This PR constrains the support added for handling unevaluated consts in polymorphization (introduced in #75260) by:
- Skipping associated constants as this causes cycle errors.
- Skipping promoted constants when they contain `Self` as this ensures `T` is used in constants of the form `<Self as Foo<T>>`.
Due to an oversight on my part, when landing #75260 and #75255, some tests started failing when polymorphization was enabled that I didn't notice until after landing - this PR fixes the regressions from #75260.
Yuki Okushi [Tue, 11 Aug 2020 07:23:45 +0000 (16:23 +0900)]
Rollup merge of #75085 - lzutao:ip_union, r=cuviper
Transmute big endian `s6_addr` and `[u16; 8]`
The old code already made the assumption to reinterpret
`Ipv6Addr` as `[u16; 8]`.
Glibc, Linux, FreeBSD, Win32 all makes this assumption.
The main motivation of using union it to better optimize code.
Godbolt: https://rust.godbolt.org/z/b4bGvo
Const is introducing unsafe when transmuting.
bors [Tue, 11 Aug 2020 04:10:39 +0000 (04:10 +0000)]
Auto merge of #74621 - LukasKalbertodt:float-docs, r=GuillaumeGomez
Improve `f32` and `f64` primitive documentation
I noticed that the docs for the primitive floats were fairly short. I first only wanted to add the IEEE specification information (compare [the reference](https://doc.rust-lang.org/reference/types/numeric.html)), but then also added some more beginner-friendly docs. Let me know what you think!
Dylan DPC [Mon, 10 Aug 2020 23:56:39 +0000 (01:56 +0200)]
Rollup merge of #75339 - RalfJung:eval-required, r=oli-obk
evaluate required_consts when pushing stack frame in Miri engine
[Just like codegen](https://github.com/rust-lang/rust/pull/70820/files#diff-32c57af5c8e23eb048f55d1e955e5cd5R194), Miri needs to make sure all `required_consts` evaluate successfully, to catch post-monomorphization errors.
While at it I also moved the const_eval error reporting logic into rustc_mir::const_eval::error; there is no reason it should be in `rustc_middle`. I kept this in a separate commit for easier reviewing.
Helps with https://github.com/rust-lang/miri/issues/1382. I will add a test on the Miri side (done now: https://github.com/rust-lang/miri/pull/1504).
r? @oli-obk
Dylan DPC [Mon, 10 Aug 2020 23:56:38 +0000 (01:56 +0200)]
Rollup merge of #75337 - davidtwco:polymorphization-75255-fixes, r=eddyb
instance: only polymorphize upvar substs
This PR restricts the substitution polymorphization added in #75255 to only apply to the tupled upvar substitution, rather than all substitutions, fixing a bunch of regressions when polymorphization is
enabled.
Due to an oversight on my part, when landing #75260 and #75255, some tests started failing when polymorphization was enabled that I didn't notice until after landing - this PR fixes the regressions from #75255. #75336 has been filed to make sure that we don't forget to try make this change again in future, as it does enable some optimisations.
Dylan DPC [Mon, 10 Aug 2020 23:56:36 +0000 (01:56 +0200)]
Rollup merge of #75316 - alexcrichton:fix-wasm-simd, r=oli-obk
Don't try to use wasm intrinsics on vectors
This commit fixes an issue with #74695 where the fptosi and fptoui
specializations on wasm were accidentally used on vector types by the
`simd_cast` intrinsic. This issue showed up as broken CI for the stdsimd
crate. Here this commit simply skips the specialization on vector kinds
flowing into `fpto{s,u}i`.
Dylan DPC [Mon, 10 Aug 2020 23:56:34 +0000 (01:56 +0200)]
Rollup merge of #75315 - Mark-Simulacrum:save-temps, r=ecstatic-morse
Avoid deleting temporary files on error
Previously if the compiler error'd, fatally, then temporary directories which
should be preserved by -Csave-temps would be deleted due to fatal compiler
errors being implemented as panics.
Dylan DPC [Mon, 10 Aug 2020 23:56:30 +0000 (01:56 +0200)]
Rollup merge of #75098 - Ryan1729:clippy-pointer-cast-lint-experiment, r=oli-obk
Clippy pointer cast lint experiment
This PR is an experiment about exposing more parts of `rustc_typeck` for use in `clippy`. In particular, the code that checks where a cast is valid or not was exposed, which necessitated exposing [`FnCtxt`](https://doc.rust-lang.org/nightly/nightly-rustc/rustc_typeck/check/struct.FnCtxt.html), and figuring out how to create an instance of that type inside `clippy`.
This was prompted by [this clippy issue](https://github.com/rust-lang/rust-clippy/issues/2064).
bors [Mon, 10 Aug 2020 23:50:39 +0000 (23:50 +0000)]
Auto merge of #74005 - estebank:type-ascription-redux, r=petrochenkov
Clean up errors in typeck and resolve
* Tweak ordering of suggestions
* Do not suggest similarly named enclosing item
* Point at item definition in foreign crates
* Add missing primary label
bors [Mon, 10 Aug 2020 15:19:46 +0000 (15:19 +0000)]
Auto merge of #74953 - JulianKnodt:master, r=lcnr
Remove restriction on type parameters preceding consts w/ feature const-generics
Removed the restriction on type parameters preceding const parameters when the feature const-generics is enabled.
Builds on #74676, which deals with unsorted generic parameters. This just lifts the check in lowering the AST to HIR that permits consts and types to be reordered with respect to each other. Lifetimes still must precede both
This change is not intended for min-const-generics, and is gated behind the `#![feature(const_generics)]`.
One thing is that it also permits type parameters without a default to come after consts, which I expected to not work, and was hoping to get more guidance on whether that should be permitted or how to prevent it otherwise.
I did not go through the RFC process for this pull request because there was prior work to get this feature added. In the previous PR that was cited, work was done to enable this change.
This commit constrains the support added for handling unevaluated consts
in polymorphization (introduced in #75260) by:
- Skipping associated constants as this causes cycle errors.
- Skipping promoted constants when they contain `Self` as this ensures
`T` is used in constants of the form `<Self as Foo<T>>`.
bors [Mon, 10 Aug 2020 05:34:53 +0000 (05:34 +0000)]
Auto merge of #75204 - cuviper:freebsd11, r=Mark-Simulacrum
Upgrade the FreeBSD toolchain to version 11.4
FreeBSD 10 reached its end-of-life in October 2018, and that toolchain
caused issues in the LLVM 11 upgrade (#73526) that are resolved with the
toolchain from FreeBSD 11.
bors [Mon, 10 Aug 2020 00:09:45 +0000 (00:09 +0000)]
Auto merge of #75351 - JohnTitor:rollup-q9udsyx, r=JohnTitor
Rollup of 8 pull requests
Successful merges:
- #74200 (Std panicking unsafe block in unsafe fn)
- #75286 (Add additional case for Path starts with)
- #75318 (Resolve `char` as a primitive even if there is a module in scope)
- #75320 (Detect likely `for foo of bar` JS syntax)
- #75328 (Cleanup E0749)
- #75344 (Rename "Important traits" to "Notable traits")
- #75348 (Move to intra-doc links in library/core/src/time.rs)
- #75350 (Do not ICE when lowering invalid extern fn with bodies)
bors [Sun, 9 Aug 2020 22:08:02 +0000 (22:08 +0000)]
Auto merge of #75345 - Aaron1011:fix/no-std-hygiene, r=petrochenkov
Remove normalization of `Span` debug output in proc-macro tests
Fixes #74800
The definition of `is_x86_feature_detected!` (and similar macros)
depends on the platform - it is produced by a `cfg_if!` invocation on
x86, and a plain `#[cfg]` on other platforms. Since it is part of the
prelude, we will end up importing different hygiene information
depending on the platform. This previously required us to avoid printing raw
`SyntaxContext` ids in any tests that uses the standard library, since
the captured output will be platform-dependent.
Previously, we replaced all `SyntaxContext` ids with "#CTXT", and the
raw `Span` lo/hi bytes with "LO..HI".
This commit adds `#![no_std]` and `extern crate std` to all proc-macro
tests that print spans. This suppresses the prelude import, while
still using lang items from `std` (which gives us a buildable binary).
With this apporach, we will only load hygiene information for things
which we explicitly import. This lets us re-add
`-Z unpretty=expanded,hygiene`, since its output can now be made stable
across all platforms.
Additionally, we use `-Z span-debug` in more places, which lets us avoid
the "LO..HI" normalization hack.
`CowBoxSymStr` is a type that either holds a `SymbolStr` (which is much
the same as a `Symbol`), or an owned string. When computing skeletons,
a `SymbolStr` is stored if the skeleton is the same as the original
string, otherwise an owned string is stored.
So, basically, `CowBoxSymStr` is a type for string interning. But we
already have one of those: `Symbol` itself. This PR removes
`CowBoxSymStr`, using `Symbol` instead. A good thing about this is that
it avoids storing `SymbolStr` values in `skeleton_map`, something that
is discouraged.
The PR also inlines and removes the `calc_skeleton()` function because
that simplifies the code.
Fix symbol ordering for confusable idents detection.
Confusable idents detection uses a type `BTreeMap<Symbol, Span>`. This is
highly dubious given that `Symbol` doesn't guarantee a meaningful order. (In
practice, it currently gives an order that mostly matches source code order.)
As a result, changes in `Symbol` representation make the
`lint-confusable-idents.rs` test fail, because this error message:
> identifier pair considered confusable between `s` and `s`
is changed to this:
> identifier pair considered confusable between `s` and `s`
and the corresponding span pointers get swapped erroneously, leading to
an incorrect "previous identifier" label.
This commit sorts the relevant symbols by span before doing the checking,
which ensures that the ident that appears first in the code will be mentioned
first in the message. The commit also extends the test slightly to be more
thorough.
bors [Sun, 9 Aug 2020 20:08:50 +0000 (20:08 +0000)]
Auto merge of #75278 - cuviper:indexmap, r=Mark-Simulacrum
Upgrade indexmap and use it more
First this upgrades `indexmap` to 1.5.1, which is now based on `hashbrown::raw::RawTable`. This means it shares a lot of the same performance characteristics for insert, lookup, etc., while keeping items in insertion order.
Then across various rustc crates, this replaces a lot of `Vec`+`HashMap` pairs with a single `IndexMap` or `IndexSet`.
Aaron Hill [Sun, 9 Aug 2020 17:36:31 +0000 (13:36 -0400)]
Remove normalization of `Span` debug output in proc-macro tests
Fixes #74800
The definition of `is_x86_feature_detected!` (and similar macros)
depends on the platform - it is produced by a `cfg_if!` invocation on
x86, and a plain `#[cfg]` on other platforms. Since it is part of the
prelude, we will end up importing different hygiene information
depending on the platform. This previously required us to avoid printing raw
`SyntaxContext` ids in any tests that uses the standard library, since
the captured output will be platform-dependent.
Previously, we replaced all `SyntaxContext` ids with "#CTXT", and the
raw `Span` lo/hi bytes with "LO..HI".
This commit adds `#![no_std]` and `extern crate std` to all proc-macro
tests that print spans. This suppresses the prelude import, while
still using lang items from `std` (which gives us a buildable binary).
With this apporach, we will only load hygiene information for things
which we explicitly import. This lets us re-add
`-Z unpretty=expanded,hygiene`, since its output can now be made stable
across all platforms.
Additionally, we use `-Z span-debug` in more places, which lets us avoid
the "LO..HI" normalization hack.
bors [Sun, 9 Aug 2020 18:19:34 +0000 (18:19 +0000)]
Auto merge of #75137 - Aaron1011:fix/hygiene-skip-expndata, r=petrochenkov
Don't serialize ExpnData for foreign crates
When we encode an ExpnId into the crate metadata, we write out the
CrateNum of the crate that 'owns' the corresponding `ExpnData`, which
is later used to decode the `ExpnData` from its owning crate.
However, we current serialize the `ExpnData` for all `ExpnIds` that we
serialize, even if the `ExpnData` was already serialized into a foreign
crate. This commit skips encoding this kind of `ExpnData`, which should
hopefully speed up metadata encoding and reduce the total metadata size.
Josh Stone [Wed, 5 Aug 2020 22:58:35 +0000 (15:58 -0700)]
Upgrade the FreeBSD toolchain to version 11.4
FreeBSD 10 reached its end-of-life in October 2018, and that toolchain
caused issues in the LLVM 11 upgrade (#73526) that are resolved with the
toolchain from FreeBSD 11.
David Wood [Sun, 9 Aug 2020 13:53:33 +0000 (14:53 +0100)]
instance: only polymorphize upvar substs
This commit restricts the substitution polymorphization added in #75255
to only apply to the tupled upvar substitution, rather than all
substitutions, fixing a bunch of regressions when polymorphization is
enabled.
Mark Rousskov [Sun, 9 Aug 2020 01:05:50 +0000 (21:05 -0400)]
Avoid deleting temporary files on error
Previously if the compiler error'd, fatally, then temporary directories which
should be preserved by -Csave-temps would be deleted due to fatal compiler
errors being implemented as panics.