Thread: make won't filter-out

I found my workaround, I edited the loslib.c file to use this instead:

Code:

typedef unsigned int lua_uint;
size_t u2a( char *str, size_t size, lua_uint val, lua_uint base, lua_uint small ) {
static const char bstr[] = "0123456789ABCDEF";
static const char sstr[] = "0123456789abcdef";
static const size_t min = (sizeof(int) * CHAR_BIT);
	const char *s = small ? sstr : bstr;
	size_t i = 0;
	if ( !str || !size || base < 2 || base > strlen(s) ) return 0;
	--size;
	if ( size < min ) {
		memset( str, 0, size );
		return 0;
	}
	while ( val > 0u ) {
		str[i++] = s[val % base];
		val /= base;
	}
	str[i] = 0;
	return i;
}
/* Avoids unistd.h, can expect sys/stat.h on all supported systems */
#include <sys/stat.h>
size_t tmpname( char *str, size_t size ) {
	size_t i;
	struct stat m = {0};
	if ( u2a( str, --size, 0, 2, 0 ) == 0 ) return 0;
	str[0] = '_';
	++str;
	do {
		i = u2a( str, size, rand(), 16, 1 );
	} while ( stat(str,&m) == 1 );
	return i;
}

static int os_tmpname (lua_State *L) {
  char buff[LUA_TMPNAMBUFSIZE];
  if ( tmpname(buff, LUA_TMPNAMBUFSIZE) == 0 )
    return luaL_error(L, "unable to generate a unique filename");
  lua_pushstring(L, buff);
  return 1;
}

I'm going to send this to the lua mailing list and hopefully it will be incorporated in their sources (with a few edits to conform to their own library stuff)

submitted

Yes but this doesn't set the compiler off when compiling, there is also the factor of it being significantly more thread safe since one doesn't know whether to free the pointer returned by tmpnam, my one you gotta provide the string buffer and size of it (to prevent eternal loop), while it cannot avoid the caveat of the file being created between returning an OK and caller attempting to create the file, it at least reduces other risks and stops the compiler complaining of a dangerous function, if you can think of a way to avoid that one caveat (without opening the file) then be my guest

Ah I must've mixed up the reference I was reading then, I searched for alternatives to access since when I tried it the compiler threw hissy fit about implicit function, I ended up choosing stat because I knew that was in one way or another available on windows too and was very likely available on any linux variant (albiet not always with all the defined variables)