rust/compiler/rustc_trait_selection/src/traits/structural_match.rs

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

209 lines
8.1 KiB
Rust
Raw Normal View History

use rustc_data_structures::fx::FxHashSet;
use rustc_hir as hir;
use rustc_middle::ty::{self, Ty, TyCtxt, TypeSuperVisitable, TypeVisitable, TypeVisitor};
use rustc_span::Span;
use std::ops::ControlFlow;
/// This method traverses the structure of `ty`, trying to find an
/// instance of an ADT (i.e. struct or enum) that doesn't implement
/// the structural-match traits, or a generic type parameter
/// (which cannot be determined to be structural-match).
///
/// The "structure of a type" includes all components that would be
/// considered when doing a pattern match on a constant of that
/// type.
///
/// * This means this method descends into fields of structs/enums,
/// and also descends into the inner type `T` of `&T` and `&mut T`
///
/// * The traversal doesn't dereference unsafe pointers (`*const T`,
/// `*mut T`), and it does not visit the type arguments of an
/// instantiated generic like `PhantomData<T>`.
///
/// The reason we do this search is Rust currently require all ADTs
/// reachable from a constant's type to implement the
/// structural-match traits, which essentially say that
/// the implementation of `PartialEq::eq` behaves *equivalently* to a
/// comparison against the unfolded structure.
///
/// For more background on why Rust has this requirement, and issues
/// that arose when the requirement was not enforced completely, see
/// Rust RFC 1445, rust-lang/rust#61188, and rust-lang/rust#62307.
pub fn search_for_structural_match_violation<'tcx>(
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
span: Span,
tcx: TyCtxt<'tcx>,
ty: Ty<'tcx>,
) -> Option<Ty<'tcx>> {
ty.visit_with(&mut Search { tcx, span, seen: FxHashSet::default(), adt_const_param: false })
.break_value()
}
/// This method traverses the structure of `ty`, trying to find any
/// types that are not allowed to be used in a const generic.
///
/// This is either because the type does not implement `StructuralEq`
/// and `StructuralPartialEq`, or because the type is intentionally
/// not supported in const generics (such as floats and raw pointers,
/// which are allowed in match blocks).
pub fn search_for_adt_const_param_violation<'tcx>(
span: Span,
tcx: TyCtxt<'tcx>,
ty: Ty<'tcx>,
) -> Option<Ty<'tcx>> {
ty.visit_with(&mut Search { tcx, span, seen: FxHashSet::default(), adt_const_param: true })
.break_value()
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
}
2019-10-25 14:47:04 +02:00
/// This implements the traversal over the structure of a given type to try to
/// find instances of ADTs (specifically structs or enums) that do not implement
/// the structural-match traits (`StructuralPartialEq` and `StructuralEq`).
struct Search<'tcx> {
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
span: Span,
tcx: TyCtxt<'tcx>,
2019-10-25 14:47:04 +02:00
/// Tracks ADTs previously encountered during search, so that
/// we will not recur on them again.
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
seen: FxHashSet<hir::def_id::DefId>,
// Additionally deny things that have been allowed in patterns,
// but are not allowed in adt const params, such as floats and
// fn ptrs.
adt_const_param: bool,
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
}
impl<'tcx> Search<'tcx> {
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
fn type_marked_structural(&self, adt_ty: Ty<'tcx>) -> bool {
adt_ty.is_structural_eq_shallow(self.tcx)
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
}
}
impl<'tcx> TypeVisitor<'tcx> for Search<'tcx> {
type BreakTy = Ty<'tcx>;
2020-11-05 17:30:39 +01:00
fn visit_ty(&mut self, ty: Ty<'tcx>) -> ControlFlow<Self::BreakTy> {
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
debug!("Search visiting ty: {:?}", ty);
2020-08-03 00:49:11 +02:00
let (adt_def, substs) = match *ty.kind() {
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
ty::Adt(adt_def, substs) => (adt_def, substs),
ty::Param(_) => {
return ControlFlow::Break(ty);
}
2020-04-11 21:02:49 +02:00
ty::Dynamic(..) => {
return ControlFlow::Break(ty);
2020-04-11 21:02:49 +02:00
}
ty::Foreign(_) => {
return ControlFlow::Break(ty);
}
2022-11-27 17:52:17 +00:00
ty::Alias(..) => {
return ControlFlow::Break(ty);
}
2021-10-18 00:00:00 +00:00
ty::Closure(..) => {
return ControlFlow::Break(ty);
2021-10-18 00:00:00 +00:00
}
ty::Generator(..) | ty::GeneratorWitness(..) => {
return ControlFlow::Break(ty);
}
ty::FnDef(..) => {
// Types of formals and return in `fn(_) -> _` are also irrelevant;
2019-10-25 14:47:04 +02:00
// so we do not recur into them via `super_visit_with`
return ControlFlow::CONTINUE;
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
}
ty::Array(_, n)
if { n.try_eval_usize(self.tcx, ty::ParamEnv::reveal_all()) == Some(0) } =>
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
{
// rust-lang/rust#62336: ignore type of contents
// for empty array.
return ControlFlow::CONTINUE;
}
ty::Bool | ty::Char | ty::Int(_) | ty::Uint(_) | ty::Str | ty::Never => {
// These primitive types are always structural match.
//
// `Never` is kind of special here, but as it is not inhabitable, this should be fine.
return ControlFlow::CONTINUE;
}
ty::FnPtr(..) => {
if !self.adt_const_param {
return ControlFlow::CONTINUE;
} else {
return ControlFlow::Break(ty);
}
}
ty::RawPtr(..) => {
if !self.adt_const_param {
// structural-match ignores substructure of
// `*const _`/`*mut _`, so skip `super_visit_with`.
//
// For example, if you have:
// ```
// struct NonStructural;
// #[derive(PartialEq, Eq)]
// struct T(*const NonStructural);
// const C: T = T(std::ptr::null());
// ```
//
// Even though `NonStructural` does not implement `PartialEq`,
// structural equality on `T` does not recur into the raw
// pointer. Therefore, one can still use `C` in a pattern.
return ControlFlow::CONTINUE;
} else {
return ControlFlow::Break(ty);
}
}
ty::Float(_) => {
if !self.adt_const_param {
return ControlFlow::CONTINUE;
} else {
return ControlFlow::Break(ty);
}
}
ty::Array(..) | ty::Slice(_) | ty::Ref(..) | ty::Tuple(..) => {
2020-05-13 11:52:22 +02:00
// First check all contained types and then tell the caller to continue searching.
return ty.super_visit_with(self);
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
}
2021-10-18 00:00:00 +00:00
ty::Infer(_) | ty::Placeholder(_) | ty::Bound(..) => {
2020-05-12 23:35:29 +02:00
bug!("unexpected type during structural-match checking: {:?}", ty);
}
ty::Error(_) => {
self.tcx.sess.delay_span_bug(self.span, "ty::Error in structural-match check");
2020-05-12 23:35:29 +02:00
// We still want to check other types after encountering an error,
// as this may still emit relevant errors.
return ControlFlow::CONTINUE;
2020-05-12 23:35:29 +02:00
}
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
};
if !self.seen.insert(adt_def.did()) {
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
debug!("Search already seen adt_def: {:?}", adt_def);
return ControlFlow::CONTINUE;
}
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
if !self.type_marked_structural(ty) {
debug!("Search found ty: {:?}", ty);
return ControlFlow::Break(ty);
Migrate from `#[structural_match]` attribute a lang-item trait. (Or more precisely, a pair of such traits: one for `derive(PartialEq)` and one for `derive(Eq)`.) ((The addition of the second marker trait, `StructuralEq`, is largely a hack to work-around `fn (&T)` not implementing `PartialEq` and `Eq`; see also issue rust-lang/rust#46989; otherwise I would just check if `Eq` is implemented.)) Note: this does not use trait fulfillment error-reporting machinery; it just uses the trait system to determine if the ADT was tagged or not. (Nonetheless, I have kept an `on_unimplemented` message on the new trait for structural_match check, even though it is currently not used.) Note also: this does *not* resolve the ICE from rust-lang/rust#65466, as noted in a comment added in this commit. Further work is necessary to resolve that and other problems with the structural match checking, especially to do so without breaking stable code (adapted from test fn-ptr-is-structurally-matchable.rs): ```rust fn r_sm_to(_: &SM) {} fn main() { const CFN6: Wrap<fn(&SM)> = Wrap(r_sm_to); let input: Wrap<fn(&SM)> = Wrap(r_sm_to); match Wrap(input) { Wrap(CFN6) => {} Wrap(_) => {} }; } ``` where we would hit a problem with the strategy of unconditionally checking for `PartialEq` because the type `for <'a> fn(&'a SM)` does not currently even *implement* `PartialEq`. ---- added review feedback: * use an or-pattern * eschew `return` when tail position will do. * don't need fresh_expansion; just add `structural_match` to appropriate `allow_internal_unstable` attributes. also fixed example in doc comment so that it actually compiles.
2019-10-17 10:54:37 +02:00
}
// structural-match does not care about the
// instantiation of the generics in an ADT (it
// instead looks directly at its fields outside
// this match), so we skip super_visit_with.
//
// (Must not recur on substs for `PhantomData<T>` cf
// rust-lang/rust#55028 and rust-lang/rust#55837; but also
// want to skip substs when only uses of generic are
// behind unsafe pointers `*const T`/`*mut T`.)
// even though we skip super_visit_with, we must recur on
// fields of ADT.
let tcx = self.tcx;
adt_def.all_fields().map(|field| field.ty(tcx, substs)).try_for_each(|field_ty| {
let ty = self.tcx.normalize_erasing_regions(ty::ParamEnv::empty(), field_ty);
debug!("structural-match ADT: field_ty={:?}, ty={:?}", field_ty, ty);
ty.visit_with(self)
})
}
}