libstdc++ bootstrap failures on sparc-sun-solaris2.8 (analyzed)

[Benjamin Kosnik]

Huh. I am not objecting to this approach. This is Gaby's baby: he needs 
to figure out how to fix it so that everybody's happy. I think it's 
gotten too complex as well.

.... passing the buck,
benjamin

> I think std_limits.h should be entirely pregenerated, based on the
> architecture, and stored in config/cpu or config/os.  (Perhaps only
> a small bit needs to be stored there, and the rest can be entirely
> generic.  For example, maybe you could just do `#define BITS 32' in
> config/cpu.  The details aren't important.)
> 
> Then gen-num-limits would never be run during any build of any
> kind.  It would only be run to generate a new file, when doing a
> new port, and then the output would be stored away.