long double and GLIBCPP_CHECK_MATH_SUPPORT

[David Edelsohn]

>>>>> Benjamin Kosnik writes:

Ben> Please don't change

Ben> GLIBCPP_CHECK_MATH_DECL_AND_LINKAGE_*

Ben> stuff.

Ben> Just change the general case: 
Ben> GLIBCPP_CHECK_BUILTIN_MATH_SUPPORT
Ben> GLIBCPP_CHECK_MATH_SUPPORT

Ben> to pre-emptively kill checking the entire long double spectrum if the
Ben> initial few aren't found. 

	Yes, that is what I initially had planned, but that makes
generating all of the HAVE_SINL, etc. macros for the pre-emptively killed
functions very cumbersome unless generating those macros can be avoided
all together.  Wrapping the range of long double tests in
GLIBCPP_CHECK_BUILTIN_MATH_SUPPORT and GLIBCPP_CHECK_MATH_SUPPORT only
works if all of the uses of those functions are wrapped by
_GLIBCPP_USE_LONG_DOUBLE in addition to the individual macros.  I was
concerned about imposing that constraint everywhere although that is the
rhythm in std_cmath.h.

	Is imposing that assumption all other places those functions may
be used in the future acceptable to you?

Thanks, David