]> git.lizzy.rs Git - rust.git/log
rust.git
5 years agoGet rid of unnecessary #![feature(core_float)].
Mazdak Farrokhzad [Sun, 19 Aug 2018 16:37:00 +0000 (18:37 +0200)]
Get rid of unnecessary #![feature(core_float)].

5 years agoDrop `identity` from prelude.
Mazdak Farrokhzad [Sun, 19 Aug 2018 16:36:18 +0000 (18:36 +0200)]
Drop `identity` from prelude.

5 years agoMerge branch 'master' into feature/core_convert_id
Mazdak Farrokhzad [Sun, 19 Aug 2018 16:34:46 +0000 (18:34 +0200)]
Merge branch 'master' into feature/core_convert_id

5 years agoAuto merge of #51131 - qnighy:unsized-locals, r=eddyb
bors [Sun, 19 Aug 2018 12:21:56 +0000 (12:21 +0000)]
Auto merge of #51131 - qnighy:unsized-locals, r=eddyb

Implement Unsized Rvalues

This PR is the first step to implement RFC1909: unsized rvalues (#48055).

## Implemented

- `Sized` is removed for arguments and local bindings. (under `#![feature(unsized_locals)]`)
- Unsized locations are allowed in MIR
- Unsized places and operands are correctly translated at codegen

## Not implemented in this PR

- Additional `Sized` checks:
  - tuple struct constructor (accidentally compiles now)
  - closure arguments at closure generation (accidentally compiles now)
  - upvars (ICEs now)
- Generating vtable for `fn method(self)` (ICEs now)
- VLAs: `[e; n]` where `n` isn't const
- Reduce unnecessary allocations

## Current status

- [x] Fix `__rust_probestack` (rust-lang-nursery/compiler-builtins#244)
  - [x] Get the fix merged
- [x] `#![feature(unsized_locals)]`
  - [x] Give it a tracking issue number
- [x] Lift sized checks in typeck and MIR-borrowck
  - [ ] <del>Forbid `A(unsized-expr)`</del> will be another PR
- [x] Minimum working codegen
- [x] Add more examples and fill in unimplemented codegen paths
- [ ] <del>Loosen object-safety rules (will be another PR)</del>
- [ ] <del>Implement `Box<FnOnce>` (will be another PR)</del>
- [ ] <del>Reduce temporaries (will be another PR)</del>

5 years agoAuto merge of #52972 - RalfJung:from_raw_parts_align, r=alexcrichton
bors [Sun, 19 Aug 2018 09:40:36 +0000 (09:40 +0000)]
Auto merge of #52972 - RalfJung:from_raw_parts_align, r=alexcrichton

debug_assert to ensure that from_raw_parts is only used properly aligned

This does not help nearly as much as I would hope because everybody uses the distributed libstd which is compiled without debug assertions. For this reason, I am not sure if this is even worth it. OTOH, this would have caught the misalignment fixed by https://github.com/rust-lang/rust/issues/42789 *if* there had been any tests actually using ZSTs with alignment >1 (we have a CI runner which has debug assertions in libstd enabled), and it seems to currently [fail in the rg testsuite](https://ci.appveyor.com/project/rust-lang/rust/build/1.0.8403/job/v7dfdcgn8ay5j6sb). So maybe it is worth it, after all.

I have seen the attribute `#[rustc_inherit_overflow_checks]` in some places, does that make it so that the *caller's* debug status is relevant? Is there a similar attribute for `debug_assert!`? That could even subsume `rustc_inherit_overflow_checks`: Something like `rustc_inherit_debug_flag` could affect *all* places that change the generated code depending on whether we are in debug or release mode. In fact, given that we have to keep around the MIR for generic functions anyway, is there ever a reason *not* to handle the debug flag that way? I guess currently we apply debug flags like `cfg` so this is dropped early during the MIR pipeline?

EDIT: I learned from @eddyb that because of how `debug_assert!` works, this is not realistic. Well, we could still have it for the rustc CI runs and then maybe, eventually, when libstd gets compiled client-side and there is both a debug and a release build... then this will also benefit users.^^

5 years agoIntegrate OperandValue::UnsizedRef into OperandValue::Ref.
Masaki Hara [Fri, 3 Aug 2018 14:50:13 +0000 (23:50 +0900)]
Integrate OperandValue::UnsizedRef into OperandValue::Ref.

5 years agoIntegrate PassMode::UnsizedIndirect into PassMode::Indirect.
Masaki Hara [Fri, 3 Aug 2018 14:32:21 +0000 (23:32 +0900)]
Integrate PassMode::UnsizedIndirect into PassMode::Indirect.

5 years agoRemove a now-unnecessary paragraph.
Masaki Hara [Thu, 12 Jul 2018 11:27:10 +0000 (20:27 +0900)]
Remove a now-unnecessary paragraph.

The paragraph described a case where we can't optimize away repetitive
dynamic stack allocation. However, as arielb1 pointed out, it can
actually optimizable by dynamically delaying the stack unwinding.

5 years agoUpdate the unstable book regarding [e; dyn n].
Masaki Hara [Thu, 12 Jul 2018 11:26:13 +0000 (20:26 +0900)]
Update the unstable book regarding [e; dyn n].

5 years agoAdd an unstable-book article about unsized_locals.
Masaki Hara [Tue, 29 May 2018 15:33:01 +0000 (00:33 +0900)]
Add an unstable-book article about unsized_locals.

5 years agoImplement simple codegen for unsized rvalues.
Masaki Hara [Mon, 28 May 2018 15:12:55 +0000 (00:12 +0900)]
Implement simple codegen for unsized rvalues.

5 years agoLift some Sized checks.
Masaki Hara [Mon, 28 May 2018 15:11:34 +0000 (00:11 +0900)]
Lift some Sized checks.

5 years agoAdd #![feature(unsized_locals)].
Masaki Hara [Mon, 28 May 2018 15:10:09 +0000 (00:10 +0900)]
Add #![feature(unsized_locals)].

5 years agoAdd Builder::array_alloca.
Masaki Hara [Mon, 28 May 2018 15:07:23 +0000 (00:07 +0900)]
Add Builder::array_alloca.

5 years agoAdd notes on unsized argument errors.
Masaki Hara [Tue, 29 May 2018 15:25:56 +0000 (00:25 +0900)]
Add notes on unsized argument errors.

5 years agoAuto merge of #52592 - eddyb:or-default, r=Mark-Simulacrum
bors [Sat, 18 Aug 2018 21:58:37 +0000 (21:58 +0000)]
Auto merge of #52592 - eddyb:or-default, r=Mark-Simulacrum

Use the new Entry::or_default method where possible.

5 years agoAuto merge of #53324 - alexreg:self_in_typedefs, r=eddyb
bors [Sat, 18 Aug 2018 19:34:24 +0000 (19:34 +0000)]
Auto merge of #53324 - alexreg:self_in_typedefs, r=eddyb

`Self` in type definitions (self_in_typedefs)

This implements the [`self_in_typedefs` feature](https://github.com/rust-lang/rfcs/blob/master/text/2300-self-in-typedefs.md) ([tracking issue 49303](https://github.com/rust-lang/rust/issues/49303)).

r? @eddyb

CC @Centril

5 years agoAdded page for feature to unstable book.
Alexander Regueiro [Tue, 14 Aug 2018 20:26:33 +0000 (21:26 +0100)]
Added page for feature to unstable book.

5 years agoAdded tests.
Alexander Regueiro [Wed, 15 Aug 2018 01:24:05 +0000 (02:24 +0100)]
Added tests.

5 years agoAdded feature gate.
Alexander Regueiro [Tue, 14 Aug 2018 17:45:44 +0000 (18:45 +0100)]
Added feature gate.

5 years agoUpdated diagnostics.
Alexander Regueiro [Tue, 14 Aug 2018 17:29:53 +0000 (18:29 +0100)]
Updated diagnostics.

5 years agoResolve `Self` within type definitions.
Alexander Regueiro [Tue, 14 Aug 2018 01:49:06 +0000 (02:49 +0100)]
Resolve `Self` within type definitions.

Currently type definitions include `struct`, `enum`, `union`, `existential type`.

5 years agoUse the new Entry::or_default method where possible.
Eduard-Mihai Burtescu [Sat, 21 Jul 2018 19:43:31 +0000 (22:43 +0300)]
Use the new Entry::or_default method where possible.

5 years agoAuto merge of #53436 - cuviper:trace_fn-stop, r=alexcrichton
bors [Sat, 18 Aug 2018 17:15:31 +0000 (17:15 +0000)]
Auto merge of #53436 - cuviper:trace_fn-stop, r=alexcrichton

std: stop backtracing when the frames are full

This is a defensive measure to mitigate the infinite unwind loop seen in #53372.  That case will still repeatedly unwind `__rust_try`, but now it will at least stop when `cx.frames` is full.

r? @alexcrichton

5 years agoAuto merge of #53286 - nagisa:cast-assumes, r=eddyb
bors [Sat, 18 Aug 2018 14:35:42 +0000 (14:35 +0000)]
Auto merge of #53286 - nagisa:cast-assumes, r=eddyb

Do not generate assumes for plain integer casts

I gave up on making anything more elegant for now.

r? @eddyb

5 years agoAuto merge of #53175 - matthewjasper:more-return-stuff, r=nikomatsakis
bors [Sat, 18 Aug 2018 11:57:46 +0000 (11:57 +0000)]
Auto merge of #53175 - matthewjasper:more-return-stuff, r=nikomatsakis

[NLL] Returns are interesting for free regions

Based on #53088 - creating now to get feedback.

Closes #51175

* Make assigning to the return type interesting.
* Use "returning this value" instead of "return" in error messages.
* Prefer one of the explanations that we have a name for to a generic interesting cause in some cases.
* Treat causes that involve the destination of a call like assignments.

5 years agoAuto merge of #52553 - Pazzaz:vecdeque-append, r=SimonSapin
bors [Sat, 18 Aug 2018 08:56:12 +0000 (08:56 +0000)]
Auto merge of #52553 - Pazzaz:vecdeque-append, r=SimonSapin

Non-naive implementation of `VecDeque.append`

Replaces the old, simple implementation with a more manual (and **unsafe** ðŸ˜±) one. I've added 1 more test and verified that it covers all 6 code paths in the function.

This new implementation was about 60% faster than the old naive one when I tried benchmarking it.

5 years agoAuto merge of #53437 - alexcrichton:fix-target-features, r=michaelwoerister
bors [Sat, 18 Aug 2018 02:50:39 +0000 (02:50 +0000)]
Auto merge of #53437 - alexcrichton:fix-target-features, r=michaelwoerister

Set more llvm function attributes for __rust_try

This shim is generated elsewhere in the compiler so this commit adds support to
ensure it goes through similar paths as the rest of the compiler to set llvm
function attributes like target features.

cc #53372

5 years agoAuto merge of #53369 - pnkfelix:partial-53351-make-some-ported-compile-fail-tests...
bors [Fri, 17 Aug 2018 23:33:31 +0000 (23:33 +0000)]
Auto merge of #53369 - pnkfelix:partial-53351-make-some-ported-compile-fail-tests-more-robust-wrt-nll, r=davidtwco

Make some ported cfail tests robust w.r.t. NLL

Updated the most glaring instances of weak tests w.r.t. NLL that came from #53196.

See also the bulletpoint list on #53351.

5 years agoAuto merge of #53356 - michaelwoerister:itlto, r=alexcrichton
bors [Fri, 17 Aug 2018 21:24:22 +0000 (21:24 +0000)]
Auto merge of #53356 - michaelwoerister:itlto, r=alexcrichton

Preliminary work for incremental ThinLTO (CGU name edition)

Bring back the first half of #52266 but hopefully without the performance regression.

5 years agoAuto merge of #50911 - petrochenkov:macuse, r=alexcrichton
bors [Fri, 17 Aug 2018 19:10:34 +0000 (19:10 +0000)]
Auto merge of #50911 - petrochenkov:macuse, r=alexcrichton

Stabilize `use_extern_macros`

Closes https://github.com/rust-lang/rust/issues/35896

5 years agoAuto merge of #53449 - frewsxcv:rollup, r=frewsxcv
bors [Fri, 17 Aug 2018 17:00:10 +0000 (17:00 +0000)]
Auto merge of #53449 - frewsxcv:rollup, r=frewsxcv

Rollup of 11 pull requests

Successful merges:

 - #52858 (Implement Iterator::size_hint for Elaborator.)
 - #53321 (Fix usage of `wasm_target_feature`)
 - #53326 ([nll] add regression test for issue #27868)
 - #53347 (rustc_resolve: don't allow paths starting with `::crate`.)
 - #53349 ([nll] add tests for #48697 and #30104)
 - #53357 (Pretty print btreemap for GDB)
 - #53358 (`{to,from}_{ne,le,be}_bytes` for unsigned integer types)
 - #53406 (Do not suggest conversion method that is already there)
 - #53407 (make more ported compile fail tests more robust w.r.t. NLL)
 - #53413 (Warn that `#![feature(rust_2018_preview)]` is implied when the edition is set to Rust 2018.)
 - #53434 (wasm: Remove --strip-debug argument to LLD)

Failed merges:

r? @ghost

5 years agoRollup merge of #53434 - alexcrichton:remove-strip-debug, r=yurydelendik
Corey Farwell [Fri, 17 Aug 2018 15:23:45 +0000 (08:23 -0700)]
Rollup merge of #53434 - alexcrichton:remove-strip-debug, r=yurydelendik

wasm: Remove --strip-debug argument to LLD

Originally added in #52887 this commit disables passing `--strip-debug` to LLD
when optimized. This bring back the original bug of emitting broken debuginfo
but currently it *also* strips the `name` section which makes it very difficult
to inspect the final binary. A real fix is happening at
https://reviews.llvm.org/D50729 and we can reevaluate once we've updated LLD to
have that commit.

5 years agoRollup merge of #53413 - eddyb:featured-in-the-latest-edition, r=varkor
Corey Farwell [Fri, 17 Aug 2018 15:23:44 +0000 (08:23 -0700)]
Rollup merge of #53413 - eddyb:featured-in-the-latest-edition, r=varkor

Warn that `#![feature(rust_2018_preview)]` is implied when the edition is set to Rust 2018.

cc @varkor @petrochenkov @joshtriplett

5 years agoRollup merge of #53407 - pnkfelix:partial-53351-make-more-ported-compile-fail-tests...
Corey Farwell [Fri, 17 Aug 2018 15:23:43 +0000 (08:23 -0700)]
Rollup merge of #53407 - pnkfelix:partial-53351-make-more-ported-compile-fail-tests-more-robust-wrt-nll, r=nikomatsakis

make more ported compile fail tests more robust w.r.t. NLL

This is similar to PR #53369, except it covers a disjoint (and much smaller) set of tests that I needed to look at more carefully before being 100% certain they were the same kind of issue.

5 years agoRollup merge of #53406 - estebank:to_string-to_string, r=michaelwoerister
Corey Farwell [Fri, 17 Aug 2018 15:23:42 +0000 (08:23 -0700)]
Rollup merge of #53406 - estebank:to_string-to_string, r=michaelwoerister

Do not suggest conversion method that is already there

Fix #53348.

5 years agoRollup merge of #53358 - SimonSapin:int-bytes, r=shepmaster
Corey Farwell [Fri, 17 Aug 2018 15:23:40 +0000 (08:23 -0700)]
Rollup merge of #53358 - SimonSapin:int-bytes, r=shepmaster

`{to,from}_{ne,le,be}_bytes` for unsigned integer types

Same as https://github.com/rust-lang/rust/pull/51919 did for signed integers.

Tracking issue: https://github.com/rust-lang/rust/issues/52963

5 years agoRollup merge of #53357 - fukatani:pretty-print-btreemap, r=michaelwoerister
Corey Farwell [Fri, 17 Aug 2018 15:23:39 +0000 (08:23 -0700)]
Rollup merge of #53357 - fukatani:pretty-print-btreemap, r=michaelwoerister

Pretty print btreemap for GDB

Merge #53112 first, please.

5 years agoRollup merge of #53349 - memoryruins:nll-tests, r=nikomatsakis
Corey Farwell [Fri, 17 Aug 2018 15:23:38 +0000 (08:23 -0700)]
Rollup merge of #53349 - memoryruins:nll-tests, r=nikomatsakis

[nll] add tests for #48697 and #30104

Adds tests for the following issues:
- #48697 ``[NLL] ICE: unexpected region for local data with reference to closure``
- #30104 ``Destructuring boxes into multiple mutable references seems broken``

r? @nikomatsakis

5 years agoRollup merge of #53347 - eddyb:no-crate-in-root, r=petrochenkov
Corey Farwell [Fri, 17 Aug 2018 15:23:36 +0000 (08:23 -0700)]
Rollup merge of #53347 - eddyb:no-crate-in-root, r=petrochenkov

rustc_resolve: don't allow paths starting with `::crate`.

cc @aturon @joshtriplett
r? @petrochenkov

5 years agoRollup merge of #53326 - memoryruins:issue-27868-test, r=nikomatsakis
Corey Farwell [Fri, 17 Aug 2018 15:23:35 +0000 (08:23 -0700)]
Rollup merge of #53326 - memoryruins:issue-27868-test, r=nikomatsakis

[nll] add regression test for issue #27868

Adds a test for #27868 ``Inconsistent evaluation order for assignment operations``

apart of #47366 ``tracking issue for bugs fixed by the MIR borrow checker or NLL``

r? @nikomatsakis

5 years agoRollup merge of #53321 - alexcrichton:wasm-target-feature, r=nikomatsakis
Corey Farwell [Fri, 17 Aug 2018 15:23:34 +0000 (08:23 -0700)]
Rollup merge of #53321 - alexcrichton:wasm-target-feature, r=nikomatsakis

Fix usage of `wasm_target_feature`

Currently usage results in:

```
error: internal compiler error: librustc_typeck/collect.rs:1928: unknown target feature gate wasm_target_feature

thread 'main' panicked at 'Box<Any>', librustc_errors/lib.rs:579:9
note: Run with `RUST_BACKTRACE=1` for a backtrace.
error: aborting due to previous error

note: the compiler unexpectedly panicked. this is a bug.

note: we would appreciate a bug report: https://github.com/rust-lang/rust/blob/master/CONTRIBUTING.md#bug-reports

note: rustc 1.30.0-nightly (d5a448b3f 2018-08-13) running on x86_64-unknown-linux-gnu

note: compiler flags: -C debuginfo=2 -C linker=/tmp/lld-shim -C incremental --crate-type lib

note: some of the compiler flags provided by cargo are hidden

error: Could not compile `coresimd`.

To learn more, run the command again with --verbose.
```

and hopefully this should fix the ICE!

5 years agoRollup merge of #52858 - frewsxcv:frewsxcv-elaborator, r=SimonSapin
Corey Farwell [Fri, 17 Aug 2018 15:23:33 +0000 (08:23 -0700)]
Rollup merge of #52858 - frewsxcv:frewsxcv-elaborator, r=SimonSapin

Implement Iterator::size_hint for Elaborator.

None

5 years agoAuto merge of #53288 - RalfJung:dropck, r=nikomatsakis
bors [Fri, 17 Aug 2018 14:16:54 +0000 (14:16 +0000)]
Auto merge of #53288 - RalfJung:dropck, r=nikomatsakis

unions are not always trivially dropable

Fixes #52786

r? @nikomatsakis

5 years agoGratuitous at byte boundaries in hex i32 literals in some doc-tests
Simon Sapin [Fri, 17 Aug 2018 12:01:57 +0000 (14:01 +0200)]
Gratuitous at byte boundaries in hex i32 literals in some doc-tests

5 years agoAuto merge of #53190 - sekineh:thumb-cortex-m, r=jamesmunns
bors [Fri, 17 Aug 2018 10:40:25 +0000 (10:40 +0000)]
Auto merge of #53190 - sekineh:thumb-cortex-m, r=jamesmunns

Add crate build test for `thumb*` targets. [IRR-2018-embedded]

## Summary

This PR adds `run-make` test that compiles `cortex-m` crate for all supported `thumb*-none-*` targets using `cargo` and stage2 `rustc`.

- Supported `thumb*-none-*` targets:
  - thumbv6m-none-eabi (Bare Cortex-M0, M0+, M1)
  - thumbv7em-none-eabi (Bare Cortex-M4, M7)
  - thumbv7em-none-eabihf (Bare Cortex-M4F, M7F, FPU, hardfloat)
  - thumbv7m-none-eabi (Bare Cortex-M3)

## How to run & Example output
I tested locally and all targets succeeded like below:
```
./x.py clean
./x.py test --target thumbv6m-none-eabi,thumbv7em-none-eabi,thumbv7em-none-eabihf,thumbv7m-none-eabi src/test/run-make
```
```
Check compiletest suite=run-make mode=run-make (x86_64-unknown-linux-gnu -> thumbv6m-none-eabi)

running 5 tests
.....
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out
```

## How to re-run

Remove `stamp` file for the test run.
```
rm build/x86_64-unknown-linux-gnu/test/run-make/thumb-none-cortex-m/stamp
```
Then run `test`
```
./x.py test --target thumbv6m-none-eabi,thumbv7em-none-eabi,thumbv7em-none-eabihf,thumbv7m-none-eabi src/test/run-make
(snip)
running 5 tests
iiii.
test result: ok. 1 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out
```

## Artifacts

You can examine the artifacts under the directory below:
```
sekineh@sekineh-VirtualBox:~/rustme10$ ls -l build/x86_64-unknown-linux-gnu/test/run-make/thumb-none-cortex-m/thumb-none-cortex-m/
total 4
drwxrwxr-x 7 sekineh sekineh 4096  8月 14 22:40 cortex-m
```
where `build/x86_64-unknown-linux-gnu/test/run-make/thumb-none-cortex-m/thumb-none-cortex-m/` is came from TMPDIR variable.

## Other notes

For `test.rs` modification, I used the same logic as:
- https://github.com/rust-lang/rust/blame/d8b3c830fbcdd14d085209a8dcc3399151f3286a/src/bootstrap/dist.rs#L652-L657
```
            if builder.no_std(target) == Some(true) {
                // the `test` doesn't compile for no-std targets
                builder.ensure(compile::Std { compiler, target });
            } else {
                builder.ensure(compile::Test { compiler, target });
            }
```
It is a useful snippet when adding `no_std` support to `src/bootstrap` code.

CC @kennytm @jamesmunns @nerdyvaishali

5 years agoFix undesirable fallout
Vadim Petrochenkov [Fri, 17 Aug 2018 10:17:39 +0000 (13:17 +0300)]
Fix undesirable fallout

compile-fail-fulldeps/proc-macro/proc-macro-attributes.rs - resolution change for derive helper attributes with the same name as derive itself
run-pass/macro-comma-support.rs - indeterminate resolutions for macros in expression positions
ui/issues/issue-49074.rs - diagnostics regression, not enough recovery to report the second error
ui/object-lifetime/object-lifetime-default.stderr - unstable diagnostics?

5 years agoStabilize `use_extern_macros`
Vadim Petrochenkov [Mon, 14 May 2018 00:22:52 +0000 (03:22 +0300)]
Stabilize `use_extern_macros`

5 years agorustc_resolve: don't allow paths starting with `::crate`.
Eduard-Mihai Burtescu [Tue, 14 Aug 2018 09:54:43 +0000 (12:54 +0300)]
rustc_resolve: don't allow paths starting with `::crate`.

5 years agoAuto merge of #53383 - nnethercote:HybridIdxSetBuf, r=nikomatsakis
bors [Fri, 17 Aug 2018 07:20:23 +0000 (07:20 +0000)]
Auto merge of #53383 - nnethercote:HybridIdxSetBuf, r=nikomatsakis

Speed up NLL with HybridIdxSetBuf.

It's a sparse-when-small but dense-when-large index set that is very
efficient for sets that (a) have few elements, (b) have large
universe_size values, and (c) are cleared frequently. Which makes it
perfect for the `gen_set` and `kill_set` sets used by the new borrow
checker.

This patch reduces `tuple-stress`'s NLL-check time by 40%, and up to 12%
for several other benchmarks. And it halves the max-rss for `keccak`,
and has smaller wins for `inflate` and `clap-rs`.

5 years agofix license
Ralf Jung [Thu, 16 Aug 2018 18:36:34 +0000 (20:36 +0200)]
fix license

5 years agoImplement Iterator::size_hint for Elaborator.
Corey Farwell [Mon, 30 Jul 2018 13:07:51 +0000 (09:07 -0400)]
Implement Iterator::size_hint for Elaborator.

5 years agoAuto merge of #53427 - eddyb:uniform-paths-diagnostics, r=varkor
bors [Fri, 17 Aug 2018 00:28:16 +0000 (00:28 +0000)]
Auto merge of #53427 - eddyb:uniform-paths-diagnostics, r=varkor

rustc_resolve: overhaul `#![feature(uniform_paths)]` error reporting.

Fixes #53408 by only considering external crates to conflict within their (type/module) namespace, *not* with the value or macro namespaces, and also by adding a special-cased error for redundant `use crate_name;` imports (without actually allowing them).
Also, all canaries for a given import are grouped into one diagnostic per namespace, in order to make block-scoped ambiguities clearer.
See changed/added tests for more details.

r? @petrochenkov cc @aturon @joshtriplett

5 years agorustc_resolve: overhaul `#![feature(uniform_paths)]` error reporting.
Eduard-Mihai Burtescu [Thu, 16 Aug 2018 12:59:33 +0000 (15:59 +0300)]
rustc_resolve: overhaul `#![feature(uniform_paths)]` error reporting.

5 years agoAuto merge of #53397 - varkor:feature-collector-expand-visitor, r=nikomatsakis
bors [Thu, 16 Aug 2018 22:11:58 +0000 (22:11 +0000)]
Auto merge of #53397 - varkor:feature-collector-expand-visitor, r=nikomatsakis

Visit all attributes for feature collection

Previously feature attributes were just collected on item-like "things" as well as exported macros and crate attributes. This ignored some places feature attributes could be specified, such as on enum variants.

Fixes https://github.com/rust-lang/rust/issues/53391.

5 years agoSet more llvm function attributes for __rust_try
Alex Crichton [Thu, 16 Aug 2018 20:19:04 +0000 (13:19 -0700)]
Set more llvm function attributes for __rust_try

This shim is generated elsewhere in the compiler so this commit adds support to
ensure it goes through similar paths as the rest of the compiler to set llvm
function attributes like target features.

cc #53372

5 years agostd: stop backtracing when the frames are full
Josh Stone [Thu, 16 Aug 2018 18:26:27 +0000 (11:26 -0700)]
std: stop backtracing when the frames are full

5 years agoAuto merge of #53433 - kennytm:rollup, r=kennytm
bors [Thu, 16 Aug 2018 17:53:29 +0000 (17:53 +0000)]
Auto merge of #53433 - kennytm:rollup, r=kennytm

Rollup of 10 pull requests

Successful merges:

 - #52946 (Documented impl From on line 367 of libserialize/json.rs)
 - #53234 (Remove Travis shutdown debug scripts, and remove CI-specific DNS settings)
 - #53313 (Two small improvements)
 - #53360 (Addressed #51602)
 - #53364 (Warn if the user tries to use GATs)
 - #53373 (Tweak unclosed delimiter parser error)
 - #53377 (std: Use target_pointer_width for BACKTRACE_ELF_SIZE)
 - #53395 (Use #[non_exhaustive] on internal enums)
 - #53399 (Tidy: ignore non-Markdown files when linting for the Unstable Book)
 - #53412 (syntax_ext: remove leftover span_err_if_not_stage0 macro.)

5 years agowasm: Remove --strip-debug argument to LLD
Alex Crichton [Thu, 16 Aug 2018 17:17:06 +0000 (10:17 -0700)]
wasm: Remove --strip-debug argument to LLD

Originally added in #52887 this commit disables passing `--strip-debug` to LLD
when optimized. This bring back the original bug of emitting broken debuginfo
but currently it *also* strips the `name` section which makes it very difficult
to inspect the final binary. A real fix is happening at
https://reviews.llvm.org/D50729 and we can reevaluate once we've updated LLD to
have that commit.

5 years agoRollup merge of #53234 - kennytm:debug-9696, r=alexcrichton
kennytm [Thu, 16 Aug 2018 16:54:45 +0000 (00:54 +0800)]
Rollup merge of #53234 - kennytm:debug-9696, r=alexcrichton

Remove Travis shutdown debug scripts, and remove CI-specific DNS settings

Since the cause of the host shutdown (travis-ci/travis-ci#4924) is found, we could revert the shutdown debug attempts to shorten the logs.

Additionally, we're pretty sure a custom DNS (added in  will not help travis-ci/travis-ci#9696, so reverting that part of #51420 to reduce CI-specific settings.

5 years agoRollup merge of #53412 - eddyb:stage-who, r=alexcrichton
kennytm [Thu, 16 Aug 2018 16:13:29 +0000 (00:13 +0800)]
Rollup merge of #53412 - eddyb:stage-who, r=alexcrichton

 syntax_ext: remove leftover span_err_if_not_stage0 macro.

I believe this is the right fix for #53380, although I'm not sure what happened.
My guess is this copy of the macro was accidentally missed when others were removed?

cc @matthewjasper @varkor (please do not put this in a rollup, in case it fails)

5 years agoRollup merge of #53399 - varkor:ignore-non-md-unstable_book, r=nikomatsakis
kennytm [Thu, 16 Aug 2018 16:13:27 +0000 (00:13 +0800)]
Rollup merge of #53399 - varkor:ignore-non-md-unstable_book, r=nikomatsakis

Tidy: ignore non-Markdown files when linting for the Unstable Book

Previously, any file would be assumed a `.md`, which is annoying when files like `.DS_STORE` slip in and then cause tidy errors.

5 years agoRollup merge of #53395 - varkor:__Nonexhaustive-to-non_exhaustive, r=shepmaster
kennytm [Thu, 16 Aug 2018 16:13:26 +0000 (00:13 +0800)]
Rollup merge of #53395 - varkor:__Nonexhaustive-to-non_exhaustive, r=shepmaster

Use #[non_exhaustive] on internal enums

This replaces `__Nonexhaustive` variants (and variants thereof)  with `#[non_exhaustive]`. These were mostly unstable previously, with the exception of the `cloudabi` enums and `Level` in proc_macro: these were `#[doc(hidden)]`, so clearly intended not to be used directly. It should be safe to replace all of these.

5 years agoRollup merge of #53377 - cuviper:pointer-elf_size, r=alexcrichton
kennytm [Thu, 16 Aug 2018 16:13:25 +0000 (00:13 +0800)]
Rollup merge of #53377 - cuviper:pointer-elf_size, r=alexcrichton

std: Use target_pointer_width for BACKTRACE_ELF_SIZE

The former code used `target.contains("64")` to detect Elf64 targets,
but this is inaccurate in a few cases:

- `s390x-unknown-linux-gnu` is 64-bit
- `sparcv9-sun-solaris` is 64-bit
- `x86_64-unknown-linux-gnux32` is 32-bit

Instead the `std` build script can use `CARGO_CFG_TARGET_POINTER_WIDTH`
to reliably detect 64-bit targets for libbacktrace.

Also update to backtrace-sys 0.1.24 for alexcrichton/backtrace-rs#122.

5 years agoRollup merge of #53373 - estebank:unclosed, r=petrochenkov
kennytm [Thu, 16 Aug 2018 16:13:24 +0000 (00:13 +0800)]
Rollup merge of #53373 - estebank:unclosed, r=petrochenkov

Tweak unclosed delimiter parser error

5 years agoRollup merge of #53364 - varkor:gat-warn-broken, r=pnkfelix
kennytm [Thu, 16 Aug 2018 16:13:23 +0000 (00:13 +0800)]
Rollup merge of #53364 - varkor:gat-warn-broken, r=pnkfelix

Warn if the user tries to use GATs

GATs are currently broken, but still accessible behind a feature gate. This leads to people attempting to use them and then immediately encountering ICEs (or other broken behaviour). Here, we emit a warning if the user tries to use any feature associated with GATs, hopefully making it obvious that ICEs and the like are expected. For the meantime, this seems better than continually getting reported errors (for example: [here](https://github.com/rust-lang/rust/issues?q=is%3Aissue+gat+is%3Aclosed) and [here](https://github.com/rust-lang/rust/issues?utf8=%E2%9C%93&q=is%3Aissue+generic_associated_types+is%3Aclosed)).

5 years agoRollup merge of #53360 - PramodBisht:issue/51602, r=estebank
kennytm [Thu, 16 Aug 2018 16:13:21 +0000 (00:13 +0800)]
Rollup merge of #53360 - PramodBisht:issue/51602, r=estebank

Addressed #51602

Fixed #51602
r? @estebank

here I have addressed the case where `in` was not expected right after `if` block. Speaking of `type ascription` I am not sure if this the best approach which I have implemented. Plus I think one more test case can be added to test `type-ascription` case, though I don't have any at this point of time. I will ping you again if all existing testcases pass.

5 years agoRollup merge of #53313 - llogiq:two-small-improvements, r=estebank
kennytm [Thu, 16 Aug 2018 16:13:20 +0000 (00:13 +0800)]
Rollup merge of #53313 - llogiq:two-small-improvements, r=estebank

Two small improvements

In `librustc_apfloat/ieee.rs`, use the iterator.[r]find methods to simplify the code. In `libserialize/json.rs`, make use of the fact that `Vec.last` on an empty `Vec` returns `None` to simplify the code to a single match.

5 years agoRollup merge of #52946 - Ajacmac:doc-impl-from, r=GuillaumeGomez
kennytm [Thu, 16 Aug 2018 16:13:19 +0000 (00:13 +0800)]
Rollup merge of #52946 - Ajacmac:doc-impl-from, r=GuillaumeGomez

Documented impl From on line 367 of libserialize/json.rs

This is for the impl From mentioned in #51430 assigned to @skade .

Hopefully I didn't miss anything/get anything wrong. I looked over another PR for another part of this same issue to see what the proper formatting was, etc.

Thanks!

5 years agoAuto merge of #53304 - dtolnay:extend, r=dtolnay
bors [Thu, 16 Aug 2018 15:44:30 +0000 (15:44 +0000)]
Auto merge of #53304 - dtolnay:extend, r=dtolnay

TokenStream::extend

Two new insta-stable impls in libproc_macro:

```rust
impl Extend<TokenTree> for TokenStream
impl Extend<TokenStream> for TokenStream
```

`proc_macro::TokenStream` already implements `FromIterator<TokenTree>` and `FromIterator<TokenStream>` so I elected to support the same input types for `Extend`.

**This commit reduces compile time of Serde derives by 60% (takes less than half as long to compile)** as measured by building our test suite:

```console
$ git clone https://github.com/serde-rs/serde
$ cd serde/test_suite
$ cargo check --tests --features proc-macro2/nightly
$ rm -f ../target/debug/deps/libtest_*.rmeta
$ time cargo check --tests --features proc-macro2/nightly
Before: 20.8 seconds
After: 8.6 seconds
```

r? @alexcrichton

5 years agoenable more tests on stage1 again
Ralf Jung [Thu, 16 Aug 2018 13:37:33 +0000 (15:37 +0200)]
enable more tests on stage1 again

5 years agoAuto merge of #53256 - ollie27:writeln, r=KodrAus
bors [Thu, 16 Aug 2018 13:29:40 +0000 (13:29 +0000)]
Auto merge of #53256 - ollie27:writeln, r=KodrAus

Don't accept non-string literals for the format string in writeln

This is to improve diagnostics.

`println` and `eprintln` were already fixed by #52394.

Fixes #30143

5 years agoRevert "Use public DNS server instead of 169.254.169.254 on CI."
kennytm [Thu, 16 Aug 2018 12:06:55 +0000 (20:06 +0800)]
Revert "Use public DNS server instead of 169.254.169.254 on CI."

This reverts commit 32e8bda4e35ada92beab7236d62520e6c6df8596.

5 years agoRevert "Auto merge of #53035 - alexcrichton:debug-travis, r=kennytm"
kennytm [Thu, 9 Aug 2018 18:13:15 +0000 (02:13 +0800)]
Revert "Auto merge of #53035 - alexcrichton:debug-travis, r=kennytm"

This reverts commit 579adf8c727861841b4819b4913385c2782977fb, reversing
changes made to c11f2d286c747b44f8e03ce10f6daa36f5d8331c.

5 years agoRevert "Auto merge of #53134 - alexcrichton:tweak-travis, r=Mark-Simulacrum"
kennytm [Thu, 9 Aug 2018 18:09:29 +0000 (02:09 +0800)]
Revert "Auto merge of #53134 - alexcrichton:tweak-travis, r=Mark-Simulacrum"

This reverts commit f9b9b6e078aa81c0faa064bf861e88001684765c, reversing
changes made to 45a9d410f934247256812eda55ca66dd6999b3de.

5 years agoAuto merge of #53173 - alexcrichton:win-arm64, r=sfackler
bors [Thu, 16 Aug 2018 11:18:12 +0000 (11:18 +0000)]
Auto merge of #53173 - alexcrichton:win-arm64, r=sfackler

Start adding an `aarch64-pc-windows-msvc` target

This commit adds the necessary definitions for target specs and such as well as
the necessary support in libstd to compile basic `aarch64-pc-windows-msvc`
binaries. The target is not currently built on CI, but it can be built locally
with:

    ./configure --target=aarch64-pc-windows-msvc --set rust.lld
    ./x.py build src/libstd --target aarch64-pc-windows-msvc

Currently this fails to build `libtest` due to a linker bug (seemingly in LLD?)
which hasn't been investigate yet. Otherwise though with libstd you can build a
hello world program (linked with LLD). I've not tried to execute it yet, but it
at least links!

Full support for this target is still a long road ahead, but this is hopefully a
good stepping stone to get started.

Points of note about this target are:

* Currently defaults to `panic=abort` as support is still landing in LLVM for
  SEH on AArch64.
* Currently defaults to LLD as a linker as I was able to get farther with it
  than I was with `link.exe`

5 years agoAuto merge of #53147 - ashtneoi:dont-suggest-ref, r=estebank
bors [Thu, 16 Aug 2018 08:45:32 +0000 (08:45 +0000)]
Auto merge of #53147 - ashtneoi:dont-suggest-ref, r=estebank

For move errors, suggest match ergonomics instead of `ref`

Partially fixes issue #52423. Also makes errors and suggestions more consistent between move-from-place and move-from-value errors.

Limitations:
- Only the first pattern in a match arm can have a "consider removing this borrow operator" suggestion.
- Suggestions don't always compile as-is (see the TODOs in the test for details).

Sorry for the really long test. I wanted to make sure I handled every case I could think of, and it turned out there were a lot of them.

Questions:
- Is there any particular applicability I should set on those suggestions?
- Are the notes about the `Copy` trait excessive?

5 years agoSpeed up NLL with `HybridIdxSetBuf`.
Nicholas Nethercote [Tue, 14 Aug 2018 00:21:24 +0000 (10:21 +1000)]
Speed up NLL with `HybridIdxSetBuf`.

`HybridIdxSetBuf` is a sparse-when-small but dense-when-large index set
that is very efficient for sets that (a) have few elements, (b) have
large `universe_size` values, and (c) are cleared frequently. Which
makes it perfect for the `gen_set` and `kill_set` sets used by the new
borrow checker.

This patch reduces the execution time of the five slowest NLL benchmarks
by 55%, 21%, 16%, 10% and 9%. It also reduces the max-rss of three
benchmarks by 53%, 33%, and 9%.

5 years agotests: prefer edition: directives to compile-flags:--edition.
Eduard-Mihai Burtescu [Thu, 16 Aug 2018 07:36:11 +0000 (10:36 +0300)]
tests: prefer edition: directives to compile-flags:--edition.

5 years agosyntax: also warn about edition "umbrella" features being implied by --edition.
Eduard-Mihai Burtescu [Wed, 15 Aug 2018 23:09:12 +0000 (02:09 +0300)]
syntax: also warn about edition "umbrella" features being implied by --edition.

5 years agoAuto merge of #53295 - estebank:on-unimplemented, r=michaelwoerister
bors [Thu, 16 Aug 2018 06:34:13 +0000 (06:34 +0000)]
Auto merge of #53295 - estebank:on-unimplemented, r=michaelwoerister

Various changes to `rustc_on_unimplemented`

- Add `from_method` and `from_desugaring` to formatting options
- Change wording of errors slightly

5 years agoSplit tests more and bless them again
ashtneoi [Thu, 16 Aug 2018 05:36:19 +0000 (22:36 -0700)]
Split tests more and bless them again

5 years agoMove tests into their own directory
ashtneoi [Thu, 16 Aug 2018 05:35:56 +0000 (22:35 -0700)]
Move tests into their own directory

5 years agoAuto merge of #53293 - petrochenkov:gramattr2, r=alexcrichton
bors [Thu, 16 Aug 2018 04:16:12 +0000 (04:16 +0000)]
Auto merge of #53293 - petrochenkov:gramattr2, r=alexcrichton

syntax: Enforce attribute grammar in the parser

Also fix feature-gating for `unrestricted_attribute_tokens` that was introduced in https://github.com/rust-lang/rust/pull/53270, but was actually broken.

cc https://github.com/rust-lang/rust/pull/50911

5 years agofix broken test
Esteban Küber [Thu, 16 Aug 2018 01:47:34 +0000 (18:47 -0700)]
fix broken test

5 years agoAuto merge of #53289 - ljedrz:improve_lexer, r=michaelwoerister
bors [Thu, 16 Aug 2018 01:08:13 +0000 (01:08 +0000)]
Auto merge of #53289 - ljedrz:improve_lexer, r=michaelwoerister

A few cleanups and minor improvements for the lexer

- improve readability by adjusting the formatting of some function signatures and adding some newlines
- reorder some functions for easier reading
- remove redundant `'static` in `const`s
- remove some explicit `return`s
- read directly to a `String` in `gather_comments_and_literals`
- change `unwrap_or!` (macro) to `unwrap_or` (function)
- move an `assert!`ion from `try_next_token` (called in a loop) to `try_real_token` after all calls to `try_next_token`
- `#[inline]` some one-liner functions
- assign directly from an `if-else` expression
- refactor a `match` to `map_or`
- add a `token::is_irrelevant` function to detect tokens that are not "`real`"

5 years agoStart adding an `aarch64-pc-windows-msvc` target
Alex Crichton [Mon, 16 Jul 2018 23:38:56 +0000 (16:38 -0700)]
Start adding an `aarch64-pc-windows-msvc` target

This commit adds the necessary definitions for target specs and such as well as
the necessary support in libstd to compile basic `aarch64-pc-windows-msvc`
binaries. The target is not currently built on CI, but it can be built locally
with:

    ./configure --target=aarch64-pc-windows-msvc --set rust.lld
    ./x.py build src/libstd --target aarch64-pc-windows-msvc

Currently this fails to build `libtest` due to a linker bug (seemingly in LLD?)
which hasn't been investigate yet. Otherwise though with libstd you can build a
hello world program (linked with LLD). I've not tried to execute it yet, but it
at least links!

Full support for this target is still a long road ahead, but this is hopefully a
good stepping stone to get started.

Points of note about this target are:

* Currently defaults to `panic=abort` as support is still landing in LLVM for
  SEH on AArch64.
* Currently defaults to LLD as a linker as I was able to get farther with it
  than I was with `link.exe`

5 years agosyntax: process all edition features before other features.
Eduard-Mihai Burtescu [Wed, 15 Aug 2018 23:06:26 +0000 (02:06 +0300)]
syntax: process all edition features before other features.

5 years agoRevert "Ignore test that fails on stage1"
Eduard-Mihai Burtescu [Wed, 15 Aug 2018 23:11:44 +0000 (02:11 +0300)]
Revert "Ignore test that fails on stage1"

This reverts commit 0e43e6fef251732f5268738eee9c258668ec7942.

5 years agosyntax_ext: remove leftover span_err_if_not_stage0 macro.
Eduard-Mihai Burtescu [Wed, 15 Aug 2018 23:10:54 +0000 (02:10 +0300)]
syntax_ext: remove leftover span_err_if_not_stage0 macro.

5 years agoRe-bless test
ashtneoi [Wed, 15 Aug 2018 22:25:53 +0000 (15:25 -0700)]
Re-bless test

5 years agoAdd files I forgot to commit earlier
ashtneoi [Wed, 15 Aug 2018 02:03:43 +0000 (19:03 -0700)]
Add files I forgot to commit earlier

5 years agoChange a method name and clarify a comment
ashtneoi [Tue, 14 Aug 2018 19:54:31 +0000 (12:54 -0700)]
Change a method name and clarify a comment

5 years agoBless tests
ashtneoi [Tue, 14 Aug 2018 19:38:37 +0000 (12:38 -0700)]
Bless tests

5 years agoTest move-into-Fn/FnMut errors too
ashtneoi [Tue, 14 Aug 2018 18:56:20 +0000 (11:56 -0700)]
Test move-into-Fn/FnMut errors too

5 years agoFix compiletest JSON error message
ashtneoi [Tue, 14 Aug 2018 06:49:21 +0000 (23:49 -0700)]
Fix compiletest JSON error message

5 years agoAllocate fewer Strings at a time
ashtneoi [Tue, 14 Aug 2018 05:01:54 +0000 (22:01 -0700)]
Allocate fewer Strings at a time

5 years agoAdd another FIXME about suggestions
ashtneoi [Tue, 14 Aug 2018 04:49:39 +0000 (21:49 -0700)]
Add another FIXME about suggestions

5 years agoBless UI tests
ashtneoi [Mon, 13 Aug 2018 23:45:40 +0000 (16:45 -0700)]
Bless UI tests

5 years agoRemove incorrect space
ashtneoi [Mon, 13 Aug 2018 23:12:35 +0000 (16:12 -0700)]
Remove incorrect space