Change interners to start preallocated with an increased capacity
Inspired by https://github.com/rust-lang/rust/issues/137005.
Added a `with_capacity` function to `InternedSet`. Changed the `CtxtInterners` to start with `InternedSets` preallocated with a capacity.
This *does* increase memory usage at very slightly(by ~1 MB at the start), altough that increase quickly disaperars for larger crates(since they require such capacity anyway).
A local perf run indicates this improves compiletimes for small crates(like `ripgrep`), without a negative effect on larger ones.
Include version number of libs being built in cargo lib metadata (esp. `librustc_driver*.so`)
Previously, on a non-stable channel, it's possible for two builds from different versioned sources (e.g. 1.84.0 vs 1.84.1) to produce a `librustc_driver*.so` with the same filename hashes. This causes problems with side-by-side installs wrt. linker search paths because 1.84.1 rustc bin and 1.84.0 rustc bin may try to link to the "same" `librustc_driver*.so` (same filename hash) but fail because the contents of the so is actually different.
We try to mitigate this by including the version number of artifacts being built via `__CARGO_DEFAULT_LIB_METADATA` (kind of an ugly hack, but I don't think cargo has a way for us to tell cargo to use a package version override).
Fixes#136701 (mitigates, really).
### Testing
Tested manually[^host] by:
```bash
$ cat src/version
1.86.0
$ ./x build library # w/ compiler profile, (non-stable) dev channel
$ lddtree build/host/stage1/bin/rustc
rustc => build/host/stage1/bin/rustc (interpreter => /lib64/ld-linux-x86-64.so.2)
librustc_driver-ea1b1b2291881cc4.so => build/host/stage1/bin/../lib/librustc_driver-ea1b1b2291881cc4.so
[...]
```
and observing that changing `src/version` to bump a point release causes `librustc_driver*.so` to have a different hash while sources are unmodified otherwise.
```bash
$ cat src/version
1.86.1
$ ./x build library # w/ compiler profile, (non-stable) dev channel
$ lddtree build/host/stage1/bin/rustc
rustc => build/host/stage1/bin/rustc (interpreter => /lib64/ld-linux-x86-64.so.2)
librustc_driver-746badadbcb74721.so => build/host/stage1/bin/../lib/librustc_driver-746badadbcb74721.so
[...]
```
cc `@clan` `@demize` could you check that if you backport this change against 1.84.{0,1} as reported in #136701, that the produced `rustc` binary works, under the context of the Gentoo build system setup?
[^host]: on a `x86_64-unknown-linux-gnu` host, no cross
Build GCC on CI
Previously, we have downloaded a specific commit of GCC and prebuilt it inside Docker using the `build-gccjit.sh` script. This PR removes that scripts and uses the bootstrap GCC step. This allows us to use the `src/gcc` submodule for determining which GCC should be built, and it also moves the logic closer to LLVM, which is also built by bootstrap.
A few things to note:
- The `sccache` option is currently in the `llvm` block, so the GCC build uses `llvm.ccache`, which is a bit weird :) We could either add `gcc.ccache`, or (what I think would be better) to just move `ccache` to the `build` section, as I don't think that it will be necessary to use ccache for LLVM, but not for GCC.
- When the GCC codegen backend is built, it needs to depend on a step that first builds GCC. This is currently done in a hacky way. The proper solution is to create a separate step for the GCC codegen backend, but that is a larger change. Let me know what you think.
r? `@onur-ozkan`
try-job: i686-msvc-1
try-job: x86_64-mingw-1
Make `#[used]` work when linking with `ld64`
To make `#[used]` work in static libraries, we use the `symbols.o` trick introduced in https://github.com/rust-lang/rust/pull/95604.
However, the linker shipped with Xcode, ld64, works a bit differently from other linkers; in particular, [it completely ignores undefined symbols by themselves](https://github.com/apple-oss-distributions/ld64/blob/ld64-954.16/src/ld/parsers/macho_relocatable_file.cpp#L2455-L2468), and only consider them if they have relocations (something something atoms something fixups, I don't know the details).
So to make the `symbols.o` file work on ld64, we need to actually insert a relocation. That's kinda cumbersome to do though, since the relocation must be valid, and hence must point to a valid piece of machine code, and is hence very architecture-specific.
Fixes https://github.com/rust-lang/rust/issues/133491, see that for investigation.
---
Another option would be to pass `-u _foo` to the final linker invocation. This has the problem that `-u` causes the linker to not be able to dead-strip the symbol, which is undesirable. (If we did this, we would possibly also want to do it by putting the arguments in a file by itself, and passing that file via ``@`,` e.g. ``@undefined_symbols.txt`,` similar to https://github.com/rust-lang/rust/issues/52699, though that [is only supported since Xcode 12](https://developer.apple.com/documentation/xcode-release-notes/xcode-12-release-notes#Linking), and I'm not sure we wanna bump that).
Various other options that are probably all undesirable as they affect link time performance:
- Pass `-all_load` to the linker.
- Pass `-ObjC` to the linker (the Objective-C support in the linker has different code paths that load more of the binary), and instrument the binaries that contain `#[used]` symbols.
- Pass `-force_load` to libraries that contain `#[used]` symbols.
Failed attempt: Embed `-u _foo` in the object file with `LC_LINKER_OPTION`, akin to https://github.com/rust-lang/rust/issues/121293. Doesn't work, both because `ld64` doesn't read that from archive members unless it already has a reason to load the member (which is what this PR is trying to make it do), and because `ld64` only support the `-l`, `-needed-l`, `-framework` and `-needed_framework` flags in there.
---
TODO:
- [x] Support all Apple architectures.
- [x] Ensure that this works regardless of the actual type of the symbol.
- [x] Write up more docs.
- [x] Wire up a few proper tests.
`@rustbot` label O-apple
librustdoc: Use `pulldown-cmark-escape` for HTML escaping
Implementation of `@notriddle` 's [suggestion](https://github.com/rust-lang/rust/pull/137274#issuecomment-2669001585).
Somewhat related to #137274 , but the two PRs should be complementary.
Local perf results look like a nice improvement! (so would love a perf run on the CI)
Emit getelementptr inbounds nuw for pointer::add()
Lower pointer::add (via intrinsic::offset with unsigned offset) to getelementptr inbounds nuw on LLVM versions that support it. This lets LLVM make use of the pre-condition that the offset addition does not wrap in an unsigned sense. Together with inbounds, this also implies that the offset is non-negative.
Fixes https://github.com/rust-lang/rust/issues/137217.
vectorcall ABI: require SSE2
According to the official docs at https://learn.microsoft.com/en-us/cpp/cpp/vectorcall, SSE2 is required for this ABI. Add a check that enforces this.
I put this together with the other checks ensuring the target features required for a function are present... however, since the ABI is known pre-monomorphization, it would be possible to do this check earlier, which would have the advantage of checking even in `cargo check`. It would have the disadvantage of spreading this code in yet more places.
The first commit just does a little refactoring of the mono-time ABI check to make it easier to add the new check.
Cc `@workingjubilee`
try-job: dist-i586-gnu-i586-i686-musl
Rollup of 9 pull requests
Successful merges:
- #135354 ([Debuginfo] Add MSVC Synthetic and Summary providers to LLDB)
- #136826 (Replace mem::zeroed with mem::MaybeUninit::uninit for large struct in Unix)
- #137194 (More const {} init in thread_local)
- #137334 (Greatly simplify lifetime captures in edition 2024)
- #137382 (bootstrap: add doc for vendor build step)
- #137423 (Improve a bit HIR pretty printer)
- #137435 (Fix "missing match arm body" suggestion involving `!`)
- #137448 (Fix bugs due to unhandled `ControlFlow` in compiler)
- #137458 (Fix missing self subst when rendering `impl Fn*<T>` with no output type)
r? `@ghost`
`@rustbot` modify labels: rollup
Improve a bit HIR pretty printer
This PR improve (a bit) the HIR pretty printer.
It does so by:
- Not printing elided lifetimes (those are not expressible in surface Rust anyway)
- And by rendering implicit self with the shorthand syntax
I also tried fixing some indentation and other things but gave up for now.
Best reviewed commit by commit.
Replace mem::zeroed with mem::MaybeUninit::uninit for large struct in Unix
As discussion in #136737.
- Replace `mem::zeroed()` with `MaybeUninit::uninit()` for `sockaddr_storage` in `accept()` and `recvfrom()` since these functions fill in the address structure
- Replace `mem::zeroed()` with `MaybeUninit::uninit()` for `pthread_attr_t` in thread-related functions since `pthread_attr_init()` initializes the structure
- Add references to man pages to document this behavior
[Debuginfo] Add MSVC Synthetic and Summary providers to LLDB
Adds handling for `tuple$<>`, `ref$<slice$2<>`, `ref$<str$>` and `enum2$<>`.
Also fixes a bug in MSVC vec/string handling where the script was unable to determine the element's type due to LLDB ignoring template arg debug information
<details>
<summary>Sample code</summary>
```rust
pub enum Number {
One = 57,
Two = 99,
}
#[repr(u8)]
pub enum Container {
First(u32),
Second { val: u64, val2: i8 },
Third,
}
...
let u8_val = b'a';
let float = 42.78000000000001;
let tuple = (u8_val, float);
let str_val = "eef";
let mut string = "freef".to_owned();
let mut_str = string.as_mut_str();
let array: [u8; 4] = [1, 2, 3, 4];
let ref_array = array.as_slice();
let mut array2: [u32; 4] = [1, 2, 3, 4];
let mut_array = array2.as_mut_slice();
let enum_val = Number::One;
let mut enum_val2 = Number::Two;
let sum_val = Container::First(15);
let sum_val_2 = Container::Second { val: 0, val2: 0 };
let sum_val_3 = Container::Third;
let non_zero = NonZeroU128::new(100).unwrap();
let large_discr = NonZeroU128::new(255);
```
</details>
Before:
After:
try-job: aarch64-apple
try-job: x86_64-msvc-1
try-job: i686-msvc-1
try-job: x86_64-mingw-1
try-job: i686-mingw
try-job: aarch64-gnu
stabilize stage management for rustc tools
https://github.com/rust-lang/rust/pull/135990 got out of control due to excessive complexity. This PR aims to achieve the same goal with a simpler approach, likely through multiple smaller PRs. I will keep the other one read-only and open as a reference for future work.
This work stabilizes the staging logic for `ToolRustc` programs, so you no longer need to handle build and target compilers separately in steps. Previously, most tools didn't do this correctly, which was causing the compiler to be built twice (e.g., `x test cargo --stage 1` would compile the stage 2 compiler before, but now it only compiles the stage 1 compiler).
I also tried to document how we should write `ToolRustc` steps as they are quite different and require more attention than other tools.
Next goal is to stabilize how stages are handled for the rustc itself. Currently, `x build --stage 1` builds the stage 1 compiler which is fine, but `x build compiler --stage 1` builds stage 2 compiler.
~~for now, r? ghost~~
Remove `NtVis` and `NtTy`
The next part of #124141. The first actual remove of `Nonterminal` variants. `NtVis` is a simple case that doesn't get much use, but `NtTy` is more complex.
r? `@petrochenkov`
Emit `trunc nuw` for unchecked shifts and `to_immediate_scalar`
- For shifts this shrinks the IR by no longer needing an `assume` while still providing the UB information
- Having this on the `i8`→`i1` truncations will hopefully help with some places that have to load `i8`s or pass those in LLVM structs without range information
Bump sccache in CI to 0.9.1
We haven't updated the used sccache version for years, it has accrued a bunch of fixes and features in the meantime. It now supports the `--show-adv-stats` flag, which gives a more detailed summary of the results of caching. And it can also cache Rust code, which could be useful in the future (https://github.com/rust-lang/rust/pull/136942 - although now there are no large wins).
It also supports caching PGO now, but since the PGO profiles are always different, it won't make any real difference.
https://github.com/rust-lang/rust/pull/133076 previously tried to update the version to 0.3 (CC `@klensy)`
r? `@marcoieni`
Organize `OsString`/`OsStr` shims
Synchronize the `bytes.rs` and `wtf8.rs` shims for `OsString`/`OsStr` so they're easier to diff between each other. This is mostly ordering items the same between the two. I tried to minimize moves and went for the average locations between the files.
With them in the same order, it is clear that `FromInner<_>` is not implemented for `bytes::Buf` and `Clone::clone_from` is not implemented for `wtf8::Buf`, but they are for the other. Fix that.
I added #[inline] to all inherent methods of the `OsString`/`OsStr` shims, because it seemed that was already the rough pattern. `bytes.rs` has more inlining than `wtf8.rs`, so I added the corresponding ones to `wtf8.rs`. Then, the common missing ones have no discernible pattern to me. They're not divided by non-allocating/allocating. Perhaps the pattern is that UTF-8 validation isn't inlined? Since these types are merely the inner values in `OsStr`/`OsString`, I put inline on all methods and let those public types dictate inlining. I have not inspected codegen or run benchmarks.
Also, touch up some (private) documentation comments.
r? ``````@ChrisDenton``````
Lint `#[must_use]` attributes applied to methods in trait impls
The `#[must_use]` attribute has no effect when applied to methods in trait implementations. This PR adds it to the unused `#[must_use]` lint, and cleans the extra attributes in portable-simd and Clippy.
CI: Stop /msys64/bin from being prepended to PATH in msys2 shell
We used to do this along time ago but we stopped doing it when we started installing msys2 manually. 4fd3cf96a1/src/ci/scripts/install-msys2.sh (L11-L13)Fixes#136795
try-job: dist-i686-mingw
Use more explicit and reliable ptr select in sort impls
Using `if ...` with the intent to avoid branches can be surprising to readers and carries the risk of turning into jumps/branches generated by some future compiler version, breaking crucial optimizations.
This commit replaces their usage with the explicit and IR annotated `bool::select_unpredictable`.
Add `MAX_LEN_UTF8` and `MAX_LEN_UTF16` Constants
This pull request adds the `MAX_LEN_UTF8` and `MAX_LEN_UTF16` constants as per #45795, gated behind the `char_max_len` feature.
The constants are currently applied in the `alloc`, `core` and `std` libraries.
coverage: Get hole spans from nested items without fully visiting them
This is a small simplification to the code that collects the spans of nested items within a function, so that those spans can be treated as “holes” to be avoided by the current function's coverage mappings.
The old code was using `nested_filter::All` to ensure that the visitor would see nested items. But we don't need the actual items themselves; we just need their spans, which we can obtain via a custom implementation of `visit_nested_item`.
This avoids the more expansive queries required by `nested_filter::All`.
Don't mention `FromResidual` on bad `?`
Unless `try_trait_v2` is enabled, don't mention that `FromResidual` isn't implemented for a specific type when the implicit `From` conversion of a `?` fails. For the end user on stable, `?` might as well be a compiler intrinsic, so we remove that note to avoid further confusion and allowing other parts of the error to be more prominent.
```
error[E0277]: `?` couldn't convert the error to `u8`
--> $DIR/bad-interconversion.rs:4:20
|
LL | fn result_to_result() -> Result<u64, u8> {
| --------------- expected `u8` because of this
LL | Ok(Err(123_i32)?)
| ------------^ the trait `From<i32>` is not implemented for `u8`
| |
| this can't be annotated with `?` because it has type `Result<_, i32>`
|
= note: the question mark operation (`?`) implicitly performs a conversion on the error value using the `From` trait
= help: the following other types implement trait `From<T>`:
`u8` implements `From<Char>`
`u8` implements `From<bool>`
```
Add a bullet point to `std::fs::copy`
I needed to copy a file but I got the following error:
```
Os { code: 2, kind: NotFound, message: "No such file or directory" }
```
After read the documentation, I though the error was generated by the `from` parameter, forgetting the `to` part. Anyway, I got the error because the parent folder of `to` didn't exist.
Even if the documentation explicitly saying `but is not limited to just these cases`, I would like to add this case because I spent 3 hours around it.
This PR just wants to put a mention about it.
Added project-specific Zed IDE settings
This repository currently has project-specific VS Code IDE settings in `.vscode` and `compiler/rustc_codegen_cranelift/.vscode`. Now there are equivalent project-specific Zed IDE settings alongside those.
This fixes `rust-analyzer` not being able to properly handle this project.
Note that:
1. The contents of `src/tools/rust-analyzer/.vscode` could not be translated to Zed, as they aren't basic IDE settings.
2. One of the VS Code settings in `.vscode` has no corresponding setting in Zed, and so this has been noted like this:
```json
"_settings_only_in_vs_code_not_yet_in_zed": {
"git.detectSubmodulesLimit": 20
},
```
improve cold_path()
#120370 added a new instrinsic `cold_path()` and used it to fix `likely` and `unlikely`
However, in order to limit scope, the information about cold code paths is only used in 2-target switch instructions. This is sufficient for `likely` and `unlikely`, but limits usefulness of `cold_path` for idiomatic rust. For example, code like this:
```
if let Some(x) = y { ... }
```
may generate 3-target switch:
```
switch y.discriminator:
0 => true branch
1 = > false branch
_ => unreachable
```
and therefore marking a branch as cold will have no effect.
This PR improves `cold_path()` to work with arbitrary switch instructions.
Note that for 2-target switches, we can use `llvm.expect`, but for multiple targets we need to manually emit branch weights. I checked Clang and it also emits weights in this situation. The Clang's weight calculation is more complex that this PR, which I believe is mainly because `switch` in `C/C++` can have multiple cases going to the same target.
León Orell Valerian Liehr
bors
Boxy
Jakub Beránek
DianQK
jyn
Jacob Pratt
Deadbeef
Matthias Krüger
许杰友 Jieyou Xu (Joe)
onur-ozkan
Ed Page