]> git.lizzy.rs Git - rust.git/log
rust.git
7 years agoUpdate link to COMPILER_TESTS.md in CONTRIBUTING.md
CrazyMerlyn [Tue, 7 Mar 2017 17:43:31 +0000 (23:13 +0530)]
Update link to COMPILER_TESTS.md in CONTRIBUTING.md

7 years agoAuto merge of #40272 - jseyfried:fix_const_macro_invocations, r=petrochenkov
bors [Mon, 6 Mar 2017 15:04:10 +0000 (15:04 +0000)]
Auto merge of #40272 - jseyfried:fix_const_macro_invocations, r=petrochenkov

macros: fix const expression invocations

Fixes #40136.
r? @nrc

7 years agoAuto merge of #40276 - topecongiro:add-test, r=petrochenkov
bors [Mon, 6 Mar 2017 09:17:06 +0000 (09:17 +0000)]
Auto merge of #40276 - topecongiro:add-test, r=petrochenkov

Add missing tests for 'E-needstest' labeled issues

This PR adds missing tests for issue #35988, #19712, ~~#18627~~, #24947, #28600 and #34751.

7 years agoAdd missing tests for 'E-needstest' labeled issues
topecongiro [Sat, 4 Mar 2017 02:46:14 +0000 (11:46 +0900)]
Add missing tests for 'E-needstest' labeled issues

7 years agoAuto merge of #40285 - estebank:issue-40221, r=frewsxcv
bors [Mon, 6 Mar 2017 03:24:09 +0000 (03:24 +0000)]
Auto merge of #40285 - estebank:issue-40221, r=frewsxcv

Fix ICE: don't use `struct_variant` on enums

Fix #40221 and add unittest.

7 years agoFix ICE: don't use `struct_variant` on enums
Esteban Küber [Mon, 6 Mar 2017 00:18:22 +0000 (21:18 -0300)]
Fix ICE: don't use `struct_variant` on enums

Fix #40221 and add unittest.

7 years agoFix const expression macro invocations.
Jeffrey Seyfried [Sun, 5 Mar 2017 08:14:21 +0000 (08:14 +0000)]
Fix const expression macro invocations.

7 years agoAuto merge of #40266 - Mark-Simulacrum:cleanup-parser, r=petrochenkov
bors [Sun, 5 Mar 2017 19:53:29 +0000 (19:53 +0000)]
Auto merge of #40266 - Mark-Simulacrum:cleanup-parser, r=petrochenkov

Inline function to avoid naming confusion.

Fixes the non-RFC requiring portion of #18394. The suggestion for a new token there probably needs to either be refiled onto rust-lang/rfcs if it's still relevant (may not be given Macros 2.0 work, not sure).

7 years agoAuto merge of #40252 - topecongiro:compile-fail-test-cfg-stmt-expr, r=petrochenkov
bors [Sun, 5 Mar 2017 17:47:06 +0000 (17:47 +0000)]
Auto merge of #40252 - topecongiro:compile-fail-test-cfg-stmt-expr, r=petrochenkov

Add compile fail test for stmt_expr_attributes

Adds a missing feature gate test for `stmt_expr_attributes`. Issue #39059.

7 years agoAdd compile fail test for stmt_expr_attributes
topecongiro [Sat, 4 Mar 2017 02:46:14 +0000 (11:46 +0900)]
Add compile fail test for stmt_expr_attributes

7 years agoInline function to avoid naming confusion.
Mark Simulacrum [Sat, 4 Mar 2017 19:37:45 +0000 (12:37 -0700)]
Inline function to avoid naming confusion.

7 years agoAuto merge of #40236 - petrochenkov:btweak, r=alexcrichton
bors [Sat, 4 Mar 2017 19:20:58 +0000 (19:20 +0000)]
Auto merge of #40236 - petrochenkov:btweak, r=alexcrichton

rustbuild: A few tweaks

Fixes https://github.com/rust-lang/rust/issues/40016
Fixes https://github.com/rust-lang/rust/issues/39507

r? @alexcrichton

7 years agoAutomate timestamp creation and build skipping for native libraries
Vadim Petrochenkov [Fri, 3 Mar 2017 17:11:04 +0000 (20:11 +0300)]
Automate timestamp creation and build skipping for native libraries

Add comments

7 years agoDo not purge LLVM build directory on rebuild
Vadim Petrochenkov [Fri, 3 Mar 2017 14:18:44 +0000 (17:18 +0300)]
Do not purge LLVM build directory on rebuild

Add some comments

7 years agoSeparate "ui-fulldeps" tests from "ui" tests
Vadim Petrochenkov [Fri, 3 Mar 2017 10:28:22 +0000 (13:28 +0300)]
Separate "ui-fulldeps" tests from "ui" tests

7 years agobootstrap.py: Report build status
Vadim Petrochenkov [Fri, 3 Mar 2017 02:27:07 +0000 (05:27 +0300)]
bootstrap.py: Report build status

Move some code from x.py to bootstrap.py

7 years agorun-make on MSVC: Do not generate object files in the source directory
Vadim Petrochenkov [Fri, 3 Mar 2017 00:09:35 +0000 (03:09 +0300)]
run-make on MSVC: Do not generate object files in the source directory

7 years agoBuild compiler-rt and sanitizers only once
Vadim Petrochenkov [Thu, 2 Mar 2017 23:15:56 +0000 (02:15 +0300)]
Build compiler-rt and sanitizers only once

7 years agoSupport combination MSVC + Ninja
Vadim Petrochenkov [Mon, 27 Feb 2017 20:39:16 +0000 (23:39 +0300)]
Support combination MSVC + Ninja

7 years agoAdd/remove `rerun-if-changed` when necessary
Vadim Petrochenkov [Thu, 23 Feb 2017 15:49:54 +0000 (18:49 +0300)]
Add/remove `rerun-if-changed` when necessary

7 years agoAuto merge of #40213 - steveklabnik:extract-reference, r=alexcrichton
bors [Sat, 4 Mar 2017 15:40:10 +0000 (15:40 +0000)]
Auto merge of #40213 - steveklabnik:extract-reference, r=alexcrichton

Move the reference into a submodule

Part of https://github.com/rust-lang/rust/issues/39588

This moves the reference out into its own submodule. Right now, it's under my own account; I would assume that it should go under `rust-lang`, but I guess there's an argument for the nursery too? I can fix that up before we commit it.

r? @alexcrichton @brson

7 years agoimport reference submodule
steveklabnik [Thu, 2 Mar 2017 23:47:06 +0000 (18:47 -0500)]
import reference submodule

7 years agoremove reference
steveklabnik [Thu, 2 Mar 2017 23:44:45 +0000 (18:44 -0500)]
remove reference

7 years agoAuto merge of #40245 - cuviper:maybe-not-pie, r=alexcrichton
bors [Sat, 4 Mar 2017 12:33:35 +0000 (12:33 +0000)]
Auto merge of #40245 - cuviper:maybe-not-pie, r=alexcrichton

Let `-Crelocation-model` better control `-pie` linking

Prior to this, if relocation model in the target options was "pic", as
most targets have it, then the user's `-Crelocation-model` had no effect
on whether `-pie` would be used.  Only `-Clink-arg=-static` would have a
further override here.

Now we use `context::get_reloc_model`, which gives precedence to the
user's option, and if it's `RelocMode::PIC` we enable `-pie`.  This is
the same test as `context::is_pie_binary`, except that folds across all
`sess.crate_types`, not just the current one.

Fixes #35061.

7 years agoAuto merge of #40242 - cramertj:fix-while-let-ribs-scope, r=petrochenkov
bors [Sat, 4 Mar 2017 10:21:30 +0000 (10:21 +0000)]
Auto merge of #40242 - cramertj:fix-while-let-ribs-scope, r=petrochenkov

Fix missing WhileLet pattern scope

Fix #40235

7 years agoAuto merge of #40207 - phil-opp:llvm-submodule-update, r=alexcrichton
bors [Sat, 4 Mar 2017 08:10:52 +0000 (08:10 +0000)]
Auto merge of #40207 - phil-opp:llvm-submodule-update, r=alexcrichton

LLVM: Update submodule to include x86-interrupt ABI patches

Update LLVM submodule to `rust-llvm-2016-10-29` branch (commit https://github.com/rust-lang/llvm/commit/50ab09fb43f038e4f824eea6cb278f560d3e8621).

7 years agoAuto merge of #40202 - jseyfried:integrate_tokenstream, r=nrc
bors [Sat, 4 Mar 2017 05:44:12 +0000 (05:44 +0000)]
Auto merge of #40202 - jseyfried:integrate_tokenstream, r=nrc

syntax: integrate `TokenStream`

Use `TokenStream` instead of `Vec<TokenTree>` in `TokenTree::Delimited` and elsewhere.
r? @nrc

7 years agoAuto merge of #40163 - arielb1:normalization-1702, r=nikomatsakis
bors [Sat, 4 Mar 2017 01:47:51 +0000 (01:47 +0000)]
Auto merge of #40163 - arielb1:normalization-1702, r=nikomatsakis

More through normalization, Feb/Mar 2017 edition

Fix a few normalization bugs.

Fixes #27901.
Fixes #28828.
Fixes #38135.
Fixes #39363.
Fixes #39367.

7 years agoAuto merge of #39917 - alexcrichton:build-cargo, r=brson
bors [Fri, 3 Mar 2017 23:26:26 +0000 (23:26 +0000)]
Auto merge of #39917 - alexcrichton:build-cargo, r=brson

rustbuild: Add support for compiling Cargo

This commit adds support to rustbuild for compiling Cargo as part of the release
process. Previously rustbuild would simply download a Cargo snapshot and
repackage it. With this change we should be able to turn off artifacts from the
rust-lang/cargo repository and purely rely on the artifacts Cargo produces here.

The infrastructure added here is intended to be extensible to other components,
such as the RLS. It won't exactly be a one-line addition, but the addition of
Cargo didn't require too much hooplah anyway.

The process for release Cargo will now look like:

* The rust-lang/rust repository has a Cargo submodule which is used to build a
  Cargo to pair with the rust-lang/rust release
* Periodically we'll update the cargo submodule as necessary on rust-lang/rust's
  master branch
* When branching beta we'll create a new branch of Cargo (as we do today), and
  the first commit to the beta branch will be to update the Cargo submodule to
  this exact revision.
* When branching stable, we'll ensure that the Cargo submodule is updated and
  then make a stable release.

Backports to Cargo will look like:

* Send a PR to cargo's master branch
* Send a PR to cargo's release branch (e.g. rust-1.16.0)
* Send a PR to rust-lang/rust's beta branch updating the submodule
* Eventually send a PR to rust-lang/rust's master branch updating the submodule

For reference, the process to add a new component to the rust-lang/rust release
would look like:

* Add `$foo` as a submodule in `src/tools`
* Add a `tool-$foo` step which compiles `$foo` with the specified compiler,
  likely mirroring what Cargo does.
* Add a `dist-$foo` step which uses `src/tools/$foo` and the `tool-$foo` output
  to create a rust-installer package for `$foo` likely mirroring what Cargo
  does.
* Update the `dist-extended` step with a new dependency on `dist-$foo`
* Update `src/tools/build-manifest` for the new component.

7 years agoLet `-Crelocation-model` better control `-pie` linking
Josh Stone [Fri, 3 Mar 2017 19:39:11 +0000 (11:39 -0800)]
Let `-Crelocation-model` better control `-pie` linking

Prior to this, if relocation model in the target options was "pic", as
most targets have it, then the user's `-Crelocation-model` had no effect
on whether `-pie` would be used.  Only `-Clink-arg=-static` would have a
further override here.

Now we use `context::get_reloc_model`, which gives precedence to the
user's option, and if it's `RelocMode::PIC` we enable `-pie`.  This is
the same test as `context::is_pie_binary`, except that folds across all
`sess.crate_types`, not just the current one.

7 years agoFix missing WhileLet pattern scope
Taylor Cramer [Fri, 3 Mar 2017 17:40:44 +0000 (09:40 -0800)]
Fix missing WhileLet pattern scope

7 years agoAuto merge of #40101 - danobi:feat_gate_test_simd, r=alexcrichton
bors [Fri, 3 Mar 2017 17:26:10 +0000 (17:26 +0000)]
Auto merge of #40101 - danobi:feat_gate_test_simd, r=alexcrichton

Add compile fail test for SIMD

This completes the missing SIMD test task for issue #39059.

7 years agorustbuild: Add support for compiling Cargo
Alex Crichton [Wed, 15 Feb 2017 23:57:06 +0000 (15:57 -0800)]
rustbuild: Add support for compiling Cargo

This commit adds support to rustbuild for compiling Cargo as part of the release
process. Previously rustbuild would simply download a Cargo snapshot and
repackage it. With this change we should be able to turn off artifacts from the
rust-lang/cargo repository and purely rely on the artifacts Cargo produces here.

The infrastructure added here is intended to be extensible to other components,
such as the RLS. It won't exactly be a one-line addition, but the addition of
Cargo didn't require too much hooplah anyway.

The process for release Cargo will now look like:

* The rust-lang/rust repository has a Cargo submodule which is used to build a
  Cargo to pair with the rust-lang/rust release
* Periodically we'll update the cargo submodule as necessary on rust-lang/rust's
  master branch
* When branching beta we'll create a new branch of Cargo (as we do today), and
  the first commit to the beta branch will be to update the Cargo submodule to
  this exact revision.
* When branching stable, we'll ensure that the Cargo submodule is updated and
  then make a stable release.

Backports to Cargo will look like:

* Send a PR to cargo's master branch
* Send a PR to cargo's release branch (e.g. rust-1.16.0)
* Send a PR to rust-lang/rust's beta branch updating the submodule
* Eventually send a PR to rust-lang/rust's master branch updating the submodule

For reference, the process to add a new component to the rust-lang/rust release
would look like:

* Add `$foo` as a submodule in `src/tools`
* Add a `tool-$foo` step which compiles `$foo` with the specified compiler,
  likely mirroring what Cargo does.
* Add a `dist-$foo` step which uses `src/tools/$foo` and the `tool-$foo` output
  to create a rust-installer package for `$foo` likely mirroring what Cargo
  does.
* Update the `dist-extended` step with a new dependency on `dist-$foo`
* Update `src/tools/build-manifest` for the new component.

7 years agoAuto merge of #40133 - arielb1:operand-lifetimes, r=eddyb
bors [Fri, 3 Mar 2017 13:12:08 +0000 (13:12 +0000)]
Auto merge of #40133 - arielb1:operand-lifetimes, r=eddyb

[MIR] improve operand lifetimes

r? @eddyb

7 years agowork around LLVM PR#32123
Ariel Ben-Yehuda [Fri, 3 Mar 2017 00:32:32 +0000 (02:32 +0200)]
work around LLVM PR#32123

7 years agofix codegen test
Ariel Ben-Yehuda [Tue, 28 Feb 2017 11:15:15 +0000 (13:15 +0200)]
fix codegen test

7 years agomake operands live to the end of their containing expression
Ariel Ben-Yehuda [Sun, 26 Feb 2017 14:21:08 +0000 (16:21 +0200)]
make operands live to the end of their containing expression

In MIR construction, operands need to live exactly until they are used,
which is during the (sub)expression that made the call to `as_operand`.

Before this PR, operands lived until the end of the temporary scope,
which was sometimes unnecessarily longer and sometimes too short.

Fixes #38669.

7 years agoAuto merge of #40189 - SimonSapin:one-width, r=alexcrichton
bors [Fri, 3 Mar 2017 09:57:57 +0000 (09:57 +0000)]
Auto merge of #40189 - SimonSapin:one-width, r=alexcrichton

Reduce std_unicode’s public API

 * Only keep one copy of the `UTF8_CHAR_WIDTH` table instead of one of each of libcore and libstd_unicode.
* Move the `utf8_char_width` function to `core::str` under the `str_internals` unstable feature.
* Remove `std_unicode::str::is_utf16`. It was only accessible through the `#[unstable]` crate std_unicode. It has never been used in the compiler or standard library since 47e7a05 added it in 2012 “for OS API interop”. It can be replaced with a one-liner:

   ```rust
   fn is_utf16(slice: &[u16]) -> bool {
       std::char::decode_utf16(s).all(|r| r.is_ok())
   }
   ```

7 years agoAuto merge of #40178 - arielb1:provide-destructors, r=eddyb
bors [Fri, 3 Mar 2017 05:16:10 +0000 (05:16 +0000)]
Auto merge of #40178 - arielb1:provide-destructors, r=eddyb

convert AdtDef::destructor to on-demand

This removes the `Cell` from `AdtDef`. Also, moving destructor validity
checking to on-demand (forced during item-type checking) ensures that
invalid destructors can't cause ICEs.

Fixes #38868.
Fixes #40132.

r? @eddyb

7 years agoAuto merge of #40152 - eddyb:order-in-rustbuild, r=alexcrichton
bors [Fri, 3 Mar 2017 03:15:52 +0000 (03:15 +0000)]
Auto merge of #40152 - eddyb:order-in-rustbuild, r=alexcrichton

rustbuild: use deterministic step ordering and respect path order on the command-line.

Restores similar behavior to `make` rules, where:
* the step order was always the same, e.g. the testsuite order in `make check`
* `make check-stage1-{cfail,rpass}` would *always* run `cfail` before `rpass`
  * `./x.py test--stage 1 src/test/{compile-fail,run-pass}` is now equivalent

r? @alexcrichton

7 years agoFix fallout in unit tests.
Jeffrey Seyfried [Thu, 2 Mar 2017 01:29:40 +0000 (01:29 +0000)]
Fix fallout in unit tests.

7 years agoFix `token::Eof` spans.
Jeffrey Seyfried [Tue, 21 Feb 2017 12:04:45 +0000 (12:04 +0000)]
Fix `token::Eof` spans.

7 years agoIntegrate `TokenStream`.
Jeffrey Seyfried [Tue, 21 Feb 2017 05:05:59 +0000 (05:05 +0000)]
Integrate `TokenStream`.

7 years agoIntroduce `syntax::parse::parser::TokenCursor`.
Jeffrey Seyfried [Mon, 20 Feb 2017 05:44:06 +0000 (05:44 +0000)]
Introduce `syntax::parse::parser::TokenCursor`.

7 years agoOptimize `syntax::tokenstream::Cursor`.
Jeffrey Seyfried [Sat, 18 Feb 2017 12:45:32 +0000 (12:45 +0000)]
Optimize `syntax::tokenstream::Cursor`.

7 years agoRemove lifetime parameter from `syntax::tokenstream::Cursor`.
Jeffrey Seyfried [Sat, 18 Feb 2017 06:18:29 +0000 (06:18 +0000)]
Remove lifetime parameter from `syntax::tokenstream::Cursor`.

7 years agoAdd compile fail test for SIMD
Daniel Xu [Sat, 25 Feb 2017 23:02:19 +0000 (17:02 -0600)]
Add compile fail test for SIMD

7 years agoAuto merge of #39927 - nikomatsakis:incr-comp-skip-borrowck-2, r=eddyb
bors [Fri, 3 Mar 2017 00:14:10 +0000 (00:14 +0000)]
Auto merge of #39927 - nikomatsakis:incr-comp-skip-borrowck-2, r=eddyb

transition borrowck to visit all **bodies** and not item-likes

This is a better structure for incremental compilation and also more compatible with the eventual borrowck mir. It also fixes #38520 as a drive-by fix.

r? @eddyb

7 years agoschedule drops on bindings only after initializing them
Ariel Ben-Yehuda [Sun, 26 Feb 2017 14:21:26 +0000 (16:21 +0200)]
schedule drops on bindings only after initializing them

This reduces the number of dynamic drops in libstd from 1141 to 899.
However, without this change, the next patch would have created much
more dynamic drops.

A basic merge unswitching hack reduced the number of dynamic drops to
644, with no effect on stack usage. I should be writing a more dedicated
drop unswitching pass.

No performance measurements.

7 years agoAuto merge of #40216 - frewsxcv:rollup, r=frewsxcv
bors [Thu, 2 Mar 2017 20:10:40 +0000 (20:10 +0000)]
Auto merge of #40216 - frewsxcv:rollup, r=frewsxcv

Rollup of 7 pull requests

- Successful merges: #39832, #40104, #40110, #40117, #40129, #40139, #40166
- Failed merges:

7 years agoRollup merge of #40166 - aidanhs:aphs-index-coerce, r=nikomatsakis
Corey Farwell [Thu, 2 Mar 2017 19:53:51 +0000 (14:53 -0500)]
Rollup merge of #40166 - aidanhs:aphs-index-coerce, r=nikomatsakis

Allow types passed to [] to coerce, like .index()

Fixes #40085

Basically steals the relevant part of [check_argument_types](https://github.com/rust-lang/rust/blob/1.15.1/src/librustc_typeck/check/mod.rs#L2653-L2672).

7 years agoRollup merge of #40139 - tedsta:fuchsia_std_process_fix, r=alexcrichton
Corey Farwell [Thu, 2 Mar 2017 19:53:49 +0000 (14:53 -0500)]
Rollup merge of #40139 - tedsta:fuchsia_std_process_fix, r=alexcrichton

std::process for fuchsia: updated to latest liblaunchpad

Our liblaunchpad changed a bit and so fuchsia's std::process impl needs to change a bit.

@raphlinus

7 years agoRollup merge of #40129 - abonander:proc_macro_bang, r=jseyfried
Corey Farwell [Thu, 2 Mar 2017 19:53:46 +0000 (14:53 -0500)]
Rollup merge of #40129 - abonander:proc_macro_bang, r=jseyfried

Implement function-like procedural macros ( `#[proc_macro]`)

Adds the `#[proc_macro]` attribute, which expects bare functions of the kind `fn(TokenStream) -> TokenStream`, which can be invoked like `my_macro!()`.

cc rust-lang/rfcs#1913, #38356

r? @jseyfried
cc @nrc

7 years agoRollup merge of #40117 - SimonSapin:to-err-is-for-the-formatter, r=alexcrichton
Corey Farwell [Thu, 2 Mar 2017 19:53:44 +0000 (14:53 -0500)]
Rollup merge of #40117 - SimonSapin:to-err-is-for-the-formatter, r=alexcrichton

Panic on errors in `format!` or `<T: Display>::to_string`

… instead of silently ignoring a result.

`fmt::Write for String` never returns `Err`, so implementations of `Display` (or other traits of that family) never should either.

Fixes #40103

7 years agoRollup merge of #40110 - benschreiber:nostackcheck, r=brson
Corey Farwell [Thu, 2 Mar 2017 19:53:43 +0000 (14:53 -0500)]
Rollup merge of #40110 - benschreiber:nostackcheck, r=brson

Made no_stack_check a stable_removed attribute

r? @brson

7 years agoRollup merge of #40104 - nagisa:mir-the-shiny, r=eddyb
Corey Farwell [Thu, 2 Mar 2017 19:53:42 +0000 (14:53 -0500)]
Rollup merge of #40104 - nagisa:mir-the-shiny, r=eddyb

[MIR] Rvalue::ty infallible + remove TypedConstVal

Feel free to r+ whenever there aren't any big bit-rot sensitive PRs in the queue.

r? @eddyb

7 years agoRollup merge of #39832 - phil-opp:x86-interrupt-calling-convention, r=nagisa
Corey Farwell [Thu, 2 Mar 2017 19:53:41 +0000 (14:53 -0500)]
Rollup merge of #39832 - phil-opp:x86-interrupt-calling-convention, r=nagisa

Add support for the x86-interrupt calling convention

This calling convention can be used for definining interrupt handlers on 32-bit and 64-bit x86 targets. The compiler then uses `iret` instead of `ret` for returning and ensures that all registers are restored to their
original values.

Usage:

```rust
extern "x86-interrupt" fn handler(stack_frame: &ExceptionStackFrame) {…}
```

for interrupts and exceptions without error code and

```rust
extern "x86-interrupt" fn handler_with_err_code(stack_frame: &ExceptionStackFrame,
                                                error_code: u64) {…}
```

for exceptions that push an error code (e.g., page faults or general protection faults). The programmer must ensure that the correct version is used for each interrupt.

For more details see the [LLVM PR][1] and the corresponding [proposal][2].

[1]: https://reviews.llvm.org/D15567
[2]: http://lists.llvm.org/pipermail/cfe-dev/2015-September/045171.html

It is also possible to implement interrupt handlers on x86 through [naked functions](https://github.com/rust-lang/rfcs/blob/master/text/1201-naked-fns.md). In fact, almost all existing Rust OS projects for x86 use naked functions for this, including [Redox](https://github.com/redox-os/kernel/blob/b9793deb59c7650f0805dea96adb6b773ad99336/arch/x86_64/src/lib.rs#L109-L147), [IntermezzOS](https://github.com/intermezzOS/kernel/blob/f959cc18c78b1ba153f3ff7039d9ecc07f397628/interrupts/src/lib.rs#L28-L72), and [blog_os](https://github.com/phil-opp/blog_os/blob/844d739379ffdea6a7ede88365ec6e21a725bbf5/src/interrupts/mod.rs#L49-L64). So support for the `x86-interrupt` calling convention isn't absolutely needed.

However, it has a number of benefits to naked functions:

- **No inline assembly needed**: [Inline assembly](https://doc.rust-lang.org/book/inline-assembly.html) is highly unstable and dangerous. It's pretty easy to mess things up. Also, it uses an arcane syntax and requires that the programmer knows x86 assembly.
- **Higher performance**: A naked wrapper function always saves _all_ registers before calling the Rust function. This isn't needed for a compiler supported calling convention, since the compiler knows which registers are clobbered by the interrupt handler. Thus, only these registers need to be saved and restored.
- **Safer interfaces**: We can write a `set_handler` function that takes a `extern "x86-interrupt" fn(&ExceptionStackFrame)` and the compiler ensures that we always use the right function type for all handler functions. This isn't possible with the `#[naked]` attribute.
- **More convenient**: Instead of writing [tons of assembly boilerplate](https://github.com/redox-os/kernel/blob/b9793deb59c7650f0805dea96adb6b773ad99336/arch/x86_64/src/lib.rs#L109-L147) and desperately trying to improve things [through macros](https://github.com/phil-opp/blog_os/blob/844d739379ffdea6a7ede88365ec6e21a725bbf5/src/interrupts/mod.rs#L17-L92), we can just write [code like this](https://github.com/phil-opp/blog_os/blob/e6a61f9507a4c4fef6fb4e3474bc596391bc97d2/src/interrupts/mod.rs#L85-L89).
- **Naked functions are unreliable**: It is allowed to use Rust code inside a naked function, which sometimes works and sometimes not. For example, [calling a function](https://github.com/redox-os/kernel/blob/b9793deb59c7650f0805dea96adb6b773ad99336/arch/x86_64/src/lib.rs#L132) through Rust code seems to work fine without function prologue, but [code declaring a variable](https://is.gd/NQYXqE) silently adds a prologue even though the function is naked (look at the generated assembly, there is a `movl` instruction before the `nop`).

**Edit**: See the [tracking issue](https://github.com/rust-lang/rust/issues/40180) for an updated list of issues.

Unfortunately, the implementation of the `x86-interrupt` calling convention in LLVM has some issues that make it unsuitable for 64-bit kernels at the moment:

- LLVM always tries to backup the `xmm` registers on 64-bit platforms even if the target doesn't support SSE. This leads to invalid opcode exceptions whenever an interrupt handler is invoked. I submitted a fix to LLVM in [D29959](https://reviews.llvm.org/D29959). The fix is really small (<10 lines), so maybe we could backport it to [Rust's LLVM fork](https://github.com/rust-lang/llvm)?. **Edit**: The fix was merged to LLVM trunk in [rL295347](https://reviews.llvm.org/rL295347). Backported in https://github.com/rust-lang/llvm/pull/63.

- On targets with SSE support, LLVM uses the `movaps` instruction for saving the `xmm` registers, which requires an alignment of 16. For handlers with error codes, however, the stack alignment is only 8, so a alignment exception occurs. This issue is tracked in [bug 26413](https://bugs.llvm.org/show_bug.cgi?id=26413). ~~Unfortunately, I don't know enough about LLVM to fix this.~~ **Edit**: Fix submitted in [D30049](https://reviews.llvm.org/D30049).

This PR adds experimental support for this calling convention under the `abi_x86_interrupt` feature gate. The implementation is very similar to #38465 and was surprisingly simple :).

There is no accepted RFC for this change. In fact, the [RFC for interrupt calling convention](https://github.com/rust-lang/rfcs/pull/1275) from 2015 was closed in favor of naked functions. However, the reactions to the recent [PR](https://github.com/rust-lang/rust/pull/38465) for a MSP430 interrupt calling convention were [in favor of experimental interrupt ABIs](https://github.com/rust-lang/rust/pull/38465#issuecomment-270015470).

- [x] Add compile-fail tests for the feature gate.
- [x] Create tracking issue for the `abi_x86_interrupt` feature (and link it in code). **Edit**: Tracking issue: #40180
- [x] Backport [rL295347](https://reviews.llvm.org/rL295347) to Rust's LLVM fork. **Edit**: Done in https://github.com/rust-lang/llvm/pull/63

@tari @steveklabnik @jackpot51 @ticki @hawkw @thepowersgang, you might be interested in this.

7 years agoAdd support for x86-interrupt calling convention
Philipp Oppermann [Tue, 14 Feb 2017 20:39:42 +0000 (21:39 +0100)]
Add support for x86-interrupt calling convention

Tracking issue: https://github.com/rust-lang/rust/issues/40180

This calling convention can be used for definining interrupt handlers on
32-bit and 64-bit x86 targets. The compiler then uses `iret` instead of
`ret` for returning and ensures that all registers are restored to their
original values.

Usage:

```
extern "x86-interrupt" fn handler(stack_frame: &ExceptionStackFrame) {…}
```

for interrupts and exceptions without error code and

```
extern "x86-interrupt" fn page_fault_handler(stack_frame: &ExceptionStackFrame,
                                             error_code: u64) {…}
```

for exceptions that push an error code (e.g., page faults or general
protection faults). The programmer must ensure that the correct version
is used for each interrupt.

For more details see the [LLVM PR][1] and the corresponding [proposal][2].

[1]: https://reviews.llvm.org/D15567
[2]: http://lists.llvm.org/pipermail/cfe-dev/2015-September/045171.html

7 years agoAuto merge of #39655 - durka:recursion-limit-suggestion, r=nikomatsakis
bors [Thu, 2 Mar 2017 17:44:17 +0000 (17:44 +0000)]
Auto merge of #39655 - durka:recursion-limit-suggestion, r=nikomatsakis

suggest doubling recursion limit in more situations

Fixes #38852.

r? @bluss

7 years agoRemove std_unicode::str::is_utf16
Simon Sapin [Wed, 1 Mar 2017 22:01:09 +0000 (23:01 +0100)]
Remove std_unicode::str::is_utf16

It was only accessible through the `#[unstable]` crate std_unicode.

It has never been used in the compiler or standard library
since 47e7a05a28c9662159af2d2e0f2b7efc13fa09cb added it in 2012
“for OS API interop”.
It can be replaced with a one-liner:

```rust
fn is_utf16(slice: &[u16]) -> bool {
    std::char::decode_utf16(s.iter().cloned()).all(|r| r.is_ok())
}
```

7 years agoAuto merge of #40206 - GuillaumeGomez:rollup, r=GuillaumeGomez
bors [Thu, 2 Mar 2017 14:38:12 +0000 (14:38 +0000)]
Auto merge of #40206 - GuillaumeGomez:rollup, r=GuillaumeGomez

Rollup of 9 pull requests

- Successful merges: #40081, #40144, #40168, #40169, #40170, #40173, #40175, #40191, #40194
- Failed merges:

7 years agoRollup merge of #40194 - letmaik:patch-1, r=steveklabnik
Guillaume Gomez [Thu, 2 Mar 2017 10:29:43 +0000 (11:29 +0100)]
Rollup merge of #40194 - letmaik:patch-1, r=steveklabnik

Fix wrong word used in book page "const and static"

7 years agoRollup merge of #40191 - topecongiro:x86-interrupt, r=steveklabnik
Guillaume Gomez [Thu, 2 Mar 2017 10:29:42 +0000 (11:29 +0100)]
Rollup merge of #40191 - topecongiro:x86-interrupt, r=steveklabnik

Add abi_x86_interrupt to the unstable book

This PR closes #40181.

7 years agoRollup merge of #40175 - d-e-s-o:fix-inconsistency-in-guessing-game-readme, r=stevekl...
Guillaume Gomez [Thu, 2 Mar 2017 10:29:41 +0000 (11:29 +0100)]
Rollup merge of #40175 - d-e-s-o:fix-inconsistency-in-guessing-game-readme, r=steveklabnik

doc: fix inconsistency in error output in guessing-game.md

The line '.expect("failed to read line");' is partly started with a
lower case 'f' and partly with an uppercase one, adding additional
spurious changes to otherwise clean diffs if each sample is
copy-and-pasted over the previous.
This change starts the string with an uppercase everywhere which is in
line with the style of the other strings.

7 years agoRollup merge of #40173 - er-1:master, r=alexcrichton
Guillaume Gomez [Thu, 2 Mar 2017 10:29:40 +0000 (11:29 +0100)]
Rollup merge of #40173 - er-1:master, r=alexcrichton

Add a reference to the dl library to the Makefile of the test issue-2…

…4445.

It prevents the test to fail on ppc64el at least.

Part of #39015

7 years agoRollup merge of #40170 - iKevinY:if-let-typo, r=frewsxcv
Guillaume Gomez [Thu, 2 Mar 2017 10:29:39 +0000 (11:29 +0100)]
Rollup merge of #40170 - iKevinY:if-let-typo, r=frewsxcv

Fix link in `if let` docs

r? @steveklabnik

7 years agoRollup merge of #40169 - MajorBreakfast:patch-8, r=steveklabnik
Guillaume Gomez [Thu, 2 Mar 2017 10:29:38 +0000 (11:29 +0100)]
Rollup merge of #40169 - MajorBreakfast:patch-8, r=steveklabnik

String docs: Add "the"

r? @steveklabnik

7 years agoRollup merge of #40168 - topecongiro:compile-fail-test-abi-ptx, r=petrochenkov
Guillaume Gomez [Thu, 2 Mar 2017 10:29:37 +0000 (11:29 +0100)]
Rollup merge of #40168 - topecongiro:compile-fail-test-abi-ptx, r=petrochenkov

Add compile fail test for abi_ptx

Issue #39059.

7 years agoRollup merge of #40144 - MajorBreakfast:patch-7, r=frewsxcv
Guillaume Gomez [Thu, 2 Mar 2017 10:29:36 +0000 (11:29 +0100)]
Rollup merge of #40144 - MajorBreakfast:patch-7, r=frewsxcv

Unit-like structs doc: Improve code sample

r? @steveklabnik

BTW it seems that
```Rust
let p = Proton {};
```
compiles without an error. That's why I didn't add it to the example. It's about consistency anyway.

7 years agoRollup merge of #40081 - GuillaumeGomez:poison-docs, r=frewsxcv
Guillaume Gomez [Thu, 2 Mar 2017 10:29:35 +0000 (11:29 +0100)]
Rollup merge of #40081 - GuillaumeGomez:poison-docs, r=frewsxcv

Add missing url in sync structs

r? @frewsxcv

7 years agoLLVM: Update submodule to include x86-interrupt ABI patches
Philipp Oppermann [Thu, 2 Mar 2017 10:28:09 +0000 (11:28 +0100)]
LLVM: Update submodule to include x86-interrupt ABI patches

7 years agonote -> help
Alex Burka [Tue, 28 Feb 2017 18:18:54 +0000 (18:18 +0000)]
note -> help

7 years agoAuto merge of #40188 - nikomatsakis:issue-40029, r=eddyb
bors [Thu, 2 Mar 2017 03:34:53 +0000 (03:34 +0000)]
Auto merge of #40188 - nikomatsakis:issue-40029, r=eddyb

inhibit enum layout optimizations under `#[repr(C)]` or `#[repr(u8)]`

Fixes #40029

7 years agoAdd abi_x86_interrupt to the unstable book
topecongiro [Wed, 1 Mar 2017 22:06:18 +0000 (07:06 +0900)]
Add abi_x86_interrupt to the unstable book

7 years agofix wrong word used (static vs const)
Maik Riechert [Wed, 1 Mar 2017 23:06:40 +0000 (23:06 +0000)]
fix wrong word used (static vs const)

7 years agoAuto merge of #39803 - brson:fpic, r=alexcrichton
bors [Wed, 1 Mar 2017 22:48:17 +0000 (22:48 +0000)]
Auto merge of #39803 - brson:fpic, r=alexcrichton

Add a test that -fPIC is applied

r? @alexcrichton Can it really be this simple? I've tested it works, but still testing that it used to fail.

7 years agoPanic on errors in `format!` or `<T: Display>::to_string`
Simon Sapin [Sun, 26 Feb 2017 18:42:11 +0000 (19:42 +0100)]
Panic on errors in `format!` or `<T: Display>::to_string`

… instead of silently ignoring a result.

`fmt::Write for String` never returns `Err`,
so implementations of `Display` (or other traits of that family)
never should either.

Fixes #40103

7 years agoOnly keep one copy of the UTF8_CHAR_WIDTH table.
Simon Sapin [Wed, 1 Mar 2017 21:41:44 +0000 (22:41 +0100)]
Only keep one copy of the UTF8_CHAR_WIDTH table.

… instead of one of each of libcore and libstd_unicode.

Move the `utf8_char_width` function to `core::str`
under the `str_internals` unstable feature.

7 years agofix tests to handle the Typeof bodies
Niko Matsakis [Wed, 1 Mar 2017 22:04:01 +0000 (17:04 -0500)]
fix tests to handle the Typeof bodies

7 years agoinhibit enum layout optimizations under `#[repr(C)]` or `#[repr(u8)]`
Niko Matsakis [Wed, 1 Mar 2017 20:22:12 +0000 (15:22 -0500)]
inhibit enum layout optimizations under `#[repr(C)]` or `#[repr(u8)]`

Fixes #40029

7 years agoconvert AdtDef::destructor to on-demand
Ariel Ben-Yehuda [Wed, 1 Mar 2017 16:42:26 +0000 (18:42 +0200)]
convert AdtDef::destructor to on-demand

This removes the Cell from AdtDef. Also, moving destructor validity
checking to on-demand (forced during item-type checking) ensures that
invalid destructors can't cause ICEs.

Fixes #38868.
Fixes #40132.

7 years agoDon't run test on darwin
Alex Crichton [Wed, 1 Mar 2017 16:08:53 +0000 (08:08 -0800)]
Don't run test on darwin

7 years agoAdd Cargo as a submodule
Alex Crichton [Wed, 15 Feb 2017 22:55:26 +0000 (14:55 -0800)]
Add Cargo as a submodule

7 years agotransform broken MIR warnings to hard ICEs
Ariel Ben-Yehuda [Wed, 1 Mar 2017 00:44:01 +0000 (02:44 +0200)]
transform broken MIR warnings to hard ICEs

We ought to do that sometime, and this PR fixes all broken MIR errors I
could find.

7 years agofix a few more typeck normalization cases
Ariel Ben-Yehuda [Wed, 1 Mar 2017 01:39:43 +0000 (03:39 +0200)]
fix a few more typeck normalization cases

I'll like @nikomatsakis or someone to look at the unsolved variable
case.

7 years agostore the normalized types of statics in MIR Lvalues
Ariel Ben-Yehuda [Wed, 1 Mar 2017 00:29:57 +0000 (02:29 +0200)]
store the normalized types of statics in MIR Lvalues

The types of statics, like all other items, are stored in the tcx
unnormalized. This is necessarily so, because
    a) Item types other than statics have generics, which can't be
normalized.
    b) Eager normalization causes undesirable on-demand dependencies.

Keeping with the principle that MIR lvalues require no normalization in
order to interpret, this patch stores the normalized type of the statics
in the Lvalue and reads it to get the lvalue type.

Fixes #39367.

7 years agomore through normalization in typeck & trans
Ariel Ben-Yehuda [Tue, 28 Feb 2017 23:30:41 +0000 (01:30 +0200)]
more through normalization in typeck & trans

Fixes #27901.
Fixes #28828.
Fixes #38135.
Fixes #39363.

7 years agodoc: fix inconsistency in error output in guessing-game.md
deso [Wed, 1 Mar 2017 13:44:50 +0000 (05:44 -0800)]
doc: fix inconsistency in error output in guessing-game.md

The line '.expect("failed to read line");' is partly started with a
lower case 'f' and partly with an uppercase one, adding additional
spurious changes to otherwise clean diffs if each sample is
copy-and-pasted over the previous.
This change starts the string with an uppercase everywhere which is in
line with the style of the other strings.

7 years agoAdd a reference to the dl library to the Makefile of the test issue-24445.
er-1 [Wed, 1 Mar 2017 11:17:27 +0000 (12:17 +0100)]
Add a reference to the dl library to the Makefile of the test issue-24445.
It prevents the test to fail on ppc64el at least.

Part of #39015

7 years agoAuto merge of #34198 - eddyb:you're-a-bad-transmute-and-you-should-feel-bad, r=nikoma...
bors [Wed, 1 Mar 2017 10:03:44 +0000 (10:03 +0000)]
Auto merge of #34198 - eddyb:you're-a-bad-transmute-and-you-should-feel-bad, r=nikomatsakis

Make transmuting from fn item types to pointer-sized types a hard error.

Closes #19925 by removing the future compatibility lint and the associated workarounds.
This is a `[breaking-change]` if you `transmute` from a function item without casting first.
For more information on how to fix your code, see https://github.com/rust-lang/rust/issues/19925.

7 years agoUnit-like structs doc: Add compile fail tag
Josef Brandl [Wed, 1 Mar 2017 09:03:07 +0000 (10:03 +0100)]
Unit-like structs doc: Add compile fail tag

7 years agoFix link in `if let` docs
Kevin Yap [Wed, 1 Mar 2017 09:01:37 +0000 (01:01 -0800)]
Fix link in `if let` docs

7 years agoString docs: Add "the"
Josef Brandl [Wed, 1 Mar 2017 08:56:52 +0000 (09:56 +0100)]
String docs: Add "the"

7 years agoAuto merge of #40167 - frewsxcv:rollup, r=frewsxcv
bors [Wed, 1 Mar 2017 07:57:09 +0000 (07:57 +0000)]
Auto merge of #40167 - frewsxcv:rollup, r=frewsxcv

Rollup of 6 pull requests

- Successful merges: #39419, #39936, #39944, #39960, #40028, #40128
- Failed merges:

7 years agoAuto merge of #39419 - jseyfried:simplify_tokentree, r=nrc
bors [Wed, 1 Mar 2017 05:58:09 +0000 (05:58 +0000)]
Auto merge of #39419 - jseyfried:simplify_tokentree, r=nrc

Simplify `TokenTree` and fix `macro_rules!` bugs

This PR
 - fixes #39390, fixes #39403, and fixes #39404 (each is a [breaking-change], see issues for examples),
 - fixes #39889,
 - simplifies and optimizes macro invocation parsing,
 - cleans up `ext::tt::transcribe`,
 - removes `tokenstream::TokenTree::Sequence` and `Token::MatchNt`,
   - instead, adds a new type `ext::tt::quoted::TokenTree` for use by `macro_rules!` (`ext::tt`)
 - removes `parser.quote_depth` and `parser.parsing_token_tree`, and
 - removes `quote_matcher!`.
   - Instead, use `quote_tokens!` and `ext::tt::quoted::parse` the result with `expect_matchers=true`.
   - I found no outside uses of `quote_matcher!` when searching Rust code on Github.

r? @nrc

7 years agoAdd compile fail test for abi_ptx
topecongiro [Wed, 1 Mar 2017 04:04:13 +0000 (13:04 +0900)]
Add compile fail test for abi_ptx

7 years agoRollup merge of #40128 - cengizIO:master, r=nikomatsakis
Corey Farwell [Wed, 1 Mar 2017 03:55:31 +0000 (22:55 -0500)]
Rollup merge of #40128 - cengizIO:master, r=nikomatsakis

Move two large error_reporting fn's to a separate file

Hello!

I tried to make `librustc/infer/error_reporting,rs` more readable by modularizing it and moving its two largest functions to a separate file.

If you have any suggestions, please send it right away! 🚀

Thanks goes to @nikomatsakis for supporting.

7 years agoRollup merge of #40028 - withoutboats:string_from_iter, r=alexcrichton
Corey Farwell [Wed, 1 Mar 2017 03:55:30 +0000 (22:55 -0500)]
Rollup merge of #40028 - withoutboats:string_from_iter, r=alexcrichton

impl FromIterator<&char> for String

7 years agoRollup merge of #39960 - lukaramu:issue-39925, r=alexcrichton
Corey Farwell [Wed, 1 Mar 2017 03:55:29 +0000 (22:55 -0500)]
Rollup merge of #39960 - lukaramu:issue-39925, r=alexcrichton

added Error and Display impl for std::ffi::FromBytesWithNulError

Fixes #39925.

This is my first PR, so I wasn't quite sure about the stability annotation.

7 years agoRollup merge of #39944 - GuillaumeGomez:associated-consts, r=frewsxcv
Corey Farwell [Wed, 1 Mar 2017 03:55:28 +0000 (22:55 -0500)]
Rollup merge of #39944 - GuillaumeGomez:associated-consts, r=frewsxcv

Improve associated constant rendering in rustdoc

Before:

<img width="1440" alt="screen shot 2017-02-19 at 00 30 51" src="https://cloud.githubusercontent.com/assets/3050060/23097697/caeed80e-f63a-11e6-98c2-5d27e4efd76d.png">

After:

<img width="1440" alt="screen shot 2017-02-19 at 00 30 39" src="https://cloud.githubusercontent.com/assets/3050060/23097698/cfb4874e-f63a-11e6-80cf-ffbf5c5c6162.png">

cc @SergioBenitez

r? @rust-lang/docs