Non-exhaustive structs may be empty
This is a follow-up to a discrepancy noticed in https://github.com/rust-lang/rust/pull/122792: today, the following struct is considered inhabited (non-empty) outside its defining crate:
```rust
#[non_exhaustive]
pub struct UninhabitedStruct {
pub never: !,
// other fields
}
```
`#[non_exhaustive]` on a struct should mean that adding fields to it isn't a breaking change. There is no way that adding fields to this struct could make it non-empty since the `never` field must stay and is inconstructible. I suspect this was implemented this way due to confusion with `#[non_exhaustive]` enums, which indeed should be considered non-empty outside their defining crate.
I propose that we consider such a struct uninhabited (empty), just like it would be without the `#[non_exhaustive]` annotation.
Code that doesn't pass today and will pass after this:
```rust
// In a different crate
fn empty_match_on_empty_struct<T>(x: UninhabitedStruct) -> T {
match x {}
}
```
This is not a breaking change.
r? ``@compiler-errors``
Remove `print::Pat` from the printing of `WitnessPat`
After the preliminary work done in #128536, we can now get rid of `print::Pat` entirely.
- First, we introduce a variant `PatKind::Print(String)`.
- Then we incrementally remove each other variant of `PatKind`, by having the relevant code produce `PatKind::Print` instead.
- Once `PatKind::Print` is the only remaining variant, it becomes easy to remove `print::Pat` and replace it with `String`.
There is more cleanup that I have in mind, but this seemed like a natural stopping point for one PR.
r? ```@Nadrieril```
Shrink `TyKind::FnPtr`.
By splitting the `FnSig` within `TyKind::FnPtr` into `FnSigTys` and `FnHeader`, which can be packed more efficiently. This reduces the size of the hot `TyKind` type from 32 bytes to 24 bytes on 64-bit platforms. This reduces peak memory usage by a few percent on some benchmarks. It also reduces cache misses and page faults similarly, though this doesn't translate to clear cycles or wall-time improvements on CI.
r? `@compiler-errors`
Preliminary cleanup of `WitnessPat` hoisting/printing
Follow-up to #128430.
The eventual goal is to remove `print::Pat` entirely, but in the course of working towards that I made so many small improvements that it seems wise to let those be reviewed/merged on their own first.
Best reviewed commit-by-commit, most of which should be pretty simple and straightforward.
r? ``@Nadrieril``
By splitting the `FnSig` within `TyKind::FnPtr` into `FnSigTys` and
`FnHeader`, which can be packed more efficiently. This reduces the size
of the hot `TyKind` type from 32 bytes to 24 bytes on 64-bit platforms.
This reduces peak memory usage by a few percent on some benchmarks. It
also reduces cache misses and page faults similarly, though this doesn't
translate to clear cycles or wall-time improvements on CI.
We can replace some tricky iterator-mutation code with a much simpler version
that uses `while let` to shrink a slice.
We also check whether a subpattern would be a wildcard _before_ hoisting it,
which will be very useful when trying to get rid of `print::PatKind` later.
The pattern-analysis code needs to print patterns, as part of its user-visible
diagnostics. But it never actually tries to print "real" patterns! Instead, it
only ever prints synthetic patterns that it has reconstructed from its own
internal represenations.
We can therefore simultaneously remove two obstacles to changing `thir::Pat`,
by having the pattern-analysis code use its own dedicated type for building
printable patterns, and then making `thir::Pat` not printable at all.
Isolate the diagnostic code that expects `thir::Pat` to be printable
Currently, `thir::Pat` implements `fmt::Display` (and `IntoDiagArg`) directly, for use by a few diagnostics.
That makes it tricky to experiment with alternate representations for THIR patterns, because the patterns currently need to be printable on their own. That immediately rules out possibilities like storing subpatterns as a `PatId` index into a central list (instead of the current directly-owned `Box<Pat>`).
This PR therefore takes an incremental step away from that obstacle, by removing `thir::Pat` from diagnostic structs in `rustc_pattern_analysis`, and hiding the pattern-printing process behind a single public `Pat::to_string` method. Doing so makes it easier to identify and update the code that wants to print patterns, and gives a place to pass in additional context in the future if necessary.
---
I'm currently not sure whether switching over to `PatId` is actually desirable or not, but I think this change makes sense on its own merits, by reducing the coupling between `thir::Pat` and the pattern-analysis error types.
match exhaustiveness: Expand or-patterns as a separate step
To compute exhaustiveness, we must expand or-patterns. Previously, we expanded them at the same time that we pushed patterns into the matrix. This made it harder to track pattern reachability, because the or-pattern itself would never show up in the matrix so we had to recover missing information.
This PR changes that: we no longer expand or-patterns as we push them into the matrix. Instead, if we find an or-pattern in the matrix we expand them in a step very much like the specialization we already do. This simplifies a bunch of things, and should greatly simplify the implementation of https://github.com/rust-lang/rust/issues/127870.
r? `@compiler-errors`
Clean up a few minor refs in `format!` macro, as it has a performance cost. Apparently the compiler is unable to inline `format!("{}", &variable)`, and does a run-time double-reference instead (format macro already does one level referencing). Inlining format args prevents accidental `&` misuse.
There are some comments describing multiple subsequent `use` items. When
the big `use` reformatting happens some of these `use` items will be
reordered, possibly moving them away from the comment. With this
additional level of formatting it's not really feasible to have comments
of this type. This commit removes them in various ways:
- merging separate `use` items when appropriate;
- inserting blank lines between the comment and the first `use` item;
- outright deletion (for comments that are relatively low-value);
- adding a separate "top-level" comment.
We also entirely skip formatting for four library files that contain
nothing but `pub use` re-exports, where reordering would be painful.
This section of code depends on `rustc_apfloat` rather than our internal
types, so this is one potential ICE that we should be able to melt now.
This also fixes some missing range and match handling in `rustc_middle`.
Most modules have such a blank line, but some don't. Inserting the blank
line makes it clearer that the `//!` comments are describing the entire
module, rather than the `use` declaration(s) that immediately follows.
We already do this for a number of crates, e.g. `rustc_middle`,
`rustc_span`, `rustc_metadata`, `rustc_span`, `rustc_errors`.
For the ones we don't, in many cases the attributes are a mess.
- There is no consistency about order of attribute kinds (e.g.
`allow`/`deny`/`feature`).
- Within attribute kind groups (e.g. the `feature` attributes),
sometimes the order is alphabetical, and sometimes there is no
particular order.
- Sometimes the attributes of a particular kind aren't even grouped
all together, e.g. there might be a `feature`, then an `allow`, then
another `feature`.
This commit extends the existing sorting to all compiler crates,
increasing consistency. If any new attribute line is added there is now
only one place it can go -- no need for arbitrary decisions.
Exceptions:
- `rustc_log`, `rustc_next_trait_solver` and `rustc_type_ir_macros`,
because they have no crate attributes.
- `rustc_codegen_gcc`, because it's quasi-external to rustc (e.g. it's
ignored in `rustfmt.toml`).
