]> git.lizzy.rs Git - rust.git/log
rust.git
2 years agoAdd missing unsafe marker.
Mara Bos [Mon, 11 Apr 2022 12:06:18 +0000 (14:06 +0200)]
Add missing unsafe marker.

This is now necessary because of deny(unsafe_op_in_unsafe_fn).

2 years agoAllow cvt_nz to be unused on some platforms.
Mara Bos [Sat, 9 Apr 2022 14:13:25 +0000 (16:13 +0200)]
Allow cvt_nz to be unused on some platforms.

2 years agoAdd #[deny(unsafe_op_in_unsafe_fn)] to thread_local!(const).
Mara Bos [Wed, 6 Apr 2022 20:22:43 +0000 (22:22 +0200)]
Add #[deny(unsafe_op_in_unsafe_fn)] to thread_local!(const).

This avoids 'unused unsafe' warnings when using this feature inside std.

2 years agoAdd debug asserts to futex ReentrantMutex impl.
Mara Bos [Wed, 6 Apr 2022 20:15:43 +0000 (22:15 +0200)]
Add debug asserts to futex ReentrantMutex impl.

2 years agoInitialize thread local with const{}.
Mara Bos [Wed, 6 Apr 2022 20:14:43 +0000 (22:14 +0200)]
Initialize thread local with const{}.

2 years agoMove current_thread_unique_ptr to the only module that uses it.
Mara Bos [Wed, 6 Apr 2022 20:12:47 +0000 (22:12 +0200)]
Move current_thread_unique_ptr to the only module that uses it.

2 years agoMake current_thread_unique_ptr work during thread destruction.
Mara Bos [Wed, 6 Apr 2022 11:50:45 +0000 (13:50 +0200)]
Make current_thread_unique_ptr work during thread destruction.

Otherwise we can't use println!() within atexit handlers etc.

2 years agoAdd futex-based ReentrantMutex on Linux.
Mara Bos [Wed, 6 Apr 2022 10:50:02 +0000 (12:50 +0200)]
Add futex-based ReentrantMutex on Linux.

2 years agoAdd current_thread_unique_ptr() in std::sys_common.
Mara Bos [Wed, 6 Apr 2022 10:49:46 +0000 (12:49 +0200)]
Add current_thread_unique_ptr() in std::sys_common.

2 years agoAuto merge of #95399 - gilescope:plan_b, r=scottmcm
bors [Tue, 12 Apr 2022 05:54:50 +0000 (05:54 +0000)]
Auto merge of #95399 - gilescope:plan_b, r=scottmcm

Faster parsing for lower numbers for radix up to 16 (cont.)

( Continuation of https://github.com/rust-lang/rust/pull/83371 )

With LingMan's change I think this is potentially ready.

2 years agoAuto merge of #95893 - luqmana:no-prepopulate-passes-tweaks, r=nikic
bors [Tue, 12 Apr 2022 03:26:53 +0000 (03:26 +0000)]
Auto merge of #95893 - luqmana:no-prepopulate-passes-tweaks, r=nikic

Respect -Z verify-llvm-ir and other flags that add extra passes when combined with -C no-prepopulate-passes in the new LLVM Pass Manager.

As part of the switch to the new LLVM Pass Manager the behaviour of flags such as `-Z verify-llvm-ir` (e.g. sanitizer, instrumentation) was modified when combined with `-C no-prepopulate-passes`. With the old PM, rustc was the one manually constructing the pipeline and respected those flags but in the new pass manager, those flags are used to build a list of callbacks that get invoked at certain extension points in the pipeline. Unfortunately, `-C no-prepopulate-passes` would skip building the pipeline altogether meaning we'd never add the corresponding passes. The fix here is to just manually invoke those callbacks as needed.

Fixes #95874

Demonstrating the current vs fixed behaviour using the bug in #95864
```console
$ rustc +nightly asm-miscompile.rs --edition 2021 --emit=llvm-ir -C no-prepopulate-passes -Z verify-llvm-ir
$ echo $?
0
$ rustc +stage1 asm-miscompile.rs --edition 2021 --emit=llvm-ir -C no-prepopulate-passes -Z verify-llvm-ir
Basic Block in function '_ZN14asm_miscompile3foo28_$u7b$$u7b$closure$u7d$$u7d$17h360e2f7eee1275c5E' does not have terminator!
label %bb1
LLVM ERROR: Broken module found, compilation aborted!
```

2 years agoDocument that DirEntry holds the directory open
Martin Pool [Tue, 12 Apr 2022 02:00:29 +0000 (19:00 -0700)]
Document that DirEntry holds the directory open

2 years agoBump stdarch
Jacob Pratt [Tue, 12 Apr 2022 01:40:31 +0000 (21:40 -0400)]
Bump stdarch

2 years ago[bootstrap] Grab the right FileCheck binary for dist when cross-compiling.
Luqman Aden [Tue, 12 Apr 2022 01:31:42 +0000 (21:31 -0400)]
[bootstrap] Grab the right FileCheck binary for dist when cross-compiling.

2 years agoAuto merge of #93408 - liangyongrui:master, r=scottmcm
bors [Tue, 12 Apr 2022 00:18:51 +0000 (00:18 +0000)]
Auto merge of #93408 - liangyongrui:master, r=scottmcm

fix Layout struct member naming style

2 years agofix broken link in coverage tools docs
AnthonyMikh [Mon, 11 Apr 2022 23:23:40 +0000 (02:23 +0300)]
fix broken link in coverage tools docs

2 years agouse `to_string` instead of `format!`
Takayuki Maeda [Sun, 10 Apr 2022 11:39:02 +0000 (20:39 +0900)]
use `to_string` instead of `format!`

2 years agoAuto merge of #95944 - Dylan-DPC:rollup-idggkrh, r=Dylan-DPC
bors [Mon, 11 Apr 2022 21:55:58 +0000 (21:55 +0000)]
Auto merge of #95944 - Dylan-DPC:rollup-idggkrh, r=Dylan-DPC

Rollup of 7 pull requests

Successful merges:

 - #95008 ([`let_chains`] Forbid `let` inside parentheses)
 - #95801 (Replace RwLock by a futex based one on Linux)
 - #95864 (Fix miscompilation of inline assembly with outputs in cases where we emit an invoke instead of call instruction.)
 - #95894 (Fix formatting error in pin.rs docs)
 - #95895 (Clarify str::from_utf8_unchecked's invariants)
 - #95901 (Remove duplicate aliases for `check codegen_{cranelift,gcc}` and fix `build codegen_gcc`)
 - #95927 (CI: do not compile libcore twice when performing LLVM PGO)

Failed merges:

r? `@ghost`
`@rustbot` modify labels: rollup

2 years agoFix line numbers
Michael Howell [Mon, 11 Apr 2022 16:26:05 +0000 (09:26 -0700)]
Fix line numbers

2 years agosimplify const params diagnostic on stable
Oliver Downard [Fri, 8 Apr 2022 18:10:07 +0000 (19:10 +0100)]
simplify const params diagnostic on stable

2 years agoRemove rule that place loads may not happen with variant index set
Jakob Degen [Sat, 9 Apr 2022 14:00:19 +0000 (10:00 -0400)]
Remove rule that place loads may not happen with variant index set

2 years agoAdd more clarifications in response to Ralf's comments
Jakob Degen [Fri, 8 Apr 2022 19:53:08 +0000 (15:53 -0400)]
Add more clarifications in response to Ralf's comments

2 years agoAddress various comments and change some details around place to value conversions
Jakob Degen [Sun, 27 Mar 2022 00:46:56 +0000 (20:46 -0400)]
Address various comments and change some details around place to value conversions

2 years agoImprove MIR phases documentation with summaries of changes
Jakob Degen [Sat, 26 Mar 2022 00:08:35 +0000 (20:08 -0400)]
Improve MIR phases documentation with summaries of changes

2 years agoAdjust MIR validator to check a few more things for terminators
Jakob Degen [Sat, 26 Mar 2022 00:00:33 +0000 (20:00 -0400)]
Adjust MIR validator to check a few more things for terminators

2 years agoImprove documentation for MIR terminators
Jakob Degen [Sat, 26 Mar 2022 00:00:16 +0000 (20:00 -0400)]
Improve documentation for MIR terminators

2 years agoImprove documentation for MIR statement kinds.
Jakob Degen [Fri, 25 Mar 2022 06:17:05 +0000 (02:17 -0400)]
Improve documentation for MIR statement kinds.

2 years agoExtend the MIR validator to check many more things around rvalues.
Jakob Degen [Fri, 25 Mar 2022 02:30:23 +0000 (22:30 -0400)]
Extend the MIR validator to check many more things around rvalues.

2 years agoAdd documentation for the semantics of MIR rvalues
Jakob Degen [Fri, 25 Mar 2022 02:29:33 +0000 (22:29 -0400)]
Add documentation for the semantics of MIR rvalues

2 years agoAdjust computation of place types to detect more invalid places
Jakob Degen [Thu, 24 Mar 2022 22:25:37 +0000 (18:25 -0400)]
Adjust computation of place types to detect more invalid places

2 years agoImprove documentation of `Place` and `Operand`
Jakob Degen [Thu, 24 Mar 2022 22:25:03 +0000 (18:25 -0400)]
Improve documentation of `Place` and `Operand`

2 years agoUse const Box::default in P::<[T]>::new
Josh Stone [Mon, 11 Apr 2022 18:44:13 +0000 (11:44 -0700)]
Use const Box::default in P::<[T]>::new

2 years agoimpl const Default for Box<[T]> and Box<str>
Josh Stone [Mon, 11 Apr 2022 18:43:21 +0000 (11:43 -0700)]
impl const Default for Box<[T]> and Box<str>

2 years agopre-push.sh: Use python3 if python is not found
Cheng XU [Wed, 6 Apr 2022 06:13:46 +0000 (23:13 -0700)]
pre-push.sh: Use python3 if python is not found

Since Python 2 has reached EOL, `python` may not be available in certain
systems (e.g., recent macOS). We should use `python3` in this case to
avoid error like `python: No such file or directory`.

2 years agoAuto merge of #95796 - bzEq:bzEq/curl-redirect, r=Dylan-DPC
bors [Mon, 11 Apr 2022 18:59:40 +0000 (18:59 +0000)]
Auto merge of #95796 - bzEq:bzEq/curl-redirect, r=Dylan-DPC

[bootstrap.py] Instruct curl to follow redirect

Some mirror RUSTUP_DIST_SERVER (like https://mirrors.sjtug.sjtu.edu.cn/rust-static) perform redirection when downloading
stage0 compiler. Curl should be able to follow that.

2 years agoRollup merge of #95927 - Kobzol:ci-pgo-libcore, r=lqd
Dylan DPC [Mon, 11 Apr 2022 18:00:46 +0000 (20:00 +0200)]
Rollup merge of #95927 - Kobzol:ci-pgo-libcore, r=lqd

CI: do not compile libcore twice when performing LLVM PGO

I forgot the delete the first compilation when modifying this file in a previous PR.

r? ```@lqd```

2 years agoRollup merge of #95901 - jyn514:remove-duplicate-aliases, r=Mark-Simulacrum
Dylan DPC [Mon, 11 Apr 2022 18:00:45 +0000 (20:00 +0200)]
Rollup merge of #95901 - jyn514:remove-duplicate-aliases, r=Mark-Simulacrum

Remove duplicate aliases for `check codegen_{cranelift,gcc}` and fix `build codegen_gcc`

* Remove duplicate aliases
    Bootstrap already allows selecting these in `PathSet::has`, which allows
    any string that matches the end of a full path.

    I found these by adding `assert!(path.exists())` in `StepDescription::paths`.
    I think ideally we wouldn't have any aliases that aren't paths, but I've held
    off on enforcing that here since it may be controversial, I'll open a separate PR.

* Add `build compiler/rustc_codegen_gcc` as an alias for `CodegenBackend`

    These paths (`_cranelift` and `_gcc`) are somewhat misleading, since they
    actually tell bootstrap to build *all* codegen backends. But this seems like
    a useful improvement in the meantime.

cc ```@bjorn3``` ```@antoyo```

2 years agoRollup merge of #95895 - CAD97:patch-2, r=Dylan-DPC
Dylan DPC [Mon, 11 Apr 2022 18:00:44 +0000 (20:00 +0200)]
Rollup merge of #95895 - CAD97:patch-2, r=Dylan-DPC

Clarify str::from_utf8_unchecked's invariants

Specifically, make it clear that it is immediately UB to pass ill-formed UTF-8 into the function. The previous wording left space to interpret that the UB only occurred when calling another function, which "assumes that `&str`s are valid UTF-8."

This does not change whether str being UTF-8 is a safety or a validity invariant. (As per previous discussion, it is a safety invariant, not a validity invariant.) It just makes it clear that valid UTF-8 is a precondition of str::from_utf8_unchecked, and that emitting an Abstract Machine fault (e.g. UB or a sanitizer error) on invalid UTF-8 is a valid thing to do.

If user code wants to create an unsafe `&str` pointing to ill-formed UTF-8, it must be done via transmutes. Also, just, don't.

Zulip discussion: https://rust-lang.zulipchat.com/#narrow/stream/136281-t-lang.2Fwg-unsafe-code-guidelines/topic/str.3A.3Afrom_utf8_unchecked.20Safety.20requirement

2 years agoRollup merge of #95894 - nyanpasu64:fix-pin-docs, r=Dylan-DPC
Dylan DPC [Mon, 11 Apr 2022 18:00:43 +0000 (20:00 +0200)]
Rollup merge of #95894 - nyanpasu64:fix-pin-docs, r=Dylan-DPC

Fix formatting error in pin.rs docs

Not sure if there's more formatting issues I missed; I kinda lost interest reading midway through.

2 years agoRollup merge of #95864 - luqmana:inline-asm-unwind-store-miscompile, r=Amanieu
Dylan DPC [Mon, 11 Apr 2022 18:00:42 +0000 (20:00 +0200)]
Rollup merge of #95864 - luqmana:inline-asm-unwind-store-miscompile, r=Amanieu

Fix miscompilation of inline assembly with outputs in cases where we emit an invoke instead of call instruction.

We ran into this bug where rustc would segfault while trying to compile certain uses of inline assembly.

Here is a simple repro that demonstrates the issue:
```rust
#![feature(asm_unwind)]

fn main() {
    let _x = String::from("string here just cause we need something with a non-trivial drop");
    let foo: u64;
    unsafe {
        std::arch::asm!(
            "mov {}, 1",
            out(reg) foo,
            options(may_unwind)
        );
    }
    println!("{}", foo);
}
```
([playground link](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=7d6641e83370d2536a07234aca2498ff))

But crucially `feature(asm_unwind)` is not actually needed and this can be triggered on stable as a result of the way async functions/generators are handled in the compiler. e.g.:

```rust
extern crate futures; // 0.3.21

async fn bar() {
    let foo: u64;
    unsafe {
        std::arch::asm!(
            "mov {}, 1",
            out(reg) foo,
        );
    }
    println!("{}", foo);
}

fn main() {
    futures::executor::block_on(bar());
}
```
([playground link](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=1c7781c34dd4a3e80ae4bd936a0c82fc))

An example of the incorrect LLVM generated:
```llvm
bb1:                                              ; preds = %start
  %1 = invoke i64 asm sideeffect alignstack inteldialect unwind "mov ${0:q}, 1", "=&r,~{dirflag},~{fpsr},~{flags},~{memory}"()
          to label %bb2 unwind label %cleanup, !srcloc !9
  store i64 %1, i64* %foo, align 8

bb2:
[...snip...]
```

The store should not be placed after the asm invoke but rather should be in the normal control flow basic block (`bb2` in this case).

[Here](https://gist.github.com/luqmana/be1af5b64d2cda5a533e3e23a7830b44) is a writeup of the investigation that lead to finding this.

2 years agoRollup merge of #95801 - m-ou-se:futex-rwlock, r=Amanieu
Dylan DPC [Mon, 11 Apr 2022 18:00:41 +0000 (20:00 +0200)]
Rollup merge of #95801 - m-ou-se:futex-rwlock, r=Amanieu

Replace RwLock by a futex based one on Linux

This replaces the pthread-based RwLock on Linux by a futex based one.

This implementation is similar to [the algorithm](https://gist.github.com/kprotty/3042436aa55620d8ebcddf2bf25668bc) suggested by `@kprotty,` but modified to prefer writers and spin before sleeping. It uses two futexes: One for the readers to wait on, and one for the writers to wait on. The readers futex contains the state of the RwLock: The number of readers, a bit indicating whether writers are waiting, and a bit indicating whether readers are waiting. The writers futex is used as a simple condition variable and its contents are meaningless; it just needs to be changed on every notification.

Using two futexes rather than one has the obvious advantage of allowing a separate queue for readers and writers, but it also means we avoid the problem a single-futex RwLock would have of making it hard for a writer to go to sleep while the number of readers is rapidly changing up and down, as the writers futex is only changed when we actually want to wake up a writer.

It always prefers writers, as we decided [here](https://github.com/rust-lang/rust/issues/93740#issuecomment-1070696128).

To be able to prefer writers, it relies on futex_wake to return the number of awoken threads to be able to handle write-unlocking while both the readers-waiting and writers-waiting bits are set. Instead of waking both and letting them race, it first wakes writers and only continues to wake the readers too if futex_wake reported there were no writers to wake up.

r? `@Amanieu`

2 years agoRollup merge of #95008 - c410-f3r:let-chains-paren, r=wesleywiser
Dylan DPC [Mon, 11 Apr 2022 18:00:40 +0000 (20:00 +0200)]
Rollup merge of #95008 - c410-f3r:let-chains-paren, r=wesleywiser

[`let_chains`] Forbid `let` inside parentheses

Parenthesizes are mostly a no-op in let chains, in other words, they are mostly ignored.

```rust
let opt = Some(Some(1i32));

if (let Some(a) = opt && (let Some(b) = a)) && b == 1 {
    println!("`b` is declared inside but used outside");
}
```

As seen above, such behavior can lead to confusion.

A proper fix or nested encapsulation would probably require research, time and a modified MIR graph so in this PR I simply denied any `let` inside parentheses. Non-let stuff are still allowed.

```rust
fn main() {
    let fun = || true;

    if let true = (true && fun()) && (true) {
        println!("Allowed");
    }
}
```

It is worth noting that `let ...`  is not an expression and the RFC did not mention this specific situation.

cc `@matthewjasper`

2 years agoFix documentation for wasm32-unknown-unknown
O01eg [Mon, 11 Apr 2022 16:16:10 +0000 (19:16 +0300)]
Fix documentation for wasm32-unknown-unknown

2 years agoupdate ui tests using opaque types in impl headers
Rémy Rakic [Thu, 22 Jul 2021 18:31:21 +0000 (20:31 +0200)]
update ui tests using opaque types in impl headers

2 years agoadd regression tests for opaque types in impl headers
Rémy Rakic [Thu, 22 Jul 2021 18:26:46 +0000 (20:26 +0200)]
add regression tests for opaque types in impl headers

2 years agoprevent opaque types from appearing in impl headers
Rémy Rakic [Thu, 22 Jul 2021 18:21:48 +0000 (20:21 +0200)]
prevent opaque types from appearing in impl headers

2 years agoAuto merge of #95125 - JakobDegen:uninit-variant-rvalue, r=oli-obk
bors [Mon, 11 Apr 2022 14:49:30 +0000 (14:49 +0000)]
Auto merge of #95125 - JakobDegen:uninit-variant-rvalue, r=oli-obk

Add new `Deinit` statement

This rvalue replaces `SetDiscriminant` for ADTs. This PR is an alternative to #94590 , which only specifies that the behavior of `SetDiscriminant` is the same as what this rvalue would do. The motivation for this change are discussed in that PR and [on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/189540-t-compiler.2Fwg-mir-opt/topic/SetDiscriminant.20and.20aggregate.20initialization.20.2394590)

r? `@oli-obk`

2 years agoRemove inlining cost of `Deinit` statements
Jakob Degen [Mon, 11 Apr 2022 14:23:33 +0000 (10:23 -0400)]
Remove inlining cost of `Deinit` statements

2 years agoAdd new `MutatatingUseContext`s for deinit and `SetDiscriminant`
Jakob Degen [Mon, 11 Apr 2022 10:04:53 +0000 (06:04 -0400)]
Add new `MutatatingUseContext`s for deinit and `SetDiscriminant`

2 years agoAdd const eval tests ensuring padding gets correctly marked as deinit on deaggregation
Jakob Degen [Tue, 5 Apr 2022 23:12:09 +0000 (19:12 -0400)]
Add const eval tests ensuring padding gets correctly marked as deinit on deaggregation

2 years agoFix tests broken by deaggregation change
Jakob Degen [Tue, 5 Apr 2022 23:02:57 +0000 (19:02 -0400)]
Fix tests broken by deaggregation change

2 years agoBless tests that broke in a trivial way due to change in deaggregation
Jakob Degen [Tue, 5 Apr 2022 22:34:42 +0000 (18:34 -0400)]
Bless tests that broke in a trivial way due to change in deaggregation

2 years agofix a bad error message for `relative paths are not supported in visibilities` error
Takayuki Maeda [Mon, 11 Apr 2022 13:15:45 +0000 (22:15 +0900)]
fix a bad error message for `relative paths are not supported in visibilities` error

2 years agoDocument semantics of `Deinit` and `SetDiscriminant` MIR statements
Jakob Degen [Tue, 5 Apr 2022 21:22:31 +0000 (17:22 -0400)]
Document semantics of `Deinit` and `SetDiscriminant` MIR statements

2 years agoAdd new `Deinit` statement kind
Jakob Degen [Tue, 5 Apr 2022 21:14:59 +0000 (17:14 -0400)]
Add new `Deinit` statement kind

2 years agoUse is_ or has_ prefix for pure `-> bool` functions.
Mara Bos [Mon, 11 Apr 2022 12:52:02 +0000 (14:52 +0200)]
Use is_ or has_ prefix for pure `-> bool` functions.

2 years agoUse compare_exchange_weak in futex rwlock implementation.
Mara Bos [Mon, 11 Apr 2022 12:29:32 +0000 (14:29 +0200)]
Use compare_exchange_weak in futex rwlock implementation.

2 years agoAdd comments to futex rwlock implementation.
Mara Bos [Mon, 11 Apr 2022 12:27:06 +0000 (14:27 +0200)]
Add comments to futex rwlock implementation.

2 years agoAdd doc comments to futex operations.
Mara Bos [Mon, 11 Apr 2022 12:26:52 +0000 (14:26 +0200)]
Add doc comments to futex operations.

2 years agoAuto merge of #95931 - matthiaskrgr:rollup-1c5zhit, r=matthiaskrgr
bors [Mon, 11 Apr 2022 11:12:26 +0000 (11:12 +0000)]
Auto merge of #95931 - matthiaskrgr:rollup-1c5zhit, r=matthiaskrgr

Rollup of 7 pull requests

Successful merges:

 - #95743 (Update binary_search example to instead redirect to partition_point)
 - #95771 (Update linker-plugin-lto.md to 1.60)
 - #95861 (Note that CI tests Windows 10)
 - #95875 (bootstrap: show available paths help text for aliased subcommands)
 - #95876 (Add a note for unsatisfied `~const Drop` bounds)
 - #95907 (address fixme for diagnostic variable name)
 - #95917 (thin_box test: import from std, not alloc)

Failed merges:

r? `@ghost`
`@rustbot` modify labels: rollup

2 years agoRollup merge of #95917 - RalfJung:thin-box-test, r=dtolnay
Matthias Krüger [Mon, 11 Apr 2022 10:06:58 +0000 (12:06 +0200)]
Rollup merge of #95917 - RalfJung:thin-box-test, r=dtolnay

thin_box test: import from std, not alloc

Importing from `alloc` makes [Miri fail](https://github.com/rust-lang/miri-test-libstd/runs/5964922742?check_suite_focus=true), probably due to the hack that we used to resolve https://github.com/rust-lang/miri-test-libstd/issues/4. There might be better ways around this, but for now this is the easiest thing to do -- no other alloc integration test is importing from `alloc::`.

2 years agoRollup merge of #95907 - compiler-errors:diag, r=Dylan-DPC
Matthias Krüger [Mon, 11 Apr 2022 10:06:57 +0000 (12:06 +0200)]
Rollup merge of #95907 - compiler-errors:diag, r=Dylan-DPC

address fixme for diagnostic variable name

quick rename

2 years agoRollup merge of #95876 - fee1-dead:note-const-drop, r=oli-obk
Matthias Krüger [Mon, 11 Apr 2022 10:06:56 +0000 (12:06 +0200)]
Rollup merge of #95876 - fee1-dead:note-const-drop, r=oli-obk

Add a note for unsatisfied `~const Drop` bounds

r? ``@oli-obk``

2 years agoRollup merge of #95875 - aswild:pr/alias-cmd-paths, r=Mark-Simulacrum
Matthias Krüger [Mon, 11 Apr 2022 10:06:55 +0000 (12:06 +0200)]
Rollup merge of #95875 - aswild:pr/alias-cmd-paths, r=Mark-Simulacrum

bootstrap: show available paths help text for aliased subcommands

Running `./x.py build -h -v` shows a list of available build targets,
but the short alias `./x.py b -h -v` does not. Fix so that the aliases
behave the same as their spelled out counterparts.

2 years agoRollup merge of #95861 - ChrisDenton:windows7-support, r=Dylan-DPC
Matthias Krüger [Mon, 11 Apr 2022 10:06:54 +0000 (12:06 +0200)]
Rollup merge of #95861 - ChrisDenton:windows7-support, r=Dylan-DPC

Note that CI tests Windows 10

Currently being [discussed on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/Windows.207).

r? `````@joshtriplett`````

2 years agoRollup merge of #95771 - str4d:update-linker-plugin-lto.md-to-1.60, r=pietroalbini
Matthias Krüger [Mon, 11 Apr 2022 10:06:53 +0000 (12:06 +0200)]
Rollup merge of #95771 - str4d:update-linker-plugin-lto.md-to-1.60, r=pietroalbini

Update linker-plugin-lto.md to 1.60

I remembered this table when I was looking into what version of LLVM 1.60.0 was using 🙂

2 years agoRollup merge of #95743 - yaahc:binary-search-clarification, r=Mark-Simulacrum
Matthias Krüger [Mon, 11 Apr 2022 10:06:52 +0000 (12:06 +0200)]
Rollup merge of #95743 - yaahc:binary-search-clarification, r=Mark-Simulacrum

Update binary_search example to instead redirect to partition_point

Inspired by discussion in the tracking issue for `Result::into_ok_or_err`: https://github.com/rust-lang/rust/issues/82223#issuecomment-1067098167

People are surprised by us not providing a `Result<T, T> -> T` conversion, and the main culprit for this confusion seems to be the `binary_search` API. We should instead redirect people to the equivalent API that implicitly does that `Result<T, T> -> T` conversion internally which should obviate the need for the `into_ok_or_err` function and give us time to work towards a more general solution that applies to all enums rather than just `Result` such as making or_patterns usable for situations like this via postfix `match`.

I choose to duplicate the example rather than simply moving it from `binary_search` to partition point because most of the confusion seems to arise when people are looking at `binary_search`. It makes sense to me to have the example presented immediately rather than requiring people to click through to even realize there is an example. If I had to put it in only one place I'd leave it in `binary_search` and remove it from `partition_point` but it seems pretty obviously relevant to `partition_point` so I figured the best option would be to duplicate it.

2 years agoAuto merge of #95758 - compiler-errors:issue-54771, r=estebank
bors [Mon, 11 Apr 2022 08:31:37 +0000 (08:31 +0000)]
Auto merge of #95758 - compiler-errors:issue-54771, r=estebank

Only suggest removing semicolon when expression is compatible with `impl Trait`

https://github.com/rust-lang/rust/issues/54771#issuecomment-476423690
> It still needs checking that the last statement's expr can actually conform to the trait, but the naïve behavior is there.

Only suggest removing a semicolon when the type behind the semicolon actually implements the trait in an RPIT `-> impl Trait`. Also upgrade the label that suggests removing the semicolon to a suggestion (should it be verbose?).

cc #54771

2 years agoCI: do not compile libcore twice when performing LLVM PGO
Jakub Beránek [Mon, 11 Apr 2022 08:05:05 +0000 (10:05 +0200)]
CI: do not compile libcore twice when performing LLVM PGO

2 years agobetter def of is signed in tests.
Giles Cope [Mon, 11 Apr 2022 06:37:53 +0000 (07:37 +0100)]
better def of is signed in tests.

2 years agofix Layout struct member naming style
liangyongrui [Mon, 11 Apr 2022 05:35:18 +0000 (13:35 +0800)]
fix Layout struct member naming style

2 years agoAuto merge of #95754 - compiler-errors:binder-assoc-ty, r=nagisa
bors [Mon, 11 Apr 2022 05:16:48 +0000 (05:16 +0000)]
Auto merge of #95754 - compiler-errors:binder-assoc-ty, r=nagisa

Better error for `for<...>` on associated type bound

With GATs just around the corner, we'll probably see more people trying out `Trait<for<'a> Assoc<'a> = ..>`.

This PR improves the syntax error slightly, and also makes it slightly easier to make this into real syntax in the future.

Feel free to push back if the reviewer thinks this should have a suggestion on how to fix it (i.e. push the `for<'a>` outside of the angle brackets), but that can also be handled in a follow-up PR.

2 years agouse find_ancestor_inside to get right span in CastCheck
Michael Goulet [Mon, 11 Apr 2022 05:15:31 +0000 (22:15 -0700)]
use find_ancestor_inside to get right span in CastCheck

2 years agoDelay a bug when we see SelfCtor in ref pattern
Michael Goulet [Mon, 11 Apr 2022 03:55:10 +0000 (20:55 -0700)]
Delay a bug when we see SelfCtor in ref pattern

2 years agothin_box test: import from std, not alloc
Ralf Jung [Mon, 11 Apr 2022 02:59:51 +0000 (22:59 -0400)]
thin_box test: import from std, not alloc

2 years agoAdd a note for unsatisfied `~const Drop` bounds
Deadbeef [Sun, 10 Apr 2022 07:36:08 +0000 (17:36 +1000)]
Add a note for unsatisfied `~const Drop` bounds

2 years agoAuto merge of #94243 - compiler-errors:compiler-flags-typo, r=Mark-Simulacrum
bors [Mon, 11 Apr 2022 00:58:22 +0000 (00:58 +0000)]
Auto merge of #94243 - compiler-errors:compiler-flags-typo, r=Mark-Simulacrum

`s/compiler-flags/compile-flags` in compiletest

Also make compiletest panic so this doesn't happen in the future! I literally always forget which it's called, so I wanted to make my life easier in the future.

Also open to the possibility of parsing both.

2 years agoFix test case for windows
Michael Howell [Mon, 11 Apr 2022 00:46:30 +0000 (17:46 -0700)]
Fix test case for windows

2 years agobetter error for binder on associated type bound
Michael Goulet [Thu, 7 Apr 2022 06:08:22 +0000 (23:08 -0700)]
better error for binder on associated type bound

2 years agoAdd a useful comment.
Nicholas Nethercote [Fri, 8 Apr 2022 04:25:37 +0000 (14:25 +1000)]
Add a useful comment.

2 years agoTweak `NamedMatch` representation.
Nicholas Nethercote [Fri, 8 Apr 2022 04:16:44 +0000 (14:16 +1000)]
Tweak `NamedMatch` representation.

The `Lrc` isn't necessary, neither is the `SmallVec`. Performance is
changed negligibly, but the new code is simpler.

2 years agoChange internal naming of macros.
Nicholas Nethercote [Fri, 8 Apr 2022 03:10:52 +0000 (13:10 +1000)]
Change internal naming of macros.

When a `macro_rules! foo { ... }` invocation is compiled the name used
is `foo`, not `macro_rules!`. This is different to all other macro
invocations, and confused me when I was inserted debugging println
statements for macro evaluation.

This commit changes it to `macro_rules` (or just `macro`), which is what
I expected. There are no externally visible changes.

2 years agoFix crate_type attribute to not warn on duplicates
Eric Huss [Sun, 10 Apr 2022 23:35:37 +0000 (16:35 -0700)]
Fix crate_type attribute to not warn on duplicates

2 years agoonly suggest removing semicolon when expr implements trait
Michael Goulet [Thu, 7 Apr 2022 08:21:35 +0000 (01:21 -0700)]
only suggest removing semicolon when expr implements trait

2 years agorustdoc: Reduce allocations in a `theme` function
Roc Yu [Sun, 10 Apr 2022 23:17:38 +0000 (19:17 -0400)]
rustdoc: Reduce allocations in a `theme` function

2 years agoAdd `build compiler/rustc_codegen_gcc` as an alias for `CodegenBackend`
Joshua Nelson [Sun, 10 Apr 2022 21:41:21 +0000 (16:41 -0500)]
Add `build compiler/rustc_codegen_gcc` as an alias for `CodegenBackend`

These paths (`_cranelift` and `_gcc`) are somewhat misleading, since they
actually tell bootstrap to build *all* codegen backends. But this seems like
a useful improvement in the meantime.

2 years agoRemove duplicate aliases for `codegen_{cranelift,gcc}`
Joshua Nelson [Sun, 10 Apr 2022 21:37:24 +0000 (16:37 -0500)]
Remove duplicate aliases for `codegen_{cranelift,gcc}`

Bootstrap already allows selecting these in `PathSet::has`, which allows
any string that matches the end of a full path.

I found these by adding `assert!(path.exists())` in `StepDescription::paths`.
I think ideally we wouldn't have any aliases that aren't paths, but I've held
off on enforcing that here since it may be controversial, I'll open a separate PR.

2 years agoAuto merge of #95889 - Dylan-DPC:rollup-1cmywu4, r=Dylan-DPC
bors [Sun, 10 Apr 2022 21:01:13 +0000 (21:01 +0000)]
Auto merge of #95889 - Dylan-DPC:rollup-1cmywu4, r=Dylan-DPC

Rollup of 7 pull requests

Successful merges:

 - #95566 (Avoid duplication of doc comments in `std::char` constants and functions)
 - #95784 (Suggest replacing `typeof(...)` with an actual type)
 - #95807 (Suggest adding a local for vector to fix borrowck errors)
 - #95849 (Check for git submodules in non-git source tree.)
 - #95852 (Fix missing space in lossy provenance cast lint)
 - #95857 (Allow multiple derefs to be splitted in deref_separator)
 - #95868 (rustdoc: Reduce allocations in a `html::markdown` function)

Failed merges:

r? `@ghost`
`@rustbot` modify labels: rollup

2 years agoClarify str::from_utf8_unchecked's invariants
Christopher Durham [Sun, 10 Apr 2022 20:04:57 +0000 (15:04 -0500)]
Clarify str::from_utf8_unchecked's invariants

Specifically, make it clear that it is immediately UB to pass ill-formed UTF-8 into the function. The previous wording left space to interpret that the UB only occurred when calling another function, which "assumes that `&str`s are valid UTF-8."

This does not change whether str being UTF-8 is a safety or a validity invariant. (As per previous discussion, it is a safety invariant, not a validity invariant.) It just makes it clear that valid UTF-8 is a precondition of str::from_utf8_unchecked, and that emitting an Abstract Machine fault (e.g. UB or a sanitizer error) on invalid UTF-8 is a valid thing to do.

If user code wants to create an unsafe `&str` pointing to ill-formed UTF-8, it must be done via transmutes. Also, just, don't.

2 years agoFix formatting error in pin.rs docs
nyanpasu64 [Sun, 10 Apr 2022 19:41:31 +0000 (12:41 -0700)]
Fix formatting error in pin.rs docs

2 years agoRespect -Z verify-llvm-ir and other flags that add extra passes when combined with...
Luqman Aden [Sun, 10 Apr 2022 19:30:35 +0000 (15:30 -0400)]
Respect -Z verify-llvm-ir and other flags that add extra passes when combined with -C no-prepopulate-passes in the new LLVM Pass Manager.

2 years agoRollup merge of #95868 - vacuus:markdown-code-blocks, r=GuillaumeGomez
Dylan DPC [Sun, 10 Apr 2022 19:03:39 +0000 (21:03 +0200)]
Rollup merge of #95868 - vacuus:markdown-code-blocks, r=GuillaumeGomez

rustdoc: Reduce allocations in a `html::markdown` function

2 years agoRollup merge of #95857 - ouz-a:mir-opt, r=oli-obk
Dylan DPC [Sun, 10 Apr 2022 19:03:38 +0000 (21:03 +0200)]
Rollup merge of #95857 - ouz-a:mir-opt, r=oli-obk

Allow multiple derefs to be splitted in deref_separator

Previously in #95649 only a single deref within projection was supported and multiple derefs caused a bunch of issues, this PR fixes those issues.

```@oli-obk``` helped a ton again ❤️

2 years agoRollup merge of #95852 - niluxv:strict-provenance-lint-fixup, r=Dylan-DPC
Dylan DPC [Sun, 10 Apr 2022 19:03:37 +0000 (21:03 +0200)]
Rollup merge of #95852 - niluxv:strict-provenance-lint-fixup, r=Dylan-DPC

Fix missing space in lossy provenance cast lint

See https://github.com/rust-lang/rust/pull/95599#discussion_r846425050

2 years agoRollup merge of #95849 - ehuss:check-submodules, r=Mark-Simulacrum
Dylan DPC [Sun, 10 Apr 2022 19:03:36 +0000 (21:03 +0200)]
Rollup merge of #95849 - ehuss:check-submodules, r=Mark-Simulacrum

Check for git submodules in non-git source tree.

People occasionally download the source from https://github.com/rust-lang/rust/releases, but those source distributions will not work because they are missing the submodules. They will get a confusing `failed to load manifest for workspace member` error.

Unfortunately AFAIK there is no way to disable the GitHub source links. This change tries to detect this scenario and provide an error message that guides them toward a solution.

Closes #95608

2 years agoRollup merge of #95807 - TaKO8Ki:suggest-local-var-for-vector, r=fee1-dead
Dylan DPC [Sun, 10 Apr 2022 19:03:35 +0000 (21:03 +0200)]
Rollup merge of #95807 - TaKO8Ki:suggest-local-var-for-vector, r=fee1-dead

Suggest adding a local for vector to fix borrowck errors

closes #95574

2 years agoRollup merge of #95784 - WaffleLapkin:typeof_cool_suggestion, r=compiler-errors
Dylan DPC [Sun, 10 Apr 2022 19:03:34 +0000 (21:03 +0200)]
Rollup merge of #95784 - WaffleLapkin:typeof_cool_suggestion, r=compiler-errors

Suggest replacing `typeof(...)` with an actual type

This PR adds suggestion to replace `typeof(...)` with an actual type of `...`, for example in case of `typeof(1)` we suggest replacing it with `i32`.

If the expression
1. Is not const (`{ let a = 1; let _: typeof(a); }`)
2. Can't be found (`let _: typeof(this_variable_does_not_exist)`)
3. Or has non-suggestable type (closure, generator, error, etc)
we don't suggest anything.

The 1 one is sad, but it's not clear how to support non-consts expressions for `typeof`.

_This PR is inspired by [this tweet]._

[this tweet]: https://twitter.com/compiler_errors/status/1511945354752638976

2 years agoRollup merge of #95566 - eduardosm:std_char_consts_and_methods, r=Mark-Simulacrum
Dylan DPC [Sun, 10 Apr 2022 19:03:34 +0000 (21:03 +0200)]
Rollup merge of #95566 - eduardosm:std_char_consts_and_methods, r=Mark-Simulacrum

Avoid duplication of doc comments in `std::char` constants and functions

For those consts and functions, only the summary is kept and a reference to the `char` associated const/method is included.

Additionaly, re-exported functions have been converted to function definitions that call the previously re-exported function. This makes it easier to add a deprecated attribute to these functions in the future.

2 years agoFIXME for diagnostic variable name
Michael Goulet [Sun, 10 Apr 2022 18:11:25 +0000 (11:11 -0700)]
FIXME for diagnostic variable name

2 years agoAuto merge of #95487 - cjgillot:menhir, r=oli-obk
bors [Sun, 10 Apr 2022 17:59:27 +0000 (17:59 +0000)]
Auto merge of #95487 - cjgillot:menhir, r=oli-obk

Avoid accessing HIR from MIR passes

`hir_owner_nodes` contains a lot of information, and the query result is typically dirty. This forces dependent queries to be re-executed needlessly.

This PR refactors some accesses to HIR to go through more targeted queries that yield the same result.

Based on https://github.com/rust-lang/rust/pull/95435 and https://github.com/rust-lang/rust/pull/95436