Re: [Mingw-users] GCC 3.1: long double bug

[<dan...@ya...>]

 --- Luke Dunstan <cod...@ho...> wrote: > 
> Hi,
> 
> I thought I'd try building Boost 1.28.0 with Mingw GCC 3.1, and so far it
> seems to work well except for part of a regression test
> (libs/config/test/limits_test.cpp) involving long doubles. The following
> program demonstrates the problem:
> 
> /////
> #include <iostream>
> #include <limits>
> 
> template <class T>
> void print_hex_val(T t, const char* name)
> {
>   const unsigned char* p = (const unsigned char*)&t;
>   std::cout << "hex value of " << name << " is: " << std::endl;
>   for (unsigned int i = 0; i < sizeof(T); ++i) {
>     if(p[i] <= 0xF)
>       std::cout << "0";
>     std::cout << std::hex << (int)p[i];
>   }
>   std::cout << std::dec << std::endl;
> }
> 
> int main()
> {
>   long double one = 1.0, zero = 0.0, inf = one / zero;
>   _iec559_consts<long double>::_rep oldconst = { 0, 0, 0, 0x8000, 0x7fff };
> 
>   print_hex_val(inf, "long double infinity");
>   print_hex_val(oldconst.ldouble_val, "{ 0, 0, 0, 0x8000, 0x7fff }");
>   print_hex_val(std::numeric_limits<long double>::infinity(), "numeric
> limits");
>   print_hex_val(_iec559_consts<long double>::_S_InfL.ldouble_val,
> "iec559.ldouble_val");
>   print_hex_val(_iec559_consts<long double>::_S_InfL.double_val,
> "iec559.double_val");
> 
>   return 0;
> }
> /////
> 
> The output is:
> 
> hex value of long double infinity is:
> 0000000000000080ff7f0000
> hex value of { 0, 0, 0, 0x8000, 0x7fff } is:
> 0000000000000080ff7f0000
> hex value of numeric limits is:
> 000000000000000000800000
> hex value of iec559.ldouble_val is:
> 0000000000000080ff7f0000
> hex value of iec559.double_val is:
> 0000000000000080
> 
> 
> It is caused by a bug in the header
> libstdc++/config/os/mingw32/bits/cpu-limits.h of GCC:
> 
> --- cpu_limits.h.old Wed Jun  5 12:08:19 2002
> +++ cpu_limits.h Wed Jun  5 12:10:01 2002
> @@ -180,13 +180,13 @@
>  #define __glibcpp_long_double_has_denorm denorm_present
>  #define __glibcpp_long_double_has_denorm_loss true
>  #define __glibcpp_long_double_infinity \
> - _iec559_consts<bool>::_S_InfL.double_val
> + _iec559_consts<bool>::_S_InfL.ldouble_val
>  #define __glibcpp_long_double_quiet_NaN \
> - _iec559_consts<bool>::_S_QNaNL.double_val
> + _iec559_consts<bool>::_S_QNaNL.ldouble_val
>  #define __glibcpp_long_double_signaling_NaN \
> - _iec559_consts<bool>::_S_SNaNL.double_val
> + _iec559_consts<bool>::_S_SNaNL.ldouble_val
>  #define __glibcpp_long_double_denorm_min \
> - _iec559_consts<bool>::_S_DenormL.double_val
> + _iec559_consts<bool>::_S_DenormL.ldouble_val
>  #define __glibcpp_long_double_is_iec559 true
>  #define __glibcpp_long_double_is_bounded true
>  #define __glibcpp_long_double_is_modulo false
> 
> 
> Looks like a copy & paste typo :)
> 

Thanks! I had that fixed in my local swamp but forgot to apply to the distro
version.
Danny


Luke Dunstan
> 
> 
> _______________________________________________________________
> 
> Don't miss the 2002 Sprint PCS Application Developer's Conference
> August 25-28 in Las Vegas -- http://devcon.sprintpcs.com/adp/index.cfm
> 
> _______________________________________________
> MinGW-users mailing list
> Min...@li...
> 
> You may change your MinGW Account Options or unsubscribe at:
> https://lists.sourceforge.net/lists/listinfo/mingw-users 

http://www.sold.com.au - The Sold.com.au Big Brand Sale
- New PCs, notebooks, digital cameras, phones and more ... Sale ends June 12