep)
{
Sym *s, *s1;
for(s = *ptop; s != slast; s = s1) {
s1 = s->prev;
if (s->r == LABEL_DECLARED) {
tcc_warning_c(warn_all)("label '%s' declared but not used", get_tok_str(s->v, NULL));
} else if (s->r == LABEL_FORWARD) {
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, s->jnext, 1);
}
}
/* remove label */
if (s->r != LABEL_GONE)
table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
if (!keep)
sym_free(s);
else
s->r = LABEL_GONE;
}
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);
}
static CachedInclude *
search_cached_include(TCCState *s1, const char *filename, int add);
static int parse_include(TCCState *s1, int do_next, int test)
{
int c, i;
CString cs;
char name[1024], buf[1024], *p;
CachedInclude *e;
cstr_new(&cs);
c = skip_spaces();
if (c == '<' || c == '\"') {
file->buf_ptr = parse_pp_string(file->buf_ptr, c == '<' ? '>' : c, &cs);
next_nomacro();
} else {
/* computed #include : concatenate tokens until result is 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);
for (;;) {
next();
p = cs.data, i = cs.size - 1;
if (i > 0
&& ((p[0] == '"' && p[i] == '"')
|| (p[0] == '<' && p[i] == '>')))
break;
if (tok == TOK_LINEFEED)
tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
cstr_cat(&cs, get_tok_str(tok, &tokc), -1);
}
c = p[0];
/* remove '<>|""' */
memmove(p, p + 1, cs.size -= 2);
}
cstr_ccat(&cs, '\0');
pstrcpy(name, sizeof name, cs.data);
cstr_free(&cs);
i = do_next ? file->include_next_index : -1;
for (;;) {
++i;
if (i == 0) {
/* check absolute include path */
if (!IS_ABSPATH(name))
continue;
buf[0] = '\0';
} else if (i == 1) {
/* search in file's dir if "header.h" */
if (c != '\"')
continue;
p = file->true_filename;
pstrncpy(buf, p, tcc_basename(p) - p);
} else {
int j = i - 2, k = j - s1->nb_include_paths;
if (k < 0)
p = s1->include_paths[j];
else if (k < s1->nb_sysinclude_paths)
p = s1->sysinclude_paths[k];
else if (test)
return 0;
else
tcc_error("include file '%s' not found", name);
pstrcpy(buf, sizeof buf, p);
pstrcat(buf, sizeof buf, "/");
}
pstrcat(buf, sizeof buf, name);
e = search_cached_include(s1, buf, 0);
if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) {
/* no need to parse the include because the 'ifndef macro'
is defined (or had #pragma once) */
#ifdef INC_DEBUG
printf("%s: skipping cached %s\n", file->filename, buf);
#endif
return 1;
}
if (tcc_open(s1, buf) >= 0)
break;
}
if (test) {
tcc_close();
} else {
if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
tcc_error("#include recursion too deep");
/* push previous file on stack */
*s1->include_stack_ptr++ = file->prev;
file->include_next_index = i;
#ifdef INC_DEBUG
printf("%s: including %s\n", file->prev->filename, file->filename);
#endif
/* update target deps */
if (s1->gen_deps) {
BufferedFile *bf = file;
while (i == 1 && (bf = bf->prev))
i = bf->include_next_index;
/* skip system include files */
if (s1->include_sys_deps || i - 2 < s1->nb_include_paths)
dynarray_add(&s1->target_deps, &s1->nb_target_deps,
tcc_strdup(buf));
}
/* add include file debug info */
tcc_debug_bincl(s1);
tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
}
return 1;
}
/* eval an expression for #if/#elif */
static int expr_preprocess(TCCState *s1)
{
int c, t;
TokenString *str;
str = tok_str_alloc();
pp_expr = 1;
while (tok != TOK_LINEFEED && tok != TOK_EOF) {
next(); /* do macro subst */
redo:
if (tok == TOK_DEFINED) {
next_nomacro();
t = tok;
if (t == '(')
next_nomacro();
if (tok < TOK_IDENT)
expect("identifier");
if (s1->run_test)
maybe_run_test(s1);
c = 0;
if (define_find(tok)
|| tok == TOK___HAS_INCLUDE
|| tok == TOK___HAS_INCLUDE_NEXT)
c = 1;
if (t == '(') {
next_nomacro();
if (tok != ')')
expect("')'");
}
tok = TOK_CINT;
tokc.i = c;
} else if (tok == TOK___HAS_INCLUDE ||
tok == TOK___HAS_INCLUDE_NEXT) {
t = tok;
next_nomacro();
if (tok != '(')
expect("(");
c = parse_include(s1, t - TOK___HAS_INCLUDE, 1);
if (tok != ')')
expect("')'");
tok = TOK_CINT;
tokc.i = c;
} else if (tok >= TOK_IDENT) {
/* if undefined macro, replace with zero, check for func-like */
t = tok;
tok = TOK_CINT;
tokc.i = 0;
tok_str_add_tok(str);
next();
if (tok == '(')
tcc_error("function-like macro '%s' is not defined",
get_tok_str(t, NULL));
goto redo;
}
tok_str_add_tok(str);
}
pp_expr = 0;
tok_str_add(str, -1); /* simulate end of file */
tok_str_add(str, 0);
/* now evaluate C constant expression */
begin_macro(str, 1);
next();
c = expr_const();
end_macro();
return c != 0;
}
/* parse after #define */
ST_FUNC void parse_define(void)
{
Sym *s, *first, **ps;
int v, t, varg, is_vaargs, spc;
int saved_parse_flags = parse_flags;
v = tok;
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();
parse_flags &= ~PARSE_FLAG_SPACES;
if (tok == '(') {
int dotid = set_idnum('.', 0);
next_nomacro();
ps = &first;
if (tok != ')') for (;;) {
varg = tok;
next_nomacro();
is_vaargs = 0;
if (varg == TOK_DOTS) {
varg = TOK___VA_ARGS__;
is_vaargs = 1;
} else if (tok == TOK_DOTS && gnu_ext) {
is_vaargs = 1;
next_nomacro();
}
if (varg < TOK_IDENT)
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 == ')')
break;
if (tok != ',' || is_vaargs)
goto bad_list;
next_nomacro();
}
parse_flags |= PARSE_FLAG_SPACES;
next_nomacro();
t = MACRO_FUNC;
set_idnum('.', dotid);
}
tokstr_buf.len = 0;
spc = 2;
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 '#' */
if (TOK_TWOSHARPS == tok) {
if (2 == spc)
goto bad_twosharp;
if (1 == spc)
--tokstr_buf.len;
spc = 3;
tok = TOK_PPJOIN;
} else if ('#' == tok) {
spc = 4;
} else if (check_space(tok, &spc)) {
goto skip;
}
tok_str_add2(&tokstr_buf, tok, &tokc);
skip:
next_nomacro();
}
parse_flags = saved_parse_flags;
if (spc == 1)
--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 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;
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);
i = s1->cached_includes_hash[h];
for(;;) {
if (i == 0)
break;
e = s1->cached_includes[i - 1];
if (0 == PATHCMP(e->filename, filename))
return e;
i = e->hash_next;
}
if (!add)
return NULL;
e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
strcpy(e->filename, filename);
e->ifndef_macro = e->once = 0;
dynarray_add(&s1->cached_includes, &s1->nb_cached_includes, e);
/* add in hash table */
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)
{
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) // push current
#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:
tcc_error("out of pack stack");
}
s1->pack_stack_ptr--;
} else {
int val = 0;
if (tok != ')') {
if (tok == TOK_ASM_push) {
next();
if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
goto stk_error;
val = *s1->pack_stack_ptr++;
if (tok != ',')
goto pack_set;
next();
}
if (tok != TOK_CINT)
goto pragma_err;
val = tokc.i;
if (val < 1 || val > 16 || (val & (val - 1)) != 0)
goto pragma_err;
next();
}
pack_set:
*s1->pack_stack_ptr = val;
}
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
tcc_warning_c(warn_unsupported)("#pragma %s 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 */
ST_FUNC void preprocess(int is_bof)
{
TCCState *s1 = tcc_state;
int c, n, saved_parse_flags;
char buf[1024], *q;
Sym *s;
saved_parse_flags = parse_flags;
parse_flags = PARSE_FLAG_PREPROCESS
| PARSE_FLAG_TOK_NUM
| PARSE_FLAG_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)
define_undef(s);
break;
case TOK_INCLUDE:
case TOK_INCLUDE_NEXT:
parse_include(s1, tok - TOK_INCLUDE, 0);
break;
case TOK_IFNDEF:
c = 1;
goto do_ifdef;
case TOK_IF:
c = expr_preprocess(s1);
goto do_if;
case TOK_IFDEF:
c = 0;
do_ifdef:
next_nomacro();
if (tok < TOK_IDENT)
tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
if (is_bof) {
if (c) {
#ifdef INC_DEBUG
printf("#ifndef %s\n", get_tok_str(tok, NULL));
#endif
file->ifndef_macro = tok;
}
}
if (define_find(tok)
|| tok == TOK___HAS_INCLUDE
|| tok == TOK___HAS_INCLUDE_NEXT)
c ^= 1;
do_if:
if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
tcc_error("memory full (ifdef)");
*s1->ifdef_stack_ptr++ = c;
goto test_skip;
case TOK_ELSE:
if (s1->ifdef_stack_ptr == s1->ifdef_stack)
tcc_error("#else without matching #if");
if (s1->ifdef_stack_ptr[-1] & 2)
tcc_error("#else after #else");
c = (s1->ifdef_stack_ptr[-1] ^= 3);
goto test_else;
case TOK_ELIF:
if (s1->ifdef_stack_ptr == s1->ifdef_stack)
tcc_error("#elif without matching #if");
c = s1->ifdef_stack_ptr[-1];
if (c > 1)
tcc_error("#elif after #else");
/* last #if/#elif expression was true: we skip */
if (c == 1) {
c = 0;
} else {
c = expr_preprocess(s1);
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)) {
preprocess_skip();
is_bof = 0;
goto redo;
}
break;
case TOK_ENDIF:
if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
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 */
if (file->ifndef_macro &&
s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
file->ifndef_macro_saved = file->ifndef_macro;
/* need to set to zero to avoid false matches if another
#ifndef at middle of file */
file->ifndef_macro = 0;
while (tok != TOK_LINEFEED)
next_nomacro();
tok_flags |= TOK_FLAG_ENDIF;
goto the_end;
}
break;
case TOK_PPNUM:
n = strtoul((char*)tokc.str.data, &q, 10);
goto _line_num;
case TOK_LINE:
next();
if (tok != TOK_CINT)
_line_err:
tcc_error("wrong #line format");
n = tokc.i;
_line_num:
next();
if (tok != TOK_LINEFEED) {
if (tok == TOK_STR) {
if (file->true_filename == file->filename)
file->true_filename = tcc_strdup(file->filename);
q = (char *)tokc.str.data;
buf[0] = 0;
if (!IS_ABSPATH(q)) {
/* prepend directory from real file */
pstrcpy(buf, sizeof buf, file->true_filename);
*tcc_basename(buf) = 0;
}
pstrcat(buf, sizeof buf, q);
tcc_debug_putfile(s1, buf);
} 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;
break;
case TOK_ERROR:
case TOK_WARNING:
q = buf;
c = skip_spaces();
while (c != '\n' && c != CH_EOF) {
if ((q - buf) < sizeof(buf) - 1)
*q++ = c;
c = ninp();
}
*q = '\0';
if (tok == TOK_ERROR)
tcc_error("#error %s", buf);
else
tcc_warning("#warning %s", buf);
break;
case TOK_PRAGMA:
pragma_parse(s1);
break;
case TOK_LINEFEED:
goto the_end;
default:
/* 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)
next_nomacro();
the_end:
parse_flags = saved_parse_flags;
}
/* evaluate escape codes in a string. */
static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
{
int c, n, i;
const uint8_t *p;
p = buf;
for(;;) {
c = *p;
if (c == '\0')
break;
if (c == '\\') {
p++;
/* escape */
c = *p;
switch(c) {
case '0': case '1': case '2': case '3':
case '4': case '5': case '6': case '7':
/* at most three octal digits */
n = c - '0';
p++;
c = *p;
if (isoct(c)) {
n = n * 8 + c - '0';
p++;
c = *p;
if (isoct(c)) {
n = n * 8 + c - '0';
p++;
}
}
c = n;
goto add_char_nonext;
case 'x': i = 0; goto parse_hex_or_ucn;
case 'u': i = 4; goto parse_hex_or_ucn;
case 'U': i = 8; goto parse_hex_or_ucn;
parse_hex_or_ucn:
p++;
n = 0;
do {
c = *p;
if (c >= 'a' && c <= 'f')
c = c - 'a' + 10;
else if (c >= 'A' && c <= 'F')
c = c - 'A' + 10;
else if (isnum(c))
c = c - '0';
else if (i > 0)
expect("more hex digits in universal-character-name");
else
goto add_hex_or_ucn;
n = n * 16 + c;
p++;
} while (--i);
if (is_long) {
add_hex_or_ucn:
c = n;
goto add_char_nonext;
}
cstr_u8cat(outstr, n);
continue;
case 'a':
c = '\a';
break;
case 'b':
c = '\b';
break;
case 'f':
c = '\f';
break;
case 'n':
c = '\n';
break;
case 'r':
c = '\r';
break;
case 't':
c = '\t';
break;
case 'v':
c = '\v';
break;
case 'e':
if (!gnu_ext)
goto invalid_escape;
c = 27;
break;
case '\'':
case '\"':
case '\\':
case '?':
break;
default:
invalid_escape:
if (c >= '!' && c <= '~')
tcc_warning("unknown escape sequence: \'\\%c\'", c);
else
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 {
#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)
cstr_ccat(outstr, '\0');
else
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_c(warn_all)("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 */
static void bn_lshift(unsigned int *bn, int shift, int or_val)
{
int i;
unsigned int v;
for(i=0;i<BN_SIZE;i++) {
v = bn[i];
bn[i] = (v << shift) | or_val;
or_val = v >> (32 - shift);
}
}
static void bn_zero(unsigned int *bn)
{
int i;
for(i=0;i<BN_SIZE;i++) {
bn[i] = 0;
}
}
/* parse number in null terminated string 'p' and return it in the
current token */
static void parse_number(const char *p)
{
int b, t, shift, frac_bits, s, exp_val, ch;
char *q;
unsigned int bn[BN_SIZE];
double d;
/* number */
q = token_buf;
ch = *p++;
t = ch;
ch = *p++;
*q++ = t;
b = 10;
if (t == '.') {
goto float_frac_parse;
} else if (t == '0') {
if (ch == 'x' || ch == 'X') {
q--;
ch = *p++;
b = 16;
} else if (tcc_state->tcc_ext && (ch == 'b' || ch == 'B')) {
q--;
ch = *p++;
b = 2;
}
}
/* parse all digits. cannot check octal numbers at this stage
because of floating point constants */
while (1) {
if (ch >= 'a' && ch <= 'f')
t = ch - 'a' + 10;
else if (ch >= 'A' && ch <= 'F')
t = ch - 'A' + 10;
else if (isnum(ch))
t = ch - '0';
else
break;
if (t >= b)
break;
if (q >= token_buf + STRING_MAX_SIZE) {
num_too_long:
tcc_error("number too long");
}
*q++ = ch;
ch = *p++;
}
if (ch == '.' ||
((ch == 'e' || ch == 'E') && b == 10) ||
((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
if (b != 10) {
/* NOTE: strtox should support that for hexa numbers, but
non ISOC99 libcs do not support it, so we prefer to do
it by hand */
/* hexadecimal or binary floats */
/* XXX: handle overflows */
*q = '\0';
if (b == 16)
shift = 4;
else
shift = 1;
bn_zero(bn);
q = token_buf;
while (1) {
t = *q++;
if (t == '\0') {
break;
} else if (t >= 'a') {
t = t - 'a' + 10;
} else if (t >= 'A') {
t = t - 'A' + 10;
} else {
t = t - '0';
}
bn_lshift(bn, shift, t);
}
frac_bits = 0;
if (ch == '.') {
ch = *p++;
while (1) {
t = ch;
if (t >= 'a' && t <= 'f') {
t = t - 'a' + 10;
} else if (t >= 'A' && t <= 'F') {
t = t - 'A' + 10;
} else if (t >= '0' && t <= '9') {
t = t - '0';
} else {
break;
}
if (t >= b)
tcc_error("invalid digit");
bn_lshift(bn, shift, t);
frac_bits += shift;
ch = *p++;
}
}
if (ch != 'p' && ch != 'P')
expect("exponent");
ch = *p++;
s = 1;
exp_val = 0;
if (ch == '+') {
ch = *p++;
} else if (ch == '-') {
s = -1;
ch = *p++;
}
if (ch < '0' || ch > '9')
expect("exponent digits");
while (ch >= '0' && ch <= '9') {
exp_val = exp_val * 10 + ch - '0';
ch = *p++;
}
exp_val = exp_val * s;
/* now we can generate the number */
/* XXX: should patch directly float number */
d = (double)bn[1] * 4294967296.0 + (double)bn[0];
d = ldexp(d, exp_val - frac_bits);
t = toup(ch);
if (t == 'F') {
ch = *p++;
tok = TOK_CFLOAT;
/* float : should handle overflow */
tokc.f = (float)d;
} else if (t == 'L') {
ch = *p++;
#ifdef TCC_USING_DOUBLE_FOR_LDOUBLE
tok = TOK_CDOUBLE;
tokc.d = d;
#else
tok = TOK_CLDOUBLE;
/* XXX: not large enough */
tokc.ld = (long double)d;
#endif
} else {
tok = TOK_CDOUBLE;
tokc.d = d;
}
} else {
/* decimal floats */
if (ch == '.') {
if (q >= token_buf + STRING_MAX_SIZE)
goto num_too_long;
*q++ = ch;
ch = *p++;
float_frac_parse:
while (ch >= '0' && ch <= '9') {
if (q >= token_buf + STRING_MAX_SIZE)
goto num_too_long;
*q++ = ch;
ch = *p++;
}
}
if (ch == 'e' || ch == 'E') {
if (q >= token_buf + STRING_MAX_SIZE)
goto num_too_long;
*q++ = ch;
ch = *p++;
if (ch == '-' || ch == '+') {
if (q >= token_buf + STRING_MAX_SIZE)
goto num_too_long;
*q++ = ch;
ch = *p++;
}
if (ch < '0' || ch > '9')
expect("exponent digits");
while (ch >= '0' && ch <= '9') {
if (q >= token_buf + STRING_MAX_SIZE)
goto num_too_long;
*q++ = ch;
ch = *p++;
}
}
*q = '\0';
t = toup(ch);
errno = 0;
if (t == 'F') {
ch = *p++;
tok = TOK_CFLOAT;
tokc.f = strtof(token_buf, NULL);
} else if (t == 'L') {
ch = *p++;
#ifdef TCC_USING_DOUBLE_FOR_LDOUBLE
tok = TOK_CDOUBLE;
tokc.d = strtod(token_buf, NULL);
#else
tok = TOK_CLDOUBLE;
tokc.ld = strtold(token_buf, NULL);
#endif
} else {
tok = TOK_CDOUBLE;
tokc.d = strtod(token_buf, NULL);
}
}
} else {
unsigned long long n, n1;
int lcount, ucount, ov = 0;
const char *p1;
/* integer number */
*q = '\0';
q = token_buf;
if (b == 10 && *q == '0') {
b = 8;
q++;
}
n = 0;
while(1) {
t = *q++;
/* no need for checks except for base 10 / 8 errors */
if (t == '\0')
break;
else if (t >= 'a')
t = t - 'a' + 10;
else if (t >= 'A')
t = t - 'A' + 10;
else
t = t - '0';
if (t >= b)
tcc_error("invalid digit");
n1 = n;
n = n * b + t;
/* detect overflow */
if (n1 >= 0x1000000000000000ULL && n / b != n1)
ov = 1;
}
/* 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);