[v3] policy-based data structures, round two

[Paolo Carlini]

David Edelsohn wrote:

>	New AIX pb_ds failures:
>  
>
Very interesting, on powerpc-aix you are *not* seeing the *other* fails, 
present on ia64 and powerp64-linux. In fact, even if the new rng testing 
infrastructure doesn't seem to me completely fine wrt 64-bit arches (see 
in particular the init and get_prob functions) that doesn't explain the 
hard Segmentation faults in that specific place (line #182 of 
twister_rand_gen.cc). The latter may well be due to a code generation 
bug, as far as I know... Humpf, a complex situation...

Paolo.