Whether floating point value Display should adhere to IEEE standard?

[crlf0710]

It was raised in #78618 that there needs conscious on whether f32/f64 Display should adhere to IEEE standard, for example outputting +0 instead of more conventional 0.

This issue is created to track this.

[programmerjake]

From what I can tell, the controversial part was having format!("{}", -0.0) produce "-0" instead of "0" as required by IEEE 754 to preserve the sign of zeros when round-tripping f32/f64 -> String -> f32/f64. Producing "+0" when formatting +0.0 is unnecessary for compliance with IEEE 754 because the sign for "0" is already parsed as positive.

[joshtriplett]

Personally, I think printing -0.0 as -0 seems reasonable. If you end up with a negative zero, it seems reasonable to support round-tripping that.

[bstrie]

Related to #72676 .

[joshtriplett]

We discussed this in the @rust-lang/libs meeting today. We had consensus that changing floating-point printing to print the negative sign on -0 seems reasonable. We don't expect that to cause a problem in practice.

Please go ahead with the PR. We would like to see that PR go through a libs FCP for confirmation when it's ready.