[v3 PATCH] PR libstdc++/60970, implement LWG 2148

[Ville Voutilainen <ville dot voutilainen at gmail dot com>]

Tested on Linux-PPC64.

The proposed resolution of the issue doesn't really say whether our
regression test for PR libstdc++/52931 should remain valid. However,
it doesn't say that we shouldn't keep it valid, either. This approach
keeps it valid, but provides support for hashing enums. It took a while
to figure out suitable jiggery-pokery to make it so, but this approach
passes the testsuite without regressions. I considered an alternative
alias-template-based approach, but while that attempt would've worked
with our current front-end, it would not have worked on clang (and
it's thus far unclear whether it was intended to work by the language
rules).

2015-07-25  Ville Voutilainen  <ville.voutilainen@gmail.com>
    PR libstdc++/60970, implement LWG 2148, hash support for
    enum types.
    * include/bits/functional_hash.h
    (__hash_enum): New.
    (hash): Derive from __hash_enum.
    * testsuite/20_util/hash/60970.cc: New.

Attachment: enum-hash.diff
Description: Text document