APPoS for short, throughout the library including in mixed precision arithmetic which should now always promote to the highest precision of any of the arguments.
See https://svn.boost.org/trac10/ticket/13264.
Fix previously unsupported situation: conversion of narrow cpp_bin_float to wider integer type.
See https://svn.boost.org/trac10/ticket/13301.
Improve test coverage for cpp_bin_floats with large exponent types, and/or dynamic allocators.
Fix issue in numeric_limits::max() that caused a complement of a signed integer number to be taken when the allocator is dynamic.
Fix heuristic in exp() function that caused a crazy number of iterations to be taken when the exponent was very large.
* Build with --hash as otherwise command line gets too long on Windows machines.
* Disable tests that generate object files too large for gcc on windows.
* Split cpp_dec_float tests up a bit to reduce object file size.
* Use boost::int46_t not ::int64_t in test case.
* Disable test_arithmetic_backend_concept.cpp with original mingw32 which has issues fixed in mingw64.
(which may be either narrower or wider than the int return type specified in C99).
Change transcendental code which uses ilogb for loop termination, to check for
a minimal value return to correctly handle the case that the value has gone to zero.
See https://svn.boost.org/trac/boost/ticket/12625
