bors [Sat, 25 Nov 2017 00:17:03 +0000 (00:17 +0000)]
Auto merge of #46008 - alexcrichton:update-llvm, r=Mark-Simulacrum
rustbuild: Update LLVM and enable ThinLTO
This commit updates LLVM to fix #45511 (https://reviews.llvm.org/D39981) and
also reenables ThinLTO for libtest now that we shouldn't hit #45768. This also
opportunistically enables ThinLTO for libstd which was previously blocked
(#45661) on test failures related to debuginfo with a presumed cause of #45511.
Alex Crichton [Fri, 24 Nov 2017 17:18:22 +0000 (09:18 -0800)]
std: Flag Windows TLS dtor symbol as #[used]
Turns out ThinLTO was internalizing this symbol and eliminating it. Worse yet if
you compiled with LTO turns out no TLS destructors would run on Windows! The
`#[used]` annotation should be a more bulletproof implementation (in the face of
LTO) of preserving this symbol all the way through in LLVM and ensuring it makes
it all the way to the linker which will take care of it.
bors [Fri, 24 Nov 2017 21:50:09 +0000 (21:50 +0000)]
Auto merge of #46111 - michaelwoerister:promote-green, r=nikomatsakis
incr.comp.: Make sure we don't lose unused green results from the query cache.
In its current implementation, the query result cache works by bulk-writing the results of all cacheable queries into a monolithic binary file on disk. Prior to this PR, we would potentially lose query results during this process because only results that had already been loaded into memory were serialized. In contrast, results that were not needed during the given compilation session were not serialized again.
This PR will do one pass over all green `DepNodes` that represent a cacheable query and execute the corresponding query in order to make sure that the query result gets loaded into memory before cache serialization.
In the future we might want to look into a serialization format the can be updated in-place so that we don't have to load unchanged results just for immediately storing them again.
bors [Fri, 24 Nov 2017 15:11:11 +0000 (15:11 +0000)]
Auto merge of #46093 - scottmcm:lower-128-mir, r=nagisa
Add a MIR pass to lower 128-bit operators to lang item calls
Runs only with `-Z lower_128bit_ops` since it's not hooked into targets yet.
This isn't really useful on its own, but the declarations for the lang items need to be in the compiler before compiler-builtins can be updated to define them, so this is part 1 of at least 3.
cc https://github.com/rust-lang/rust/issues/45676 @est31 @nagisa
Alex Crichton [Wed, 15 Nov 2017 13:16:23 +0000 (05:16 -0800)]
rustbuild: Update LLVM and enable ThinLTO
This commit updates LLVM to fix #45511 (https://reviews.llvm.org/D39981) and
also reenables ThinLTO for libtest now that we shouldn't hit #45768. This also
opportunistically enables ThinLTO for libstd which was previously blocked
(#45661) on test failures related to debuginfo with a presumed cause of #45511.
bors [Fri, 24 Nov 2017 01:44:19 +0000 (01:44 +0000)]
Auto merge of #45942 - Menschenkindlein:master, r=estebank
Add hints for the case of confusing enum with its variants
A solution for https://github.com/rust-lang/rust/issues/43871. When one uses an enum in a place that accepts variants (e.g., `Option(result)` instead of `Some(result)`), suggest one of this enum's variants.
Make float::from_bits transmute (and update the documentation to reflect this).
The current implementation/documentation was made to avoid sNaN because of
potential safety issues implied by old/bad LLVM documentation. These issues
aren't real, so we can just make the implementation transmute (as permitted
by the existing documentation of this method).
Also the documentation didn't actually match the behaviour: it said we may
change sNaNs, but in fact we canonicalized *all* NaNs.
Also an example in the documentation was wrong: it said we *always* change
sNaNs, when the documentation was explicitly written to indicate it was
implementation-defined.
This makes to_bits and from_bits perfectly roundtrip cross-platform, except
for one caveat: although the 2008 edition of IEEE-754 specifies how to
interpet the signaling bit, earlier editions didn't. This lead to some platforms
picking the opposite interpretation, so all signaling NaNs on x86/ARM are quiet
on MIPS, and vice-versa.
NaN-boxing is a fairly important optimization, while we don't even guarantee
that float operations properly preserve signalingness. As such, this seems like
the more natural strategy to take (as opposed to trying to mangle the signaling
bit on a per-platform basis).
Guillaume Gomez [Thu, 23 Nov 2017 20:51:09 +0000 (21:51 +0100)]
Rollup merge of #45635 - virgil-palanciuc:master, r=kennytm
Add test for #44953
Added the requested test - trying to see if it passes; my local build fails, but not sure why - the nightly shows this output, but in my build the compilation error changed.
bors [Thu, 23 Nov 2017 10:46:02 +0000 (10:46 +0000)]
Auto merge of #46054 - nikomatsakis:nll-master-to-rust-master-1, r=arielb1
typeck aggregate rvalues in MIR type checker
This branch is an attempt to land content by @spastorino and @Nashenas88 that was initially landed on nll-master while we waited for https://github.com/rust-lang/rust/pull/45825 to land.
The biggest change it contains is that it extends the MIR type-checker to also type-check MIR aggregate rvalues (at least partially). Specifically, it checks that the operands provided for each field have the right type.
It does not yet check that their well-formedness predicates are met. That is https://github.com/rust-lang/rust/issues/45827. It also does not check other kinds of rvalues (that is https://github.com/rust-lang/rust/issues/45959). @spastorino is working on those issues now.
bors [Thu, 23 Nov 2017 08:20:33 +0000 (08:20 +0000)]
Auto merge of #46051 - cramertj:in-band-lifetimes, r=nikomatsakis
Implement in-band lifetime bindings
TODO (perhaps in a future PR): Should we ban explicit instantiation of generics with in-band lifetimes, or is it uncontroversial to just append them to the end of the lifetimes list?
This PR allows a user to filter items present in the save-analysis data by setting the `reachable_only` config option. This option is intended for use by the new rustdoc. The PR isn't quite finished, because it's dependent on a new release of rls-data, but I want to make sure that the approach is valid.
https://github.com/nrc/rls-analysis/issues/79 mentions that `pub use` might need to be handled, but my thinking is that the consumer of the analysis data would be able to infer which imports are `pub use`, and which items are only reachable through `pub use`, so that doesn't need to be handled here.
Niko Matsakis [Wed, 22 Nov 2017 17:06:19 +0000 (12:06 -0500)]
modify inherent impls test to indicate `TypeckTables` do not change
I also added some comments explaining what is going on. In short, the
changes in question do not, in fact, affect the`TypeckTables` in any
semantic way. However, altering the order of lowering can cause it
appear to affect the `TypeckTables`: if we lower generics before the
body, then the `HirId` for things in the body will be affected. In
this case, we are now lowering the generics etc
*after* the body, so the hash no longer changes. This seems good.
bors [Wed, 22 Nov 2017 07:27:54 +0000 (07:27 +0000)]
Auto merge of #46040 - zilbuz:mir-misc, r=nikomatsakis
MIR-borrowck: Some minor fixes
- Remove parens when printing dereference (fix #45185)
- Change argument type of `autoderef` to `bool`
- Change argument type of `field_index` to `Field`
bors [Tue, 21 Nov 2017 22:52:19 +0000 (22:52 +0000)]
Auto merge of #45879 - nikomatsakis:nll-kill-cyclic-closures, r=arielb1
move closure kind, signature into `ClosureSubsts`
Instead of using side-tables, store the closure-kind and signature in the substitutions themselves. This has two key effects:
- It means that the closure's type changes as inference finds out more things, which is very nice.
- As a result, it avoids the need for the `freshen_closure_like` code (though we still use it for generators).
- It avoids cyclic closures calls.
- These were never meant to be supported, precisely because they make a lot of the fancy inference that we do much more complicated. However, due to an oversight, it was previously possible -- if challenging -- to create a setup where a closure *directly* called itself (see e.g. #21410).
We have to see what the effect of this change is, though. Needs a crater run. Marking as [WIP] until that has been assessed.
kennytm [Tue, 21 Nov 2017 17:13:05 +0000 (01:13 +0800)]
Rollup merge of #46155 - SimonSapin:stylo, r=kennytm
Revert servo to upstream in cargotest
This is a follow-up to https://github.com/rust-lang/rust/pull/45225#issuecomment-345503017 now that upstream has adjusted: https://github.com/servo/servo/pull/19316
kennytm [Tue, 21 Nov 2017 17:13:00 +0000 (01:13 +0800)]
Rollup merge of #46103 - zackmdavis:dead_code_lint_should_say_never_constructed_for_variants, r=arielb1
dead code lint to say "never constructed" for variants
As reported in #19140, #44083, and #44565, some users were confused when
the dead-code lint reported an enum variant to be "unused" when it was
matched on (but not constructed). This wording change makes it clearer
that the lint is in fact checking for construction.
We continue to say "used" for all other items (it's tempting to say
"called" for functions and methods, but this turns out not to be
correct: functions can be passed as arguments and the dead-code lint
isn't special-casing that or anything).
kennytm [Tue, 21 Nov 2017 17:12:58 +0000 (01:12 +0800)]
Rollup merge of #46050 - sunfishcode:read_to_end, r=sfackler
Optimize `read_to_end`.
This patch makes `read_to_end` use Vec's memory-growth pattern rather
than using a custom pattern.
This has some interesting effects:
- If memory is reserved up front, `read_to_end` can be faster, as it
starts reading at the buffer size, rather than always starting at 32
bytes. This speeds up file reading by 2x in one of my use cases.
- It can reduce the number of syscalls when reading large files.
Previously, `read_to_end` would settle into a sequence of 8192-byte
reads. With this patch, the read size follows Vec's allocation
pattern. For example, on a 16MiB file, it can do 21 read syscalls
instead of 2057. In simple benchmarks of large files though, overall
speed is still dominated by the actual I/O.
- A downside is that Read implementations that don't implement
`initializer()` may see increased memory zeroing overhead.
I benchmarked this on a variety of data sizes, with and without
preallocated buffers. Most benchmarks see no difference, but reading
a small/medium file with a pre-allocated buffer is faster.
kennytm [Tue, 21 Nov 2017 17:12:57 +0000 (01:12 +0800)]
Rollup merge of #46031 - Keruspe:cargofmt, r=Mark-Simulacrum
rustbuild: distribute cargo-fmt alongside rustfmt
Not sure whether we want that nor if it's the right way to do so, but it feels quite weird to have rustfmt without cargo-fmt. Or are there other plans wrt that?
Simon Sapin [Tue, 21 Nov 2017 13:21:24 +0000 (14:21 +0100)]
Revert servo to upstream in cargotest
This is a follow-up to https://github.com/rust-lang/rust/pull/45225#issuecomment-345503017
now that upstream has adjusted https://github.com/servo/servo/pull/19316
bors [Tue, 21 Nov 2017 10:00:18 +0000 (10:00 +0000)]
Auto merge of #45701 - cramertj:impl-trait-this-time, r=eddyb
impl Trait Lifetime Handling
This PR implements the updated strategy for handling `impl Trait` lifetimes, as described in [RFC 1951](https://github.com/rust-lang/rfcs/blob/master/text/1951-expand-impl-trait.md) (cc #42183).
With this PR, the `impl Trait` desugaring works as follows:
```rust
fn foo<T, 'a, 'b, 'c>(...) -> impl Foo<'a, 'b> { ... }
// desugars to
exists type MyFoo<ParentT, 'parent_a, 'parent_b, 'parent_c, 'a, 'b>: Foo<'a, 'b>;
fn foo<T, 'a, 'b, 'c>(...) -> MyFoo<T, 'static, 'static, 'static, 'a, 'b> { ... }
```
All of the in-scope (parent) generics are listed as parent generics of the anonymous type, with parent regions being replaced by `'static`. Parent regions referenced in the `impl Trait` return type are duplicated into the anonymous type's generics and mapped appropriately.
One case came up that wasn't specified in the RFC: it's possible to write a return type that contains multiple regions, neither of which outlives the other. In that case, it's not clear what the required lifetime of the output type should be, so we generate an error.
There's one remaining FIXME in one of the tests: `-> impl Foo<'a, 'b> + 'c` should be able to outlive both `'a` and `'b`, but not `'c`. Currently, it can't outlive any of them. @nikomatsakis and I have discussed this, and there are some complex interactions here if we ever allow `impl<'a, 'b> SomeTrait for AnonType<'a, 'b> { ... }`, so the plan is to hold off on this until we've got a better idea of what the interactions are here.
bors [Tue, 21 Nov 2017 06:42:14 +0000 (06:42 +0000)]
Auto merge of #45545 - durka:macro-backtrace, r=nrc
show macro backtrace with -Z flag
Fixes #39413 by adding a facility to restore the "old school" macro expansion backtraces (previously removed in https://github.com/rust-lang/rust/commit/61865384b8fa6d79d2b36cbd7c899eaf15f4aeea).
The restored functionality is accessed through the flag `-Z external-macro-backtrace`. Errors showing the truncated backtraces will suggest this flag.
<br/><br/>
Running without env var (note: first error is from remote macro, second from local macro):
<details>
```
$ cargo +custom run
Compiling b v0.1.0
error: expected one of `!`, `.`, `::`, `;`, `?`, `{`, `}`, or an operator, found `error`
--> src/main.rs:12:5
|
12 | a!();
| ^^^^^
| |
| expected one of 8 possible tokens here
| unexpected token
|
= note: this error originates in a macro outside of the current crate (run with RUST_MACRO_BACKTRACE=1 for more info)
error: expected one of `!`, `.`, `::`, `;`, `?`, `{`, `}`, or an operator, found `error`
--> src/main.rs:7:16
|
7 | syntax error;
| -^^^^^ unexpected token
| |
| expected one of 8 possible tokens here
...
13 | b!();
| ----- in this macro invocation
error: aborting due to 2 previous errors
error: Could not compile `b`.
To learn more, run the command again with --verbose.
```
</details>
The output is the same as today, except for an addition to the note which aids discoverability of the new environment variable.
<br/><br/>
Running _with_ env var:
<details>
```
$ RUST_MACRO_BACKTRACE=1 cargo +custom run
Compiling b v0.1.0
error: expected one of `!`, `.`, `::`, `;`, `?`, `{`, `}`, or an operator, found `error`
--> <a macros>:1:72
|
1 | ( ) => { a ! ( @ ) } ; ( @ ) => { a ! ( @ @ ) } ; ( @ @ ) => { syntax error ;
| -^^^^^ unexpected token
| |
| expected one of 8 possible tokens here
src/main.rs:12:5: 12:10 note: in this expansion of a! (defined in <a macros>)
<a macros>:1:11: 1:20 note: in this expansion of a! (defined in <a macros>)
<a macros>:1:36: 1:47 note: in this expansion of a! (defined in <a macros>)
error: expected one of `!`, `.`, `::`, `;`, `?`, `{`, `}`, or an operator, found `error`
--> src/main.rs:7:16
|
7 | syntax error;
| -^^^^^ unexpected token
| |
| expected one of 8 possible tokens here
src/main.rs:12:5: 12:10 note: in this expansion of a! (defined in <a macros>)
<a macros>:1:11: 1:20 note: in this expansion of a! (defined in <a macros>)
<a macros>:1:36: 1:47 note: in this expansion of a! (defined in <a macros>)
error: expected one of `!`, `.`, `::`, `;`, `?`, `{`, `}`, or an operator, found `error`
--> src/main.rs:7:16
|
7 | syntax error;
| -^^^^^ unexpected token
| |
| expected one of 8 possible tokens here
src/main.rs:13:5: 13:10 note: in this expansion of b! (defined in src/main.rs)
src/main.rs:4:13: 4:18 note: in this expansion of b! (defined in src/main.rs)
src/main.rs:5:14: 5:20 note: in this expansion of b! (defined in src/main.rs)
error: aborting due to 2 previous errors
error: Could not compile `b`.
To learn more, run the command again with --verbose.
```
</details>
The output is hard to read, but better than nothing (and it's exactly what we used to have before the infamous `fix_multispans_in_std_macros`).
<br/><br/>
Wishlist:
- Save the actual source of macros in crate metadata, not just AST, so the output can be improved
- Hopefully this would allow line numbers in the trace as well
- Show the actual macro invocations in the traces
bors [Tue, 21 Nov 2017 03:03:28 +0000 (03:03 +0000)]
Auto merge of #45039 - QuietMisdreavus:doc-spotlight, r=GuillaumeGomez,QuietMisdreavus
show in docs whether the return type of a function impls Iterator/Read/Write
Closes #25928
This PR makes it so that when rustdoc documents a function, it checks the return type to see whether it implements a handful of specific traits. If so, it will print the impl and any associated types. Rather than doing this via a whitelist within rustdoc, i chose to do this by a new `#[doc]` attribute parameter, so things like `Future` could tap into this if desired.
### Known shortcomings
~~The printing of impls currently uses the `where` class over the whole thing to shrink the font size relative to the function definition itself. Naturally, when the impl has a where clause of its own, it gets shrunken even further:~~ (This is no longer a problem because the design changed and rendered this concern moot.)
The lookup currently just looks at the top-level type, not looking inside things like Result or Option, which renders the spotlights on Read/Write a little less useful:
<details><summary>`File::{open, create}` don't have spotlight info (pic of old design)</summary>
All three of the initially spotlighted traits are generically implemented on `&mut` references. Rustdoc currently treats a `&mut T` reference-to-a-generic as an impl on the reference primitive itself. `&mut Self` counts as a generic in the eyes of rustdoc. All this combines to create this lovely scene on `Iterator::by_ref`:
<details><summary>`Iterator::by_ref` spotlights Iterator, Read, and Write (pic of old design)</summary>
kennytm [Mon, 20 Nov 2017 19:14:42 +0000 (03:14 +0800)]
Rollup merge of #46088 - vitiral:read_doc, r=steveklabnik
add doc for doing `Read` from `&str`
This information can be found on [stackoverflow](https://stackoverflow.com/questions/32674905/pass-string-to-function-taking-read-trait) but I think it would be beneficial if it was documented in the `Read` trait itself.
I had an *extremely* hard time finding this information, and "mocking" a reader with a string is an EXTREMELY common thing (I believe).
kennytm [Mon, 20 Nov 2017 19:14:41 +0000 (03:14 +0800)]
Rollup merge of #46082 - Enet4:mutex_from, r=sfackler
impl From for Mutex and RwLock
I felt that these implementations were missing, because doing `x.into()` works for other smart containers (such as `RefCell`), and in general I would say that the conversion makes sense.