Add const to constructors by a feature

[fmiguelgarcia]

Resolves Constructor generated new should be marked const

Synopsis

Added a new const feature which generates const fn on Constructor derives.

Solution

Checklist

• Documentation is updated
• Tests are added/updated IMPORTANT: I am not sure if new tests are added to the CI.
• [CHANGELOG entry][l:1] is added

[tyranron]

Should this really be feature-gated? I cannot imagine the situation where const will be harmful on constructors.

[fmiguelgarcia]

Should this really be feature-gated? I cannot imagine the situation where const will be harmful on constructors.

Just in case you are using a legacy compiler with no support of const in Generics.
Having said that, I think const feature should be enable by default, WDYT?

att. @tyranron

[tyranron]

@fmiguelgarcia well, once we release 1.0 (next release) we will have MSRV set to 1.65, so you won't be able to use legacy compilers 🤷‍♂️

[JelteF]

Yes, given we're going dropping support for old rust versions in the next release I think this shouldn't be behind a feature, but should simply be the only behaviour.

[JelteF]

Also, I feel like the same should be done for the IsVariant and Unwrap derives.

[tyranron]

@JelteF IsVariant yes, but Unwrap definitely not, as const at the moment is not very clever with destructor evaluation checks, so almost any struct/enum destruction in const leads to compilation error. See rust-lang/rust#60964 and rust-lang/rust#73255 for details.

[fmiguelgarcia]

@tyranron I've used the const feature to enable const in Unwrap (I've just read your comment).
IsVariant and Constructor now uses const fn.

Do you want to remove const feature completely? Or Do you want to keep it and support const in Unwrap?
att. @JelteF @tyranron

[tyranron]

@fmiguelgarcia I'd vote for removing const feature completely. We always may constify Unwrap once feature(const_precise_live_drops) is stabilized. It's not something that vital, deserving to squat-around it with an additional Cargo feature.

[JelteF]

This doesn't have to be a separate variable, I'm guessing this is a leftover from the feature.

[JelteF]

Something is weird about the github diff, I'm not sure why but it shows changes introduced in other PRs. You might want to try squashing your commits together and rebasing on top of master.

[JelteF]

I think this is almost good to merge, just a few things are needed:

1. Remove const from unwrap due to the bugs (so only const for constructor and isvariant for now)
2. Remove the const feature and simply always use const.
3. fix clippy errors.

[JelteF]

@fmiguelgarcia do you have time to do that? This is one of the final things left on the 1.0.0 milestone.