Floats in configuration probably shouldn't round to integers if theres precision loss

[nagisa]

Currently when you Config::get an integer and config has a float specified, the config-rs crate will transparently lose the precision, potentially leading to unexpected behaviour. For instance:

frame_rate = 59.94

will transparently round up to 60, potentially giving wrong video. There are probably many more instances where loss of precision can give very invalid results.

I believe the relevant code is probably this line.

I saw a number of FIXMEs in the codebase around not succeeding when integer wraparound occurs. We most likely can utilize something similar here – if converting results in loss of precision, fail.

[matthiasbeyer]

Agreed. Also, the crate stores u8 in u64 (and so on), which should be fixed as well (mentioning it here because it falls into the same category of issues).