bors [Sun, 31 Jan 2021 10:01:15 +0000 (10:01 +0000)]
Auto merge of #81478 - sivadeilra:windows_dll_imports, r=m-ou-se
Resolve DLL imports at CRT startup, not on demand
On Windows, libstd uses GetProcAddress to locate some DLL imports, so
that libstd can run on older versions of Windows. If a given DLL import
is not present, then libstd uses other behavior (such as fallback
implementations).
This commit uses a feature of the Windows CRT to do these DLL imports
during module initialization, before main() (or DllMain()) is called.
This is the ideal time to resolve imports, because the module is
effectively single-threaded at that point; no other threads can
touch the data or code of the module that is being initialized.
This avoids several problems. First, it makes the cost of performing
the DLL import lookups deterministic. Right now, the DLL imports are
done on demand, which means that application threads _might_ have to
do the DLL import during some time-sensitive operation. This is a
small source of unpredictability. Since threads can race, it's even
possible to have more than one thread running the same redundant
DLL lookup.
This commit also removes using the heap to allocate strings, during
the DLL lookups.
bors [Sun, 31 Jan 2021 07:03:09 +0000 (07:03 +0000)]
Auto merge of #81327 - RalfJung:codegen-no-const-fail, r=oli-obk
codegen: assume constants cannot fail to evaluate
https://github.com/rust-lang/rust/pull/80579 landed, so we can finally remove this old hack from codegen and instead assume that consts never fail to evaluate. :)
Jonas Schievink [Sun, 31 Jan 2021 00:47:46 +0000 (01:47 +0100)]
Rollup merge of #81562 - the8472:improve-inplaceiterable-docs, r=sfackler
Clarify that InPlaceIterable guarantees extend to all advancing iterator methods.
A documentation update that should answer a question that came up in [this zulip discussion](https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/Safety.20guarantees.20of.20InPlaceIterable/near/223743336)
Note the second image is scrolled to the back (right), I added some padding for the text block (not the code block) to make it more comfortable to read since the last character is stuck to the last character.
Jonas Schievink [Sun, 31 Jan 2021 00:47:38 +0000 (01:47 +0100)]
Rollup merge of #81491 - jryans:rustdoc-deref-ice-81395, r=GuillaumeGomez
Balance sidebar `Deref` cycle check with main content
The `Deref` cycle checks added as part of #80653 were "unbalanced" in the sense
that the main content code path checks for cycles _before_ descending, while the
sidebar checks _after_. Checking _before_ is correct, so this changes the
sidebar path to match the main content path.
Jonas Schievink [Sun, 31 Jan 2021 00:47:36 +0000 (01:47 +0100)]
Rollup merge of #81484 - Kogia-sima:perf/optimize-udiv_1e19, r=nagisa
Optimize decimal formatting of 128-bit integers
## Description
This PR optimizes the `udivmod_1e19` function, which is used for formatting 128-bit integers, based on the algorithm provided in \[1\]. This optimization improves performance of formatting 128-bit integers, especially on 64-bit architectures. It also slightly reduces the output binary size.
## Assembler comparison
https://godbolt.org/z/YrG5zY
## Performance
#### previous results
```
test fmt::write_u128_max ... bench: 552 ns/iter (+/- 4)
test fmt::write_u128_min ... bench: 125 ns/iter (+/- 2)
```
#### new results
```
test fmt::write_u128_max ... bench: 205 ns/iter (+/- 13)
test fmt::write_u128_min ... bench: 129 ns/iter (+/- 5)
```
## Reference
\[1\] T. Granlund and P. Montgomery, “Division by Invariant Integers Using Multiplication” in Proc. of the SIGPLAN94 Conference on Programming Language Design and Implementation, 1994, pp. 61–72
Jonas Schievink [Sun, 31 Jan 2021 00:47:34 +0000 (01:47 +0100)]
Rollup merge of #81472 - Aaron1011:fix/revert-cursor-clone, r=petrochenkov
Clone entire `TokenCursor` when collecting tokens
Reverts PR #80830
Fixes taiki-e/pin-project#312
We can have an arbitrary number of `None`-delimited group frames pushed
on the stack due to proc-macro invocations, which can legally be exited.
Attempting to account for this would add a lot of complexity for a tiny
performance gain, so let's just use the original strategy.
Jonas Schievink [Sun, 31 Jan 2021 00:47:31 +0000 (01:47 +0100)]
Rollup merge of #81198 - dtolnay:partialeq, r=m-ou-se
Remove requirement that forces symmetric and transitive PartialEq impls to exist
### Counterexample of symmetry:
If you [have](https://docs.rs/proc-macro2/1.0.24/proc_macro2/struct.Ident.html#impl-PartialEq%3CT%3E) an impl like:
```rust
impl<T> PartialEq<T> for Ident
where
T: ?Sized + AsRef<str>
```
then Rust will not even allow the symmetric impl to exist.
```console
error[E0210]: type parameter `T` must be covered by another type when it appears before the first local type (`Ident`)
--> src/main.rs:9:6
|
9 | impl<T> PartialEq<Ident> for T where T: ?Sized + AsRef<str> {
| ^ type parameter `T` must be covered by another type when it appears before the first local type (`Ident`)
|
= note: implementing a foreign trait is only possible if at least one of the types for which it is implemented is local, and no uncovered type parameters appear before that first local type
= note: in this case, 'before' refers to the following order: `impl<..> ForeignTrait<T1, ..., Tn> for T0`, where `T0` is the first and `Tn` is the last
```
<br>
### Counterexample of transitivity:
Consider these two existing impls from `regex` and `clap`:
```rust
// regex
/// An inline representation of `Option<char>`.
pub struct Char(u32);
impl PartialEq<char> for Char {
fn eq(&self, other: &char) -> bool {
self.0 == *other as u32
}
}
```
impl PartialEq<char> for KeyType {
fn eq(&self, rhs: &char) -> bool {
match self {
KeyType::Short(c) => c == rhs,
_ => false,
}
}
}
```
It's nice to be able to add `PartialEq<proc_macro::Punct> for char` in libproc_macro (https://github.com/rust-lang/rust/pull/80595), but it makes no sense to force an `impl PartialEq<Punct> for Char` and `impl PartialEq<Punct> for KeyType` in `regex` and `clap` in code that otherwise has nothing to do with proc macros.
This stabilizes the `slice_fill_with` feature for Rust 1.51, following the stabilization of `slice_fill` in 1.50. This was requested by libs team members in https://github.com/rust-lang/rust/pull/79213.
This PR also adds the "memset" alias for `slice::fill_with`, mirroring the alias set on the `slice::fill` sibling API. This will ensure someone looking for "memset" will find both variants.
This is unblocked now that `min_const_generics` has been stabilized in https://github.com/rust-lang/rust/pull/79135.
This PR does *not* include the corresponding `IntoIterator` impl, which is https://github.com/rust-lang/rust/pull/65819. Instead, an iterator can be constructed through the `new` method.
`new` would become unnecessary when `IntoIterator` is implemented and might be deprecated then, although it will stay stable.
Jonas Schievink [Sun, 31 Jan 2021 00:47:23 +0000 (01:47 +0100)]
Rollup merge of #80279 - Yaulendil:str-as-mut, r=m-ou-se
Implement missing `AsMut<str>` for `str`
Allows `&mut str` to be taken by a Generic which requires `T` such that `T: AsMut<str>`. Motivating example:
```rust
impl<'i, T> From<T> for StructImmut<'i> where
T: AsRef<str> + 'i,
{
fn from(asref: T) -> Self {
let string: &str = asref.as_ref();
// ...
}
}
impl<'i, T> From<T> for StructMut<'i> where
T: AsMut<str> + 'i,
{
fn from(mut asmut: T) -> Self {
let string: &mut str = asmut.as_mut();
// ...
}
}
```
The Immutable form of this structure can be constructed by `StructImmut::from(s)` where `s` may be a `&String` or a `&str`, because `AsRef<str>` is implemented for `str`. However, the mutable form of the structure can be constructed in the same way **only** with a `&mut String`, and **not** with a `&mut str`.
This change does have some precedent, because as can be seen in [the Implementors](https://doc.rust-lang.org/std/convert/trait.AsMut.html#implementors), `AsMut<[T]>` is implemented for `[T]` as well as for `Vec<T>`, but `AsMut<str>` is implemented only for `String`. This would complete the symmetry.
As a trait implementation, this should be immediately stable.
Jonas Schievink [Sun, 31 Jan 2021 00:47:22 +0000 (01:47 +0100)]
Rollup merge of #80053 - gilescope:include-ignore, r=m-ou-se
stabilise `cargo test -- --include-ignored`
stabilise `cargo test -- --include-ignored`
On stable there's no way to run ignored tests as well as the normal tests.
An example use case where stabilising this would help:
Exercism has some initial tests and then some additional ignored tests that people run currently with --ignore but currently they can't run all the tests in one go without being on nightly. It would be a little more ergonomic if this flag was stablilised.
( Fixes #65770 )
I built with ./x.py build -i library/test - but as libtest is a dylib is there an easy way to invoke it manually to check it's working as expected? (I've updated the automated tests.)
Stabilizes `Arc::{incr,decr}_strong_count`, enabling unsafely incrementing an decrementing the Arc strong count directly with fewer gotchas. This API was first introduced on nightly six months ago, and has not seen any changes since. The initial PR showed two existing pieces of code that would benefit from this API, and included a change inside the stdlib to use this.
Given the small surface area, predictable use, and no changes since introduction, I'd like to propose we stabilize this.
## Links
* [Initial implementation](https://github.com/rust-lang/rust/pull/70733)
* [Motivation from #68700](https://github.com/rust-lang/rust/pull/68700#discussion_r396169064)
* [Real world example in an executor](https://docs.rs/extreme/666.666.666666/src/extreme/lib.rs.html#13)
bors [Sat, 30 Jan 2021 16:49:45 +0000 (16:49 +0000)]
Auto merge of #6654 - flip1995:no_lazy_static_regex, r=flip1995
No lazy static regex
r? `@llogiq`
#6500
regex is unnecessary for this lint (https://github.com/rust-lang/rust-clippy/pull/6500#discussion_r558555071)
lazy_static is unnecessary. The std lazy feature should be used instead.
bors [Sat, 30 Jan 2021 16:41:05 +0000 (16:41 +0000)]
Auto merge of #81453 - jumbatm:clashing-extern-decl-perf, r=nagisa
clashing_extern_declarations: Use symbol interning to avoid string alloc.
Use symbol interning as a hack to avoid allocating a string for every symbol name we store in the seen set. This hopefully addresses the minor perf regression described in https://github.com/rust-lang/rust/pull/80009#issuecomment-763526902.
J. Ryan Stinnett [Thu, 28 Jan 2021 22:03:20 +0000 (22:03 +0000)]
Balance sidebar `Deref` cycle check with main content
The `Deref` cycle checks added as part of #80653 were "unbalanced" in the sense
that the main content code path checks for cycles _before_ descending, while the
sidebar checks _after_. Checking _before_ is correct, so this changes the
sidebar path to match the main content path.
bors [Sat, 30 Jan 2021 07:41:19 +0000 (07:41 +0000)]
Auto merge of #81521 - nikic:update-dist-various, r=Mark-Simulacrum
Update dist-various to Ubuntu 20.04
This updates the dist-various-1 and dist-various-2 images to Ubuntu
20.04. This requires some adjustments:
* `DEBIAN_FRONTEND=noninteractive` required for apt install.
* `team-gcc-argm-embedded` PPA does not support focal. However,
we can simply use the distro-provided `gcc-arm-none-eabi`. Per
the comment, the PPA was only used to get a newer version.
* rumprun has to be updated to avoid a linker error.
* We need to build rumrun with `NOGCCERROR`, which disables use
of `-Werror` and allows building with a newer compiler.
* We need to install `libtinfo5`, which appears to be a dependency
of the clang used during the fuchsia build.
* We need to switch to `g++-8` rather than `g++-7`, as at least
`g++-7-arm-linux-gnueabi` is not available on focal.
* We need to upgrade to GCC 6.5 for the Solaris build, as GCC 6.4
does not support the newer libisl version.
Yuki Okushi [Sat, 30 Jan 2021 04:37:04 +0000 (13:37 +0900)]
Rollup merge of #81520 - jyn514:rustc2, r=Mark-Simulacrum
Don't clone LLVM submodule when download-ci-llvm is set
Previously, `downloading_llvm` would check `self.build` while it was
still an empty string, and think it was always false. This fixes the
check.
This addresses the worst part of https://github.com/rust-lang/rust/issues/76653. There are still some large submodules being downloaded (in particular, `rustc-by-example` is 146 MB, and all the submodules combined are 311 MB), but this is a lot better than the whopping 1.4 GB before.
Yuki Okushi [Sat, 30 Jan 2021 04:36:56 +0000 (13:36 +0900)]
Rollup merge of #81501 - calebcartwright:update-rustfmt, r=sanxiyn
update rustfmt to v1.4.34
Short summary: Various formatting fixes (several const generic related) and introduction of `imports_granularity` config option
Long summary copied from changelog:
#### Changed
- `merge_imports` configuration has been deprecated in favor of the new `imports_granularity` option. Any existing usage of `merge_imports` will be automatically mapped to the corresponding value on `imports_granularity` with a warning message printed to encourage users to update their config files.
#### Added
- New `imports_granularity` option has been added which succeeds `merge_imports`. This new option supports several additional variants which allow users to merge imports at different levels (crate or module), and even flatten imports to have a single use statement per item. ([PR rust-lang/rustfmt#4634](https://github.com/rust-lang/rustfmt/pull/4634), [PR rust-lang/rustfmt#4639](https://github.com/rust-lang/rustfmt/pull/4639))
See the section on the configuration site for more information
https://rust-lang.github.io/rustfmt/?version=v1.4.33&search=#imports_granularity
#### Fixed
- Fix erroneous removal of `const` keyword on const trait impl ([rust-lang/rustfmt#4084](https://github.com/rust-lang/rustfmt/issues/4084))
- Fix incorrect span usage wit const generics in supertraits ([rust-lang/rustfmt#4204](https://github.com/rust-lang/rustfmt/issues/4204))
- Use correct span for const generic params ([rust-lang/rustfmt#4263](https://github.com/rust-lang/rustfmt/issues/4263))
- Correct span on const generics to include type bounds ([rust-lang/rustfmt#4310](https://github.com/rust-lang/rustfmt/issues/4310))
- Idempotence issue on blocks containing only empty statements ([rust-lang/rustfmt#4627](https://github.com/rust-lang/rustfmt/issues/4627) and [#3868](https://github.com/rust-lang/rustfmt/issues/3868))
- Fix issue with semicolon placement on required functions that have a trailing comment that ends in a line-style comment before the semicolon ([rust-lang/rustfmt#4646](https://github.com/rust-lang/rustfmt/issues/4646))
- Avoid shared interned cfg_if symbol since rustfmt can re-initialize the rustc_ast globals on multiple inputs ([rust-lang/rustfmt#4656](https://github.com/rust-lang/rustfmt/issues/4656))
- Don't insert trailing comma on (base-less) rest in struct literals within macros ([rust-lang/rustfmt#4675](https://github.com/rust-lang/rustfmt/issues/4675))
Yuki Okushi [Sat, 30 Jan 2021 04:36:53 +0000 (13:36 +0900)]
Rollup merge of #81495 - camelid:rustdoc-output_format-optional, r=GuillaumeGomez
rustdoc: Remove unnecessary optional
Previously, the HTML output format was represented by both
`Some(OutputFormat::Html)` and `None` so there's no need to have an
optional. Instead, `OutputFormat::Html` is explicitly the default and we
no longer have a "tri-state enum".
Yuki Okushi [Sat, 30 Jan 2021 04:36:50 +0000 (13:36 +0900)]
Rollup merge of #81468 - est31:cfg_version, r=petrochenkov
cfg(version): treat nightlies as complete
This PR makes cfg(version) treat the nightlies
for version 1.n.0 as 1.n.0, even though that nightly
version might not have all stabilizations and features
of the released 1.n.0. This is done for greater
convenience for people who want to test a newly
stabilized feature on nightly, or in other words,
give newly stabilized features as many eyeballs
as possible.
For users who wish to pin nightlies, this commit adds
a -Z assume-incomplete-release option that they can
enable if they run into any issues due to this change.
Implements the suggestion in https://github.com/rust-lang/rust/issues/64796#issuecomment-640851454
Yuki Okushi [Sat, 30 Jan 2021 04:36:46 +0000 (13:36 +0900)]
Rollup merge of #81291 - sexxi-goose:fix-struct-update-functional-record-update-syntax-error, r=nikomatsakis
Support FRU pattern with `[feature(capture_disjoint_fields)]`
In case of a functional record update syntax for creating a structure, `ExprUseVisitor` to only detect the precise use of some of the field in the `..x` part of the syntax. However, when we start building MIR, we
1. First, build the place for `x`
2. and then, add precise field projections so that only some parts of `x` end up getting read.
When `capture_disjoint_fields` is enabled, and FRU is used within a closure `x` won't be completely captured, and therefore the first step will fail. This PR updates `mir_build` to create a place builder in the first step and then create place from the builder only after applying the field projection.
Yuki Okushi [Sat, 30 Jan 2021 04:36:44 +0000 (13:36 +0900)]
Rollup merge of #80959 - jhpratt:unsigned_abs-stabilization, r=m-ou-se
Stabilize `unsigned_abs`
Resolves #74913.
This PR stabilizes the `i*::unsigned_abs()` method, which returns the absolute value of an integer _as its unsigned equivalent_. This has the advantage that it does not overflow on `i*::MIN`.
I have gone ahead and used this in a couple locations throughout the repository.
Yuki Okushi [Sat, 30 Jan 2021 04:36:43 +0000 (13:36 +0900)]
Rollup merge of #80886 - RalfJung:stable-raw-ref-macros, r=m-ou-se
Stabilize raw ref macros
This stabilizes `raw_ref_macros` (https://github.com/rust-lang/rust/issues/73394), which is possible now that https://github.com/rust-lang/rust/issues/74355 is fixed.
However, as I already said in https://github.com/rust-lang/rust/issues/73394#issuecomment-751342185, I am not particularly happy with the current names of the macros. So I propose we also change them, which means I am proposing to stabilize the following in `core::ptr`:
```rust
pub macro const_addr_of($e:expr) {
&raw const $e
}
Yuki Okushi [Sat, 30 Jan 2021 04:36:41 +0000 (13:36 +0900)]
Rollup merge of #80562 - nagisa:nagisa/bools-are-unsigned, r=eddyb
Consider Scalar to be a bool only if its unsigned
This seems right, given that conceptually bools are unsigned, but the
implications of this change may have more action at distance that I'm
not sure how to exhaustively consider.
For instance there are a number of cases where code attaches range
metadata if `scalar.is_bool()` holds. Supposedly it would no longer be
attached to the `repr(i8)` enums? Though I'm not sure why booleans are
being special-cased here in the first place...
This patch adds the `core::stream` submodule and implements `core::stream::Stream` in accordance with [RFC2996](https://github.com/rust-lang/rfcs/pull/2996). The RFC hasn't been merged yet, but as requested by the libs team in https://github.com/rust-lang/rfcs/pull/2996#issuecomment-725696389 I'm filing this PR to get the ball rolling.
## Documentatation
The docs in this PR have been adapted from [`std::iter`](https://doc.rust-lang.org/std/iter/index.html), [`async_std::stream`](https://docs.rs/async-std/1.7.0/async_std/stream/index.html), and [`futures::stream::Stream`](https://docs.rs/futures/0.3.8/futures/stream/trait.Stream.html). Once this PR lands my plan is to follow this up with PRs to add helper methods such as `stream::repeat` which can be used to document more of the concepts that are currently missing. That will allow us to cover concepts such as "infinite streams" and "laziness" in more depth.
## Feature gate
The feature gate for `Stream` is `stream_trait`. This matches the `#[lang = "future_trait"]` attribute name. The intention is that only the APIs defined in RFC2996 will use this feature gate, with future additions such as `stream::repeat` using their own feature gates. This is so we can ensure a smooth path towards stabilizing the `Stream` trait without needing to stabilize all the APIs in `core::stream` at once. But also don't start expanding the API until _after_ stabilization, as was the case with `std::future`.
__edit:__ the feature gate has been changed to `async_stream` to match the feature gate proposed in the RFC.
## Conclusion
This PR introduces `core::stream::{Stream, Next}` and re-exports it from `std` as `std::stream::{Stream, Next}`. Landing `Stream` in the stdlib has been a mult-year process; and it's incredibly exciting for this to finally happen!
bors [Sat, 30 Jan 2021 01:02:18 +0000 (01:02 +0000)]
Auto merge of #81398 - bugadani:rustdoc-perf, r=GuillaumeGomez
rustdoc tweaking
* Reuse memory
* simplify `next_def_id`, avoid multiple hashing and unnecessary lookups
* remove `all_fake_def_ids`, use the global map instead (probably not a good step toward parallelization, though...)
* convert `add_deref_target` to iterative implementation
* use `ArrayVec` where we know the max number of elements
* minor touchups here and there
* avoid building temporary vectors that get appended to other vectors
At most places I may or may not be doing the compiler's job is this PR.
bors [Sat, 30 Jan 2021 01:02:18 +0000 (01:02 +0000)]
rustdoc tweaking
* Reuse memory
* simplify `next_def_id`, avoid multiple hashing and unnecessary lookups
* remove `all_fake_def_ids`, use the global map instead (probably not a good step toward parallelization, though...)
* convert `add_deref_target` to iterative implementation
* use `ArrayVec` where we know the max number of elements
* minor touchups here and there
* avoid building temporary vectors that get appended to other vectors
At most places I may or may not be doing the compiler's job is this PR.
This seems right, given that conceptually bools are unsigned, but the
implications of this change may have more action at distance that I'm
not sure how to exhaustively consider.
For instance there are a number of cases where code attaches range
metadata if `scalar.is_bool()` holds. Supposedly it would no longer be
attached to the `repr(i8)` enums? Though I'm not sure why booleans are
being special-cased here in the first place...
Arlie Davis [Wed, 27 Jan 2021 20:16:03 +0000 (12:16 -0800)]
Resolve DLL imports at CRT startup, not on demand
On Windows, libstd uses GetProcAddress to locate some DLL imports, so
that libstd can run on older versions of Windows. If a given DLL import
is not present, then libstd uses other behavior (such as fallback
implementations).
This commit uses a feature of the Windows CRT to do these DLL imports
during module initialization, before main() (or DllMain()) is called.
This is the ideal time to resolve imports, because the module is
effectively single-threaded at that point; no other threads can
touch the data or code of the module that is being initialized.
This avoids several problems. First, it makes the cost of performing
the DLL import lookups deterministic. Right now, the DLL imports are
done on demand, which means that application threads _might_ have to
do the DLL import during some time-sensitive operation. This is a
small source of unpredictability. Since threads can race, it's even
possible to have more than one thread running the same redundant
DLL lookup.
This commit also removes using the heap to allocate strings, during
the DLL lookups.
Nikita Popov [Fri, 29 Jan 2021 14:08:17 +0000 (15:08 +0100)]
Update dist-various to Ubuntu 20.04
This updates the dist-various-1 and dist-various-2 images to Ubuntu
20.04. This requires some adjustments:
* `DEBIAN_FRONTEND=noninteractive` required for apt install.
* `team-gcc-argm-embedded` PPA does not support focal. However,
we can simply use the distro-provided `gcc-arm-none-eabi`. Per
the comment, the PPA was only used to get a newer version.
* rumprun has to be updated to avoid a linker error.
* We need to build rumrun with `NOGCCERROR`, which disables use
of `-Werror` and allows building with a newer compiler.
* We need to install `libtinfo5`, which appears to be a dependency
of the clang used during the fuchsia build.
* We need to switch to `g++-8` rather than `g++-7`, as at least
`g++-7-arm-linux-gnueabi` is not available on focal.
* We need to upgrade to GCC 6.5 for the Solaris build, as GCC 6.4
does not support the newer libisl version.
bors [Fri, 29 Jan 2021 13:10:09 +0000 (13:10 +0000)]
Auto merge of #81470 - tmiasko:remove-allocations, r=matthewjasper
Avoid memory allocation when removing dead blocks
Use `reachable_as_bitset` to reuse a bitset from the traversal rather
than allocating it seprately. Additionally check if there are any
unreachable blocks before proceeding.
bors [Fri, 29 Jan 2021 10:11:16 +0000 (10:11 +0000)]
Auto merge of #81419 - rylev:canocalize-extern-entries, r=petrochenkov
Pre-canoncalize ExternLocation::ExactPaths
This stores pre-canacolized paths inside `ExternLocation::ExactPaths` so that we don't need to canoncalize them every time we want to compare them to source lib paths.
est31 [Thu, 28 Jan 2021 08:24:55 +0000 (09:24 +0100)]
Treat nightlies for a version as complete
This commit makes cfg(version) treat the nightlies
for version 1.n.0 as 1.n.0, even though that nightly
version might not have all stabilizations and features
of the released 1.n.0. This is done for greater
convenience for people who want to test a newly
stabilized feature on nightly.
For users who wish to pin nightlies, this commit adds
a -Z assume-incomplete-release option that they can
enable if there are any issues due to this change.
bors [Fri, 29 Jan 2021 04:06:38 +0000 (04:06 +0000)]
Auto merge of #81073 - ssomers:btree_owned_root_vs_dying, r=Mark-Simulacrum
BTreeMap: prevent tree from ever being owned by non-root node
This introduces a new marker type, `Dying`, which is used to note trees which are in the process of deallocation. On such trees, some fields may be in an inconsistent state as we are deallocating the tree. Unfortunately, there's not a great way to express conditional unsafety, so the methods for traversal can cause UB if not invoked correctly, but not marked as such. This is not a regression from the previous state, but rather isolates the destructive methods to solely being called on the dying state.
Camelid [Fri, 29 Jan 2021 02:00:07 +0000 (18:00 -0800)]
rustdoc: Remove unnecessary optional
Previously, the HTML output format was represented by both
`Some(OutputFormat::Html)` and `None` so there's no need to have an
optional. Instead, `OutputFormat::Html` is explicitly the default and we
no longer have a "tri-state enum".
bors [Fri, 29 Jan 2021 00:58:43 +0000 (00:58 +0000)]
Auto merge of #81493 - JohnTitor:rollup-sa4m4zh, r=JohnTitor
Rollup of 10 pull requests
Successful merges:
- #79570 (rustc: Stabilize `-Zrun-dsymutil` as `-Csplit-debuginfo`)
- #79819 (Add `SEMICOLON_IN_EXPRESSIONS_FROM_MACROS` lint)
- #79991 (rustdoc: Render HRTB correctly for bare functions)
- #80215 (Use -target when linking binaries for Mac Catalyst)
- #81158 (Point to span of upvar making closure FnMut)
- #81176 (Improve safety of `LateContext::qpath_res`)
- #81287 (Split rustdoc JSON types into separately versioned crate)
- #81306 (Fuse inner iterator in FlattenCompat and improve related tests)
- #81333 (clean up some const error reporting around promoteds)
- #81459 (Fix rustdoc text selection for page titles)
Retry of PR #81397 due to merge conflicts confusing me. Recreated the same changes as in that PR, but now when I run `./x.py doc library/std`, no changes I make to `src/librustdoc/html/render/mod.rs` to the file are reflected in the built doc files, even if I delete the `build` folder. I'm guessing there's some cache I'm missing?
Yuki Okushi [Fri, 29 Jan 2021 00:17:32 +0000 (09:17 +0900)]
Rollup merge of #81176 - camsteffen:qpath-res, r=oli-obk
Improve safety of `LateContext::qpath_res`
This is my first rustc code change, inspired by hacking on clippy!
The first change is to clear cached `TypeckResults` from `LateContext` when visiting a nested item. I took a hint from [here](https://github.com/rust-lang/rust/blob/5e91c4ecc09312d8b63d250a432b0f3ef83f1df7/compiler/rustc_privacy/src/lib.rs#L1300).
Clippy has a `qpath_res` util function to avoid a possible ICE in `LateContext::qpath_res`. But the docs of `LateContext::qpath_res` promise no ICE. So this updates the `LateContext` method to keep its promises, and removes the util function.
Related: rust-lang/rust-clippy#4545
CC ````````````@eddyb```````````` since you've done related work
CC ````````````@flip1995```````````` FYI
Yuki Okushi [Fri, 29 Jan 2021 00:17:32 +0000 (09:17 +0900)]
Rollup merge of #81176 - camsteffen:qpath-res, r=oli-obk
Improve safety of `LateContext::qpath_res`
This is my first rustc code change, inspired by hacking on clippy!
The first change is to clear cached `TypeckResults` from `LateContext` when visiting a nested item. I took a hint from [here](https://github.com/rust-lang/rust/blob/5e91c4ecc09312d8b63d250a432b0f3ef83f1df7/compiler/rustc_privacy/src/lib.rs#L1300).
Clippy has a `qpath_res` util function to avoid a possible ICE in `LateContext::qpath_res`. But the docs of `LateContext::qpath_res` promise no ICE. So this updates the `LateContext` method to keep its promises, and removes the util function.
Related: rust-lang/rust-clippy#4545
CC ````````````@eddyb```````````` since you've done related work
CC ````````````@flip1995```````````` FYI
Yuki Okushi [Fri, 29 Jan 2021 00:17:29 +0000 (09:17 +0900)]
Rollup merge of #80215 - visigoth:issue-80202-fix, r=estebank
Use -target when linking binaries for Mac Catalyst
When running `rustc` with `-target x86_64-apple-ios-macabi`, the linker
eventually gets run with `-arch x86_64`, because the linker back end splits the
LLVM target triple and uses the first token as the target architecture. However,
this does not work for the Mac Catalyst ABI, which is a separate target from
Darwin.
Specifying the full target triple with `-target` allows Mac Catalyst binaries to
link and run.