fix formatting floating-point values without type and precision provided

[fsb4000]

Fixes #4319

libcxx\test\std\utilities\format\format.functions\locale-specific_form.pass.cpp is still failing after this fix.

Format string   {:$<11.6Lg}
Expected output 1,234.57$$$
Actual output   1,234.57$$$$

I guess, it's a different bug, isn't it?

[frederick-vs-ja]

libcxx\test\std\utilities\format\format.functions\locale-specific_form.pass.cpp is still failing.

Format string   {:$<11.6Lg}
Expected output 1,234.57$$$
Actual output   1,234.57$$$$

I guess, it's a different bug, isn't it?

Looks like #4316.

[StephanTLavavej]

Thanks! I looked into using <charconv>'s internal convention of using chars_format{} to signal "plain shortest" with an in-band special value, but the following logic in <format> would require additional adjustments and I felt that it was more risky than your fix.

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

[StephanTLavavej]

Thanks for fixing this libcxx failure - all shall love shortest round-trip and despair! ❤️ 😹 🛠️