Remove some usages of `guess_head_span`
No need to pass things through `guess_head_span` if they already point to the head span.
Only major change is that we point to the head span of `enum`s on some errors now, which I prefer.
r? `@cjgillot`
Move abstract const to middle
Moves AbstractConst (and all associated methods) to rustc middle for use in `rustc_infer`.
This allows for const resolution in infer to use abstract consts to walk consts and check if
they are resolvable.
This attempts to resolve the issue where `Foo<{ concrete const }, generic T>` is incorrectly marked as conflicting, and is independent from the other issue where nested abstract consts must be resolved.
r? `@lcnr`
Implement `for<>` lifetime binder for closures
This PR implements RFC 3216 ([TI](https://github.com/rust-lang/rust/issues/97362)) and allows code like the following:
```rust
let _f = for<'a, 'b> |a: &'a A, b: &'b B| -> &'b C { b.c(a) };
// ^^^^^^^^^^^--- new!
```
cc ``@Aaron1011`` ``@cjgillot``
don't use `commit_if_ok` during `higher_ranked_sub`
This snapshot doesn't really do anything useful for us, especially once we deal with placeholder outlive bounds during trait solving.
I guess that currently the idea is that `higher_ranked_sub` could cause a later `leak_check` to fail even if the combine operation isn't actually relevant. But really, using combine outside of snapshot and ignoring its result is wrong anyways, as it can constrain inference variables.
r? rust-lang/types
don't succeed `evaluate_obligation` query if new opaque types were registered
fixes#98608fixes#98604
The root cause of all this is that in type flag computation we entirely ignore nongeneric things like struct fields and the signature of function items. So if a flag had to be set for a struct if it is set for a field, that will only happen if the field is generic, as only the generic parameters are checked.
I now believe we cannot use type flags to handle opaque types. They seem like the wrong tool for this.
Instead, this PR replaces the previous logic by adding a new variant of `EvaluatedToOk`: `EvaluatedToOkModuloOpaqueTypes`, which says that there were some opaque types that got hidden types bound, but that binding may not have been legal (because we don't know if the opaque type was in its defining scope or not).
Make TAIT behave exactly like RPIT
fixes https://github.com/rust-lang/rust/issues/96552
This makes type-alias-impl-trait behave like return-position-impl-trait. Unfortunately it also causes some cases to stop compiling due to "needing type annotations" and makes panicking cause fallback for the hidden type to `()`.
All of these are addressable, but we should probably address them for RPIT and TAIT together
r? ``@lcnr``
Avoid some `&str` to `String` conversions with `MultiSpan::push_span_label`
This patch removes some`&str` to `String` conversions with `MultiSpan::push_span_label`.
Clean up arg mismatch diagnostic, generalize tuple wrap suggestion
This is based on top of #97542, so just look at the last commit which contains the relevant changes.
1. Remove `final_arg_types` which was one of the last places we were using raw (`usize`) indices instead of typed indices in the arg mismatch suggestion code.
2. Improve the tuple wrap suggestion, now we suggest things like `call(a, b, c, d)` -> `call(a, (b, c), d)` 😺
3. Folded in fix#98645
Reverse folder hierarchy
#91318 introduced a trait for infallible folders distinct from the fallible version. For some reason (completely unfathomable to me now that I look at it with fresh eyes), the infallible trait was a supertrait of the fallible one: that is, all fallible folders were required to also be infallible. Moreover the `Error` associated type was defined on the infallible trait! It's so absurd that it has me questioning whether I was entirely sane.
This trait reverses the hierarchy, so that the fallible trait is a supertrait of the infallible one: all infallible folders are required to also be fallible (which is a trivial blanket implementation). This of course makes much more sense! It also enables the `Error` associated type to sit on the fallible trait, where it sensibly belongs.
There is one downside however: folders expose a `tcx` accessor method. Since the blanket fallible implementation for infallible folders only has access to a generic `F: TypeFolder`, we need that trait to expose such an accessor to which we can delegate. Alternatively it's possible to extract that accessor into a separate `HasTcx` trait (or similar) that would then be a supertrait of both the fallible and infallible folder traits: this would ensure that there's only one unambiguous `tcx` method, at the cost of a little additional boilerplate. If desired, I can submit that as a separate PR.
r? ````@jackh726````
fix universes in the NLL type tests
In the NLL code, we were not accommodating universes in the
`type_test` logic.
Fixes#98095.
r? `@compiler-errors`
This breaks some tests, however, so the purpose of this branch is more explanatory and perhaps to do a crater run.
Point at return expression for RPIT-related error
Certainly this needs some diagnostic refining, but I wanted to show that it was possible first and foremost. Not sure if this is the right approach. Open to feedback.
Fixes#80583
lub: don't bail out due to empty binders
allows for the following to compile. The equivalent code using `struct Wrapper<'upper>(fn(&'upper ());` already compiles on stable.
```rust
let _: fn(&'upper ()) = match v {
true => lt_in_fn::<'a>(),
false => lt_in_fn::<'b>(),
};
```
see https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=7034a677190110941223cafac6632f70 for a complete example
r? ```@rust-lang/types```
#91318 introduced a trait for infallible folders distinct from the fallible version. For some reason (completely unfathomable to me now that I look at it with fresh eyes), the infallible trait was a supertrait of the fallible one: that is, all fallible folders were required to also be infallible. Moreover the `Error` associated type was defined on the infallible trait! It's so absurd that it has me questioning whether I was entirely sane.
This trait reverses the hierarchy, so that the fallible trait is a supertrait of the infallible one: all infallible folders are required to also be fallible (which is a trivial blanket implementation). This of course makes much more sense! It also enables the `Error` associated type to sit on the fallible trait, where it sensibly belongs.
There is one downside however: folders expose a `tcx` accessor method. Since the blanket fallible implementation for infallible folders only has access to a generic `F: TypeFolder`, we need that trait to expose such an accessor to which we can delegate. Alternatively it's possible to extract that accessor into a separate `HasTcx` trait (or similar) that would then be a supertrait of both the fallible and infallible folder traits: this would ensure that there's only one unambiguous `tcx` method, at the cost of a little additional boilerplate. If desired, I can submit that as a separate PR.
r? @jackh726
The code now accepts `Binder<OutlivesPredicate>`
instead of just `OutlivesPredicate` and thus exercises
the new, generalized `IfEqBound` codepaths. Note though
that we never *produce* Binder<OutlivesPredicate>, so we
are only testing a subset of those codepaths that excludes
actual higher-ranked outlives bounds.
Make `ExprKind::Closure` a struct variant.
Simple refactor since we both need it to introduce additional fields in `ExprKind::Closure`.
r? ``@Aaron1011``
Rename the `ConstS::val` field as `kind`.
And likewise for the `Const::val` method.
Because its type is called `ConstKind`. Also `val` is a confusing name
because `ConstKind` is an enum with seven variants, one of which is
called `Value`. Also, this gives consistency with `TyS` and `PredicateS`
which have `kind` fields.
The commit also renames a few `Const` variables from `val` to `c`, to
avoid confusion with the `ConstKind::Value` variant.
r? `@BoxyUwU`
Remove RegionckMode in favor of calling new skip_region_resolution
Simple cleanup. We can skip a bunch of stuff for places where NLL does the region checking, so skip earlier.
r? rust-lang/types
And likewise for the `Const::val` method.
Because its type is called `ConstKind`. Also `val` is a confusing name
because `ConstKind` is an enum with seven variants, one of which is
called `Value`. Also, this gives consistency with `TyS` and `PredicateS`
which have `kind` fields.
The commit also renames a few `Const` variables from `val` to `c`, to
avoid confusion with the `ConstKind::Value` variant.
Mention `infer::Trace` methods on `infer::At` methods' docs
I missed that you could do `infcx.at(...).trace(...).eq(a, b)` when `a` and `b` dont implement `ToTrace` but does implement `Relate` these docs would have helped see that 😅
This commit makes type folding more like the way chalk does it.
Currently, `TypeFoldable` has `fold_with` and `super_fold_with` methods.
- `fold_with` is the standard entry point, and defaults to calling
`super_fold_with`.
- `super_fold_with` does the actual work of traversing a type.
- For a few types of interest (`Ty`, `Region`, etc.) `fold_with` instead
calls into a `TypeFolder`, which can then call back into
`super_fold_with`.
With the new approach, `TypeFoldable` has `fold_with` and
`TypeSuperFoldable` has `super_fold_with`.
- `fold_with` is still the standard entry point, *and* it does the
actual work of traversing a type, for all types except types of
interest.
- `super_fold_with` is only implemented for the types of interest.
Benefits of the new model.
- I find it easier to understand. The distinction between types of
interest and other types is clearer, and `super_fold_with` doesn't
exist for most types.
- With the current model is easy to get confused and implement a
`super_fold_with` method that should be left defaulted. (Some of the
precursor commits fixed such cases.)
- With the current model it's easy to call `super_fold_with` within
`TypeFolder` impls where `fold_with` should be called. The new
approach makes this mistake impossible, and this commit fixes a number
of such cases.
- It's potentially faster, because it avoids the `fold_with` ->
`super_fold_with` call in all cases except types of interest. A lot of
the time the compile would inline those away, but not necessarily
always.
Remove migrate borrowck mode
Closes#58781Closes#43234
# Stabilization proposal
This PR proposes the stabilization of `#![feature(nll)]` and the removal of `-Z borrowck`. Current borrow checking behavior of item bodies is currently done by first infering regions *lexically* and reporting any errors during HIR type checking. If there *are* any errors, then MIR borrowck (NLL) never occurs. If there *aren't* any errors, then MIR borrowck happens and any errors there would be reported. This PR removes the lexical region check of item bodies entirely and only uses MIR borrowck. Because MIR borrowck could never *not* be run for a compiled program, this should not break any programs. It does, however, change diagnostics significantly and allows a slightly larger set of programs to compile.
Tracking issue: #43234
RFC: https://github.com/rust-lang/rfcs/blob/master/text/2094-nll.md
Version: 1.63 (2022-06-30 => beta, 2022-08-11 => stable).
## Motivation
Over time, the Rust borrow checker has become "smarter" and thus allowed more programs to compile. There have been three different implementations: AST borrowck, MIR borrowck, and polonius (well, in progress). Additionally, there is the "lexical region resolver", which (roughly) solves the constraints generated through HIR typeck. It is not a full borrow checker, but does emit some errors.
The AST borrowck was the original implementation of the borrow checker and was part of the initially stabilized Rust 1.0. In mid 2017, work began to implement the current MIR borrow checker and that effort ompleted by the end of 2017, for the most part. During 2018, efforts were made to migrate away from the AST borrow checker to the MIR borrow checker - eventually culminating into "migrate" mode - where HIR typeck with lexical region resolving following by MIR borrow checking - being active by default in the 2018 edition.
In early 2019, migrate mode was turned on by default in the 2015 edition as well, but with MIR borrowck errors emitted as warnings. By late 2019, these warnings were upgraded to full errors. This was followed by the complete removal of the AST borrow checker.
In the period since, various errors emitted by the MIR borrow checker have been improved to the point that they are mostly the same or better than those emitted by the lexical region resolver.
While there do remain some degradations in errors (tracked under the [NLL-diagnostics tag](https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3ANLL-diagnostics), those are sufficiently small and rare enough that increased flexibility of MIR borrow check-only is now a worthwhile tradeoff.
## What is stabilized
As said previously, this does not fundamentally change the landscape of accepted programs. However, there are a [few](https://github.com/rust-lang/rust/issues?q=is%3Aopen+is%3Aissue+label%3ANLL-fixed-by-NLL) cases where programs can compile under `feature(nll)`, but not otherwise.
There are two notable patterns that are "fixed" by this stabilization. First, the `scoped_threads` feature, which is a continutation of a pre-1.0 API, can sometimes emit a [weird lifetime error](https://github.com/rust-lang/rust/issues/95527) without NLL. Second, actually seen in the standard library. In the `Extend` impl for `HashMap`, there is an implied bound of `K: 'a` that is available with NLL on but not without - this is utilized in the impl.
As mentioned before, there are a large number of diagnostic differences. Most of them are better, but some are worse. None are serious or happen often enough to need to block this PR. The biggest change is the loss of error code for a number of lifetime errors in favor of more general "lifetime may not live long enough" error. While this may *seem* bad, the former error codes were just attempts to somewhat-arbitrarily bin together lifetime errors of the same type; however, on paper, they end up being roughly the same with roughly the same kinds of solutions.
## What isn't stabilized
This PR does not completely remove the lexical region resolver. In the future, it may be possible to remove that (while still keeping HIR typeck) or to remove it together with HIR typeck.
## Tests
Many test outputs get updated by this PR. However, there are number of tests specifically geared towards NLL under `src/test/ui/nll`
## History
* On 2017-07-14, [tracking issue opened](https://github.com/rust-lang/rust/issues/43234)
* On 2017-07-20, [initial empty MIR pass added](https://github.com/rust-lang/rust/pull/43271)
* On 2017-08-29, [RFC opened](https://github.com/rust-lang/rfcs/pull/2094)
* On 2017-11-16, [Integrate MIR type-checker with NLL](https://github.com/rust-lang/rust/pull/45825)
* On 2017-12-20, [NLL feature complete](https://github.com/rust-lang/rust/pull/46862)
* On 2018-07-07, [Don't run AST borrowck on mir mode](https://github.com/rust-lang/rust/pull/52083)
* On 2018-07-27, [Add migrate mode](https://github.com/rust-lang/rust/pull/52681)
* On 2019-04-22, [Enable migrate mode on 2015 edition](https://github.com/rust-lang/rust/pull/59114)
* On 2019-08-26, [Don't downgrade errors on 2015 edition](https://github.com/rust-lang/rust/pull/64221)
* On 2019-08-27, [Remove AST borrowck](https://github.com/rust-lang/rust/pull/64790)
rewrite error handling for unresolved inference vars
Pretty much completely rewrites `fn emit_inference_failure_err`.
This new setup should hopefully be easier to extend and is already a lot better when looking for generic arguments.
Because this is a rewrite there are still some parts which are lacking, these are tracked in #94483 and will be fixed in later PRs.
r? `@estebank` `@petrochenkov`
Diagnose anonymous lifetimes errors more uniformly between async and regular fns
Async fns and regular fns are desugared differently. For the former, we create a generic parameter at HIR level. For the latter, we just create an anonymous region for typeck.
I plan to migrate regular fns to the async fn desugaring.
Before that, this PR attempts to merge the diagnostics for both cases.
r? ```@estebank```
Finish bumping stage0
It looks like the last time had left some remaining cfg's -- which made me think
that the stage0 bump was actually successful. This brings us to a released 1.62
beta though.
This now brings us to cfg-clean, with the exception of check-cfg-features in bootstrap;
I'd prefer to leave that for a separate PR at this time since it's likely to be more tricky.
cc https://github.com/rust-lang/rust/pull/97147#issuecomment-1132845061
r? `@pietroalbini`
It looks like the last time had left some remaining cfg's -- which made me think
that the stage0 bump was actually successful. This brings us to a released 1.62
beta though.
Lifetime variance fixes for rustc
#97287 migrates rustc to a `Ty` type that is invariant over its lifetime `'tcx`, so I need to fix a bunch of places that assume that `Ty<'a>` and `Ty<'b>` can be unified by shortening both to some common lifetime.
This is doable, since many lifetimes are already `'tcx`, so all this PR does is be a bit more explicit that elided lifetimes are actually `'tcx`.
Split out from #97287 so the compiler team can review independently.
Add EarlyBinder
Chalk has no concept of `Param` (e0ade19d13/chalk-ir/src/lib.rs (L579)) or `ReEarlyBound` (e0ade19d13/chalk-ir/src/lib.rs (L1308)). Everything is just "bound" - the equivalent of rustc's late-bound. It's not completely clear yet whether to move everything to the same time of binder in rustc or add `Param` and `ReEarlyBound` in Chalk.
Either way, tracking when we have or haven't already substituted out these in rustc can be helpful.
As a first step, I'm just adding a `EarlyBinder` newtype that is required to call `subst`. I also add a couple "transparent" `bound_*` wrappers around a couple query that are often immediately substituted.
r? `@nikomatsakis`
Begin fixing all the broken doctests in `compiler/`
Begins to fix#95994.
All of them pass now but 24 of them I've marked with `ignore HELP (<explanation>)` (asking for help) as I'm unsure how to get them to work / if we should leave them as they are.
There are also a few that I marked `ignore` that could maybe be made to work but seem less important.
Each `ignore` has a rough "reason" for ignoring after it parentheses, with
- `(pseudo-rust)` meaning "mostly rust-like but contains foreign syntax"
- `(illustrative)` a somewhat catchall for either a fragment of rust that doesn't stand on its own (like a lone type), or abbreviated rust with ellipses and undeclared types that would get too cluttered if made compile-worthy.
- `(not-rust)` stuff that isn't rust but benefits from the syntax highlighting, like MIR.
- `(internal)` uses `rustc_*` code which would be difficult to make work with the testing setup.
Those reason notes are a bit inconsistently applied and messy though. If that's important I can go through them again and try a more principled approach. When I run `rg '```ignore \(' .` on the repo, there look to be lots of different conventions other people have used for this sort of thing. I could try unifying them all if that would be helpful.
I'm not sure if there was a better existing way to do this but I wrote my own script to help me run all the doctests and wade through the output. If that would be useful to anyone else, I put it here: https://github.com/Elliot-Roberts/rust_doctest_fixing_tool
Only crate root def-ids don't have a parent, and in majority of cases the argument of `DefIdTree::parent` cannot be a crate root.
So we now panic by default in `parent` and introduce a new non-panicing function `opt_parent` for cases where the argument can be a crate root.
Same applies to `local_parent`/`opt_local_parent`.
Erase type params when suggesting fully qualified path
When suggesting the use of a fully qualified path for a method call that
is ambiguous because it has multiple candidates, erase type params in
the resulting code, as they would result in an error when applied. We
replace them with `_` in the output to rely on inference. There might be
cases where this still produces slighlty incomplete suggestions, but it
otherwise produces many more errors in relatively common cases.
Fix#96292
Change `span_suggestion` (and variants) to take `impl ToString` rather
than `String` for the suggested code, as this simplifies the
requirements on the diagnostic derive.
Signed-off-by: David Wood <david.wood@huawei.com>
Because NLL borrowck is run after typeck, `in_progress_typeck_results`
was always `None` which was preventing the retrieval of the span to which
the suggestion is suppose to add the lifetime bound.
We now manually pass the `LocalDefId` owner to `construct_generic_bound_failure`
so that under NLL, we give the owner id of the current body.
When suggesting the use of a fully qualified path for a method call that
is ambiguous because it has multiple candidates, erase type params in
the resulting code, as they would result in an error when applied. We
replace them with `_` in the output to rely on inference. There might be
cases where this still produces slighlty incomplete suggestions, but it
otherwise produces many more errors in relatively common cases.
Fix#96292
Better method call error messages
Rebase/continuation of #71827
~Based on #92360~
~Based on #93118~
There's a decent description in #71827 that I won't copy here (for now at least)
In addition to rebasing, I've tried to restore most of the original suggestions for invalid arguments. Unfortunately, this does make some of the errors a bit verbose. To fix this will require a bit of refactoring to some of the generalized error suggestion functions, and I just don't have the time to go into it right now.
I think this is in a state that the error messages are overall better than before without a reduction in the suggestions given.
~I've tried to split out some of the easier and self-contained changes into separate commits (mostly in #92360, but also one here). There might be more than can be done here, but again just lacking time.~
r? `@estebank` as the original reviewer of #71827
This attempts to bring better error messages to invalid method calls, by applying some heuristics to identify common mistakes.
The algorithm is inspired by Levenshtein distance and longest common sub-sequence. In essence, we treat the types of the function, and the types of the arguments you provided as two "words" and compute the edits to get from one to the other.
We then modify that algorithm to detect 4 cases:
- A function input is missing
- An extra argument was provided
- The type of an argument is straight up invalid
- Two arguments have been swapped
- A subset of the arguments have been shuffled
(We detect the last two as separate cases so that we can detect two swaps, instead of 4 parameters permuted.)
It helps to understand this argument by paying special attention to terminology: "inputs" refers to the inputs being *expected* by the function, and "arguments" refers to what has been provided at the call site.
The basic sketch of the algorithm is as follows:
- Construct a boolean grid, with a row for each argument, and a column for each input. The cell [i, j] is true if the i'th argument could satisfy the j'th input.
- If we find an argument that could satisfy no inputs, provided for an input that can't be satisfied by any other argument, we consider this an "invalid type".
- Extra arguments are those that can't satisfy any input, provided for an input that *could* be satisfied by another argument.
- Missing inputs are inputs that can't be satisfied by any argument, where the provided argument could satisfy another input
- Swapped / Permuted arguments are identified with a cycle detection algorithm.
As each issue is found, we remove the relevant inputs / arguments and check for more issues. If we find no issues, we match up any "valid" arguments, and start again.
Note that there's a lot of extra complexity:
- We try to stay efficient on the happy path, only computing the diagonal until we find a problem, and then filling in the rest of the matrix.
- Closure arguments are wrapped in a tuple and need to be unwrapped
- We need to resolve closure types after the rest, to allow the most specific type constraints
- We need to handle imported C functions that might be variadic in their inputs.
I tried to document a lot of this in comments in the code and keep the naming clear.
Stabilize `derive_default_enum`
This stabilizes `#![feature(derive_default_enum)]`, as proposed in [RFC 3107](https://github.com/rust-lang/rfcs/pull/3107) and tracked in #87517. In short, it permits you to `#[derive(Default)]` on `enum`s, indicating what the default should be by placing a `#[default]` attribute on the desired variant (which must be a unit variant in the interest of forward compatibility).
```````@rustbot``````` label +S-waiting-on-review +T-lang
Cached stable hash cleanups
r? `@nnethercote`
Add a sanity assertion in debug mode to check that the cached hashes are actually the ones we get if we compute the hash each time.
Add a new data structure that bundles all the hash-caching work to make it easier to re-use it for different interned data structures
This commit updates the signatures of all diagnostic functions to accept
types that can be converted into a `DiagnosticMessage`. This enables
existing diagnostic calls to continue to work as before and Fluent
identifiers to be provided. The `SessionDiagnostic` derive just
generates normal diagnostic calls, so these APIs had to be modified to
accept Fluent identifiers.
In addition, loading of the "fallback" Fluent bundle, which contains the
built-in English messages, has been implemented.
Each diagnostic now has "arguments" which correspond to variables in the
Fluent messages (necessary to render a Fluent message) but no API for
adding arguments has been added yet. Therefore, diagnostics (that do not
require interpolation) can be converted to use Fluent identifiers and
will be output as before.
`MultiSpan` contains labels, which are more complicated with the
introduction of diagnostic translation and will use types from
`rustc_errors` - however, `rustc_errors` depends on `rustc_span` so
`rustc_span` cannot use types like `DiagnosticMessage` without
dependency cycles. Introduce a new `rustc_error_messages` crate that can
contain `DiagnosticMessage` and `MultiSpan`.
Signed-off-by: David Wood <david.wood@huawei.com>
Better suggestions for `Fn`-family trait selection errors
1. Suppress suggestions to add `std::ops::Fn{,Mut,Once}` bounds when a type already implements `Fn{,Mut,Once}`
2. Add a note that points out that a type does in fact implement `Fn{,Mut,Once}`, but the arguments vary (either by number or by actual arguments)
3. Add a note that points out that a type does in fact implement `Fn{,Mut,Once}`, but not the right one (e.g. implements `FnMut`, but `Fn` is required).
Fixes#95147
Lazy type-alias-impl-trait take two
### user visible change 1: RPIT inference from recursive call sites
Lazy TAIT has an insta-stable change. The following snippet now compiles, because opaque types can now have their hidden type set from wherever the opaque type is mentioned.
```rust
fn bar(b: bool) -> impl std::fmt::Debug {
if b {
return 42
}
let x: u32 = bar(false); // this errors on stable
99
}
```
The return type of `bar` stays opaque, you can't do `bar(false) + 42`, you need to actually mention the hidden type.
### user visible change 2: divergence between RPIT and TAIT in return statements
Note that `return` statements and the trailing return expression are special with RPIT (but not TAIT). So
```rust
#![feature(type_alias_impl_trait)]
type Foo = impl std::fmt::Debug;
fn foo(b: bool) -> Foo {
if b {
return vec![42];
}
std::iter::empty().collect() //~ ERROR `Foo` cannot be built from an iterator
}
fn bar(b: bool) -> impl std::fmt::Debug {
if b {
return vec![42]
}
std::iter::empty().collect() // Works, magic (accidentally stabilized, not intended)
}
```
But when we are working with the return value of a recursive call, the behavior of RPIT and TAIT is the same:
```rust
type Foo = impl std::fmt::Debug;
fn foo(b: bool) -> Foo {
if b {
return vec![];
}
let mut x = foo(false);
x = std::iter::empty().collect(); //~ ERROR `Foo` cannot be built from an iterator
vec![]
}
fn bar(b: bool) -> impl std::fmt::Debug {
if b {
return vec![];
}
let mut x = bar(false);
x = std::iter::empty().collect(); //~ ERROR `impl Debug` cannot be built from an iterator
vec![]
}
```
### user visible change 3: TAIT does not merge types across branches
In contrast to RPIT, TAIT does not merge types across branches, so the following does not compile.
```rust
type Foo = impl std::fmt::Debug;
fn foo(b: bool) -> Foo {
if b {
vec![42_i32]
} else {
std::iter::empty().collect()
//~^ ERROR `Foo` cannot be built from an iterator over elements of type `_`
}
}
```
It is easy to support, but we should make an explicit decision to include the additional complexity in the implementation (it's not much, see a721052457cf513487fb4266e3ade65c29b272d2 which needs to be reverted to enable this).
### PR formalities
previous attempt: #92007
This PR also includes #92306 and #93783, as they were reverted along with #92007 in #93893fixes#93411fixes#88236fixes#89312fixes#87340fixes#86800fixes#86719fixes#84073fixes#83919fixes#82139fixes#77987fixes#74282fixes#67830fixes#62742fixes#54895
