bors [Sun, 31 Oct 2021 09:07:09 +0000 (09:07 +0000)]
Auto merge of #90434 - matthiaskrgr:rollup-xbn393a, r=matthiaskrgr
Rollup of 8 pull requests
Successful merges:
- #89446 (Add paragraph to ControlFlow docs to menion it works with the ? operator (#88715))
- #89677 (Stabilize `is_symlink()` for `Metadata` and `Path`)
- #89833 (Add #[must_use] to Rc::downgrade)
- #89835 (Add #[must_use] to expensive computations)
- #89839 (Add #[must_use] to mem/ptr functions)
- #89897 (Add #[must_use] to remaining core functions)
- #89951 (Stabilize `option_result_unwrap_unchecked`)
- #90427 (Add #[must_use] to alloc functions that would leak memory)
Matthias Krüger [Sun, 31 Oct 2021 08:20:27 +0000 (09:20 +0100)]
Rollup merge of #90427 - jkugelman:must-use-alloc-leak, r=joshtriplett
Add #[must_use] to alloc functions that would leak memory
As [requested](https://github.com/rust-lang/rust/pull/89899#issuecomment-955600779) by `@joshtriplett.`
> Please do go ahead and add the ones whose only legitimate use for ignoring the return value is leaking memory. (In a separate PR please.) I think it's sufficiently error-prone to call something like alloc and ignore the result that it's legitimate to require `let _ =` for that.
I added `realloc` myself. Clippy ignored it because of its `mut` argument.
Matthias Krüger [Sun, 31 Oct 2021 08:20:26 +0000 (09:20 +0100)]
Rollup merge of #89897 - jkugelman:must-use-core, r=joshtriplett
Add #[must_use] to remaining core functions
I've run out of compelling reasons to group functions together across crates so I'm just going to go module-by-module. This is everything remaining from the `core` crate.
Matthias Krüger [Sun, 31 Oct 2021 08:20:25 +0000 (09:20 +0100)]
Rollup merge of #89839 - jkugelman:must-use-mem-ptr-functions, r=joshtriplett
Add #[must_use] to mem/ptr functions
There's a lot of low-level / unsafe stuff here. Are there legit use cases for ignoring any of these return values?
* No regressions in `./x.py test --stage 1 library/std src/tools/clippy`.
* One regression in `./x.py test --stage 1 src/test/ui`. Fixed.
* I am unable to run `./x.py doc` on my machine so I'll need to wait for the CI to verify doctests pass. I eyeballed all the adjacent tests and they all look okay.
Matthias Krüger [Sun, 31 Oct 2021 08:20:24 +0000 (09:20 +0100)]
Rollup merge of #89835 - jkugelman:must-use-expensive-computations, r=joshtriplett
Add #[must_use] to expensive computations
The unifying theme for this commit is weak, admittedly. I put together a list of "expensive" functions when I originally proposed this whole effort, but nobody's cared about that criterion. Still, it's a decent way to bite off a not-too-big chunk of work.
Given the grab bag nature of this commit, the messages I used vary quite a bit. I'm open to wording changes.
For some reason clippy flagged four `BTreeSet` methods but didn't say boo about equivalent ones on `HashSet`. I stared at them for a while but I can't figure out the difference so I added the `HashSet` ones in.
Matthias Krüger [Sun, 31 Oct 2021 08:20:22 +0000 (09:20 +0100)]
Rollup merge of #89677 - maxwase:is-symlink-stabilization, r=joshtriplett
Stabilize `is_symlink()` for `Metadata` and `Path`
I'm not fully sure about `since` version, correct me if I'm wrong
Needs update after stabilization: [cargo-test-support](https://github.com/rust-lang/cargo/blob/8063672238a5b6c3a901c0fc17f3164692d0be85/crates/cargo-test-support/src/paths.rs#L202)
Matthias Krüger [Sat, 30 Oct 2021 22:33:26 +0000 (00:33 +0200)]
Rollup merge of #90404 - mkroening:hermit-kernel-no-os, r=joshtriplett
hermitkernel-target: Set OS to "none"
For our kernel targets, we should not set OS, as the kernel runs bare
metal without a circular dependency on std.
This also prepares us for unifying with
https://github.com/rust-lang/rust/pull/89062. This patch requires
libhermit-rs to change a `cfg`s from `target_os = "hermit"` to `target_os
= "none"`.
Matthias Krüger [Sat, 30 Oct 2021 22:33:24 +0000 (00:33 +0200)]
Rollup merge of #89899 - jkugelman:must-use-alloc, r=joshtriplett
Add #[must_use] to remaining alloc functions
I've run out of compelling reasons to group functions together across crates so I'm just going to go module-by-module. This is everything remaining from the `alloc` crate.
I ignored these because they might be used to purposefully leak memory... or other allocator shenanigans? I dunno. I'll add them if y'all tell me to.
bors [Sat, 30 Oct 2021 19:31:47 +0000 (19:31 +0000)]
Auto merge of #90422 - GuillaumeGomez:rollup-s1mdag0, r=GuillaumeGomez
Rollup of 5 pull requests
Successful merges:
- #90156 (Remove underlines from non-top docblocks.)
- #90183 (Show all Deref implementations recursively)
- #90202 (Improve and test cross-crate hygiene)
- #90375 (Use `is_global` in `candidate_should_be_dropped_in_favor_of`)
- #90399 (Skipping verbose diagnostic suggestions when calling .as_ref() on type not implementing AsRef)
Guillaume Gomez [Sat, 30 Oct 2021 18:30:27 +0000 (20:30 +0200)]
Rollup merge of #90375 - yanok:master, r=lcnr
Use `is_global` in `candidate_should_be_dropped_in_favor_of`
This manifistated in #90195 with compiler being unable to keep
one candidate for a trait impl, if where is a global impl and more
than one trait bound in the where clause.
Before #87280 `candidate_should_be_dropped_in_favor_of` was using
`TypeFoldable::is_global()` that was enough to discard the two
`ParamCandidate`s. But #87280 changed it to use
`TypeFoldable::is_known_global()` instead, which is pessimistic, so
now the compiler drops the global impl instead (because
`is_known_global` is not sure) and then can't decide between the
two `ParamCandidate`s.
Switching it to use `is_global` again solves the issue.
Guillaume Gomez [Sat, 30 Oct 2021 18:30:27 +0000 (20:30 +0200)]
Rollup merge of #90202 - matthewjasper:xcrate-hygiene, r=petrochenkov
Improve and test cross-crate hygiene
- Decode the parent expansion for traits and enums in `rustc_resolve`, this was already being used for resolution in typeck
- Avoid suggesting importing names with def-site hygiene, since it's often not useful
- Add more tests
Guillaume Gomez [Sat, 30 Oct 2021 18:30:26 +0000 (20:30 +0200)]
Rollup merge of #90183 - GuillaumeGomez:recurse-deref, r=jyn514
Show all Deref implementations recursively
Fixes #87783.
This is a re-implementation of #80653, so taking the original PR comment:
This changes `rustdoc` to recursively follow `Deref` targets so that methods from all levels are added to the rendered output. This implementation displays the methods from all levels in the expanded state with separate sections for each level.
Noah Lev [Fri, 29 Oct 2021 03:48:48 +0000 (20:48 -0700)]
rustdoc: Remove `Crate.src` and instead compute it on-demand
It is only used in one place; `src` was about a third of `Crate`'s total
size; `Crate` is frequently moved by-value; and `src` can be easily
computed on-demand.
bors [Sat, 30 Oct 2021 16:22:49 +0000 (16:22 +0000)]
Auto merge of #90205 - mati865:link-modifiers-in-rustc, r=petrochenkov
Repace use of `static_nobundle` with `native_link_modifiers` within Rust codebase
This fixes warnings when building Rust and running tests:
```
warning: library kind `static-nobundle` has been superseded by specifying `-bundle` on library kind `static`. Try `static:-bundle`
warning: `rustc_llvm` (lib) generated 2 warnings (1 duplicate)
```
bors [Sat, 30 Oct 2021 13:06:51 +0000 (13:06 +0000)]
Auto merge of #90416 - matthiaskrgr:rollup-55lzqng, r=matthiaskrgr
Rollup of 8 pull requests
Successful merges:
- #89876 (Make most std::ops traits const on numeric types)
- #90371 (Fix incorrect doc link)
- #90374 (Unify titles in rustdoc book doc attributes chapter)
- #90377 (Make `core::slice::from_raw_parts[_mut]` const)
- #90395 (Restrict liveness of mutable borrow of inner infcx in ConstInferUnifier::consts)
- #90396 (Prevent type flags assertions being thrown in default_anon_const_substs if errors occurred)
- #90402 (Add a few query descriptions)
- #90412 (Remove unnecessary `macro_use`s in rustdoc)
Matthias Krüger [Sat, 30 Oct 2021 12:37:01 +0000 (14:37 +0200)]
Rollup merge of #90377 - WaffleLapkin:const_slice_from_raw_parts, r=oli-obk
Make `core::slice::from_raw_parts[_mut]` const
Responses to #90012 seem to allow ``@rust-lang/wg-const-eval`` to decide on use of `const_eval_select`, so we can make `core::slice::from_raw_parts[_mut]` const :)
---
This PR marks the following APIs as const:
```rust
// core::slice
pub const unsafe fn from_raw_parts<'a, T>(data: *const T, len: usize) -> &'a [T];
pub const unsafe fn from_raw_parts_mut<'a, T>(data: *mut T, len: usize) -> &'a mut [T];
```
---
Matthias Krüger [Sat, 30 Oct 2021 12:36:58 +0000 (14:36 +0200)]
Rollup merge of #89876 - AlexApps99:const_ops, r=oli-obk
Make most std::ops traits const on numeric types
This PR makes existing implementations of `std::ops` traits (`Add`, `Sub`, etc) [`impl const`](https://github.com/rust-lang/rust/issues/67792) where possible.
This affects:
- All numeric primitives (`u*`, `i*`, `f*`)
- `NonZero*`
- `Wrapping`
This is under the `rustc_const_unstable` feature `const_ops`.
I will write tests once I know what can and can't be kept for the final version of this PR.
Since this is my first PR to rustc (and hopefully one of many), please give me feedback on how to better handle the PR process wherever possible. Thanks
bors [Sat, 30 Oct 2021 07:21:21 +0000 (07:21 +0000)]
Auto merge of #89174 - ChrisDenton:automatic-verbatim-paths, r=dtolnay
Automatically convert paths to verbatim for filesystem operations that support it
This allows using longer paths without the user needing to `canonicalize` or manually prefix paths. If the path is already verbatim then this has no effect.
bors [Sat, 30 Oct 2021 04:12:40 +0000 (04:12 +0000)]
Auto merge of #88268 - GuillaumeGomez:generics-search-index, r=notriddle,camelid,jyn514
rustdoc: Fix generics generation in search index
The generics were not added to the search index as they should, instead they were added as arguments. I used this opportunity to allow generics to have generics themselves (will come in very handy for my current rewrite of the search engine!).
bors [Fri, 29 Oct 2021 21:50:46 +0000 (21:50 +0000)]
Auto merge of #90346 - ferrocene:pa-short-circuit, r=oli-obk
Replace some operators in libcore with their short-circuiting equivalents
In libcore there are a few occurrences of bitwise operators used in boolean expressions instead of their short-circuiting equivalents. This makes it harder to perform some kinds of source code analysis over libcore, for example [MC/DC] code coverage (a requirement in safety-critical environments).
This PR aims to remove as many bitwise operators in boolean expressions from libcore as possible, without any performance regression and without other changes. This means not all bitwise operators are removed, only the ones that don't have any difference with their short-circuiting counterparts. This already simplifies achieving MC/DC coverage, and the other functions can be changed in future PRs.
The PR is best reviewed commit-by-commit, and each commit has the resulting assembly in the message.
## Checked integer methods
These methods recently switched to bitwise operators in PRs https://github.com/rust-lang/rust/pull/89459 and https://github.com/rust-lang/rust/pull/89351. I confirmed bitwise operators are needed in most of the functions, except these two:
* `{integer}::checked_div` ([Godbolt link (nightly)](https://rust.godbolt.org/z/17efh5jPc))
* `{integer}::checked_rem` ([Godbolt link (nightly)](https://rust.godbolt.org/z/85qGWc94K))
`@tspiteri` already mentioned this was the case in https://github.com/rust-lang/rust/pull/89459#issuecomment-932728384, but opted to also switch those two to bitwise operators for consistency. As that makes MC/DC analysis harder this PR proposes switching those two back to short-circuiting operators.
## `{unsigned_ints}::carrying_add`
[Godbolt link (1.56.0)](https://rust.godbolt.org/z/vG9vx8x48)
In this instance replacing the `|` with `||` produces the exact same assembly when optimizations are enabled, so switching to the short-circuiting operator shouldn't have any impact.
## `{unsigned_ints}::borrowing_sub`
[Godbolt link (1.56.0)](https://rust.godbolt.org/z/asEfKaGE4)
In this instance replacing the `|` with `||` produces the exact same assembly when optimizations are enabled, so switching to the short-circuiting operator shouldn't have any impact.
## String UTF-8 validation
[Godbolt link (1.56.0)](https://rust.godbolt.org/z/a4rEbTvvx)
In this instance replacing the `|` with `||` produces practically the same assembly, with the two operands for the "or" swapped:
```asm
; Old
mov rax, qword ptr [rdi + rdx + 8]
or rax, qword ptr [rdi + rdx]
test rax, r9
je .LBB0_7
; New
mov rax, qword ptr [rdi + rdx]
or rax, qword ptr [rdi + rdx + 8]
test rax, r8
je .LBB0_7
```
Martin Kröning [Fri, 29 Oct 2021 16:06:43 +0000 (18:06 +0200)]
hermitkernel-target: Set OS to "none"
For our kernel targets, we should not set OS, as the kernel runs bare
metal without a circular dependency on std.
This also prepares us for unifying with
https://github.com/rust-lang/rust/pull/89062. This patch requires
libhermit-rs to change a `cfg`s from `target_os = "hermit"` to `target_os
= "none"`.
bors [Fri, 29 Oct 2021 15:31:37 +0000 (15:31 +0000)]
Auto merge of #90389 - camelid:rustdoc-rayon, r=jyn514
rustdoc: Switch to mainline rayon
The rustc fork of rayon integrates with Cargo's jobserver to limit the
amount of parallelism. However, rustdoc's use case is concurrent I/O,
which is not CPU-heavy, so it should be able to use mainline rayon.
bors [Fri, 29 Oct 2021 12:21:09 +0000 (12:21 +0000)]
Auto merge of #90373 - tmiasko:union-qualification, r=oli-obk
Use type based qualification for unions
Union field access is currently qualified based on the qualification of
a value previously assigned to the union. At the same time, every union
access transmutes the content of the union, which might result in a
different qualification.
For example, consider constants A and B as defined below, under the
current rules neither contains interior mutability, since a value used
in the initial assignment did not contain `UnsafeCell` constructor.
```rust
#![feature(untagged_unions)]
union U { i: u32, c: std::cell::Cell<u32> }
const A: U = U { i: 0 };
const B: std::cell::Cell<u32> = unsafe { U { i: 0 }.c };
```
To avoid the issue, the changes here propose to consider the content of
a union as opaque and use type based qualification for union types.
bors [Fri, 29 Oct 2021 08:38:39 +0000 (08:38 +0000)]
Auto merge of #90214 - tmiasko:indirect-mutation-qualif, r=ecstatic-morse,oli-obk
Consider indirect mutation during const qualification dataflow
Previously a local would be qualified if either one of two separate data
flow computations indicated so. First determined if a local could
contain the qualif, but ignored any forms of indirect mutation. Second
determined if a local could be mutably borrowed (and so indirectly
mutated), but which in turn ignored the qualif.
The end result was incorrect because the effect of indirect mutation was
effectivelly ignored in the all but the final stage of computation.
In the new implementation the indirect mutation is directly incorporated
into the qualif data flow. The local variable becomes immediately
qualified once it is mutably borrowed and borrowed place type can
contain the qualif.
In general we will now reject additional programs, program that were
prevously unintentionally accepted.
There are also some cases which are now accepted but were previously
rejected, because previous implementation didn't consider whether
borrowed place could have the qualif under the consideration.
bors [Fri, 29 Oct 2021 01:50:08 +0000 (01:50 +0000)]
Auto merge of #90363 - camelid:build-impl-perf, r=jyn514
Improve perf measurements of `build_extern_trait_impl`
Before, it was only measuring one callsite of `build_impl`, and it
incremented the call count even if `build_impl` returned early because
the `did` was already inlined.
Now, it measures all calls, minus calls that return early.
Noah Lev [Fri, 29 Oct 2021 01:19:01 +0000 (18:19 -0700)]
rustdoc: Switch to mainline rayon
The rustc fork of rayon integrates with Cargo's jobserver to limit the
amount of parallelism. However, rustdoc's use case is concurrent I/O,
which is not CPU-heavy, so it should be able to use mainline rayon.
bors [Thu, 28 Oct 2021 19:34:01 +0000 (19:34 +0000)]
Auto merge of #90281 - xldenis:public-borrow-set, r=nikomatsakis
Add BorrowSet to public api
This PR adds `BorrowSet` to the public api so that verification tools can obtain the activation and reservation points of two phase borrows without having to redo calculations themselves (and thus potentially differently from rustc).
Turns out we already can obtain `MoveData` thanks to the public `HasMoveData` trait, so constructing a `BorrowSet` should not provide much of an issue. However, I can't speak to the soundness of this approach, is it safe to take an under-approximation of `MoveData`?
bors [Thu, 28 Oct 2021 16:03:13 +0000 (16:03 +0000)]
Auto merge of #90218 - JakobDegen:adt_significant_drop_fix, r=nikomatsakis
Fixes incorrect handling of ADT's drop requirements
Fixes #90024 and a bunch of duplicates.
The main issue was just that the contract of `NeedsDropTypes::adt_components` was inconsistent; the list of types it might return were the generic parameters themselves or the fields of the ADT, depending on the nature of the drop impl. This meant that the caller could not determine whether a `.subst()` call was still needed on those types; it called `.subst()` in all cases, and this led to ICEs when the returned types were the generic params.
First contribution of more than a few lines, so feedback definitely appreciated.
Ilya Yanok [Thu, 28 Oct 2021 12:49:46 +0000 (12:49 +0000)]
Use `is_global` in `candidate_should_be_dropped_in_favor_of`
This manifistated in #90195 with compiler being unable to keep
one candidate for a trait impl, if where is a global impl and more
than one trait bound in the where clause.
Before #87280 `candidate_should_be_dropped_in_favor_of` was using
`TypeFoldable::is_global()` that was enough to discard the two
`ParamCandidate`s. But #87280 changed it to use
`TypeFoldable::is_known_global()` instead, which is pessimistic, so
now the compiler drops the global impl instead (because
`is_known_global` is not sure) and then can't decide between the
two `ParamCandidate`s.
Switching it to use `is_global` again solves the issue.
Noah Lev [Thu, 28 Oct 2021 03:06:48 +0000 (20:06 -0700)]
Improve perf measurements of `build_extern_trait_impl`
Before, it was only measuring one callsite of `build_impl`, and it
incremented the call count even if `build_impl` returned early because
the `did` was already inlined.
Now, it measures all calls, minus calls that return early.
Tomasz Miąsko [Thu, 28 Oct 2021 00:00:00 +0000 (00:00 +0000)]
Use type based qualification for unions
Union field access is currently qualified based on the qualification of
a value previously assigned to the union. At the same time, every union
access transmutes the content of the union, which might result in a
different qualification.
For example, consider constants A and B as defined below, under the
current rules neither contains interior mutability, since a value used
in the initial assignment did not contain `UnsafeCell` constructor.
```rust
#![feature(untagged_unions)]
union U { i: u32, c: std::cell::Cell<u32> }
const A: U = U { i: 0 };
const B: std::cell::Cell<u32> = unsafe { U { i: 0 }.c };
```
To avoid the issue, the changes here propose to consider the content of
a union as opaque and use type based qualification for union types.