lua-users home
lua-l archive

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index]

Hello all,
to continue my previous post,
in str_format_helper(lua 5.0.2) the code
case 'c':  case 'd':  case 'i': {
  sprintf(buff, form, luaL_checkint(L, arg));
uses checkint macro, which typecast to int which is bad for me - I loose
the most significant 32 bits.

Hope this helps,

On Wed, 14 Dec 2005 15:15:28 +0200, Todor Totev <> wrote:

Hello all,
I found a problem with string.format function which exists in both 5.0 & 5.1. I'm building Lua using 64-bit integers as lua_Number. My platform is Win32 with
Microsoft VisualStudio 2003 (VisualC v13.10)

The problem is that str_format_helper(in 5.0) and scan_format(in 5.1)
firmly believe that lua_Number is double.
The first problem is that when they encounter %x and %d they generate
%x or %d in 'form' argument as format specifiers for sprintf.
In my case these must be converted to %I64x or %I64d.

Second problem is that str_format function
typecasts lua_Numbers impropertly:
when integer number representation is requested, the upper 32 bits are lost:

   case 'o':  case 'u':  case 'x':  case 'X': {
     sprintf(buff, form, (unsigned int)(luaL_checknumber(L, arg)));

Note the (unsigned int) typecast. In my case ommiting the typecast will fix the problem. Also, when the user request floating point representation, the code does NOT have a typecast:

   case 'e':  case 'E': case 'f':
   case 'g': case 'G': {
     sprintf(buff, form, luaL_checknumber(L, arg));

This also is problem for my setup. As a patch I added a typecast to double.

I'm not experienced with programming portable C code so I cannot suggest a proper patch. For first problem I think it is best to generate LUA_NUMBER_FMT in 'form' argument of
But I cannot suggest a nice solution to the casting issue.

Best regards,
Todor Totev