Rollup of 17 pull requests
Successful merges:
- #91032 (Introduce drop range tracking to generator interior analysis)
- #92856 (Exclude "test" from doc_auto_cfg)
- #92860 (Fix errors on blanket impls by ignoring the children of generated impls)
- #93038 (Fix star handling in block doc comments)
- #93061 (Only suggest adding `!` to expressions that can be macro invocation)
- #93067 (rustdoc mobile: fix scroll offset when jumping to internal id)
- #93086 (Add tests to ensure that `let_chains` works with `if_let_guard`)
- #93087 (Fix src/test/run-make/raw-dylib-alt-calling-convention)
- #93091 (⬆ chalk to 0.76.0)
- #93094 (src/test/rustdoc-json: Check for `struct_field`s in `variant_tuple_struct.rs`)
- #93098 (Show a more informative panic message when `DefPathHash` does not exist)
- #93099 (rustdoc: auto create output directory when "--output-format json")
- #93102 (Pretty printer algorithm revamp step 3)
- #93104 (Support --bless for pp-exact pretty printer tests)
- #93114 (update comment for `ensure_monomorphic_enough`)
- #93128 (Add script to prevent point releases with same number as existing ones)
- #93136 (Backport the 1.58.1 release notes to master)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Introduce drop range tracking to generator interior analysis
This PR addresses cases such as this one from #57478:
```rust
struct Foo;
impl !Send for Foo {}
let _: impl Send = || {
let guard = Foo;
drop(guard);
yield;
};
```
Previously, the `generator_interior` pass would unnecessarily include the type `Foo` in the generator because it was not aware of the behavior of `drop`. We fix this issue by introducing a drop range analysis that finds portions of the code where a value is guaranteed to be dropped. If a value is dropped at all suspend points, then it is no longer included in the generator type. Note that we are using "dropped" in a generic sense to include any case in which a value has been moved. That is, we do not only look at calls to the `drop` function.
There are several phases to the drop tracking algorithm, and we'll go into more detail below.
1. Use `ExprUseVisitor` to find values that are consumed and borrowed.
2. `DropRangeVisitor` uses consume and borrow information to gather drop and reinitialization events, as well as build a control flow graph.
3. We then propagate drop and reinitialization information through the CFG until we reach a fix point (see `DropRanges::propagate_to_fixpoint`).
4. When recording a type (see `InteriorVisitor::record`), we check the computed drop ranges to see if that value is definitely dropped at the suspend point. If so, we skip including it in the type.
## 1. Use `ExprUseVisitor` to find values that are consumed and borrowed.
We use `ExprUseVisitor` to identify the places where values are consumed. We track both the `hir_id` of the value, and the `hir_id` of the expression that consumes it. For example, in the expression `[Foo]`, the `Foo` is consumed by the array expression, so after the array expression we can consider the `Foo` temporary to be dropped.
In this process, we also collect values that are borrowed. The reason is that the MIR transform for generators conservatively assumes anything borrowed is live across a suspend point (see `rustc_mir_transform::generator::locals_live_across_suspend_points`). We match this behavior here as well.
## 2. Gather drop events, reinitialization events, and control flow graph
After finding the values of interest, we perform a post-order traversal over the HIR tree to find the points where these values are dropped or reinitialized. We use the post-order index of each event because this is how the existing generator interior analysis refers to the position of suspend points and the scopes of variables.
During this traversal, we also record branching and merging information to handle control flow constructs such as `if`, `match`, and `loop`. This is necessary because values may be dropped along some control flow paths but not others.
## 3. Iterate to fixed point
The previous pass found the interesting events and locations, but now we need to find the actual ranges where things are dropped. Upon entry, we have a list of nodes ordered by their position in the post-order traversal. Each node has a set of successors. For each node we additionally keep a bitfield with one bit per potentially consumed value. The bit is set if we the value is dropped along all paths entering this node.
To compute the drop information, we first reverse the successor edges to find each node's predecessors. Then we iterate through each node, and for each node we set its dropped value bitfield to the intersection of all incoming dropped value bitfields.
If any bitfield for any node changes, we re-run the propagation loop again.
## 4. Ignore dropped values across suspend points
At this point we have a data structure where we can ask whether a value is guaranteed to be dropped at any post order index for the HIR tree. We use this information in `InteriorVisitor` to check whether a value in question is dropped at a particular suspend point. If it is, we do not include that value's type in the generator type.
Note that we had to augment the region scope tree to include all yields in scope, rather than just the last one as we did before.
r? `@nikomatsakis`
improve `_` constants in item signature handling
removing the "type" from the error messages does slightly worsen the error messages for types, but figuring out whether the placeholder is for a type or a constant and correctly dealing with that seemed fairly difficult to me so I took the easy way out ✨ Imo the error message is still clear enough.
r? `@BoxyUwU` cc `@estebank`
Point at correct argument when async fn output type lifetime disagrees with signature
Fixes most of #74256.
## Problems fixed
This PR fixes a couple of related problems in the error reporting code.
### Highlighting the wrong argument
First, the error reporting code was looking at the desugared return type of an `async fn` to decide which parameter to highlight. For example, a function like
```rust
async fn async_fn(self: &Struct, f: &u32) -> &u32
{ f }
```
desugars to
```rust
async fn async_fn<'a, 'b>(self: &'a Struct, f: &'b u32)
-> impl Future<Output = &'a u32> + 'a + 'b
{ f }
```
Since `f: &'b u32` is returned but the output type is `&'a u32`, the error would occur when checking that `'a: 'b`.
The reporting code would look to see if the "offending" lifetime `'b` was included in the return type, and because the code was looking at the desugared future type, it was included. So it defaulted to reporting that the source of the other lifetime `'a` (the `self` type) was the problem, when it was really the type of `f`. (Note that if it had chosen instead to look at `'a` first, it too would have been included in the output type, and it would have arbitrarily reported the error (correctly this time) on the type of `f`.)
Looking at the actual future type isn't useful for this reason; it captures all input lifetimes. Using the written return type for `async fn` solves this problem and results in less confusing error messages for the user.
This isn't a perfect fix, unfortunately; writing the "manually desugared" form of the above function still results in the wrong parameter being highlighted. Looking at the output type of every `impl Future` return type doesn't feel like a very principled approach, though it might work. The problem would remain for function signatures that look like the desugared one above but use different traits. There may be deeper changes required to pinpoint which part of each type is conflicting.
### Lying about await point capture causing lifetime conflicts
The second issue fixed by this PR is the unnecessary complexity in `try_report_anon_anon_conflict`. It turns out that the root cause I suggested in https://github.com/rust-lang/rust/issues/76547#issuecomment-692863608 wasn't really the root cause. Adding special handling to report that a variable was captured over an await point only made the error messages less correct and pointed to a problem other than the one that actually occurred.
Given the above discussion, it's easy to see why: `async fn`s capture all input lifetimes in their return type, so holding an argument across an await point should never cause a lifetime conflict! Removing the special handling simplified the code and improved the error messages (though they still aren't very good!)
## Future work
* Fix error reporting on the "desugared" form of this code
* Get the `suggest_adding_lifetime_params` suggestion firing on these examples
* cc #42703, I think
r? `@estebank`
We previously weren't tracking partial re-inits while being too
aggressive around partial drops. With this change, we simply ignore
partial drops, which is the safer, more conservative choice.
This changes drop range analysis to handle uninhabited return types such
as `!`. Since these calls to these functions do not return, we model
them as ending in an infinite loop.
This reduces the amount of work done, especially in later iterations,
by only processing nodes whose predecessors changed in the previous
iteration, or earlier in the current iteration. This also has the side
effect of completely ignoring all unreachable nodes.
The refactoring mainly keeps the separation between the modules clearer.
For example, process_deferred_edges function moved to cfg_build.rs since
that is really part of building the CFG, not finding the fixpoint.
Also, we use PostOrderId instead of usize in a lot more places now.
Splits drop_ranges into drop_ranges::record_consumed_borrow,
drop_ranges::cfg_build, and drop_ranges::cfg_propagate. The top level
drop_ranges module has an entry point that does all the coordination of
the other three phases, using code original in generator_interior.
All tests pass now! The issue was that we weren't handling all edges
correctly, but now they are handled consistently.
This includes code to dump a graphviz file for the CFG we built for drop
tracking.
Also removes old DropRanges tests.
This adds support for branching and merging control flow and uses this
to correctly handle the case where a value is dropped in one branch of
an if expression but not another.
There are other cases we need to handle, which will come in follow up
patches.
Issue #57478
This is needed to handle cases like `[a, b.await, c]`. `ExprUseVisitor`
considers `a` to be consumed when it is passed to the array, but the
array is not quite live yet at that point. This means we were missing
the `a` value across the await point. Attributing drops to the parent
expression means we do not consider the value consumed until the
consuming expression has finished.
Issue #57478
The main change needed to make this work is to do a pessimistic over-
approximation for AssignOps. The existing ScopeTree analysis in
region.rs works by doing both left to right and right to left order and
then choosing the most conservative ordering. This behavior is needed
because AssignOp's evaluation order depends on whether it is a primitive
type or an overloaded operator, which runs as a method call.
This change mimics the same behavior as region.rs in
generator_interior.rs.
Issue #57478
This change adds the basic infrastructure for tracking drop ranges in
generator interior analysis, which allows us to exclude dropped types
from the generator type.
Not yet complete, but many of the async/await and generator tests pass.
The main missing piece is tracking branching control flow (e.g. around
an `if` expression). The patch does include support, however, for
multiple yields in th e same block.
Issue #57478
Improve SIMD casts
* Allows `simd_cast` intrinsic to take `usize` and `isize`
* Adds `simd_as` intrinsic, which is the same as `simd_cast` except for saturating float-to-int conversions (matching the behavior of `as`).
cc `@workingjubilee`
Fix ICEs related to `Deref<Target=[T; N]>` on newtypes
1. Stash a const infer's type into the canonical var during canonicalization, so we can recreate the fresh const infer with that same type.
For example, given `[T; _]` we know `_` is a `usize`. If we go from infer => canonical => infer, we shouldn't forget that variable is a usize.
Fixes#92626Fixes#83704
2. Don't stash the autoderef'd slice type that we get from method lookup, but instead recreate it during method confirmation. We need to do this because the type we receive back after picking the method references a type variable that does not exist after probing is done.
Fixes#92637
... A better solution for the second issue would be to actually _properly_ implement `Deref` for `[T; N]` instead of fixing this autoderef hack to stop leaking inference variables. But I actually looked into this, and there are many complications with const impls.
Replace use of `ty()` on term and use it in more places. This will allow more flexibility in the
future, but slightly worried it allows items which are consts which only accept types.
Implement `#[rustc_must_implement_one_of]` attribute
This PR adds a new attribute — `#[rustc_must_implement_one_of]` that allows changing the "minimal complete definition" of a trait. It's similar to GHC's minimal `{-# MINIMAL #-}` pragma, though `#[rustc_must_implement_one_of]` is weaker atm.
Such attribute was long wanted. It can be, for example, used in `Read` trait to make transitions to recently added `read_buf` easier:
```rust
#[rustc_must_implement_one_of(read, read_buf)]
pub trait Read {
fn read(&mut self, buf: &mut [u8]) -> Result<usize> {
let mut buf = ReadBuf::new(buf);
self.read_buf(&mut buf)?;
Ok(buf.filled_len())
}
fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> {
default_read_buf(|b| self.read(b), buf)
}
}
impl Read for Ty0 {}
//^ This will fail to compile even though all `Read` methods have default implementations
// Both of these will compile just fine
impl Read for Ty1 {
fn read(&mut self, buf: &mut [u8]) -> Result<usize> { /* ... */ }
}
impl Read for Ty2 {
fn read_buf(&mut self, buf: &mut ReadBuf<'_>) -> Result<()> { /* ... */ }
}
```
For now, this is implemented as an internal attribute to start experimenting on the design of this feature. In the future we may want to extend it:
- Allow arbitrary requirements like `a | (b & c)`
- Allow multiple requirements like
- ```rust
#[rustc_must_implement_one_of(a, b)]
#[rustc_must_implement_one_of(c, d)]
```
- Make it appear in rustdoc documentation
- Change the syntax?
- Etc
Eventually, we should make an RFC and make this (or rather similar) attribute public.
---
I'm fairly new to compiler development and not at all sure if the implementation makes sense, but at least it passes tests :)
