Make gccrs recognize negative_impls

[badumbatish]

just a draft pr for #2962

[badumbatish]

I think i got gccrs to recognize the feature negative_impl but I'm not sure how to make it so that when user do sth like this

trait ExampleTrait {}

impl !ExampleTrait for i32 {}

it fails to compile instead of continuing with compilation because we haven't implemented them or if the user didn't do #![feature(negative_impls)] in the beginning of the .rs file

@P-E-P any pointers?

[P-E-P]

it fails to compile instead of continuing with compilation because we haven't implemented them or if the user didn't do #![feature(negative_impls)] in the beginning of the .rs file

@badumbatish You are in the right direction! There is a visitor in gcc/rust/checks/error/rust-feature-gate.cc which match the AST against available features. There is a member as_exclam in TraitImpl, if you meet one in the visitor, you need to error out and that's it.

Do not forget to add some tests! At least one with the feature enabled and one without it.

[badumbatish]

hmm i think i might be cooking something

[badumbatish]

hmm i got down the case where if a user attempts to do impl !ExampleTrait for i32 {} we error out with // { dg-error "negative_impls are not yet implemented" } but I don't think the rustc compiler is doing this https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=b63a98a5146d438ded89b9dac762be48

Also with the case that we did use #![feature(negative_impls)], we should also error out sth? https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=1f3c2bd54f30a5931c02a2a4abcac08b

I'm also not sure how this is relate to core mentioned in #2962, there seems to not have an enum state with "nightly" or sth for me to use on this

[CohenArthur]

hmm i got down the case where if a user attempts to do impl !ExampleTrait for i32 {} we error out with // { dg-error "negative_impls are not yet implemented" } but I don't think the rustc compiler is doing this https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=b63a98a5146d438ded89b9dac762be48

the error is not exactly the same but that's okay - it erros out and that's what matters :D

Also with the case that we did use #![feature(negative_impls)], we should also error out sth? https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=1f3c2bd54f30a5931c02a2a4abcac08b

we don't have a nightly and stable versions of the compiler at the moment, so you can just assume that everything is "nightly". so we handle having #![feature] at all times, and you don't need to do any particular things regarding nightly vs stable. so the right behavior is to not error out if the feature gate was enabled, like rustc does when run in nightly

I'm also not sure how this is relate to core mentioned in #2962, there seems to not have an enum state with "nightly" or sth for me to use on this

[CohenArthur]

nice job!

[badumbatish]

Ill squash everything down to 1 commit after this rebase from master passes

[P-E-P]

@badumbatish Could you rewrite your commit message please ?

[CohenArthur]

looks great! well done!