Pietro Albini [Sat, 22 Sep 2018 07:56:41 +0000 (09:56 +0200)]
Rollup merge of #54420 - nnethercote:PackedRWU-Vec, r=nikomatsakis
Compress `Liveness` data some more.
Profiling shows that the `(reader, writer, used)` triples used by
liveness analysis almost always have invalid `reader` and `writer`
fields. We can take advantage of this knowledge to use a compressed
representation for them, falling back to a secondary table for the
uncommon cases.
This change reduces instruction counts on numerous benchmarks, the best
by 16%. It also reduces max-rss on numerous benchmarks, the best by 38%.
The patch also renames these triples from `Users` to `RWU`, because it's
confusing having a type whose name is plural and then used within
vectors whose names are also plural.
Pietro Albini [Sat, 22 Sep 2018 07:56:40 +0000 (09:56 +0200)]
Rollup merge of #54415 - petrochenkov:norollback, r=estebank
parser: Tweak function parameter parsing to avoid rollback on succesfull path
Since rollback is not perfect and may e.g. leave non-fatal errors after it, we need to make sure compilation fails if it happens.
So in particular case of `fn parse_arg_general` we need to parse the "good" `TYPE` first and only then rollback and recover erroneous `PAT: TYPE` if necessary.
Found when working on https://github.com/rust-lang/rfcs/pull/2544#issuecomment-423293222.
Pietro Albini [Sat, 22 Sep 2018 07:56:39 +0000 (09:56 +0200)]
Rollup merge of #54413 - memoryruins:deref-error-twice, r=estebank
Add UI test for deref recursion limit printing twice
Closes #38940
Does ``NOTE`` in the test need to be changed to ``HELP`` if its in the stderr?
``help: consider adding a `#![recursion_limit="20"]` attribute to your crate``
It doesn't appear to complaining locally that the line isn't set to ``HELP`` in the test, and the guide says
> HELP and SUGGESTION*
> * Note: SUGGESTION must follow immediately after HELP.
Pietro Albini [Sat, 22 Sep 2018 07:56:35 +0000 (09:56 +0200)]
Rollup merge of #54402 - bgermann:master, r=alexcrichton
Use no_default_libraries for all NetBSD flavors
The no_default_libraries was introduced in #28578 because the
NetBSD-based rumprun needed to disable the link flag.
This moves the definition to be used by all NetBSD linker flavors to
close #49627.
A different solution would be adding -lc but as there is no platform
with explicit -lc, this approach is used.
Pietro Albini [Sat, 22 Sep 2018 07:56:28 +0000 (09:56 +0200)]
Rollup merge of #54280 - japaric:no-cas-for-thumbv6, r=alexcrichton
remove (more) CAS API from Atomic* types where not natively supported
closes #54276
In PR #51953 I made the Atomic* types available on targets like thumbv6m and
msp430 with the intention of *only* exposing the load and store API on those
types -- the rest of the API doesn't work on those targets because the are no
native instructions to implement CAS loops.
Unfortunately, it seems I didn't properly cfg away all the CAS API on those
targets, as evidenced in #54276. This PR amends the issue by removing the rest
of the CAS API.
This is technically a breaking change because *libraries* that were using this
API and were being compiled for e.g. thumbv6m-none-eabi will stop compiling.
However, using those libraries (before this change) in programs (binaries) would
lead to linking errors when compiled for e.g. thumbv6m so this change
effectively shifts a linker error in binaries to a compiler error in libraries.
On a side note: extending the Atomic API is a bit error prone because of these
non-cas targets. Unless the author of the change is aware of these targets and
properly uses `#[cfg(atomic = "cas")]` they could end up exposing new CAS API on
these targets. I can't think of a test to check that an API is not present on
some target, but we could extend the `tidy` tool to check that *all* newly added
atomic API has the `#[cfg(atomic = "cas")]` attribute unless it's whitelisted in
`tidy` then the author of the change would have to verify if the API can be used
on non-cas targets.
In any case, I'd like to plug this hole ASAP. We can revisit testing in a
follow-up issue / PR.
Pietro Albini [Sat, 22 Sep 2018 07:56:24 +0000 (09:56 +0200)]
Rollup merge of #53652 - oconnor663:copy_in_place, r=alexcrichton
define copy_within on slices
This is a safe wrapper around `ptr::copy`, for regions within a single slice. Previously, safe in-place copying was only available as a side effect of `Vec::drain`.
I've wanted this API a couple times in the past, and I figured I'd just whip up a PR to help discuss it. It's possible something like this exists elsewhere and I just missed it. It might also be a big enough addition to warrant an RFC, I'm not sure.
Auto merge of #54223 - pnkfelix:issue-53764-uiify-run-pass, r=nikomatsakis
`ui`ify run-pass
This addresses the remainder of #53764 by first converting `src/test/run-pass` into another `ui`-style test suite, and then turning on `compare-mode=nll` for that new suite.
After this lands, we can address #54047 for the short term by moving all the `src/test/ui/run-pass` tests back to `src/test/run-pass`.
(Longer term, the compiler team's current (hypothetical sketch of a) plan (see [1][], [2][]) is unify all the tests by embedding these meta-properties like "// run-pass` into their headers explicitly and dropping the significance of their location on the file system.)
Alex Crichton [Thu, 20 Sep 2018 21:37:53 +0000 (14:37 -0700)]
Update Cargo
This commit brings in a few Cargo updates
* Updates Cargo with experimental HTTP/2 support - a post on the forums
will be made about testing this when available.
* Bumps Cargo's own version number
* In the case of `derive-same-struct`, it seemed cleaner to add the
output than to try to modify the macro itself (which is where the
output is coming from).
* In the case of `custom-derive-partial-eq`, it was just easier to add
the output than to attempt to port the test to use a procedural
macro.
Provide way for ui tests to opt out of having their output checked.
Namely, this adds support for:
* `// dont-check-compiler-stdout`, and
* `// dont-check-compiler-stderr`.
Obviously almost all ui tests wont want to opt into these, since the whole point
of a ui test is to check the compiler ui. However, since this PR is converting
run-pass into (another set of) ui tests, these header options make sense in that
context.
(Also this puts us into a better position for eventually turning
*every* test suite into a ui test suite, by making ui-ness the default
and forcing tests to opt out explicitly.)
Profiling shows that the `(reader, writer, used)` triples used by
liveness analysis almost always have invalid `reader` and `writer`
fields. We can take advantage of this knowledge to use a compressed
representation for them, falling back to a secondary table for the
uncommon cases.
This change reduces instruction counts on numerous benchmarks, the best
by 16%. It also reduces max-rss on numerous benchmarks, the best by 38%.
The patch also renames these triples from `Users` to `RWU`, because it's
confusing having a type whose name is plural and then used within
vectors whose names are also plural.
Auto merge of #54399 - alexcrichton:fix-bug, r=steveklabnik
std: Check for overflow in `str::repeat`
This commit fixes a buffer overflow issue in the standard library
discovered by Scott McMurray where if a large number was passed to
`str::repeat` it may cause and out of bounds write to the buffer of a `Vec`.
This bug was accidentally introduced in #48657 when optimizing the
`str::repeat` function. The bug affects stable Rust releases 1.26.0 to
1.29.0. We plan on backporting this fix to create a 1.29.1 release, and
the 1.30.0 release onwards will include this fix.
The fix in this commit is to introduce a deterministic panic in the case of
capacity overflow. When repeating a slice where the resulting length is larger
than the address space, there’s no way it can succeed anyway!
The standard library and surrounding libraries were briefly checked to see if
there were othere instances of preallocating a vector with a calculation that
may overflow. No instances of this bug (out of bounds write due to a calculation
overflow) were found at this time.
Note that this commit is the first steps towards fixing this issue,
we'll be making a formal post to the Rust security list once these
commits have been merged.
- #52813 (Duration div mul extras)
- #53470 (Warn about metadata loader errors)
- #54233 (Remove LLVM 3.9 workaround.)
- #54257 (Switch wasm math symbols to their original names)
- #54258 (Enable fatal warnings for the wasm32 linker)
- #54266 (Update LLVM to fix "bool" arguments on PPC32)
- #54290 (Switch linker for aarch64-pc-windows-msvc from LLD to MSVC)
- #54292 (Suggest array indexing when tuple indexing on an array)
- #54295 (A few cleanups and minor improvements to rustc/traits)
- #54298 (miri: correctly compute expected alignment for field)
- #54333 (Update The Book to latest)
- #54337 (Remove unneeded clone() from tests in librustdoc)
- #54346 (rustc: future-proof error reporting for polymorphic constants in types.)
- #54362 (Pass --batch to gdb)
- #54367 (Add regression test for thread local static mut borrows)
The no_default_libraries was introduced in #28578 because the
NetBSD-based rumprun needed to disable the link flag.
This moves the definition to be used by all NetBSD linker flavors to
close #49627.
A different solution would be adding -lc but as there is no platform
with explicit -lc, this approach is used.
Alex Crichton [Thu, 20 Sep 2018 16:31:14 +0000 (09:31 -0700)]
std: Check for overflow in `str::repeat`
This commit fixes a buffer overflow issue in the standard library
discovered by Scott McMurray where if a large number was passed to
`str::repeat` it may cause and out of bounds write to the buffer of a `Vec`.
This bug was accidentally introduced in #48657 when optimizing the
`str::repeat` function. The bug affects stable Rust releases 1.26.0 to
1.29.0. We plan on backporting this fix to create a 1.29.1 release, and
the 1.30.0 release onwards will include this fix.
The fix in this commit is to introduce a deterministic panic in the case of
capacity overflow. When repeating a slice where the resulting length is larger
than the address space, there’s no way it can succeed anyway!
The standard library and surrounding libraries were briefly checked to see if
there were othere instances of preallocating a vector with a calculation that
may overflow. No instances of this bug (out of bounds write due to a calculation
overflow) were found at this time.
Note that this commit is the first steps towards fixing this issue,
we'll be making a formal post to the Rust security list once these
commits have been merged.
Rollup merge of #54367 - spastorino:add-thread-local-static-borrow-test, r=pnkfelix
Add regression test for thread local static mut borrows
FIXME(#54366) - We probably shouldn't allow `#[thread_local] static mut` to get a `'static` lifetime, but for now, we should at least test the behavior that `rustc` currently has.
Rollup merge of #54362 - tromey:travis-gdb-batch-mode, r=nikomatsakis
Pass --batch to gdb
In one of my travis builds, I was surprised to find that the gdb
pager was in use and caused travis to time out. Adding `--batch`
to the gdb invocation will disable the pager. Note that the
`-ex q` is retained, to make sure gdb exits with status 0, just in
case `set -e` is in effect somehow.
Rollup merge of #54346 - eddyb:constant-horror, r=nikomatsakis
rustc: future-proof error reporting for polymorphic constants in types.
Currently, we have 3 categories of positions where a constant can be used (`const` and associated `const` can be considered "aliases" for an expression):
* runtime - if the function is polymorphic, we could even just warn and emit a panic
* `static` - always monomorphic, so we can error at definition site
* type-system - **must** *enforce* evaluation success where it was written
That last one is the tricky one, because we can't easily turn *the presence* a type with an erroring const into a runtime panic, and we'd have to do post-monomorphization errors (which we'd rather avoid).
<hr/>
The solution we came up with, as part of the plans for const generics, is to require successful evaluation wherever a constant shows up in a type (currently in array lengths, and values for const parameters in the future), *through* the WF system, which means that in certain situations (e.g. function signatures) we can assume evaluation *will* succeed, and require it of users (e.g. callers) instead (we've been doing this for lifetime bounds, for a long time now, and it's pretty ergonomic).
So once we do sth about #43408, this example *should* work, by propagating the responsability, to callers of `foo::<X>`, of proving `std::mem::size_of::<X>()` succeeds (and those callers can do the same).
```rust
pub fn foo<T>(_: [u8; std::mem::size_of::<T>()]) {}
```
But this one *shouldn't*, as there is nothing in the signature/bounds to indicate it:
```rust
pub fn bar<T>() {
let _: [u8; std::mem::size_of::<T>()];
}
```
<hr/>
I've come across some bit of code in the compiler that ignores const-evaluation errors *even when* they come from a constant in a type, and I've added an ICE *only when* there are no other reported errors (e.g. it's fine to ignore evaluation errors if the constant doesn't even type-check).
Right now, rustdoc pulls all its impl information by scanning a crate's HIR for any items it finds. However, it doesn't recurse into anything other than modules, preventing it from seeing trait impls that may be inside things like functions or consts. Thanks to https://github.com/rust-lang/rust/pull/53002, now these items actually *exist* for rustdoc to see, but they still weren't getting collected for display.
But there was a secret. Whenever we pull in an item from another crate, we don't have any of its impls in the local HIR, so instead we ask the compiler for *everything* and filter out after the fact. This process is only triggered if there's a cross-crate re-export in the crate being documented, which can sometimes leave this info out of the docs. This PR instead moves this collection into an early pass, which occurs immediately after crate cleaning, so that that collection occurs regardless. In addition, by including the HIR's own `trait_impls` in addition to the existing `all_trait_implementations` calls, we can collect all these tricky trait impls without having to scan for them!
Rollup merge of #54295 - ljedrz:cleanups_rustc_traits, r=nikomatsakis
A few cleanups and minor improvements to rustc/traits
It's a little bigger than usual, so bear with me ^^:
- introduce `TyCtxt::all_impls` and use it to avoid inefficiently allocating push loops
- modify `ArgKind::from_expected_ty` to take an `Option<Span>` argument to make it more versatile
- replace `ArgKind::Arg("_".to_owned(), "_".to_owned())` with `ArgKind::empty`
- move early `return`s earlier where possible
- if all branches of a `match` end with the same expression, move it after it
- change a hacky `match` expression to an `if else` chain
- move the `push` out from a push loop closure to reduce the number of allocations
- correct the vector size for `pretty_predicates` (under `specialize`)
- take advantage of the fact that `if else` is an expression
- prefer `cloned()` to `map(|&x| x)` and `map(|x| *x)`
- prefer `vec![x; y.len()]` to `y.map(|_| x).collect()`
- use `unwrap_or_else` instead of `match` where applicable
- use `if let` instead of `match` when only one branch matters
- prefer `to_owned` to `to_string` for string literals
- remove explicit `return`s
- remove superfluous braces
- whitespace fixes
- several other minor improvements
Rollup merge of #54290 - michaelwoerister:aarch64-back-to-msvc, r=alexcrichton
Switch linker for aarch64-pc-windows-msvc from LLD to MSVC
The MSVC linker does not seem to have the same problems with Rust symbols that LLD currently has on Windows (see https://github.com/rust-lang/rust/issues/54190#issuecomment-421288625). This PR makes MSVC the default linker for `aarch64-pc-windows-msvc`.
Rollup merge of #54258 - alexcrichton:lld-fatal-warnings, r=eddyb
Enable fatal warnings for the wasm32 linker
Historically LLD has emitted warnings for various reasons but all the bugs have
since been fixed (yay!) and by enabling fatal warnings we should be able to head
off bugs like #53390 sooner.
Rollup merge of #54257 - alexcrichton:wasm-math-symbols, r=TimNN
Switch wasm math symbols to their original names
The names `Math_*` were given to help undefined symbol messages indicate how to
implement them, but these are all implemented in compiler-rt now so there's no
need to rename them! This change should make it so wasm binaries by default, no
matter the math symbols used, will not have unresolved symbols.
Rollup merge of #53470 - bjorn3:warn_metadata_errors, r=alexcrichton
Warn about metadata loader errors
Output when writing corrupting to libcore.rlib
```
warning: no metadata found: failed to read rlib metadata in '/Users/bjorn/Documents/rust_fork/build/x86_64-apple-darwin/stage1-std/x86_64-apple-darwin/release/deps/libcore-857d662d379c5d0c.rlib': File too small to be an archive
Rollup merge of #52813 - newpavlov:duration_mul_div_extras, r=alexcrichton
Duration div mul extras
Successor of #52556.
This PR adds the following `impl`s:
- `impl Mul<Duration> for u32` (to allow `10*SECOND` in addition to `SECOND*10`)
- `impl Mul<f64> for Duration` (to allow `2.5*SECOND` vs `2*SECOND + 500*MILLISECOND`)
- `impl Mul<Duration> for f64`
- `impl MulAssign<f64> for Duration`
- `impl Div<f64> for Duration`
- `impl DivAssign<f64> for Duration`
- `impl Div<Duration> for Duration` (`Output = f64`, can be useful e.g. for `duration/MINUTE`)
`f64` is chosen over `f32` to minimize rounding errors. (52 bits fraction precision vs `Duration`'s ~94 bit)
- clean/inline.rs needs this map to fill in traits when inlining
- fold.rs needs this map to allow passes to fold trait items
- html/render.rs needs this map to seed the Cache.traits map of all
known traits
The first two are the real problem, since `DocFolder` only operates on
`clean::Crate` but `clean/inline.rs` only sees the `DocContext`. The
introduction of early passes means that these two now exist at the same
time, so they need to share ownership of the map. Even better, the use
of `Crate` in a rustc thread pool means that it needs to be Sync, so it
can't use `Lrc<Lock>` to manually activate thread-safety.
`parking_lot` is reused from elsewhere in the tree to allow use of its
`ReentrantMutex`, as the relevant parts of rustdoc are still
single-threaded and this allows for easier use in that context.
Jack O'Connor [Thu, 23 Aug 2018 22:22:53 +0000 (18:22 -0400)]
define copy_within on slices
This is a safe wrapper around ptr::copy, for regions within a single
slice. Previously, safe in-place copying was only available as a side
effect of Vec::drain.
Auto merge of #54301 - alexcrichton:update-curl, r=Mark-Simulacrum
Update some `*-sys` dependencies of Cargo/RLS
This is intended to help solve #54206 on nightly where the RLS on MinGW is
having build issues with accidentally building a `curl` library which links to
pthread symbols on Windows (where it should use native mutex locking instead).
The build system for these `*-sys` crates have all been rewritten to be based on
`cc` to bypass native build systems and platform detection to make sure we
configure them correctly.
Alex Crichton [Mon, 17 Sep 2018 16:52:02 +0000 (09:52 -0700)]
Update some `*-sys` dependencies of Cargo/RLS
This is intended to help solve #54206 on nightly where the RLS on MinGW is
having build issues with accidentally building a `curl` library which links to
pthread symbols on Windows (where it should use native mutex locking instead).
The build system for these `*-sys` crates have all been rewritten to be based on
`cc` to bypass native build systems and platform detection to make sure we
configure them correctly.
Auto merge of #54211 - nnethercote:keccak-Liveness-memory, r=nikomatsakis
Split `Liveness::users` into three.
This reduces memory usage on some benchmarks because no space is wasted
for padding. For a `check-clean` build of `keccak` it reduces `max-rss`
by 20%.
r? @nikomatsakis, but I want to do a perf run. Locally, I had these results:
- instructions: slight regression
- max-rss: big win on "Clean" builds
- faults: big win on "Clean" and "Nll" builds
- wall-time: small win on "Clean" and "Nll" builds
So I want to see how a different machine compares.
Currently, `BitSet` doesn't actually know its own domain size; it just
knows how many words it contains. To improve things, this commit makes
the following changes.
- It changes `BitSet` and `SparseBitSet` to store their own domain size,
and do more precise bounds and same-size checks with it. It also
changes the signature of `BitSet::to_string()` (and puts it within
`impl ToString`) now that the domain size need not be passed in from
outside.
- It uses `derive(RustcDecodable, RustcEncodable)` for `BitSet`. This
required adding code to handle `PhantomData` in `libserialize`.
- As a result, it removes the domain size from `HybridBitSet`, making a
lot of that code nicer.
- Both set_up_to() and clear_above() were overly general, working with
arbitrary sizes when they are only needed for the domain size. The
commit removes the former, degeneralizes the latter, and removes the
(overly general) tests.
- Changes `GrowableBitSet::grow()` to `ensure()`, fixing a bug where a
(1-based) domain size was confused with a (0-based) element index.
- Changes `BitMatrix` to store its row count, and do more precise bounds
checks with it.
- Changes `ty_params` in `select.rs` from a `BitSet` to a
`GrowableBitSet` because it repeatedly failed the new, more precise
bounds checks. (Changing the type was simpler than computing an
accurate domain size.)
Auto merge of #54174 - parched:park, r=alexcrichton
Fix `thread` `park`/`unpark` synchronization
Previously the code below would not be guaranteed to exit when the
second unpark took the `return, // already unparked` path because there
was no write to synchronize with a read in `park`.
EDIT: doesn't actually require third thread
```
use std::sync::atomic::{AtomicBool, Ordering};
use std::thread::{current, spawn, park};
while !FLAG.load(Ordering::Relaxed) {
park();
}
}
```
I have some other ideas on how to improve the performance of `park` and `unpark` using fences, avoiding any atomic RMW when the state is already `NOTIFIED`, and also how to avoid calling `notify_one` without the mutex locked. But I need to write some micro benchmarks first, so I'll submit those changes at a later date if they prove to be faster.
Fixes https://github.com/rust-lang/rust/issues/53366 I hope.
Tom Tromey [Wed, 19 Sep 2018 15:46:22 +0000 (09:46 -0600)]
Pass --batch to gdb
In one of my travis builds, I was surprised to find that the gdb
pager was in use and caused travis to time out. Adding `--batch`
to the gdb invocation will disable the pager. Note that the
`-ex q` is retained, to make sure gdb exits with status 0, just in
case `set -e` is in effect somehow.
projects / rust.git / log