Store static initializers in metadata instead of the MIR of statics.
This means that adding generic statics would be even more difficult, as we can't evaluate statics from other crates anymore, but the subtle issue I have encountered make me think that having this be an explicit problem is better.
The issue is that
```rust
static mut FOO: &mut u32 = &mut 42;
static mut BAR = unsafe { FOO };
```
gets different allocations, instead of referring to the same one. This is also true for non-static mut, but promotion makes `static FOO: &u32 = &42;` annoying to demo.
Fixes https://github.com/rust-lang/rust/issues/61345
## Why is this being done?
In order to ensure all crates see the same nested allocations (which is the last issue that needs fixing before we can stabilize [`const_mut_refs`](https://github.com/rust-lang/rust/issues/57349)), I am working on creating anonymous (from the Rust side, to LLVM it's like a regular static item) static items for the nested allocations in a static. If we evaluate the static item in a downstream crate again, we will end up duplicating its nested allocations (and in some cases, like the `match` case, even duplicate the main allocation).
Enforce coroutine-closure layouts are identical
Enforce that for an async closure, the by-ref and by-move coroutine layouts are identical. This is just a sanity check to make sure that optimizations aren't doing anything fishy.
r? oli-obk
Make sure `tcx.create_def` also depends on the forever red node, instead of just `tcx.at(span).create_def`
oversight from https://github.com/rust-lang/rust/pull/119136
Not actually an issue, because all uses of `tcx.create_def` were in the resolver, which is `eval_always`, but still good to harden against future uses of `create_def`
cc `@petrochenkov` `@WaffleLapkin`
Make sure `tcx.create_def` also depends on the forever red node, instead of just `tcx.at(span).create_def`
oversight from https://github.com/rust-lang/rust/pull/119136
Not actually an issue, because all uses of `tcx.create_def` were in the resolver, which is `eval_always`, but still good to harden against future uses of `create_def`
cc `@petrochenkov` `@WaffleLapkin`
Extend documentation for `Ty::to_opt_closure_kind` method
This API was... surprising to use. With a little extra documentation, the weirdness can be reduced quite a lot. :)
r? `@compiler-errors`
Uplift `TypeVisitableExt` into `rustc_type_ir`
This uplifts `TypeVisitableExt` into `rustc_type_ir` so it can be used in an interner-agnostic way. It also moves some `TypeSuperVisitable` bounds onto `Interner` since we don't expect to support libraries that have types which aren't foldable by default.
This restores a couple of asserts in the canonicalizer code we uplifted, and also makes it so that we can use type-flags-based helpers in the solver code, which I'm interested in uplifting.
r? lcnr
Fully stop using the HIR in trait impl checks
At least I hope I found all happy path usages. I'll need to check if I can figure out a way to make queries declare that they don't access the HIR except in error paths
Properly handle `async` block and `async fn` in `if` exprs without `else`
When encountering a tail expression in the then arm of an `if` expression without an `else` arm, account for `async fn` and `async` blocks to suggest `return`ing the value and pointing at the return type of the `async fn`.
We now also account for AFIT when looking for the return type to point at.
Fix#115405.
Merge `impl_polarity` and `impl_trait_ref` queries
Hopefully this is perf neutral. I want to finish https://github.com/rust-lang/rust/pull/120835 and stop using the HIR in `coherent_trait`, which should then give us a perf improvement.
It's only has a single remaining purpose: to ensure that a diagnostic is
printed when `trimmed_def_paths` is used. It's an annoying mechanism:
weak, with odd semantics, badly named, and gets in the way of other
changes.
This commit replaces it with a simpler `must_produce_diag` mechanism,
getting rid of a diagnostic `Level` along the way.
Dejargonize `subst`
In favor of #110793, replace almost every occurence of `subst` and `substitution` from rustc codes, but they still remains in subtrees under `src/tools/` like clippy and test codes (I'd like to replace them after this)
Fix async closures in CTFE
First commit renames `is_coroutine_or_closure` into `is_closure_like`, because `is_coroutine_or_closure_or_coroutine_closure` seems confusing and long.
Second commit fixes some forgotten cases where we want to handle `TyKind::CoroutineClosure` the same as closures and coroutines.
The test exercises the change to `ValidityVisitor::aggregate_field_path_elem` which is the source of #120946, but not the change to `UsedParamsNeedSubstVisitor`, though I feel like it's not that big of a deal. Let me know if you'd like for me to look into constructing a test for the latter, though I have no idea what it'd look like (we can't assert against `TooGeneric` anywhere?).
Fixes#120946
r? oli-obk cc ``@RalfJung``
When encountering a tail expression in the then arm of an `if` expression
without an `else` arm, account for `async fn` and `async` blocks to
suggest `return`ing the value and pointing at the return type of the
`async fn`.
We now also account for AFIT when looking for the return type to point at.
Fix#115405.
Rollup of 11 pull requests
Successful merges:
- #120765 (Reorder diagnostics API)
- #120833 (More internal emit diagnostics cleanups)
- #120899 (Gracefully handle non-WF alias in `assemble_alias_bound_candidates_recur`)
- #120917 (Remove a bunch of dead parameters in functions)
- #120928 (Add test for recently fixed issue)
- #120933 (check_consts: fix duplicate errors, make importance consistent)
- #120936 (improve `btree_cursors` functions documentation)
- #120944 (Check that the ABI of the instance we are inlining is correct)
- #120956 (Clean inlined type alias with correct param-env)
- #120962 (Add myself to library/std review)
- #120972 (fix ICE for deref coercions with type errors)
r? `@ghost`
`@rustbot` modify labels: rollup
Assert that params with the same *index* have the same *name*
Found this bug when trying to build libcore with the new solver, since it will canonicalize two params with the same index into *different* placeholders if those params differ by name.
Print kind of coroutine closure
Make sure that we print "async closure" when we have an async closure, rather than calling it generically a ["coroutine-closure"](https://github.com/rust-lang/rust/pull/120361).
Fixes#120886
r? oli-obk
Remove the FIXME and keep `CRATE_HIR_ID` being its own parent.
This scheme turned out to be more practical than having an `Option` on closer inspection.
Also make `hir_owner_parent` more readable.
improve normalization of `Pointee::Metadata`
This PR makes it so that `<Wrapper<Tail> as Pointee>::Metadata` is normalized to `<Tail as Pointee>::Metadata` if we don't know `Wrapper<Tail>: Sized`. With that, the trait solver can prove projection predicates like `<Wrapper<Tail> as Pointee>::Metadata == <Tail as Pointee>::Metadata`, which makes it possible to use the metadata APIs to cast between the tail and the wrapper:
```rust
#![feature(ptr_metadata)]
use std::ptr::{self, Pointee};
fn cast_same_meta<T: ?Sized, U: ?Sized>(ptr: *const T) -> *const U
where
T: Pointee<Metadata = <U as Pointee>::Metadata>,
{
let (thin, meta) = ptr.to_raw_parts();
ptr::from_raw_parts(thin, meta)
}
struct Wrapper<T: ?Sized>(T);
fn cast_to_wrapper<T: ?Sized>(ptr: *const T) -> *const Wrapper<T> {
cast_same_meta(ptr)
}
```
Previously, this failed to compile:
```
error[E0271]: type mismatch resolving `<Wrapper<T> as Pointee>::Metadata == <T as Pointee>::Metadata`
--> src/lib.rs:16:5
|
15 | fn cast_to_wrapper<T: ?Sized>(ptr: *const T) -> *const Wrapper<T> {
| - found this type parameter
16 | cast_same_meta(ptr)
| ^^^^^^^^^^^^^^ expected `Wrapper<T>`, found type parameter `T`
|
= note: expected associated type `<Wrapper<T> as Pointee>::Metadata`
found associated type `<T as Pointee>::Metadata`
= note: an associated type was expected, but a different one was found
```
(Yes, you can already do this with `as` casts. But using functions is so much ✨ *safer* ✨, because you can't change the metadata on accident.)
---
This PR essentially changes the built-in impls of `Pointee` from this:
```rust
// before
impl Pointee for u8 {
type Metadata = ();
}
impl Pointee for [u8] {
type Metadata = usize;
}
// ...
impl Pointee for Wrapper<u8> {
type Metadata = ();
}
impl Pointee for Wrapper<[u8]> {
type Metadata = usize;
}
// ...
// This impl is only selected if `T` is a type parameter or unnormalizable projection or opaque type.
fallback impl<T: ?Sized> Pointee for Wrapper<T>
where
Wrapper<T>: Sized
{
type Metadata = ();
}
// This impl is only selected if `T` is a type parameter or unnormalizable projection or opaque type.
fallback impl<T /*: Sized */> Pointee for T {
type Metadata = ();
}
```
to this:
```rust
// after
impl Pointee for u8 {
type Metadata = ();
}
impl Pointee for [u8] {
type Metadata = usize;
}
// ...
impl<T: ?Sized> Pointee for Wrapper<T> {
// in the old solver this will instead project to the "deep" tail directly,
// e.g. `Wrapper<Wrapper<T>>::Metadata = T::Metadata`
type Metadata = <T as Pointee>::Metadata;
}
// ...
// This impl is only selected if `T` is a type parameter or unnormalizable projection or opaque type.
fallback impl<T /*: Sized */> Pointee for T {
type Metadata = ();
}
```
Invert diagnostic lints.
That is, change `diagnostic_outside_of_impl` and `untranslatable_diagnostic` from `allow` to `deny`, because more than half of the compiler has been converted to use translated diagnostics.
This commit removes more `deny` attributes than it adds `allow` attributes, which proves that this change is warranted.
r? ````@davidtwco````
Toggle assert_unsafe_precondition in codegen instead of expansion
The goal of this PR is to make some of the unsafe precondition checks in the standard library available in debug builds. Some UI tests are included to verify that it does that.
The diff is large, but most of it is blessing mir-opt tests and I've also split up this PR so it can be reviewed commit-by-commit.
This PR:
1. Adds a new intrinsic, `debug_assertions` which is lowered to a new MIR NullOp, and only to a constant after monomorphization
2. Rewrites `assume_unsafe_precondition` to check the new intrinsic, and be monomorphic.
3. Skips codegen of the `assume` intrinsic in unoptimized builds, because that was silly before but with these checks it's *very* silly
4. The checks with the most overhead are `ptr::read`/`ptr::write` and `NonNull::new_unchecked`. I've simply added `#[cfg(debug_assertions)]` to the checks for `ptr::read`/`ptr::write` because I was unable to come up with any (good) ideas for decreasing their impact. But for `NonNull::new_unchecked` I found that the majority of callers can use a different function, often a safe one.
Yes, this PR slows down the compile time of some programs. But in our benchmark suite it's never more than 1% icount, and the average icount change in debug-full programs is 0.22%. I think that is acceptable for such an improvement in developer experience.
https://github.com/rust-lang/rust/issues/120539#issuecomment-1922687101
Remove unused args from functions
`#[instrument]` suppresses the unused arguments from a function, *and* suppresses unused methods too! This PR removes things which are only used via `#[instrument]` calls, and fixes some other errors (privacy?) that I will comment inline.
It's possible that some of these arguments were being passed in for the purposes of being instrumented, but I am unconvinced by most of them.
Introduce `enter_forall` to supercede `instantiate_binder_with_placeholders`
r? `@lcnr`
Long term we'd like to experiment with decrementing the universe count after "exiting" binders so that we do not end up creating infer vars in non-root universes even when they logically reside in the root universe. The fact that we dont do this currently results in a number of issues in the new trait solver where we consider goals to be ambiguous because otherwise it would require lowering the universe of an infer var. i.e. the goal `?x.0 eq <T as Trait<?y.1>>::Assoc` where the alias is rigid would not be able to instantiate `?x` with the alias as there would be a universe error.
This PR is the first-ish sort of step towards being able to implement this as eventually we would want to decrement the universe in `enter_forall`. Unfortunately its Difficult to actually implement decrementing universes nicely so this is a separate step which moves us closer to the long term goal ✨
improve pretty printing for associated items in trait objects
* Don't print a binder in front of associated items, because it's not valid syntax.
* e.g. print `dyn for<'a> Trait<'a, Assoc = &'a u8>` instead of `dyn for<'a> Trait<'a, for<'a> Assoc = &'a u8>`.
* Don't print associated items that are implied by a supertrait bound.
* e.g. if we have `trait Sub: Super<Assoc = u8> {}`, then just print `dyn Sub` instead of `dyn Sub<Assoc = u8>`.
I've added the test in the first commit, so you can see the diff of the compiler output in the second commit.
Reconstify `Add`
r? project-const-traits
I'm not happy with the ui test changes (or failures because I did not bless them and include the diffs in this PR). There is at least some bugs I need to look and try fix:
1. A third duplicated diagnostic when a consumer crate that does not have `effects` enabled has a trait selection error for an upstream const_trait trait. See tests/ui/ufcs/ufcs-qpath-self-mismatch.rs.
2. For some reason, making `Add` a const trait would stop us from suggesting `T: Add` when we try to add two `T`s without that bound. See tests/ui/suggestions/issue-97677.rs
Record coroutine kind in coroutine generics
Oops, added a new substitution (the "kind" ty) to coroutines but forgot to record it in the `generics_of`. I'm surprised I left this out of the coroutine-closure PR -- I thought I made this change; I possibly rebased it out by accident.
Fixes#120732
r? oli-obk
MirPass: make name more const
Continues #120161, this time applied to `MirPass` instead of `MirLint`, locally shaves few (very few) instructions off.
r? ``@cjgillot``
Don't expect early-bound region to be local when reporting errors in RPITIT well-formedness
The implicit lifetime in the example code gets replaced with `ReError`, which fails a `sub_regions` check in the lexical region solver. Error reporting ends up calling `is_suitable_region` on an early bound region in the *trait* definition. This causes an ICE because we `expect_local()`.
This is kind of a bad explanation, but this code just makes diagnostics reporting a bit more gracefully fallible. If the reviewer wants a thorough investigation of exactly where we get this region outlives obligation, I can write one up. Doesn't really seem worth it, though, imo.
Fixes#120638Fixes#120648
update indirect structural match lints to match RFC and to show up for dependencies
This is a large step towards implementing https://github.com/rust-lang/rfcs/pull/3535.
We currently have five lints related to "the structural match situation":
- nontrivial_structural_match
- indirect_structural_match
- pointer_structural_match
- const_patterns_without_partial_eq
- illegal_floating_point_literal_pattern
This PR concerns the first 3 of them. (The 4th already is set up to show for dependencies, and the 5th is removed by https://github.com/rust-lang/rust/pull/116284.) nontrivial_structural_match is being removed as per the RFC; the other two are enabled to show up in dependencies.
Fixes https://github.com/rust-lang/rust/issues/73448 by removing the affected analysis.
That is, change `diagnostic_outside_of_impl` and
`untranslatable_diagnostic` from `allow` to `deny`, because more than
half of the compiler has be converted to use translated diagnostics.
This commit removes more `deny` attributes than it adds `allow`
attributes, which proves that this change is warranted.
make matching on NaN a hard error, and remove the rest of illegal_floating_point_literal_pattern
These arms would never be hit anyway, so the pattern makes little sense. We have had a future-compat lint against float matches in general for a *long* time, so I hope we can get away with immediately making this a hard error.
This is part of implementing https://github.com/rust-lang/rfcs/pull/3535.
Closes https://github.com/rust-lang/rust/issues/41620 by removing the lint.
https://github.com/rust-lang/reference/pull/1456 updates the reference to match.
Move predicate, region, and const stuff into their own modules in middle
This PR mostly moves things around, and in a few cases adds some `ty::` to the beginning of names to avoid one-off imports.
I don't mean this to be the most *thorough* move/refactor. I just generally wanted to begin to split up `ty/mod.rs` and `ty/sty.rs` which are huge and hard to distinguish, and have a lot of non-ty stuff in them.
r? lcnr
it works when a non-const context that does not enable effects
calls into a const effects-enabled trait. We'd simply suggest the
non-const trait bound in this case consistent to its fallback.
hir: Remove the generic type parameter from `MaybeOwned`
It's only ever used with a reference to `OwnerInfo` as an argument.
Follow up to https://github.com/rust-lang/rust/pull/120346.
Deduplicate more sized errors on call exprs
Change the implicit `Sized` `Obligation` `Span` for call expressions to include the whole expression. This aids the existing deduplication machinery to reduce the number of errors caused by a single unsized expression.
The query accept arbitrary DefIds, not just owner DefIds.
The return can be an `Option` because if there are no nodes, then it doesn't matter whether it's due to NonOwner or Phantom.
Also rename the query to `opt_hir_owner_nodes`.
Because it's almost always static.
This makes `impl IntoDiagnosticArg for DiagnosticArgValue` trivial,
which is nice.
There are a few diagnostics constructed in
`compiler/rustc_mir_build/src/check_unsafety.rs` and
`compiler/rustc_mir_transform/src/errors.rs` that now need symbols
converted to `String` with `to_string` instead of `&str` with `as_str`,
but that' no big deal, and worth it for the simplifications elsewhere.
Error codes are integers, but `String` is used everywhere to represent
them. Gross!
This commit introduces `ErrCode`, an integral newtype for error codes,
replacing `String`. It also introduces a constant for every error code,
e.g. `E0123`, and removes the `error_code!` macro. The constants are
imported wherever used with `use rustc_errors::codes::*`.
With the old code, we have three different ways to specify an error code
at a use point:
```
error_code!(E0123) // macro call
struct_span_code_err!(dcx, span, E0123, "msg"); // bare ident arg to macro call
\#[diag(name, code = "E0123")] // string
struct Diag;
```
With the new code, they all use the `E0123` constant.
```
E0123 // constant
struct_span_code_err!(dcx, span, E0123, "msg"); // constant
\#[diag(name, code = E0123)] // constant
struct Diag;
```
The commit also changes the structure of the error code definitions:
- `rustc_error_codes` now just defines a higher-order macro listing the
used error codes and nothing else.
- Because that's now the only thing in the `rustc_error_codes` crate, I
moved it into the `lib.rs` file and removed the `error_codes.rs` file.
- `rustc_errors` uses that macro to define everything, e.g. the error
code constants and the `DIAGNOSTIC_TABLES`. This is in its new
`codes.rs` file.
ScopeTree: remove destruction_scopes as unused
last usages removed by https://github.com/rust-lang/rust/pull/116170
Unused, but still presented in memory at `t-gmax` (in DHAT termonology)
Remove unused/unnecessary features
~~The bulk of the actual code changes here is replacing try blocks with equivalent closures. I'm not entirely sure that's a good idea since it may have perf impact, happy to revert if that's the case/the change is unwanted.~~
I also removed a lot of `recursion_limit = "256"` since everything seems to build fine without that and most don't have any comment justifying it.
remove StructuralEq trait
The documentation given for the trait is outdated: *all* function pointers implement `PartialEq` and `Eq` these days. So the `StructuralEq` trait doesn't really seem to have any reason to exist any more.
One side-effect of this PR is that we allow matching on some consts that do not implement `Eq`. However, we already allowed matching on floats and consts containing floats, so this is not new, it is just allowed in more cases now. IMO it makes no sense at all to allow float matching but also sometimes require an `Eq` instance. If we want to require `Eq` we should adjust https://github.com/rust-lang/rust/pull/115893 to check for `Eq`, and rule out float matching for good.
Fixes https://github.com/rust-lang/rust/issues/115881
Remove coroutine info when building coroutine drop body
Coroutine drop shims are not themselves coroutines, so erase the "`coroutine`" field from the body so that helper fns like `yield_ty` and `coroutine_kind` properly return `None` for the drop shim.
Rollup of 10 pull requests
Successful merges:
- #119305 (Add `AsyncFn` family of traits)
- #119389 (Provide more context on recursive `impl` evaluation overflow)
- #119895 (Remove `track_errors` entirely)
- #120230 (Assert that a single scope is passed to `for_scope`)
- #120278 (Remove --fatal-warnings on wasm targets)
- #120292 (coverage: Dismantle `Instrumentor` and flatten span refinement)
- #120315 (On E0308 involving `dyn Trait`, mention trait objects)
- #120317 (pattern_analysis: Let `ctor_sub_tys` return any Iterator they want)
- #120318 (pattern_analysis: Reuse most of the `DeconstructedPat` `Debug` impl)
- #120325 (rustc_data_structures: use either instead of itertools)
r? `@ghost`
`@rustbot` modify labels: rollup
Remove `track_errors` entirely
follow up to https://github.com/rust-lang/rust/pull/119869
r? `@matthewjasper`
There are some diagnostic changes adding new diagnostics or not emitting some anymore. We can improve upon that in follow-up work imo.
Modify GenericArg and Term structs to use strict provenance rules
This is the first PR to solve issue #119217 . In this PR, I have modified the GenericArg struct to use the `NonNull` struct as the pointer instead of `NonZeroUsize`. The change were tested by running `./x test compiler/rustc_middle`.
Resolves https://github.com/rust-lang/rust/issues/119217
r? `@WaffleLapkin`
Remove all ConstPropNonsense
We track all locals and projections on them ourselves within the const propagator and only use the InterpCx to actually do some low level operations or read from constants (via `OpTy` we get for said constants).
This helps moving the const prop lint out from the normal pipeline and running it just based on borrowck information. This in turn allows us to make progress on https://github.com/rust-lang/rust/pull/108730#issuecomment-1875557745
there are various follow up cleanups that can be done after this PR (e.g. not matching on Rvalue twice and doing binop checks twice), but lets try landing this one first.
r? `@RalfJung`
Change the implicit `Sized` `Obligation` `Span` for call expressions to
include the whole expression. This aids the existing deduplication
machinery to reduce the number of errors caused by a single unsized
expression.
Rollup of 9 pull requests
Successful merges:
- #112806 (Small code improvements in `collect_intra_doc_links.rs`)
- #119766 (Split tait and impl trait in assoc items logic)
- #120139 (Do not normalize closure signature when building `FnOnce` shim)
- #120160 (Manually implement derived `NonZero` traits.)
- #120171 (Fix assume and assert in jump threading)
- #120183 (Add `#[coverage(off)]` to closures introduced by `#[test]` and `#[bench]`)
- #120195 (add several resolution test cases)
- #120259 (Split Diagnostics for Uncommon Codepoints: Add List to Display Characters Involved)
- #120261 (Provide structured suggestion to use trait objects in some cases of `if` arm type divergence)
r? `@ghost`
`@rustbot` modify labels: rollup
Do not normalize closure signature when building `FnOnce` shim
It is not necessary to normalize the closure signature when building an `FnOnce` shim for an `Fn`/`FnMut` closure. That closure shim is just calling `FnMut::call_mut(&mut self)` anyways.
It's also somewhat sketchy that we were ever doing this to begin with, since we're normalizing with a `ParamEnv::reveal_all()` param-env, which is definitely not right with possibly polymorphic substs.
This cuts out a tiny bit of unnecessary work in `Instance::resolve` and simplifies the signature because now we can unconditionally return an `Instance`.
A bunch of random modifications
r? oli-obk
Kitchen sink of changes that I didn't know where to put elsewhere. Documentation tweaks mostly, but also removing some unreachable code and simplifying the pretty printing for closures/coroutines.
const-eval interning: get rid of type-driven traversal
This entirely replaces our const-eval interner, i.e. the code that takes the final result of a constant evaluation from the local memory of the const-eval machine to the global `tcx` memory. The main goal of this change is to ensure that we can detect mutable references that sneak into this final value -- this is something we want to reject for `static` and `const`, and while const-checking performs some static analysis to ensure this, I would be much more comfortable stabilizing const_mut_refs if we had a dynamic check that sanitizes the final value. (This is generally the approach we have been using on const-eval: do a static check to give nice errors upfront, and then do a dynamic check to be really sure that the properties we need for soundness, actually hold.)
We can do this now that https://github.com/rust-lang/rust/pull/118324 landed and each pointer comes with a bit (completely independent of its type) storing whether mutation is permitted through this pointer or not.
The new interner is a lot simpler than the old one: previously we did a complete type-driven traversal to determine the mutability of all memory we see, and then a second pass to intern any leftover raw pointers. The new interner simply recursively traverses the allocation holding the final result, and all allocations reachable from it (which can be determined from the raw bytes of the result, without knowing anything about types), and ensures they all get interned. The initial allocation is interned as immutable for `const` and pomoted and non-interior-mutable `static`; all other allocations are interned as immutable for `static`, `const`, and promoted. The main subtlety is justifying that those inner allocations may indeed be interned immutably, i.e., that mutating them later would anyway already be UB:
- for promoteds, we rely on the analysis that does promotion to ensure that this is sound.
- for `const` and `static`, we check that all pointers in the final result that point to things that are new (i.e., part of this const evaluation) are immutable, i.e., were created via `&<expr>` at a non-interior-mutable type. Mutation through immutable pointers is UB so we are free to intern that memory as immutable.
Interning raises an error if it encounters a dangling pointer or a mutable pointer that violates the above rules.
I also extended our type-driven const validity checks to ensure that `&mut T` in the final value of a const points to mutable memory, at least if `T` is not zero-sized. This catches cases of people turning `&i32` into `&mut i32` (which would still be considered a read-only pointer). Similarly, when these checks encounter an `UnsafeCell`, they are checking that it lives in mutable memory. (Both of these only traverse the newly created values; if those point to other consts/promoteds, the check stops there. But that's okay, we don't have to catch all the UB.) I co-developed this with the stricter interner changes but I can split it out into a separate PR if you prefer.
This PR does have the immediate effect of allowing some new code on stable, for instance:
```rust
const CONST_RAW: *const Vec<i32> = &Vec::new() as *const _;
```
Previously that code got rejected since the type-based interner didn't know what to do with that pointer. It's a raw pointer, we cannot trust its type. The new interner does not care about types so it sees no issue with this code; there's an immutable pointer pointing to some read-only memory (storing a `Vec<i32>`), all is good. Accepting this code pretty much commits us to non-type-based interning, but I think that's the better strategy anyway.
This PR also leads to slightly worse error messages when the final value of a const contains a dangling reference. Previously we would complete interning and then the type-based validation would detect this dangling reference and show a nice error saying where in the value (i.e., in which field) the dangling reference is located. However, the new interner cannot distinguish dangling references from dangling raw pointers, so it must throw an error when it encounters either of them. It doesn't have an understanding of the value structure so all it can say is "somewhere in this constant there's a dangling pointer". (Later parts of the compiler don't like dangling pointers/references so we have to reject them either during interning or during validation.) This could potentially be improved by doing validation before interning, but that's a larger change that I have not attempted yet. (It's also subtle since we do want validation to use the final mutability bits of all involved allocations, and currently it is interning that marks a bunch of allocations as immutable -- that would have to still happen before validation.)
`@rust-lang/wg-const-eval` I hope you are okay with this plan. :)
`@rust-lang/lang` paging you in since this accepts new code on stable as explained above. Please let me know if you think FCP is necessary.
