basic_types.hpp: Cleanup; Add constexpr 'enum class endian', 'pointer_cast()' and 'bit_cast()' and have all byte-order conversion and get/set functions be of constexpr; Add generalized template [get|put]_value(..) and to_hex_string(..)

Cleanup
- Split basic_types.hpp -> basic_types.hpp + byte_util.hpp + int_types.hpp + string_util.hpp

- Moved nsize_t, snsize_t to int_types.hpp and
  using 'uint_fast32_t' and 'int_fast32_t' as natural types.

- Renamed cpp_lang_macros.hpp -> cpp_lang_util.hpp

+++

Add constexpr 'enum class endian', 'pointer_cast()' and 'bit_cast()'
and have all byte-order conversion and get/set functions be of constexpr.

- Exposing '__builtin_bit_cast(Dest_type, arg)' via
  'constexpr bool is_builtin_bit_cast_available()' and type traits.

- Adding constexpr bit_cast<>() template for C++17 (a C++20 std),
  allowing constexpr type conversion using '__builtin_bit_cast(Dest_type, arg)',
  if the latter is available.

- Adding constexpr pointer_cast<>() template,
  allowing constexpr pointer type conversion.
  Either using '__builtin_bit_cast(Dest_type, arg)' or reinterpret_cast<>().

- Add constexpr 'enum class endian' API,
  providing compile-time C++ endian evaluation w/o predefined macros.
  Inspired by C++20.

- Replace linux bswap_[16,32,64] with either __builtin_bswap[16,32,64] or own const definition,
  both allowing constexpr

- Add unified overloaded 'constexpr bswap(uint[16,32,64,128,192,256]_t const &)'
  using constexpr endian API.

- Have all [get|put]_<type>(..) operations be of constexpr,
  using pointer_cast<>() instead of plain reinterpret_cast<>()
  and the new 'constexpr bswap(..)' methods.

+++

Add generalized template [get|put]_value(..) and to_hex_string(..)

- Add generalized template 'constexpr T [get|put]_value(..) {}' for std::is_standard_layout_v<T>

- Add generalized template 'inline to_hex_string(T const &) {}' for std::is_standard_layout_v<T>

+++

All of the above is covered by unit test 'test_basictypecon.cpp' '-std=c++17':
- gcc    8.3.0 on arm32, arm64: __builtin_bit_cast() not available
- gcc   10.2.1 on amd64:        __builtin_bit_cast() not available
- clang  9.0.1 on amd64, arm64: __builtin_bit_cast() is available
- clang 11.0.1 on amd64:        __builtin_bit_cast() is available

Full build time (user) incl unit tests (C++ and Java) on GCC
- amd64 gcc 11.0.1, 32 cores: 1m38s
- arm64 gcc 8.3.0, 4 cores: 13m30s
- arm32 gcc 8.3.0, 4 cores: 17m58s

1 files changed, 1 insertions, 1 deletions

diff --git a/src/debug.cpp b/src/debug.cpp
index 6ef7306..dd51186 100644
--- a/src/debug.cpp
+++ b/src/debug.cpp
@@ -55,7 +55,7 @@ std::string jau::get_backtrace(const bool skip_anon_frames, const jau::snsize_t
             continue;
         }
         bool append_line;
-        snprintf(cstr, sizeof(cstr), "%3u: ", frame);
+        snprintf(cstr, sizeof(cstr), "%3zu: ", (size_t)frame);
         std::string line(cstr);
 
         unw_get_reg(&cursor, UNW_REG_IP, &ip); // instruction pointer (pc)