Tobias Bucher [Mon, 15 Aug 2016 23:11:33 +0000 (01:11 +0200)]
Implement reading and writing atomically at certain offsets
These functions allow to read from and write to a file in one atomic
action from multiple threads, avoiding the race between the seek and the
read.
The functions are named `{read,write}_at` on non-Windows (which don't
change the file cursor), and `seek_{read,write}` on Windows (which
change the file cursor).
bors [Sat, 8 Oct 2016 19:36:11 +0000 (12:36 -0700)]
Auto merge of #37027 - alexcrichton:less-deps-in-bootstrap, r=japaric
rustbuild: Optimize build times slightly
As the entry point for building the Rust compiler, a good user experience hinges
on this compiling quickly to get to the meat of the problem. To that end use
`#[cfg]`-specific dependencies to avoid building Windows crates on Unix and drop
the `regex` crate for now which was easily replacable with some string
searching.
Alex Crichton [Fri, 7 Oct 2016 19:26:45 +0000 (12:26 -0700)]
rustbuild: Optimize build times slightly
As the entry point for building the Rust compiler, a good user experience hinges
on this compiling quickly to get to the meat of the problem. To that end use
`#[cfg]`-specific dependencies to avoid building Windows crates on Unix and drop
the `regex` crate for now which was easily replacable with some string
searching.
Rollup merge of #37016 - alexcrichton:workspaces, r=japaric
Leverage Cargo workspaces in rustbuild
This is a continuation of https://github.com/rust-lang/rust/pull/36032 which implements the change to use `cargo metadata` to learn about the crate graph.
bors [Sat, 8 Oct 2016 05:51:04 +0000 (22:51 -0700)]
Auto merge of #37014 - nnethercote:avoid-pat_to_string, r=arielb1
Avoid some `pat_to_string` calls.
`ty_of_method_or_bare_fn` currently calls `pat_to_string` even when it
doesn't need to. Fixing this avoids hundreds of large allocations (e.g.
171,600 bytes, in `mk_printer`) in several of rustc-benchmarks.
Alex Crichton [Fri, 7 Oct 2016 06:30:38 +0000 (23:30 -0700)]
rustbuild: Use `cargo metadata` to learn about DAG
This updates the commit to use workspaces to use `cargo metadata` instead of
hardcoded lists about what to test. This should help us be resilient to updates
in the future on behalf of the crate DAG and minimize the amount of files that
need to be touched.
Ahmed Charles [Fri, 2 Sep 2016 08:55:29 +0000 (01:55 -0700)]
Use workspaces and switch to a single Cargo.lock.
This involves hacking the code used to run cargo test on various
packages, because it reads Cargo.lock to determine which packages should
be tested. This change implements a blacklist, since that will catch new
crates when they are added in the future.
bors [Fri, 7 Oct 2016 18:38:27 +0000 (11:38 -0700)]
Auto merge of #36981 - alexcrichton:catch-unwind-for-tests, r=sfackler
std: Minor cleanup to libtest
* Don't spawn two threads for all tests, just one now that `catch_unwind` is
stable.
* Remove usage of the unstable `box` keyword
* Remove usage of the unstable `FnBox` trait
bors [Fri, 7 Oct 2016 14:58:27 +0000 (07:58 -0700)]
Auto merge of #36945 - alexcrichton:proc-macro-rename, r=nrc
rustc: Rename rustc_macro to proc_macro
This commit blanket renames the `rustc_macro` infrastructure to `proc_macro`,
which reflects the general consensus of #35900. A follow up PR to Cargo will be
required to purge the `rustc-macro` name as well.
`ty_of_method_or_bare_fn` currently calls `pat_to_string` even when it
doesn't need to. Fixing this avoids hundreds of large allocations (e.g.
171,600 bytes, in `mk_printer`) in several of the rustc-benchmarks.
Alex Crichton [Mon, 3 Oct 2016 16:49:39 +0000 (09:49 -0700)]
rustc: Rename rustc_macro to proc_macro
This commit blanket renames the `rustc_macro` infrastructure to `proc_macro`,
which reflects the general consensus of #35900. A follow up PR to Cargo will be
required to purge the `rustc-macro` name as well.
Alex Crichton [Wed, 5 Oct 2016 16:50:15 +0000 (09:50 -0700)]
std: Minor cleanup to libtest
* Don't spawn two threads for all tests, just one now that `catch_unwind` is
stable.
* Remove usage of the unstable `box` keyword
* Remove usage of the unstable `FnBox` trait
Jonathan Turner [Thu, 6 Oct 2016 15:35:44 +0000 (08:35 -0700)]
Rollup merge of #36977 - Razican:bufreader_typo_fix, r=alexcrichton
Fixed small typo in `BufRead` comments
`BufRead` comments, in the `Seek` trait implementation, was talking about allocating 8 *ebibytes*. It was a typo, the correct unit is *exbibytes*, since *ebibytes* don't even exist. The calculation is correct, though.
Jonathan Turner [Thu, 6 Oct 2016 15:35:43 +0000 (08:35 -0700)]
Rollup merge of #36972 - nastevens:fix-rustbuild-per-target-musl-root, r=alexcrichton
Fix rustbuild per-target musl root
In #36292, support was added to target musl libc for ARM targets using rustbuild. Specifically, that change allowed the addition of per-target `musl-root` options in the rustbuild `config.toml` so that multiple targets depending on musl could be built. However, that implementation contained a couple of omissions: the `musl-root` option was added to the config, but was never added to the TOML parsing, and therefore was not actually being loaded from `config.toml`. This PR rectifies that.
Using these changes and a heavily modified version of the buildbot Docker container, I have been able to build rust targeting `armv7-unknown-linux-musleabihf` and have successfully run the binaries on a Raspberry Pi 3. I'm also planning to test `arm-unknown-linux-musleabi` and `arm-unknown-linux-musleabihf` systems, but have no reason to believe that this change would not simply work on those targets.
Jonathan Turner [Thu, 6 Oct 2016 15:35:41 +0000 (08:35 -0700)]
Rollup merge of #36832 - ParkHanbum:master, r=michaelwoerister
Add testcase for issue-32948
issue-32948 is similar to issue-32554.
issue-32948 : Symbol names for monomorphized trait impls are not stable across crates
issue-32554 : Symbol names for generics are not stable across crates
so, I append issue-32948's testcase to issue-32554's testcase.
thanks!
bors [Thu, 6 Oct 2016 01:10:34 +0000 (18:10 -0700)]
Auto merge of #36893 - apasel422:issue-32114, r=alexcrichton
Restore `DISCONNECTED` state in `oneshot::Packet::send`
Closes #32114
I'm not sure if this is the best approach, but the current action of swapping `DISCONNECTED` with `DATA` seems wrong. Additionally, it is strange that the `send` method (and others in the `oneshot` module) takes `&mut self` despite performing atomic operations, as this requires extra discipline to avoid data races and lets us use methods like `AtomicUsize::get_mut` instead of methods that require a memory ordering.
Razican [Wed, 5 Oct 2016 11:18:45 +0000 (13:18 +0200)]
Fixed small typo in `BufRead` comments
`BufRead` comments, in the `Seek` trait implementation, was talking
about allocating 8 *ebibytes*. It was a typo, the correct unit is
*exbibytes*, since *ebibytes* don't even exist. The calculation is
correct, though.
Nick Stevens [Wed, 5 Oct 2016 01:22:51 +0000 (20:22 -0500)]
rustbuild: Fix panic message when musl-root not set
The previous panic message delivered when a musl target was specified
but musl-root was not specified incorrectly instructed the user to add
the musl-root key to the "build" section of config.toml. The key
actually needs to be added to the "rust" section.
In #36292, support was added to target musl libc for ARM targets using
rustbuild. Specifically, that change allowed the addition of per-target
"musl-root" options in the rustbuild config.toml so that multiple
targets depending on musl could be built. However, that implementation
contained a couple of omissions: the musl-root option was added to the
config, but was never added to the TOML parsing, and therefore was not
actually being loaded from config.toml. This commit rectifies that and
allows successful building of musl-based ARM targets.
bors [Tue, 4 Oct 2016 23:30:30 +0000 (16:30 -0700)]
Auto merge of #36814 - petrochenkov:def, r=eddyb
Refactoring/bugfixing around definitions for struct/variant constructors
https://github.com/rust-lang/rust/commit/d917c364ad0edfa441e5c219da1b00511b976789 separates definitions for struct/variant constructors living in value namespace from struct/variant type definitions.
https://github.com/rust-lang/rust/commit/adfb37827b3a52a83dd11d5781e5b492714a5d4c fixes cross-crate resolution of reexports reexporting half-items, like struct constructors without struct type or types without constructor. Such reexports can appear due to glob shadowing.
Resolution now is not affected by the order in which items and reexports are decoded from metadata (cc https://github.com/rust-lang/rust/issues/31337#issuecomment-183996263). `try_define` is not used during building reduced graph anymore.
500 lines of this PR are tests for this exotic situation, the remaining line diff count is actually negative! :)
https://github.com/rust-lang/rust/commit/c695d0c8756a87c0708b97b7e277b6a3f4712b97 (and partially https://github.com/rust-lang/rust/commit/aabf132de04643602ec17b6abab9e276366d9c6d) moves most of pattern resolution checks from typeck to resolve (except those checking for associated items), uses the same wording for pattern resolution error messages from both typeck and resolve and makes the messages more precise.
https://github.com/rust-lang/rust/commit/11e3524e5afa4ac1b04aece67fb683f078e63f37 fixes seemingly incorrectly set `NON_ZERO_SIZED` attributes for struct/variant ctors in const eval.
https://github.com/rust-lang/rust/commit/4586fea2531054d9b25f8663f1ba4b32b07c11c2 eliminates `ty::VariantKind` in favor of `def::CtorKind`. The logic is that variant kinds are irrelevant for types, they make sense only when we deal with constructor functions/constants. Despite that `VariantDefData` still keeps a copy of `CtorKind`, but it's used only for various kinds of pretty-printing (and for storing in metadata).
https://github.com/rust-lang/rust/commit/aabf132de04643602ec17b6abab9e276366d9c6d is mostly a cleanup of various impossible or improperly used definitions, and other small definition cleanups.
bors [Tue, 4 Oct 2016 20:23:09 +0000 (13:23 -0700)]
Auto merge of #36933 - alexcrichton:less-neon-again, r=eddyb
rustc: Try again to disable NEON on armv7 linux
This is a follow-up to #35814 which apparently didn't disable it hard enough. It
looks like LLVM's default armv7 target enables NEON so we'd otherwise have to
pass `-neon`, but we're already enabling armv7 with `+v7` supposedly, so let's
try just telling LLVM that the armv7 target is arm and then enable features
selectively.