Alex Crichton [Thu, 19 Jan 2017 23:49:34 +0000 (15:49 -0800)]
Rollup merge of #39118 - jseyfried:token_tree_based_parser, r=nrc
Refactor the parser to consume token trees
This is groundwork for efficiently parsing attribute proc macro invocations, bang macro invocations, and `TokenStream`-based attributes and fragment matchers.
This improves parsing performance by 8-15% and expansion performance by 0-5% on a sampling of the compiler's crates.
Alex Crichton [Thu, 19 Jan 2017 23:49:33 +0000 (15:49 -0800)]
Rollup merge of #39114 - alexcrichton:fix-osx-image, r=brson
travis: Tweak OSX image configuration
Somewhere between https://travis-ci.org/rust-lang/rust/jobs/192352185 and
https://travis-ci.org/rust-lang/rust/jobs/192440181 it looks like our
configuration for a newer OSX image was lost as LLDB has reverted itself back to
350. This fix appeared to work for the libc crate so let's see if we can
configure it to work for the rust repo as well.
Alex Crichton [Thu, 19 Jan 2017 23:49:32 +0000 (15:49 -0800)]
Rollup merge of #39112 - alexcrichton:i586-msvc, r=nikomatsakis
appveyor: Test/Dist i586 MSVC
This is a target that we're shipping today, so this commit adds this matrix
entry to AppVeyor. This reuses the existing i686 MSVC matrix entry as it's
currently finishing about a half hour under two hours, which should hopefully
give it enough extra time to run this test suite.
This commit doesn't compile custom toolchains like our current cross-image does,
but instead compiles musl manually and then compiles libunwind manually (like
x86_64) for use for the ARM targets and just uses openwrt toolchains for the
mips targets.
Alex Crichton [Thu, 19 Jan 2017 23:49:23 +0000 (15:49 -0800)]
Rollup merge of #38842 - abonander:proc_macro_attribute, r=jseyfried
Implement `#[proc_macro_attribute]`
This implements `#[proc_macro_attribute]` as described in https://github.com/rust-lang/rfcs/pull/1566
The following major (hopefully non-breaking) changes are included:
* Refactor `proc_macro::TokenStream` to use `syntax::tokenstream::TokenStream`.
* `proc_macro::tokenstream::TokenStream` no longer emits newlines between items, this can be trivially restored if desired
* `proc_macro::TokenStream::from_str` does not try to parse an item anymore, moved to `impl MultiItemModifier for CustomDerive` with more informative error message
* Implement `#[proc_macro_attribute]`, which expects functions of the kind `fn(TokenStream, TokenStream) -> TokenStream`
* Reactivated `#![feature(proc_macro)]` and gated `#[proc_macro_attribute]` under it
* `#![feature(proc_macro)]` and `#![feature(custom_attribute)]` are mutually exclusive
* adding `#![feature(proc_macro)]` makes the expansion pass assume that any attributes that are not built-in, or introduced by existing syntax extensions, are proc-macro attributes
* Fix `feature_gate::find_lang_feature_issue()` to not use `unwrap()`
* This change wasn't necessary for this PR, but it helped debugging a problem where I was using the wrong feature string.
* Move "completed feature gate checking" pass to after "name resolution" pass
* This was necessary for proper feature-gating of `#[proc_macro_attribute]` invocations when the `proc_macro` feature flag isn't set.
bors [Fri, 20 Jan 2017 09:33:10 +0000 (09:33 +0000)]
Auto merge of #38955 - estebank:highlighted-diags, r=nikomatsakis
Teach Diagnostics to highlight text
Support styled `Diagnostic` output:
<img width="469" alt="mismatched types error with colorized types in the note" src="https://cloud.githubusercontent.com/assets/1606434/21871227/93a84198-d815-11e6-88b1-0ede3c7e28ef.png">
bors [Thu, 19 Jan 2017 21:15:12 +0000 (21:15 +0000)]
Auto merge of #39136 - keeperofdakeys:proc-macro-doc-test, r=alexcrichton
Stop warning when doc testing proc macro crates
Fixes #39064
Add the test option to the session struct that is passed
to phase_2_configure_and_expand function inside the
rustdoc test module.
This prevents the warning code from triggering when
parsing proc_macro_derive attributes, just like when
`--test` is normally invoked.
This change makes the warning disappear, but I'm not sure what else it might change. So this early PR is mainly to run the test suite, and to get feedback.
Guillaume Gomez [Thu, 19 Jan 2017 10:56:09 +0000 (11:56 +0100)]
Rollup merge of #39150 - birkenfeld:slice-doc, r=GuillaumeGomez
collections: update docs of slice get() and friends
Resubmit of #38216.
r? @GuillaumeGomez
BTW, instead of closing a PR just because it is old and the team member who offered to fix it up did not have the time to do so, why not ping them instead? (cc @alexcrichton)
bors [Tue, 17 Jan 2017 21:52:03 +0000 (21:52 +0000)]
Auto merge of #37972 - bluss:iter-find-is-on-a-roll, r=sfackler
Improve the slice iterator's searching methods
Improve all, any, find, position, rposition by explicitly unrolling the loop for the slice iterators.
- Introduce a few extension methods and functions for raw pointers make the new code easy to express
- Introduce helper methods `search_while, rsearch_while` that generalize all the searching methods
LLVM doesn't unroll the loop in `.find()` by default (clang is the same), so performance benefits a lot from explicit unrolling here. An iterator method without conditional exits (like `.fold()`) does not need this on the other hand.
One of the raw pointer extension methods is `fn post_inc(&mut self) -> Self` which is the rustic equivalent of “`ptr++`”, and it is a nice way to express the raw pointer loop (see commit 3).
Specific development notes about `search_while`: I tried both computing an end pointer "rounded" to 4, as well as the `ptrdistance >= 4` loop condition, ptrdistance was better. I tried handling the last 0-3 elements unrolled or with a while loop, the loop was better.
bors [Tue, 17 Jan 2017 18:22:20 +0000 (18:22 +0000)]
Auto merge of #39109 - michaelwoerister:incr-comp-cache-cleanup, r=nikomatsakis
incr.comp.: Delete orphaned work-products.
The new partitioning scheme uncovered a hole in our incr. comp. cache directory garbage collection. So far, we relied on unneeded work products being deleted during the initial cache invalidation phase. However, we the new scheme, we get object files/work products that only contain code from upstream crates. Sometimes this code is not needed anymore (because all callers have been removed from the source) but because nothing that actually influences the contents of these work products had changed, we never deleted them from disk.
bors [Tue, 17 Jan 2017 12:31:15 +0000 (12:31 +0000)]
Auto merge of #39110 - petrochenkov:sum, r=eddyb
Merge ObjectSum and PolyTraitRef in AST/HIR + some other refactoring
`ObjectSum` and `PolyTraitRef` are the same thing (list of bounds), they exist separately only due to parser quirks. The second commit merges them.
The first commit replaces `Path` with `Ty` in (not yet supported) equality predicates. They are parsed as types anyway and arbitrary types can always be disguised as paths using aliases, so this doesn't add any new functionality.
The third commit uses `Vec` instead of `P<[T]>` in AST. AST is not immutable like HIR and `Vec`s are more convenient for it, unnecessary conversions are also avoided.
The last commit renames `parse_ty_sum` (which is used for parsing types in general) into `parse_ty`, and renames `parse_ty` (which is used restricted contexts where `+` is not permitted due to operator priorities or other reasons) into `parse_ty_no_plus`.
This is the first part of https://github.com/rust-lang/rust/issues/39085#issuecomment-272743755 and https://github.com/rust-lang/rust/issues/39080 focused on data changes and mechanical renaming, I'll submit a PR with parser changes a bit later.
bors [Tue, 17 Jan 2017 08:01:04 +0000 (08:01 +0000)]
Auto merge of #39050 - alexcrichton:travis-nightly, r=brson
travis: Pass --release-channel=nightly on deploy
This commit passes the `--release-channel=nightly` flag to all images which have
the `DEPLOY` flag set. This means that we'll name artifacts and the compiler
appropriately.
This reworks a bit how arguments are passed, but for now doesn't change what's
already being passed. Eventually we'll want to avoid enabling debug assertions
and llvm assertions for *all* releases, but I figure we can tackle that a little
bit more down the road.
Alex Crichton [Fri, 13 Jan 2017 23:35:58 +0000 (15:35 -0800)]
travis: Pass --release-channel=nightly on deploy
This commit passes the `--release-channel=nightly` flag to all images which have
the `DEPLOY` flag set. This means that we'll name artifacts and the compiler
appropriately.
This reworks a bit how arguments are passed, but for now doesn't change what's
already being passed. Eventually we'll want to avoid enabling debug assertions
and llvm assertions for *all* releases, but I figure we can tackle that a little
bit more down the road.
This commit doesn't compile custom toolchains like our current cross-image does,
but instead compiles musl manually and then compiles libunwind manually (like
x86_64) for use for the ARM targets and just uses openwrt toolchains for the
mips targets.
Alex Crichton [Tue, 17 Jan 2017 02:43:38 +0000 (18:43 -0800)]
travis: Tweak OSX image configuration
Somewhere between https://travis-ci.org/rust-lang/rust/jobs/192352185 and
https://travis-ci.org/rust-lang/rust/jobs/192440181 it looks like our
configuration for a newer OSX image was lost as LLDB has reverted itself back to
350. This fix appeared to work for the libc crate so let's see if we can
configure it to work for the rust repo as well.
Alex Crichton [Tue, 17 Jan 2017 00:49:49 +0000 (16:49 -0800)]
appveyor: Test/Dist i586 MSVC
This is a target that we're shipping today, so this commit adds this matrix
entry to AppVeyor. This reuses the existing i686 MSVC matrix entry as it's
currently finishing about a half hour under two hours, which should hopefully
give it enough extra time to run this test suite.
Guillaume Gomez [Mon, 16 Jan 2017 21:48:24 +0000 (22:48 +0100)]
Rollup merge of #39084 - sphela:book-update-patterns, r=steveklabnik
An update to patterns documentation
As it is written the current pattern page creates a lot of confusion, even for someone with previous rust experience. It's so hard because it introduces an entirely new language feature without explaining. Someone could update it within the span of a few minutes by just explaining the newly introduced feature.
```rust
match c {
x => println!("x: {} c: {}", x, c),
}
```
No where in the book up to this point has it explained that identifiers match patterns with just a name create an irrefutable pattern. The page uses this feature without explanation, it just assumes that readers would immediately understand it. To confuse the issue even further the topic uses this feature to explain shadowing, placing two x's from different scopes and different meanings without ever explaining why there is shadowing.
What follows comes across as utterly nonsensical given everything the reader would know about Rust about this point:
```rust
the result:
x: c c: c
x: x
```
x is c? What? Yes even if you understand that x here is not the x in the previous scope why would x equal 'c' here? What previous chapter explained this? The previous chapter on 'matching' only mentions the catch all '_' and never in any shape or form mentioned that a name here creates an irrefutable pattern and binds a value.
There are numerous examples of people not understanding this section, not finding answers and looking for them online about `x: c c: c`:
And a [google search for `rust x: c c: c`](https://www.google.com/search?q=rust+%22x:+c+c:+c%22) finds many more people being tripped up, including people who speak a language other than English. I am confident that this page has resulted in questions on the irc channel more than once. Given rust already has a pretty big learning curve I recommend this be fixed.
I was asked to create PR from where I made this same case in the [rust book repository issue](https://github.com/rust-lang/book/issues/316) (I didn't realize this was a separate project).
bors [Mon, 16 Jan 2017 20:23:09 +0000 (20:23 +0000)]
Auto merge of #39094 - nagisa:i128-fix-endianness, r=eddyb
(Shot at) Fix endian bugs in i128 intrinsic impls
Attempt to fix the endianness issues on big-endian machines such as power pc. Could not test if it actually makes stuff work on the powerpc, because setting up cross-compiler for ppc seems to be nigh-impossible on arch.
bors [Mon, 16 Jan 2017 09:16:44 +0000 (09:16 +0000)]
Auto merge of #39071 - est31:master, r=GuillaumeGomez
Mark safe_suggestion and pushpop_unsafe as removed in feature_gate.rs
This removes two features from feature_gate.rs: `safe_suggestion` and `pushpop_unsafe`. Both had been removed in other places already, but were forgotten to be removed from feature_gate.rs.
bors [Mon, 16 Jan 2017 05:17:39 +0000 (05:17 +0000)]
Auto merge of #38806 - comex:lint-attr-fix, r=nrc
Fix lint attributes on non-item nodes.
Currently, late lint checking uses two HIR visitors: LateContext and
IdVisitor. IdVisitor only overrides visit_id, and for each node searches
for builtin lints previously added to the session; LateContext overrides
a number of methods, and runs late lints. When LateContext encounters an
item, it first has IdVisitor walk everything in it except nested items
(OnlyBodies), then recurses into it itself - i.e. there are two separate
walks.
Aside from apparently being unnecessary, this separation prevents lint
attributes (allow/deny/warn) on non-item HIR nodes from working
properly. Test case:
```rust
// generates warning without this change
fn main() { #[allow(unreachable_code)] loop { break; break; } }
```
LateContext contains logic to merge attributes seen into the current lint
settings while walking (with_lint_attrs), but IdVisitor does not. So
such attributes will affect late lints (because they are called from
LateContext), and if the node contains any items within it, they will
affect builtin lints within those items (because that IdVisitor is run
while LateContext is within the attributed node), but otherwise the
attributes will be ignored for builtin lints.
This change simply removes IdVisitor and moves its visit_id into
LateContext itself. Hopefully this doesn't break anything...
Also added walk calls to visit_lifetime and visit_lifetime_def
respectively, so visit_lifetime_def will recurse into the lifetime and
visit_lifetime will recurse into the name. In principle this could
confuse lint plugins. This is "necessary" because walk_lifetime calls
visit_id on the lifetime; of course, an alternative would be directly
calling visit_id (which would require manually iterating over the
lifetimes in visit_lifetime_def), but that seems less clean.