rust/src/librustc_traits/implied_outlives_bounds.rs

//! Provider for the `implied_outlives_bounds` query.
//! Do not call this query directory. See [`rustc::traits::query::implied_outlives_bounds`].

use rustc::hir;
use rustc::infer::InferCtxt;
use rustc::infer::canonical::{self, Canonical};
use rustc::traits::{TraitEngine, TraitEngineExt};
use rustc::traits::query::outlives_bounds::OutlivesBound;
use rustc::traits::query::{CanonicalTyGoal, Fallible, NoSolution};
use rustc::ty::{self, Ty, TyCtxt, TypeFoldable};
use rustc::ty::outlives::Component;
use rustc::ty::query::Providers;
use rustc::ty::wf;
use smallvec::{SmallVec, smallvec};
use syntax::source_map::DUMMY_SP;
use rustc::traits::FulfillmentContext;

crate fn provide(p: &mut Providers<'_>) {
    *p = Providers {
        implied_outlives_bounds,
        ..*p
    };
}

fn implied_outlives_bounds<'tcx>(
    tcx: TyCtxt<'tcx>,
    goal: CanonicalTyGoal<'tcx>,
) -> Result<
    &'tcx Canonical<'tcx, canonical::QueryResponse<'tcx, Vec<OutlivesBound<'tcx>>>>,
    NoSolution,
> {
    tcx.infer_ctxt()
       .enter_canonical_trait_query(&goal, |infcx, _fulfill_cx, key| {
           let (param_env, ty) = key.into_parts();
           compute_implied_outlives_bounds(&infcx, param_env, ty)
       })
}

fn compute_implied_outlives_bounds<'tcx>(
    infcx: &InferCtxt<'_, 'tcx>,
    param_env: ty::ParamEnv<'tcx>,
    ty: Ty<'tcx>,
) -> Fallible<Vec<OutlivesBound<'tcx>>> {
    let tcx = infcx.tcx;

    // Sometimes when we ask what it takes for T: WF, we get back that
    // U: WF is required; in that case, we push U onto this stack and
    // process it next. Currently (at least) these resulting
    // predicates are always guaranteed to be a subset of the original
    // type, so we need not fear non-termination.
    let mut wf_types = vec![ty];

    let mut implied_bounds = vec![];

    let mut fulfill_cx = FulfillmentContext::new();

    while let Some(ty) = wf_types.pop() {
        // Compute the obligations for `ty` to be well-formed. If `ty` is
        // an unresolved inference variable, just substituted an empty set
        // -- because the return type here is going to be things we *add*
        // to the environment, it's always ok for this set to be smaller
        // than the ultimate set. (Note: normally there won't be
        // unresolved inference variables here anyway, but there might be
        // during typeck under some circumstances.)
        let obligations =
            wf::obligations(infcx, param_env, hir::DUMMY_HIR_ID, ty, DUMMY_SP).unwrap_or(vec![]);

        // N.B., all of these predicates *ought* to be easily proven
        // true. In fact, their correctness is (mostly) implied by
        // other parts of the program. However, in #42552, we had
        // an annoying scenario where:
        //
        // - Some `T::Foo` gets normalized, resulting in a
        //   variable `_1` and a `T: Trait<Foo=_1>` constraint
        //   (not sure why it couldn't immediately get
        //   solved). This result of `_1` got cached.
        // - These obligations were dropped on the floor here,
        //   rather than being registered.
        // - Then later we would get a request to normalize
        //   `T::Foo` which would result in `_1` being used from
        //   the cache, but hence without the `T: Trait<Foo=_1>`
        //   constraint. As a result, `_1` never gets resolved,
        //   and we get an ICE (in dropck).
        //
        // Therefore, we register any predicates involving
        // inference variables. We restrict ourselves to those
        // involving inference variables both for efficiency and
        // to avoids duplicate errors that otherwise show up.
        fulfill_cx.register_predicate_obligations(
            infcx,
            obligations
                .iter()
                .filter(|o| o.predicate.has_infer_types())
                .cloned(),
        );

        // From the full set of obligations, just filter down to the
        // region relationships.
        implied_bounds.extend(obligations.into_iter().flat_map(|obligation| {
            assert!(!obligation.has_escaping_bound_vars());
            match obligation.predicate {
                ty::Predicate::Trait(..) |
                ty::Predicate::Subtype(..) |
                ty::Predicate::Projection(..) |
                ty::Predicate::ClosureKind(..) |
                ty::Predicate::ObjectSafe(..) |
                ty::Predicate::ConstEvaluatable(..) => vec![],

                ty::Predicate::WellFormed(subty) => {
                    wf_types.push(subty);
                    vec![]
                }

                ty::Predicate::RegionOutlives(ref data) => match data.no_bound_vars() {
                    None => vec![],
                    Some(ty::OutlivesPredicate(r_a, r_b)) => {
                        vec![OutlivesBound::RegionSubRegion(r_b, r_a)]
                    }
                },

                ty::Predicate::TypeOutlives(ref data) => match data.no_bound_vars() {
                    None => vec![],
                    Some(ty::OutlivesPredicate(ty_a, r_b)) => {
                        let ty_a = infcx.resolve_vars_if_possible(&ty_a);
                        let mut components = smallvec![];
                        tcx.push_outlives_components(ty_a, &mut components);
                        implied_bounds_from_components(r_b, components)
                    }
                },
            }
        }));
    }

    // Ensure that those obligations that we had to solve
    // get solved *here*.
    match fulfill_cx.select_all_or_error(infcx) {
        Ok(()) => Ok(implied_bounds),
        Err(_) => Err(NoSolution),
    }
}

/// When we have an implied bound that `T: 'a`, we can further break
/// this down to determine what relationships would have to hold for
/// `T: 'a` to hold. We get to assume that the caller has validated
/// those relationships.
fn implied_bounds_from_components(
    sub_region: ty::Region<'tcx>,
    sup_components: SmallVec<[Component<'tcx>; 4]>,
) -> Vec<OutlivesBound<'tcx>> {
    sup_components
        .into_iter()
        .filter_map(|component| {
            match component {
                Component::Region(r) =>
                    Some(OutlivesBound::RegionSubRegion(sub_region, r)),
                Component::Param(p) =>
                    Some(OutlivesBound::RegionSubParam(sub_region, p)),
                Component::Projection(p) =>
                    Some(OutlivesBound::RegionSubProjection(sub_region, p)),
                Component::EscapingProjection(_) =>
                // If the projection has escaping regions, don't
                // try to infer any implied bounds even for its
                // free components. This is conservative, because
                // the caller will still have to prove that those
                // free components outlive `sub_region`. But the
                // idea is that the WAY that the caller proves
                // that may change in the future and we want to
                // give ourselves room to get smarter here.
                    None,
                Component::UnresolvedInferenceVariable(..) =>
                    None,
            }
        })
        .collect()
}