[v3] C++0x type_traits post-oxford additions

[Benjamin Kosnik <bkoz at redhat dot com>]

Here are some of the new traits voted in at Oxford, as per my previous 
posts plus suggestions, plus fixes when it was actually tested.

I'm hoping that now I can get out of Paolo's way WRT the rest of the 
traits fixes. :)

I'm not attempting any grand structural changes with this patch. 
However, hopefully the structure for adding in the C++0x traits 
definitions alongside the TR1 traits definitions is now clear. OK?

There are four new failures with this patch, relating to the injection 
of TR1 type traits in C++0x. They look like this:

FAIL: 20_util/headers/type_traits/types_std_c++0x_neg.cc  (test for 
errors, line
 27)
FAIL: 20_util/headers/type_traits/types_std_c++0x_neg.cc  (test for 
errors, line
 28)
FAIL: 20_util/headers/type_traits/types_std_c++0x_neg.cc  (test for 
errors, line
 29)
FAIL: 20_util/headers/type_traits/types_std_c++0x_neg.cc  (test for 
errors, line
 30)

Don't be alarmed: they should go away soon. (xfailing an expected fail 
seems not to work.)

I should also add in a negative test like this for headers/cmath/ to 
catch special math injections. This is going to wait a bit.

In addition, some of the make_signed/make_unsigned tests will fail right 
now, due to the fact that is_signed/is_unsigned is using the TR1 
definition which isn't valid C++0x behavior for floating point types. 
(xfailing for run tests that give an unexpected compile error stumped 
me. Instead I just commented this part out for the time being. When 
is_signed/is_unsigned works, these can be activated again.)

tested x86/linux

-benjamin

Attachment: p.20070503-1.bz2
Description: application/bzip