Try filtering out non-const impls when we expect const impls #87375

fee1-dead · 2021-07-22T16:05:06Z

TL;DR: Associated types on const impls are now bounded; we now disallow calling a const function with bounds when the specified type param only has a non-const impl.

r? @oli-obk

rust-highfive · 2021-07-22T16:05:08Z

Some changes occurred in src/tools/clippy.

cc @rust-lang/clippy

compiler/rustc_mir/src/transform/check_consts/validation.rs

compiler/rustc_trait_selection/src/traits/fulfill.rs

compiler/rustc_mir/src/borrow_check/type_check/mod.rs

oli-obk · 2021-07-22T16:46:06Z

compiler/rustc_trait_selection/src/traits/auto_trait.rs

@@ -285,6 +285,7 @@ impl AutoTraitFinder<'tcx> {
                def_id: trait_did,
                substs: infcx.tcx.mk_substs_trait(ty, &[]),
            },
+            constness: hir::Constness::NotConst,


similarly here, do we want Const if we are in a const context?

compiler/rustc_trait_selection/src/traits/auto_trait.rs

compiler/rustc_trait_selection/src/traits/select/mod.rs

oli-obk · 2021-07-22T16:53:02Z

compiler/rustc_traits/src/chalk/lowering.rs

@@ -702,7 +702,7 @@ impl<'tcx> LowerInto<'tcx, Option<chalk_solve::rust_ir::QuantifiedInlineBound<Ru
        let (predicate, binders, _named_regions) =
            collect_bound_vars(interner, interner.tcx, self.kind());
        match predicate {
-            ty::PredicateKind::Trait(predicate, _) => Some(chalk_ir::Binders::new(
+            ty::PredicateKind::Trait(predicate) => Some(chalk_ir::Binders::new(
                binders,
                chalk_solve::rust_ir::InlineBound::TraitBound(
                    predicate.trait_ref.lower_into(interner),


I think we need to start tracking this information in chalk, too

oli-obk · 2021-07-22T16:59:10Z

only got half way through, will look more tomorrow

compiler/rustc_typeck/src/check/compare_method.rs

fee1-dead · 2021-07-26T04:01:08Z

Aaaah why do we have these cursed tests of const + async

compiler/rustc_trait_selection/src/traits/query/evaluate_obligation.rs

oli-obk · 2021-07-26T13:56:36Z

@bors try @rust-timer queue

rust-timer · 2021-07-26T13:56:37Z

Awaiting bors try build completion.

@rustbot label: +S-waiting-on-perf

oli-obk · 2021-08-13T11:00:29Z

library/alloc/src/collections/btree/map.rs

-impl<K, V> BTreeMap<K, V> {
+// N.B. bound not on function to avoid unnecessary
+// const bound on `Ord`: see #87375
+impl<K: Ord, V> BTreeMap<K, V> {


this should not make a difference at all. All bounds on the impl must apply for functions defined inside, too.

if this truly fixes things, that is a bug. Instead we should use a ?const Ord bound

Hmmmm, are you saying bounds on impls should be treated as const bounds when the functions inside them are const? That would break a lot of existing code I think.

impl<T: Clone> MyType { fn a() { ... }; // T: Clone const fn b() { ... }; // T: const Clone? }

Only code using feature gates, stable code already doesn't allow this: https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=428faddc9d2bce802ab7423fdfa7db33

Wow, I really didn't know this. But can we use ?const inside liballoc? That feature is marked as incomplete.

hmm... I guess we are just going to break ppl using btreemap in const contexts? Or try to remove the bound in a separate PR first?

From the crater run I think not many people are using these functions. I don't think we should remove the bound because it might require more work when re-adding the bound back. I'm fine with breaking BTree{Map,Set}::new for now.

oli-obk · 2021-08-14T09:49:28Z

@bors r+ rollup=never

bors · 2021-08-14T09:49:29Z

📌 Commit 74627c1 has been approved by oli-obk

ghost · 2021-08-14T10:10:22Z

compiler/rustc_trait_selection/src/traits/select/mod.rs

@@ -293,6 +316,18 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
        self.infcx.tcx
    }

+    /// returns `true` if the predicate is considered `const` to


Suggested change

/// returns `true` if the predicate is considered `const` to

/// Returns `true` if the predicate is considered `const` to

I will change this after this lands or on this PR if CI fails

ghost · 2021-08-14T10:16:26Z

compiler/rustc_infer/src/traits/engine.rs

    fn select_where_possible(
        &mut self,
        infcx: &InferCtxt<'_, 'tcx>,
    ) -> Result<(), Vec<FulfillmentError<'tcx>>>;

+    // FIXME this should not provide a default body for chalk as chalk should be updated


Suggested change

// FIXME this should not provide a default body for chalk as chalk should be updated

// FIXME(fee1-dead): this should not provide a default body for chalk as chalk should be updated

ghost · 2021-08-14T10:16:48Z

compiler/rustc_traits/src/type_op.rs

+    fn prove_predicate(&mut self, mut predicate: Predicate<'tcx>) {
+        if let PredicateKind::Trait(mut tr) = predicate.kind().skip_binder() {
+            if let hir::Constness::Const = tr.constness {
+                // FIXME check if we actually want to prove const predicates inside AscribeUserType


Suggested change

// FIXME check if we actually want to prove const predicates inside AscribeUserType

// FIXME(fee1-dead): check if we actually want to prove const predicates inside AscribeUserType

ghost · 2021-08-14T10:17:03Z

library/alloc/tests/const_fns.rs

+#[allow(dead_code)]
+pub const MY_STRING: String = String::new();
+
+// FIXME remove this struct once we put `K: ?const Ord` on BTreeMap::new.


Suggested change

// FIXME remove this struct once we put `K: ?const Ord` on BTreeMap::new.

// FIXME(fee1-dead): remove this struct once we put `K: ?const Ord` on BTreeMap::new.

ghost · 2021-08-14T10:17:19Z

library/alloc/tests/lib.rs

@@ -23,6 +23,10 @@
 #![feature(slice_partition_dedup)]
 #![feature(vec_spare_capacity)]
 #![feature(string_remove_matches)]
+#![feature(const_btree_new)]
+#![feature(const_trait_impl)]
+// FIXME remove this when const_trait_impl is not incomplete anymore


Suggested change

// FIXME remove this when const_trait_impl is not incomplete anymore

// FIXME(fee1-dead): remove this when const_trait_impl is not incomplete anymore

bors · 2021-08-14T12:06:37Z

⌛ Testing commit 74627c1 with merge 136eaa1...

bors · 2021-08-14T14:52:21Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing 136eaa1 to master...

rust-highfive assigned oli-obk Jul 22, 2021

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jul 22, 2021

This comment has been minimized.

Sign in to view

oli-obk reviewed Jul 22, 2021

View reviewed changes

compiler/rustc_mir/src/transform/check_consts/validation.rs Outdated Show resolved Hide resolved

oli-obk reviewed Jul 22, 2021

View reviewed changes

compiler/rustc_trait_selection/src/traits/fulfill.rs Show resolved Hide resolved

oli-obk reviewed Jul 22, 2021

View reviewed changes

compiler/rustc_mir/src/borrow_check/type_check/mod.rs Show resolved Hide resolved

oli-obk reviewed Jul 22, 2021

View reviewed changes

compiler/rustc_trait_selection/src/traits/auto_trait.rs Show resolved Hide resolved

oli-obk reviewed Jul 22, 2021

View reviewed changes

compiler/rustc_trait_selection/src/traits/select/mod.rs Show resolved Hide resolved

oli-obk reviewed Jul 22, 2021

View reviewed changes

This comment has been minimized.

Sign in to view

oli-obk reviewed Jul 24, 2021

View reviewed changes

compiler/rustc_typeck/src/check/compare_method.rs Outdated Show resolved Hide resolved

fee1-dead force-pushed the move-constness-to-traitpred branch from 2ea3e1d to 663dfe6 Compare July 24, 2021 11:13

This comment has been minimized.

Sign in to view

fee1-dead force-pushed the move-constness-to-traitpred branch from 6e9dbcb to 52d8288 Compare July 26, 2021 03:16

This comment has been minimized.

Sign in to view

fee1-dead force-pushed the move-constness-to-traitpred branch from 9023ae5 to 5f19f12 Compare July 26, 2021 06:48

This comment has been minimized.

Sign in to view

fee1-dead marked this pull request as ready for review July 26, 2021 09:29

This comment has been minimized.

Sign in to view

oli-obk reviewed Jul 26, 2021

View reviewed changes

compiler/rustc_trait_selection/src/traits/query/evaluate_obligation.rs Outdated Show resolved Hide resolved

This comment has been minimized.

Sign in to view

oli-obk reviewed Aug 13, 2021

View reviewed changes

This comment has been minimized.

Sign in to view

fee1-dead force-pushed the move-constness-to-traitpred branch from fdc54d1 to 48a4868 Compare August 13, 2021 11:23

This comment has been minimized.

Sign in to view

Fix Cargo.lock and ui test

74627c1

fee1-dead force-pushed the move-constness-to-traitpred branch from 48a4868 to 74627c1 Compare August 13, 2021 11:31

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 14, 2021

ghost reviewed Aug 14, 2021

View reviewed changes

bors added the merged-by-bors This PR was explicitly merged by bors. label Aug 14, 2021

bors merged commit 136eaa1 into rust-lang:master Aug 14, 2021

rustbot added this to the 1.56.0 milestone Aug 14, 2021

Aaron1011 mentioned this pull request Aug 15, 2021

Copy not implemented for (u32, u32) error #88047

Closed

fee1-dead deleted the move-constness-to-traitpred branch August 16, 2021 11:56

steffahn mentioned this pull request Aug 27, 2021

False negative for "the trait bound K: Ord is not satisfied" #88244

Closed

JohnTitor added a commit to JohnTitor/rust-semverver that referenced this pull request Oct 13, 2021

Rustup to rust-lang/rust#87375

2d603a2

	/// returns `true` if the predicate is considered `const` to
	/// Returns `true` if the predicate is considered `const` to

	// FIXME this should not provide a default body for chalk as chalk should be updated
	// FIXME(fee1-dead): this should not provide a default body for chalk as chalk should be updated

	// FIXME check if we actually want to prove const predicates inside AscribeUserType
	// FIXME(fee1-dead): check if we actually want to prove const predicates inside AscribeUserType

	// FIXME remove this struct once we put `K: ?const Ord` on BTreeMap::new.
	// FIXME(fee1-dead): remove this struct once we put `K: ?const Ord` on BTreeMap::new.

	// FIXME remove this when const_trait_impl is not incomplete anymore
	// FIXME(fee1-dead): remove this when const_trait_impl is not incomplete anymore

Try filtering out non-const impls when we expect const impls #87375

Try filtering out non-const impls when we expect const impls #87375

Conversation

fee1-dead commented Jul 22, 2021 • edited

rust-highfive commented Jul 22, 2021

This comment has been minimized.

Choose a reason for hiding this comment

Choose a reason for hiding this comment

oli-obk commented Jul 22, 2021

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

fee1-dead commented Jul 26, 2021

This comment has been minimized.

This comment has been minimized.

This comment has been minimized.

oli-obk commented Jul 26, 2021

rust-timer commented Jul 26, 2021

This comment has been minimized.

This comment has been minimized.

Choose a reason for hiding this comment

fee1-dead Aug 13, 2021 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fee1-dead Aug 13, 2021 • edited

Choose a reason for hiding this comment

This comment has been minimized.

This comment has been minimized.

oli-obk commented Aug 14, 2021

bors commented Aug 14, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bors commented Aug 14, 2021

bors commented Aug 14, 2021

fee1-dead commented Jul 22, 2021 •

edited

fee1-dead Aug 13, 2021 •

edited

fee1-dead Aug 13, 2021 •

edited