Rollup of 6 pull requests
Successful merges:
- #100211 (Refuse to codegen an upstream static.)
- #100277 (Simplify format_args builtin macro implementation.)
- #100483 (Point to generic or arg if it's the self type of unsatisfied projection predicate)
- #100506 (change `InlineAsmCtxt` to not talk about `FnCtxt`)
- #100534 (Make code slightly more uniform)
- #100566 (Use `create_snapshot_for_diagnostic` instead of `clone` for `Parser`)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
debuginfo: Generalize C++-like encoding for enums.
The updated encoding should be able to handle niche layouts where more than one variant has fields (as introduced in https://github.com/rust-lang/rust/pull/94075).
The new encoding is more uniform as there is no structural difference between direct-tag, niche-tag, and no-tag layouts anymore. The only difference between those cases is that the "dataful" variant in a niche-tag enum will have a `(start, end)` pair denoting the tag range instead of a single value.
The new encoding now also supports 128-bit tags, which occur in at least some standard library types. These tags are represented as `u64` pairs so that debuggers (which don't always have support for 128-bit integers) can reliably deal with them. The downside is that this adds quite a bit of complexity to the encoding and especially to the corresponding NatVis.
The new encoding seems to increase the size of (x86_64-pc-windows-msvc) debuginfo by 10-15%. The size of binaries is not affected (release builds were built with `-Cdebuginfo=2`, numbers are in kilobytes):
EXE | before | after | relative
-- | -- | -- | --
cargo (debug) | 40453 | 40450 | +0%
ripgrep (debug) | 10275 | 10273 | +0%
cargo (release) | 16186 | 16185 | +0%
ripgrep (release) | 4727 | 4726 | +0%
PDB | before | after | relative
-- | -- | -- | --
cargo (debug) | 236524 | 261412 | +11%
ripgrep (debug) | 53140 | 59060 | +11%
cargo (release) | 148516 | 169620 | +14%
ripgrep (release) | 10676 | 11804 | +11%
Given that the new encoding is more general, this is to be expected. Only platforms using C++-like debuginfo are affected -- which currently is only `*-pc-windows-msvc`.
*TODO*
- [x] Properly update documentation
- [x] Add regression tests for new optimized enum layouts as introduced by #94075.
r? `@wesleywiser`
Just moving code around so that triagebot can ping relevant parties when
translation logic is modified.
Signed-off-by: David Wood <david.wood@huawei.com>
change `InlineAsmCtxt` to not talk about `FnCtxt`
wip for https://github.com/rust-lang/compiler-team/issues/529. this currently uses both the `FnCtxt` and is used by `check_mod_item_types`. This should be the only thing blocking that MCP afaict.
I am still unsure whether `rustc_hir_typeck` should depend on `rustc_hir_analysis` to use the `InlineAsmCtxt`. I think that's the best solution for now, so that's what I will go for
r? `@compiler-errors`
Point to generic or arg if it's the self type of unsatisfied projection predicate
We do this for `TraitPredicate`s in `point_at_type_arg_instead_of_call_if_possible` and `point_at_arg_instead_of_call_if_possible`, so also do it for `ProjectionPredicate`.
Improves spans for a lot of unit tests.
Simplify format_args builtin macro implementation.
Instead of a FxHashMap<Symbol, (usize, Span)> for the named arguments, this now includes the name and span in the elements of the Vec<FormatArg> directly. The FxHashMap still exists to look up the index, but no longer contains the span. Looking up the name or span of an argument is now trivial and does not need the map anymore.
Visit attributes in more places.
This adds 3 loosely related changes (I can split PRs if desired):
- Attribute checking on pattern struct fields.
- Attribute checking on struct expression fields.
- Lint level visiting on pattern struct fields, struct expression fields, and generic parameters.
There are still some lints which ignore lint levels in various positions. This is a consequence of how the lints themselves are implemented. For example, lint levels on associated consts don't work with `unused_braces`.
`Parser::parse_bottom_expr` currently constructs an empty `attrs` and
then passes it to a large number of other functions. This makes the code
harder to read than it should be, because it's not clear that many
`attrs` arguments are always empty.
This commit removes `attrs` and the passing, simplifying a lot of
functions. The commit also renames `Parser::mk_expr` (which takes an
`attrs` argument) as `mk_expr_with_attrs`, and introduces a new
`mk_expr` which creates an expression with no attributes, which is the
more common case.
rustc_target: Update some old naming around self contained linking
The "fallback" naming pre-dates introduction of `-Clink-self-contained`.
Noticed when reviewing https://github.com/rust-lang/rust/pull/99500.
This PR doesn't break any json target spec, but supporting per-linker-flavor startup objects needed by https://github.com/rust-lang/rust/pull/99500 will break them, so maybe next time I'll remove the compatibility names.
orphan check: rationalize our handling of constants
cc `@rust-lang/types` `@rust-lang/project-const-generics` on whether you agree with this reasoning.
r? types
These use the same names as LLVM and is_riscv_feature_detected!:
- zba (address generation instructions)
- zbb (basic bit manipulation)
- zbc (carry-less multiplication)
- zbs (single-bit manipulation)
Argument type error improvements
Motivated by this interesting code snippet:
```rust
#[derive(Copy, Clone)]
struct Wrapper<T>(T);
fn foo(_: fn(i32), _: Wrapper<i32>) {}
fn f(_: u32) {}
fn main() {
let w = Wrapper::<isize>(1isize);
foo(f, w);
}
```
Which currently errors like:
```
error[E0308]: arguments to this function are incorrect
--> src/main.rs:10:5
|
10 | foo(f, w);
| ^^^ - - expected `i32`, found `isize`
| |
| expected `i32`, found `u32`
|
= note: expected fn pointer `fn(i32)`
found fn item `fn(u32) {f}`
= note: expected struct `Wrapper<i32>`
found struct `Wrapper<isize>`
note: function defined here
--> src/main.rs:4:4
|
4 | fn foo(_: fn(i32), _: Wrapper<i32>) {}
| ^^^ ---------- ---------------
```
Specifically, that double `expected .. found ..` which is very difficult to correlate to the types in the arguments. Also, the fact that "expected `i32`, found `isize`" and the other argument mismatch label don't even really explain what's going on here.
After this PR:
```
error[E0308]: arguments to this function are incorrect
--> $DIR/two-mismatch-notes.rs:10:5
|
LL | foo(f, w);
| ^^^
|
note: expected fn pointer, found fn item
--> $DIR/two-mismatch-notes.rs:10:9
|
LL | foo(f, w);
| ^
= note: expected fn pointer `fn(i32)`
found fn item `fn(u32) {f}`
note: expected struct `Wrapper`, found a different struct `Wrapper`
--> $DIR/two-mismatch-notes.rs:10:12
|
LL | foo(f, w);
| ^
= note: expected struct `Wrapper<i32>`
found struct `Wrapper<isize>`
note: function defined here
--> $DIR/two-mismatch-notes.rs:4:4
|
LL | fn foo(_: fn(i32), _: Wrapper<i32>) {}
| ^^^ ---------- ---------------
error: aborting due to previous error
For more information about this error, try `rustc --explain E0308`.
```
Yeah, it's a bit verbose, but much clearer IMO.
---
Open to discussions about how this could be further improved. Motivated by `@jyn514's` [tweet](https://mobile.twitter.com/joshuayn514/status/1558042020601634816) here.
wf: correctly `shallow_resolve` consts
`shallow_resolve` on `InferConst` is always a noop. this is mostly irrelevant as inference vars should already be resolved at most - if not all - call sites. Haven't actually looked too deeply into whether this was a problem.
Erase regions better in `promote_candidate`
Use `tcx.erase_regions` instead of manually walking through the substs.... this also makes the code slightly simpler 🙈Fixes#100360Fixes#89851
