Fixed undefined behavior with log10 cast of fractional doubles

[mikael-s-persson]

Fixed undefined behavior with log10 cast of fractional doubles.
Detected by UBSan

What does this change do?

Fixed undefined behavior with log10 cast of fractional doubles. log10 for values less than 1 result in a negative result, static-casting to size_t is undefined.

Is it related to an exisiting bug report or feature request?

No. Example error:

external/tomlplusplus~/toml.hpp:16978:41: runtime error: -2.05916 is outside the range of representable values of type 'unsigned long'
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior external/tomlplusplus~/toml.hpp:16978:41

Pre-merge checklist

• I've read CONTRIBUTING.md
• I've rebased my changes against the current HEAD of origin/master (if necessary)
• [n/a] I've added new test cases to verify my change
• I've regenerated toml.hpp (how-to)
• [n/a] I've updated any affected documentation
• [n/a] I've rebuilt and run the tests with at least one of:
  • [n/a master is broken] Clang 8 or higher
  • [n/a master is broken] GCC 8 or higher
  • MSVC 19.20 (Visual Studio 2019) or higher
• [n/a] I've added my name to the list of contributors in README.md

[marzer]

Oooh, nice find, thanks! 🎉

[towerpark]

I got the following warning with the change from this PR in one of my development environments (Clang 10 & libstdc++ 8):

.../tomlplusplus/include/toml++/impl/toml_formatter.inl:94:41:
warning: using integer absolute value function 'abs' when argument is of floating
point type [-Wabsolute-value]
    return weight + static_cast<size_t>(abs(log10(val))) + 1u;
                                        ^
.../tomlplusplus/include/toml++/impl/toml_formatter.inl:94:41: note: use function
'std::abs' instead
    return weight + static_cast<size_t>(abs(log10(val))) + 1u;
                                        ^~~
                                        std::abs

However, my other environment with a newer toolchain (Clang 16 & libstdc++ 10) does not complain. I didn't dig in the environments to find out the root cause, but I think it's probably because the newer environment brings the C++ overloads into the global namespace by something like using std::abs;, whereas the older one doesn't do that, leaving abs() resolved to the C version (int abs(int) from stdlib.h) .

If older toolchains are not unsupported, it's better to prefix the abs() function with std::.