]> git.lizzy.rs Git - rust.git/log
rust.git
3 years agoAuto merge of #84507 - crlf0710:codegen_nonlocal_main_wrapper, r=nagisa
bors [Mon, 10 May 2021 00:42:31 +0000 (00:42 +0000)]
Auto merge of #84507 - crlf0710:codegen_nonlocal_main_wrapper, r=nagisa

Add primary marker on codegen unit and generate main wrapper on primary codegen.

This is the codegen part of changes extracted from #84062.

This add a marker called `primary` on each codegen units, where exactly one codegen unit will be `primary = true` at a time. This specific codegen unit will take charge of generating `main` wrapper when `main` is imported from a foreign crate after the implementation of RFC 1260.

cc #28937

I'm not sure who should i ask for review for codegen changes, so feel free to reassign.
r? `@nagisa`

3 years agomove logic to a function
hamidreza kalbasi [Sun, 9 May 2021 23:01:31 +0000 (03:31 +0430)]
move logic to a function

3 years agoUpdate rustdoc test
Guillaume Gomez [Sat, 8 May 2021 18:20:37 +0000 (20:20 +0200)]
Update rustdoc test

3 years agoEnd toggle migration
Guillaume Gomez [Sat, 8 May 2021 12:21:57 +0000 (14:21 +0200)]
End toggle migration

3 years agoAuto merge of #83800 - xobs:impl-16351-nightly, r=nagisa
bors [Sun, 9 May 2021 22:01:26 +0000 (22:01 +0000)]
Auto merge of #83800 - xobs:impl-16351-nightly, r=nagisa

Add default search path to `Target::search()`

The function `Target::search()` accepts a target triple and returns a `Target` struct defining the requested target.

There is a `// FIXME 16351: add a sane default search path?` comment that indicates it is desirable to include some sort of default. This was raised in https://github.com/rust-lang/rust/issues/16351 which was closed without any resolution.

https://github.com/rust-lang/rust/pull/31117 was proposed, however that has platform-specific logic that is unsuitable for systems without `/etc/`.

This patch implements the suggestion raised in https://github.com/rust-lang/rust/issues/16351#issuecomment-180878193 where a `target.json` file may be placed in `$(rustc --print sysroot)/lib/rustlib/<target-triple>/target.json`. This allows shipping a toolchain distribution as a single file that gets extracted to the sysroot.

3 years agoFix `Step` feature flag, make tidy lint more useful to find things like this
ltdk [Sun, 9 May 2021 21:13:22 +0000 (17:13 -0400)]
Fix `Step` feature flag, make tidy lint more useful to find things like this

3 years agoImplement @jackh726's suggestions
Fabian Wolff [Sun, 9 May 2021 20:31:49 +0000 (22:31 +0200)]
Implement @jackh726's suggestions

3 years agoMake unchecked_{add,sub,mul} inherent methods unstably const
ltdk [Sun, 9 May 2021 00:18:44 +0000 (20:18 -0400)]
Make unchecked_{add,sub,mul} inherent methods unstably const

3 years agoAuto merge of #85065 - Mark-Simulacrum:preds-no-alloc, r=jackh726
bors [Sun, 9 May 2021 19:01:05 +0000 (19:01 +0000)]
Auto merge of #85065 - Mark-Simulacrum:preds-no-alloc, r=jackh726

Avoid predecessors having Drop impls

Should be a simple win - but let's make sure with perf.

3 years agoRemove outdated FIXME for download-rustc
Joshua Nelson [Sun, 9 May 2021 16:24:59 +0000 (12:24 -0400)]
Remove outdated FIXME for download-rustc

3 years agoAuto merge of #83894 - nikic:newpm, r=nagisa
bors [Sun, 9 May 2021 16:19:21 +0000 (16:19 +0000)]
Auto merge of #83894 - nikic:newpm, r=nagisa

Improve support for NewPM

This adds various missing bits of support for NewPM and allows us to successfully run stage 2 tests with NewPM enabled.

This does not yet enable NewPM by default, as there are still known issue on LLVM 12 (such as a weak fat LTO pipeline). The plan is to make the switch after we update to LLVM 13.

3 years agoImprove "panic message is not a string literal" warning
Paul Trojahn [Fri, 7 May 2021 17:09:15 +0000 (19:09 +0200)]
Improve "panic message is not a string literal" warning

This warning always referenced panic! even in case of an
assert. Related to #84656

3 years agomore erroneous-const tests
Ralf Jung [Sun, 9 May 2021 14:07:00 +0000 (16:07 +0200)]
more erroneous-const tests

3 years agoAuto merge of #83596 - jyn514:session-dead-code, r=oli-obk
bors [Sun, 9 May 2021 13:38:17 +0000 (13:38 +0000)]
Auto merge of #83596 - jyn514:session-dead-code, r=oli-obk

Remove dead or useless code from Session

This is a more principled follow-up to https://github.com/rust-lang/rust/pull/83185#discussion_r601753839.

- Rename `Parser::span_fatal_err` -> `Parser::span_err`
- Remove some unnecessary uses of `struct_span_fatal`
- Make `Diagnostic::span_fatal` unconditionally raise an error
- Add `impl Deref<Target = Handler>` for Session and remove all functions that are exactly the same as their Handler counterparts
- Note why `Handler::fatal` is different from `Sesssion::fatal`
- Remove unused `opt_span_warn` function

r? `@oli-obk` or `@estebank`

3 years agoensure failing promoteds in const/static bodies are handled correctly
Ralf Jung [Sun, 9 May 2021 12:56:19 +0000 (14:56 +0200)]
ensure failing promoteds in const/static bodies are handled correctly

3 years agoremove const_fn feature gate
Ralf Jung [Fri, 7 May 2021 16:56:16 +0000 (18:56 +0200)]
remove const_fn feature gate

3 years agoAuto merge of #85093 - camelid:remove-fake-expect_local, r=GuillaumeGomez
bors [Sun, 9 May 2021 11:17:17 +0000 (11:17 +0000)]
Auto merge of #85093 - camelid:remove-fake-expect_local, r=GuillaumeGomez

Remove `FakeDefId::expect_local()`

This function returned a fake `DefIndex`, with no indication that it was
fake, when it was provided with a `FakeDefId::Fake`. Every use of the
function uses the returned `DefIndex` in a call to
`tcx.local_def_id_to_hir_id()`, which I'm pretty sure would panic if it
were given a fake `DefIndex`.

I removed the function and replaced all calls to it with a call to
`expect_real()` followed by `DefId::expect_local()` (that's a function
on the *real* `DefId`).

3 years agoFix typo
hi-rustin [Sun, 9 May 2021 04:24:58 +0000 (12:24 +0800)]
Fix typo

3 years agoRemove unused `opt_span_warn` function
Joshua Nelson [Sun, 28 Mar 2021 03:00:34 +0000 (23:00 -0400)]
Remove unused `opt_span_warn` function

3 years agoNote why `Handler::fatal` is different from `Sesssion::fatal`
Joshua Nelson [Sun, 28 Mar 2021 02:58:25 +0000 (22:58 -0400)]
Note why `Handler::fatal` is different from `Sesssion::fatal`

3 years agoMake `Diagnostic::span_fatal` unconditionally raise an error
Joshua Nelson [Sun, 28 Mar 2021 02:45:01 +0000 (22:45 -0400)]
Make `Diagnostic::span_fatal` unconditionally raise an error

It had no callers which didn't immediately call `raise()`, and this
unifies the behavior with `Session`.

3 years agoRemove some unnecessary uses of `struct_span_fatal`
Joshua Nelson [Sun, 28 Mar 2021 02:48:21 +0000 (22:48 -0400)]
Remove some unnecessary uses of `struct_span_fatal`

All of them immediately called `emit()` then `raise()`, so they could
just call `span_fatal` directly.

3 years agoRename `Parser::span_fatal_err` -> `Parser::span_err`
Joshua Nelson [Sun, 28 Mar 2021 02:46:50 +0000 (22:46 -0400)]
Rename `Parser::span_fatal_err` -> `Parser::span_err`

The name was misleading, it wasn't actually a fatal error.

3 years agoAdd primary marker on codegen unit to take charge of main_wrapper for non-local cases.
Charles Lew [Sat, 24 Apr 2021 05:16:34 +0000 (13:16 +0800)]
Add primary marker on codegen unit to take charge of main_wrapper for non-local cases.

3 years agoTry to fix issue 68049
hamidreza kalbasi [Sun, 9 May 2021 00:56:59 +0000 (05:26 +0430)]
Try to fix issue 68049

3 years agoAuto merge of #83089 - Aaron1011:bootstrap-semi, r=Mark-Simulacrum
bors [Sat, 8 May 2021 23:11:45 +0000 (23:11 +0000)]
Auto merge of #83089 - Aaron1011:bootstrap-semi, r=Mark-Simulacrum

Enable `-W semicolon_in_expressions_from_macros` in bootstrap

Now that stdarch has been updated, we can do this without breaking the
build.

3 years agoAdd `#[track_caller]` to `FakeDefId::expect_real()`
Camelid [Sat, 8 May 2021 22:14:21 +0000 (15:14 -0700)]
Add `#[track_caller]` to `FakeDefId::expect_real()`

Now, in the case that the function is not inlined, the panic location
will be the caller's location, which is more helpful since the panic is
not `expect_real()`'s fault.

3 years agoRemove `FakeDefId::expect_local()`
Camelid [Sat, 8 May 2021 22:03:15 +0000 (15:03 -0700)]
Remove `FakeDefId::expect_local()`

This function returned a fake `DefIndex`, with no indication that it was
fake, when it was provided with a `FakeDefId::Fake`. Every use of the
function uses the returned `DefIndex` in a call to
`tcx.local_def_id_to_hir_id()`, which I'm pretty sure would panic if it
were given a fake `DefIndex`.

I removed the function and replaced all calls to it with a call to
`expect_real()` followed by `DefId::expect_local()` (that's a function
on the *real* `DefId`).

3 years agoEnable `-W semicolon_in_expressions_from_macros` in bootstrap
Aaron Hill [Sat, 13 Mar 2021 17:46:33 +0000 (12:46 -0500)]
Enable `-W semicolon_in_expressions_from_macros` in bootstrap

Now that stdarch has been updated, we can do this without breaking the
build.

3 years agoAuto merge of #83278 - Amanieu:bump_stdarch, r=Mark-Simulacrum
bors [Sat, 8 May 2021 18:41:16 +0000 (18:41 +0000)]
Auto merge of #83278 - Amanieu:bump_stdarch, r=Mark-Simulacrum

Bump stdarch submodule

Major changes:
- More AVX-512 intrinsics.
- More ARM & AArch64 NEON intrinsics.
- Updated unstable WASM intrinsics to latest draft standards.
- Intrinsics that previously used `#[rustc_args_required_const]` now use const generics. See #83167 for more details.
- `std_detect` is now a separate crate instead of a submodule of `std`.

3 years agoBump stdarch submodule
Amanieu d'Antras [Tue, 13 Apr 2021 03:32:11 +0000 (04:32 +0100)]
Bump stdarch submodule

3 years agoError on conflicting `#[doc(inline)]`/`#[doc(no_inline)]` attributes
LeSeulArtichaut [Sat, 8 May 2021 14:57:10 +0000 (16:57 +0200)]
Error on conflicting `#[doc(inline)]`/`#[doc(no_inline)]` attributes

3 years agoEmit `invalid_doc_attributes` warnings in more cases
LeSeulArtichaut [Sat, 8 May 2021 14:25:00 +0000 (16:25 +0200)]
Emit `invalid_doc_attributes` warnings in more cases

3 years agoAuto merge of #85072 - hyd-dev:miri, r=RalfJung
bors [Sat, 8 May 2021 12:11:01 +0000 (12:11 +0000)]
Auto merge of #85072 - hyd-dev:miri, r=RalfJung

Update Miri

Fixes #85027

r? `@RalfJung`

3 years agoUpdate Miri
hyd-dev [Sat, 8 May 2021 11:53:15 +0000 (19:53 +0800)]
Update Miri

3 years agoAuto merge of #84470 - jyn514:lints, r=GuillaumeGomez
bors [Sat, 8 May 2021 09:46:18 +0000 (09:46 +0000)]
Auto merge of #84470 - jyn514:lints, r=GuillaumeGomez

rustdoc: Stop special casing `broken_intra_doc_links` unnecessarily

3 years agoUpdate LLVM submodule
Nikita Popov [Sat, 8 May 2021 09:01:09 +0000 (11:01 +0200)]
Update LLVM submodule

This updates the LLVM submodule with recent LLVM 12.x fixes. In
particular, it resolves an assertion failure when targeting
AArch64 at O0.

3 years agoSupport -C passes in NewPM
Nikita Popov [Mon, 5 Apr 2021 13:37:11 +0000 (15:37 +0200)]
Support -C passes in NewPM

And report an error if parsing the additional pass pipeline fails.
Threading through the error accounts for most of the changes here.

3 years agoDon't force -O1 with ThinLTO
Nikita Popov [Mon, 5 Apr 2021 09:12:45 +0000 (11:12 +0200)]
Don't force -O1 with ThinLTO

This doesn't seem to be necessary anymore, although I don't know
at which point or why that changed.

Forcing -O1 makes some tests fail under NewPM, because NewPM also
performs inlining at -O1, so it ends up performing much more
optimization in practice than before.

3 years agoDrop -opt-bisect-limit=0 flag from test
Nikita Popov [Sun, 4 Apr 2021 19:57:03 +0000 (21:57 +0200)]
Drop -opt-bisect-limit=0 flag from test

This causes an assertion failure under NewPM, because it also ends
up disabling the NameAnonGlobals pass.

Instead pass -Copt-level=0 to disable optimizations. If that should
be insufficient, we can use -C no-prepopulate-passes.

3 years agoExplicitly register GCOV profiling pass as well
Nikita Popov [Mon, 5 Apr 2021 09:26:48 +0000 (11:26 +0200)]
Explicitly register GCOV profiling pass as well

3 years agoExplicitly register instrprof pass
Nikita Popov [Mon, 5 Apr 2021 08:45:04 +0000 (10:45 +0200)]
Explicitly register instrprof pass

Don't use "passes" for this purpose, explicitly insert it into
the correct place in the pipeline instead.

3 years agoMake -Z new-llvm-pass-manager an Option<bool>
Nikita Popov [Sun, 4 Apr 2021 19:38:34 +0000 (21:38 +0200)]
Make -Z new-llvm-pass-manager an Option<bool>

To allow it to have an LLVM version dependent default.

3 years agoAuto merge of #85059 - ehuss:update-cargo, r=ehuss
bors [Sat, 8 May 2021 05:50:00 +0000 (05:50 +0000)]
Auto merge of #85059 - ehuss:update-cargo, r=ehuss

Update cargo

7 commits in f3e13226d6d17a2bc5f325303494b43a45f53b7f..e51522ab3db23b0d8f1de54eb1f0113924896331
2021-04-30 21:50:27 +0000 to 2021-05-07 21:29:52 +0000
- Add CARGO_TARGET_TMPDIR env var for integration tests & benches (rust-lang/cargo#9375)
- Bump to 0.55.0, update changelog (rust-lang/cargo#9464)
- Some updates to the unstable documentation (rust-lang/cargo#9457)
- Add CARGO_PROFILE_&lt;name&gt;_SPLIT_DEBUGINFO to env docs. (rust-lang/cargo#9456)
- Add `report` subcommand. (rust-lang/cargo#9438)
- Respect Cargo.toml `[package.exclude]` even not in a git repo. (rust-lang/cargo#9186)
- Document the other crates in the codebase in the contrib guide. (rust-lang/cargo#9439)

3 years agoAuto merge of #85058 - Xanewok:update-rls, r=Xanewok
bors [Sat, 8 May 2021 02:57:40 +0000 (02:57 +0000)]
Auto merge of #85058 - Xanewok:update-rls, r=Xanewok

Update RLS

This mostly just includes https://github.com/rust-lang/rls/commit/e33f4e68496b296dedb100e297dc4451f169b2b3 so that this fixes #85055 (clippy-related breakage).

3 years agoAvoid predecessors having Drop impls
Mark Rousskov [Sat, 8 May 2021 01:00:03 +0000 (21:00 -0400)]
Avoid predecessors having Drop impls

3 years agoAddress comments
hi-rustin [Sat, 8 May 2021 01:02:22 +0000 (09:02 +0800)]
Address comments

3 years agoAuto merge of #85057 - Dylan-DPC:rollup-efaseq2, r=Dylan-DPC
bors [Sat, 8 May 2021 00:37:38 +0000 (00:37 +0000)]
Auto merge of #85057 - Dylan-DPC:rollup-efaseq2, r=Dylan-DPC

Rollup of 6 pull requests

Successful merges:

 - #76808 (Improve diagnostics for functions in `struct` definitions)
 - #84887 (Remove SpanInterner::get)
 - #85034 (fix null pointer error messages)
 - #85038 (Don't stop running rustdoc-gui tests at first failure)
 - #85044 (Use `path.exists()` instead of `fs::metadata(path).is_ok()`)
 - #85052 (rustdoc: Link to the docs on namespaces when an unknown disambiguator is found)

Failed merges:

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

3 years agoUpdate cargo
Eric Huss [Fri, 7 May 2021 23:51:52 +0000 (16:51 -0700)]
Update cargo

3 years agoUpdate RLS
Igor Matuszewski [Fri, 7 May 2021 23:10:51 +0000 (01:10 +0200)]
Update RLS

3 years agoRollup merge of #85052 - jyn514:disambiguator, r=camelid
Dylan DPC [Fri, 7 May 2021 23:06:27 +0000 (01:06 +0200)]
Rollup merge of #85052 - jyn514:disambiguator, r=camelid

rustdoc: Link to the docs on namespaces when an unknown disambiguator is found

This was reverted in https://github.com/rust-lang/rust/pull/84950; this re-lands the changes, but without different behavior depending on the channel.

r? `@camelid` cc `@pietroalbini`

3 years agoRollup merge of #85044 - ChrisDenton:file-exists, r=jackh726
Dylan DPC [Fri, 7 May 2021 23:06:26 +0000 (01:06 +0200)]
Rollup merge of #85044 - ChrisDenton:file-exists, r=jackh726

Use `path.exists()` instead of `fs::metadata(path).is_ok()`

It's more explicit and potentially allows platforms to optimize the existence check.

3 years agoRollup merge of #85038 - GuillaumeGomez:dont-stop-first-rustdoc-gui-failure, r=Mark...
Dylan DPC [Fri, 7 May 2021 23:06:25 +0000 (01:06 +0200)]
Rollup merge of #85038 - GuillaumeGomez:dont-stop-first-rustdoc-gui-failure, r=Mark-Simulacrum

Don't stop running rustdoc-gui tests at first failure

I just realized that before this PR, the rustdoc-gui test suite was stopping at the first failure, which isn't very convenient. All tests are now running and if one failed, it returns an error at the end once all tests have run.

r? `@Mark-Simulacrum`

3 years agoRollup merge of #85034 - RalfJung:null-ptr, r=oli-obk
Dylan DPC [Fri, 7 May 2021 23:06:24 +0000 (01:06 +0200)]
Rollup merge of #85034 - RalfJung:null-ptr, r=oli-obk

fix null pointer error messages

I got the `match` in https://github.com/rust-lang/rust/pull/84972 wrong, this fixes that.
Also avoid redundant "inbounds test failed:" prefix.

Should fix https://github.com/rust-lang/rust/issues/85027

r? `@oli-obk`

3 years agoRollup merge of #84887 - jyn514:index-span, r=Xanewok
Dylan DPC [Fri, 7 May 2021 23:06:23 +0000 (01:06 +0200)]
Rollup merge of #84887 - jyn514:index-span, r=Xanewok

Remove SpanInterner::get

- It's used exactly once, so it's trivial to replace
- It doesn't match the normal convention for containers: normally
`get()` returns an option and indexing panics. Instead `SpanInterner::get()` panics
  and there's no indexing operation available.

3 years agoRollup merge of #76808 - LeSeulArtichaut:diagnose-functions-struct, r=jackh726
Dylan DPC [Fri, 7 May 2021 23:06:22 +0000 (01:06 +0200)]
Rollup merge of #76808 - LeSeulArtichaut:diagnose-functions-struct, r=jackh726

Improve diagnostics for functions in `struct` definitions

Tries to implement #76421.
This is probably going to need unit tests, but I wanted to hear from review all the cases tests should cover.

I'd like to follow up with the "mechanically applicable suggestion here that adds an impl block" step, but I'd need guidance. My idea for now would be to try to parse a function, and if that succeeds, create a dummy `ast::Item` impl block to then format it using `pprust`. Would that be a viable approach? Is there a better alternative?

r? `@matklad` cc `@estebank`

3 years agoFix duplicate unknown lint errors
Cameron Steffen [Fri, 7 May 2021 19:53:02 +0000 (14:53 -0500)]
Fix duplicate unknown lint errors

3 years agoAuto merge of #83901 - cjgillot:hash-attr, r=wesleywiser
bors [Fri, 7 May 2021 21:56:31 +0000 (21:56 +0000)]
Auto merge of #83901 - cjgillot:hash-attr, r=wesleywiser

Integrate attributes as part of the crate hash

r? `@Aaron1011`

3 years agorustdoc: Link to the docs on namespaces when an unknown disambiguator is found
Joshua Nelson [Mon, 5 Apr 2021 00:02:32 +0000 (20:02 -0400)]
rustdoc: Link to the docs on namespaces when an unknown disambiguator is found

3 years agoAllow checking miri and RLS with `x.py check src/tools/{miri,rls}`
Joshua Nelson [Fri, 7 May 2021 20:06:01 +0000 (16:06 -0400)]
Allow checking miri and RLS with `x.py check src/tools/{miri,rls}`

3 years agoAlso take unions and enums into account
LeSeulArtichaut [Fri, 7 May 2021 20:49:47 +0000 (22:49 +0200)]
Also take unions and enums into account

3 years agoAdd basic test
LeSeulArtichaut [Thu, 17 Sep 2020 16:44:09 +0000 (18:44 +0200)]
Add basic test

3 years agoImprove diagnostics for functions in `struct` definitions
LeSeulArtichaut [Wed, 16 Sep 2020 21:10:05 +0000 (23:10 +0200)]
Improve diagnostics for functions in `struct` definitions

3 years agoHash attributes as part of the crate hash.
Camille GILLOT [Mon, 5 Apr 2021 20:09:49 +0000 (22:09 +0200)]
Hash attributes as part of the crate hash.

3 years agoAuto merge of #84980 - flip1995:clippyup, r=Manishearth
bors [Fri, 7 May 2021 19:28:03 +0000 (19:28 +0000)]
Auto merge of #84980 - flip1995:clippyup, r=Manishearth

Update Clippy

Bi-weekly Clippy update.

r? `@Manishearth`

3 years agoFix suggestions for missing return type lifetime parameters
Fabian Wolff [Fri, 7 May 2021 17:44:32 +0000 (19:44 +0200)]
Fix suggestions for missing return type lifetime parameters

3 years agoUse `path.exists()` instead of `fs::metadata(path).is_ok()`
Chris Denton [Fri, 7 May 2021 16:32:40 +0000 (17:32 +0100)]
Use `path.exists()` instead of `fs::metadata(path).is_ok()`

It's more explicit and allows platforms to optimize the existence check.

3 years agoAuto merge of #85036 - Dylan-DPC:rollup-4qfabo3, r=Dylan-DPC
bors [Fri, 7 May 2021 16:24:46 +0000 (16:24 +0000)]
Auto merge of #85036 - Dylan-DPC:rollup-4qfabo3, r=Dylan-DPC

Rollup of 9 pull requests

Successful merges:

 - #84254 (illumos should put libc last in library search order)
 - #84442 (Unify rustc and rustdoc parsing of `cfg()`)
 - #84655 (Cleanup of `wasm`)
 - #84866 (linker: Avoid library duplication with `/WHOLEARCHIVE`)
 - #84930 (rename LLVM target for RustyHermit)
 - #84991 (rustc: Support Rust-specific features in -Ctarget-feature)
 - #85029 (SGX mutex is movable)
 - #85030 (Rearrange SGX split module files)
 - #85033 (some further small cleanups)

Failed merges:

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

3 years agobless 32bit
Ralf Jung [Fri, 7 May 2021 15:24:50 +0000 (17:24 +0200)]
bless 32bit

3 years agoDon't stop running rustdoc-gui tests at first failure
Guillaume Gomez [Fri, 7 May 2021 14:51:00 +0000 (16:51 +0200)]
Don't stop running rustdoc-gui tests at first failure

3 years agoRollup merge of #85033 - lcnr:nits2, r=jackh726
Dylan DPC [Fri, 7 May 2021 14:19:25 +0000 (16:19 +0200)]
Rollup merge of #85033 - lcnr:nits2, r=jackh726

some further small cleanups

3 years agoRollup merge of #85030 - jethrogb:jb/sgx-rearrange-files, r=nagisa
Dylan DPC [Fri, 7 May 2021 14:19:24 +0000 (16:19 +0200)]
Rollup merge of #85030 - jethrogb:jb/sgx-rearrange-files, r=nagisa

Rearrange SGX split module files

In #75979 several inlined modules were split out into multiple files.
This PR keeps the multiple files but moves a few things around to
organize things in a coherent way.

3 years agoRollup merge of #85029 - jethrogb:jb/sgx-movable-mutex, r=m-ou-se
Dylan DPC [Fri, 7 May 2021 14:19:23 +0000 (16:19 +0200)]
Rollup merge of #85029 - jethrogb:jb/sgx-movable-mutex, r=m-ou-se

SGX mutex is movable

r? ``@m-ou-se``

3 years agoRollup merge of #84991 - alexcrichton:target-feature-remap, r=nagisa
Dylan DPC [Fri, 7 May 2021 14:19:22 +0000 (16:19 +0200)]
Rollup merge of #84991 - alexcrichton:target-feature-remap, r=nagisa

rustc: Support Rust-specific features in -Ctarget-feature

Since the beginning of time the `-Ctarget-feature` flag on the command
line has largely been passed unmodified to LLVM. Afterwards, though, the
`#[target_feature]` attribute was stabilized and some of the names in
this attribute do not match the corresponding LLVM name. This is because
Rust doesn't always want to stabilize the exact feature name in LLVM for
the equivalent functionality in Rust. This creates a situation, however,
where in Rust you'd write:

    #[target_feature(enable = "pclmulqdq")]
    unsafe fn foo() {
        // ...
    }

but on the command line you would write:

    RUSTFLAGS="-Ctarget-feature=+pclmul" cargo build --release

This difference is somewhat odd to deal with if you're a newcomer and
the situation may be made worse with upcoming features like [WebAssembly
SIMD](https://github.com/rust-lang/rust/issues/74372) which may be more
prevalent.

This commit implements a mapping to translate requests via
`-Ctarget-feature` through the same name-mapping functionality that's
present for attributes in Rust going to LLVM. This means that
`+pclmulqdq` will work on x86 targets where as previously it did not.

I've attempted to keep this backwards-compatible where the compiler will
just opportunistically attempt to remap features found in
`-Ctarget-feature`, but if there's something it doesn't understand it
gets passed unmodified to LLVM just as it was before.

3 years agoRollup merge of #84930 - hermitcore:target, r=nagisa
Dylan DPC [Fri, 7 May 2021 14:19:21 +0000 (16:19 +0200)]
Rollup merge of #84930 - hermitcore:target, r=nagisa

rename LLVM target for RustyHermit

- RustyHermit is a library operating system, where the user-
  and the kernel-space use the same target
- by a mistake a previous patch changes the target to an incorect value
- this merge request revert the previous changes

3 years agoRollup merge of #84866 - petrochenkov:wholesome, r=Mark-Simulacrum
Dylan DPC [Fri, 7 May 2021 14:19:20 +0000 (16:19 +0200)]
Rollup merge of #84866 - petrochenkov:wholesome, r=Mark-Simulacrum

linker: Avoid library duplication with `/WHOLEARCHIVE`

Looks like in #72785 I misinterpreted how the `link.exe`'s `/WHOLEARCHIVE` flag works.

It's not necessary to write `mylib /WHOLEARCHIVE:mylib` to mark `mylib` as whole archive, `/WHOLEARCHIVE:mylib` alone is enough.

https://docs.microsoft.com/en-us/cpp/build/reference/wholearchive-include-all-library-object-files?view=msvc-160

3 years agoRollup merge of #84655 - CDirkx:wasm, r=m-ou-se
Dylan DPC [Fri, 7 May 2021 14:19:20 +0000 (16:19 +0200)]
Rollup merge of #84655 - CDirkx:wasm, r=m-ou-se

Cleanup of `wasm`

Some more cleanup of `sys`, this time `wasm`

- Reuse `unsupported::args` (functionally equivalent implementation, just an empty iterator).
- Split out `atomics` implementation of `wasm::thread`, the non-`atomics` implementation is reused from `unsupported`.
- Move all of the `atomics` code to a separate directory `wasm/atomics`.

````@rustbot```` label: +T-libs-impl
r? ````@m-ou-se````

3 years agoRollup merge of #84442 - jyn514:doc-cfg, r=petrochenkov
Dylan DPC [Fri, 7 May 2021 14:19:19 +0000 (16:19 +0200)]
Rollup merge of #84442 - jyn514:doc-cfg, r=petrochenkov

Unify rustc and rustdoc parsing of `cfg()`

This extracts a new `parse_cfg` function that's used between both.

- Treat `#[doc(cfg(x), cfg(y))]` the same as `#[doc(cfg(x)]
  #[doc(cfg(y))]`. Previously it would be completely ignored.
- Treat `#[doc(inline, cfg(x))]` the same as `#[doc(inline)]
  #[doc(cfg(x))]`. Previously, the cfg would be ignored.
- Pass the cfg predicate through to rustc_expand to be validated

Technically this is a breaking change, but doc_cfg is still nightly so I don't think it matters.

Fixes https://github.com/rust-lang/rust/issues/84437.

r? `````````@petrochenkov`````````

3 years agoRollup merge of #84254 - jclulow:illumos-link-order, r=petrochenkov
Dylan DPC [Fri, 7 May 2021 14:19:18 +0000 (16:19 +0200)]
Rollup merge of #84254 - jclulow:illumos-link-order, r=petrochenkov

illumos should put libc last in library search order

Under some conditions, the toolchain will produce a sequence of linker
arguments that result in a NEEDED list that puts libc before libgcc_s;
e.g.,

    [0]  NEEDED            0x2046ba            libc.so.1
    [1]  NEEDED            0x204723            libm.so.2
    [2]  NEEDED            0x204736            libsocket.so.1
    [3]  NEEDED            0x20478b            libumem.so.1
    [4]  NEEDED            0x204763            libgcc_s.so.1

Both libc and libgcc_s provide an unwinder implementation, but libgcc_s
provides some extra symbols upon which Rust directly depends.  If libc
is first in the NEEDED list we will find some of those symbols in libc
but others in libgcc_s, resulting in undefined behaviour as the two
implementations do not use compatible interior data structures.

This solution is not perfect, but is the simplest way to produce correct
binaries on illumos for now.

3 years agofix null pointer error messages
Ralf Jung [Fri, 7 May 2021 13:11:18 +0000 (15:11 +0200)]
fix null pointer error messages

3 years agoRearrange SGX split module files
Jethro Beekman [Fri, 7 May 2021 11:40:43 +0000 (13:40 +0200)]
Rearrange SGX split module files

In #75979 several inlined modules were split out into multiple files.
This PR keeps the multiple files but moves a few things around to
organize things in a coherent way.

3 years agoSGX mutex is movable
Jethro Beekman [Fri, 7 May 2021 11:21:38 +0000 (13:21 +0200)]
SGX mutex is movable

3 years agofulfillment typo
lcnr [Fri, 7 May 2021 10:17:58 +0000 (12:17 +0200)]
fulfillment typo

3 years agowfcheck fcx <-> tcx cleanup
lcnr [Fri, 7 May 2021 10:15:38 +0000 (12:15 +0200)]
wfcheck fcx <-> tcx cleanup

3 years agoAuto merge of #84797 - richkadel:cover-unreachable-statements, r=tmandry
bors [Fri, 7 May 2021 10:06:40 +0000 (10:06 +0000)]
Auto merge of #84797 - richkadel:cover-unreachable-statements, r=tmandry

Report coverage `0` of dead blocks

Fixes: #84018
With `-Z instrument-coverage`, coverage reporting of dead blocks
(for example, blocks dropped because a conditional branch is dropped,
based on const evaluation) is now supported.

If `instrument-coverage` is enabled, `simplify::remove_dead_blocks()`
finds all dropped coverage `Statement`s and adds their `code_region`s as
`Unreachable` coverage `Statement`s to the `START_BLOCK`, so they are
still included in the coverage map.

Check out the resulting changes in the test coverage reports in this PR (in [commit 1](https://github.com/rust-lang/rust/pull/84797/commits/0b0d293c7c46bdadf80e5304a667e34c53c0cf7e)).

r? `@tmandry`
cc: `@wesleywiser`

3 years agoBTree: no longer copy keys and values before dropping them
Stein Somers [Mon, 3 May 2021 09:54:15 +0000 (11:54 +0200)]
BTree: no longer copy keys and values before dropping them

3 years agorename LLVM target for RustyHermit
Stefan Lankes [Tue, 4 May 2021 23:03:50 +0000 (01:03 +0200)]
rename LLVM target for RustyHermit

RustyHermit ist is a library operating system. In this case, we link a static library as kernel to the application. The final result is a bootable application. The library and the application have to use the same target. Currently, the targets are different (see also https://github.com/rust-lang/rust/blob/master/compiler/rustc_target/src/spec/x86_64_unknown_hermit.rs). Consequently, this commit change the LLVM target to 'hermit'.

This kernel spec is needed to disable the usage of FPU registers, which are not allowed in kernel space. In contrast to Linux, everything is running in ring 0 and also in the same address space.

Signed-off-by: Stefan Lankes <slankes@eonerc.rwth-aachen.de>
3 years agoAuto merge of #85022 - JohnTitor:rollup-2q4706v, r=JohnTitor
bors [Fri, 7 May 2021 07:44:53 +0000 (07:44 +0000)]
Auto merge of #85022 - JohnTitor:rollup-2q4706v, r=JohnTitor

Rollup of 9 pull requests

Successful merges:

 - #84779 (Add support for --test-args to cargotest)
 - #84781 (Don't check bootstrap artifacts by default)
 - #84787 (bump deps)
 - #84815 (Update coverage docs and command line help)
 - #84875 (Removes unneeded check of `#[no_coverage]` in mapgen)
 - #84897 (Coverage instruments closure bodies in macros (not the macro body))
 - #84911 (Retry clang+llvm download)
 - #84972 (CTFE inbounds-error-messages tweak)
 - #84990 (Sort rustdoc-gui tests)

Failed merges:

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

3 years agoarray-to-raw-elem cast: test that Retag covers entire array
Ralf Jung [Fri, 7 May 2021 07:36:45 +0000 (09:36 +0200)]
array-to-raw-elem cast: test that Retag covers entire array

3 years agoRollup merge of #84990 - GuillaumeGomez:sort-rustdoc-gui-tests, r=Mark-Simulacrum
Yuki Okushi [Fri, 7 May 2021 06:20:30 +0000 (15:20 +0900)]
Rollup merge of #84990 - GuillaumeGomez:sort-rustdoc-gui-tests, r=Mark-Simulacrum

Sort rustdoc-gui tests

The rustdoc-gui tests were randomly run. Not really a big issue but I prefer the tests to be sorted.

3 years agoRollup merge of #84972 - RalfJung:null-ptr-msg, r=oli-obk
Yuki Okushi [Fri, 7 May 2021 06:20:29 +0000 (15:20 +0900)]
Rollup merge of #84972 - RalfJung:null-ptr-msg, r=oli-obk

CTFE inbounds-error-messages tweak

* use CheckInAllocMsg::PointerArithmeticTest for ptr_offset error
* nicer errors for some null pointer cases

r? `@oli-obk`

3 years agoRollup merge of #84911 - Mark-Simulacrum:retry-clang, r=pietroalbini
Yuki Okushi [Fri, 7 May 2021 06:20:28 +0000 (15:20 +0900)]
Rollup merge of #84911 - Mark-Simulacrum:retry-clang, r=pietroalbini

Retry clang+llvm download

We've been seeing a pretty high rate of spurious network failures (e.g., openssl
connection reset by peer). Not clear why, but let's add a retry.

r? `@pietroalbini`

3 years agoRollup merge of #84897 - richkadel:cover-closure-macros, r=tmandry
Yuki Okushi [Fri, 7 May 2021 06:20:26 +0000 (15:20 +0900)]
Rollup merge of #84897 - richkadel:cover-closure-macros, r=tmandry

Coverage instruments closure bodies in macros (not the macro body)

Fixes: #84884
This solution might be considered a compromise, but I think it is the
better choice.

The results in the `closure.rs` test correctly resolve all test cases
broken as described in #84884.

One test pattern (in both `closure_macro.rs` and
`closure_macro_async.rs`) was also affected, and removes coverage
statistics for the lines inside the closure, because the closure
includes a macro. (The coverage remains at the callsite of the macro, so
we lose some detail, but there isn't a perfect choice with macros.

Often macro implementations are split across the macro and the callsite,
and there doesn't appear to be a single "right choice" for which body
should be covered. For the current implementation, we can't do both.

The callsite is most likely to be the preferred site for coverage.

r? `@tmandry`
cc: `@wesleywiser`

3 years agoRollup merge of #84875 - richkadel:no-coverage-dont-check-unused, r=tmandry
Yuki Okushi [Fri, 7 May 2021 06:20:25 +0000 (15:20 +0900)]
Rollup merge of #84875 - richkadel:no-coverage-dont-check-unused, r=tmandry

Removes unneeded check of `#[no_coverage]` in mapgen

There is an anticipated feature request to support a compiler flag that
only adds coverage for specific files (or perhaps mods). As I thought
about where that change would need to be supported, I realized that
checking the attribute in mapgen (for unused functions) was unnecessary.
The unused functions are only synthesized if they have MIR coverage, and
functions with the `no_coverage` attribute will not have been
instrumented with MIR coverage statements in the first place.

New tests confirm this.

Also, while adding tests, I updated resolved comments and FIXMEs in
other tests, and expanded comments and tests on one remaining issue that
is still not resolved.

r? `@tmandry`
cc: `@wesleywiser`

3 years agoRollup merge of #84815 - richkadel:coverage-docs-update-2021-05, r=tmandry
Yuki Okushi [Fri, 7 May 2021 06:20:24 +0000 (15:20 +0900)]
Rollup merge of #84815 - richkadel:coverage-docs-update-2021-05, r=tmandry

Update coverage docs and command line help

r? `@tmandry`
cc: `@wesleywiser`

3 years agoRollup merge of #84787 - klensy:deps-bump, r=Mark-Simulacrum
Yuki Okushi [Fri, 7 May 2021 06:20:23 +0000 (15:20 +0900)]
Rollup merge of #84787 - klensy:deps-bump, r=Mark-Simulacrum

bump deps

openssl vulns:
https://rustsec.org/advisories/RUSTSEC-2021-0056.html
https://rustsec.org/advisories/RUSTSEC-2021-0055.html

openssl-src v111.12.0+1.1.1h -> v111.15.0+1.1.1k

not latest, but newer
openssl-sys v0.9.58 -> v0.9.61
openssl v0.10.30 -> v0.10.33

curl vulns: https://curl.se/docs/vuln-7.74.0.html

curl v0.4.34 -> v0.4.36
curl-sys v0.4.39+curl-7.74.0 -> v0.4.42+curl-7.76.0

3 years agoRollup merge of #84781 - jyn514:check-bootstrap, r=Mark-Simulacrum
Yuki Okushi [Fri, 7 May 2021 06:20:22 +0000 (15:20 +0900)]
Rollup merge of #84781 - jyn514:check-bootstrap, r=Mark-Simulacrum

Don't check bootstrap artifacts by default

Bootstrap has to build successfully or this won't run, so all it checks
are the tests, which are uncommon to change.

Fixes https://github.com/rust-lang/rust/issues/76624.

3 years agoRollup merge of #84779 - jyn514:cargotest-args, r=Mark-Simulacrum
Yuki Okushi [Fri, 7 May 2021 06:20:21 +0000 (15:20 +0900)]
Rollup merge of #84779 - jyn514:cargotest-args, r=Mark-Simulacrum

Add support for --test-args to cargotest

This allows running a single test without having to wait for all tests
to complete.

Closes https://github.com/rust-lang/rust/issues/77260.

3 years agoshrinking the deprecated method span
hi-rustin [Fri, 7 May 2021 02:41:04 +0000 (10:41 +0800)]
shrinking the deprecated method span

3 years agoAuto merge of #85014 - Dylan-DPC:rollup-jzpbkdu, r=Dylan-DPC
bors [Fri, 7 May 2021 01:16:08 +0000 (01:16 +0000)]
Auto merge of #85014 - Dylan-DPC:rollup-jzpbkdu, r=Dylan-DPC

Rollup of 11 pull requests

Successful merges:

 - #84409 (Ensure TLS destructors run before thread joins in SGX)
 - #84500 (Add --run flag to compiletest)
 - #84728 (Add test for suggestion to borrow unsized function parameters)
 - #84734 (Add `needs-unwind` and beginning of support for testing `panic=abort` std to compiletest)
 - #84755 (Allow using `core::` in intra-doc links within core itself)
 - #84871 (Disallows `#![feature(no_coverage)]` on stable and beta (using standard crate-level gating))
 - #84872 (Wire up tidy dependency checks for cg_clif)
 - #84896 (Handle incorrect placement of parentheses in trait bounds more gracefully)
 - #84905 (CTFE engine: rename copy → copy_intrinsic, move to intrinsics.rs)
 - #84953 (Remove unneeded call to with_default_session_globals in rustdoc highlight)
 - #84987 (small nits)

Failed merges:

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