gcc/libstdc++-v3 ChangeLog configure.target co ...

[Richard Earnshaw]

> | Log message:
> | 	* config/cpu/arm/cpu_limits.h: New file.
> | 	* configure.target: Use config/cpu/arm for XScale and StrongARM
> | 	configurations.
> 
> Hi Richard,
> 
>   We're no longer using __glibcpp_long_double_bits in C++'s <limits>.
> We're getting that info from the compiler through __LONG_DOUBLE_BIT__.

Argh!  Extended build cycle times...

> However, what would be helpful is means to know whether a floating
> point type uses a hidden bit.  For a 32-bit or 64-bit wide
> floating-point type, that is a non-issue.  However, practice varies
> greatly for the extended floating-point types.

The ARM ABI doesn't use an extended type for long double, it maps directly 
onto double.  That is

	long double === double

R.