Strange decimal literal behaviour

[dzarda]

Describe the bug:

I'm seeing a weird representation of a _c literal decimal value such as 0.22_c. It seems turn into a massive value in the rep.

To Reproduce:

#include <cnl/all.h>
using namespace cnl::literals;
using fix24_8 = cnl::scaled_integer<int32_t, cnl::power<-8>>;

fix24_8 a = 0.22_c; // rep 4608
fix24_8 b = 0.22; // rep 56

Expected behavior:

Both reps should be 56.

Screenshots: Double literal:

Custom literal:

Desktop (please complete the following information):

• OS: Win 10
• Compiler version GCC 9.3.1
• Compiler flags used -DNDEBUG -std=c++17 -Iinclude
• CNL version (or Git branch/SHA) https://github.com/maxgerhardt/cnl.git#v1.1.7-pio

Link to Compiler Explorer:

https://godbolt.org/z/vTjdz5sah

[johnmcfarlane]

Unfortunately, the UDL, ""_c does not work with fractions. C++17 is only allowed to use integral NTTPs (non-type template parameters) and as 0.22 isn't even representable in binary floating-point, it's out of the running.

Sorry that it compiles at all! (It doesn't under v2 branch. Action: deprecate ""_c.)