bors [Mon, 1 Oct 2018 03:24:02 +0000 (03:24 +0000)]
Auto merge of #54675 - alexcrichton:defaultlibs, r=varkor
rust: Add a `-C default-linker-libraries` option
This commit adds a new codegen option for the compiler which disables
rustc's passing of `-nodefaultlibs` by default on relevant platforms.
Sometimes Rust is linked with C code which fails to link with
`-nodefaultlibs` and is unnecessarily onerous to get linking correctly
with `-nodefaultlibs`.
An example of this is that when you compile C code with sanitizers and
then pass `-fsanitize=address` to the linker, it's incompatible with
`-nodefaultlibs` also being passed to the linker.
In these situations it's easiest to turn off Rust's default passing of
`-nodefaultlibs`, which was more ideological to start with than
anything! Preserving the default is somewhat important but having this
be opt-in shouldn't cause any breakage.
bors [Mon, 1 Oct 2018 00:51:19 +0000 (00:51 +0000)]
Auto merge of #54662 - matklad:once-perf, r=alexcrichton
Fix Once perf regression
Because `call_once` is generic, but `is_completed` is not, we need
`#[inline]` annotation to allow LLVM to inline `is_completed` into
`call_once` in downstream crates.
cc https://github.com/rust-lang/rust/pull/53027/files#r221418859
Auto merge of #53255 - orium:fix-bug-overflow-send, r=arielb1
Add a per-tree error cache to the obligation forest
This implements part of what @nikomatsakis mentioned in https://github.com/rust-lang/rust/pull/30533#issuecomment-170705871:
> 1. If you find that a new obligation is a duplicate of one already in the tree, the proper processing is:
> * if that other location is your parent, you should abort with a cycle error (or accept it, if coinductive)
> * if that other location is not an ancestor, you can safely ignore the new obligation
In particular it implements the "if that other location is your parent accept it, if coinductive" part. This fixes #40827.
I have to say that I'm not 100% confident that this is rock solid. This is my first pull request :tada:, and I didn't know anything about the trait resolver before this. In particular I'm not totally sure that comparing predicates is enough (for instance, do we need to compare `param_env` as well?). Also, I'm not sure what @nikomatsakis mentions [here](https://github.com/rust-lang/rust/issues/30977#issue-127091096), but it might be something that affects this PR:
> In particular, I am wary of getting things wrong around inference variables! We can always add things to the set in their current state, and if unifications occur then the obligation is just kind of out-of-date, but I want to be sure we don't accidentally fail to notice that something is our ancestor. I decided this was subtle enough to merit its own PR.
Anyway, go ahead and review :slightly_smiling_face:.
Ref #30977.
# Performance
We are now copying vectors around, so I decided to do some benchmarking. A simple benchmark shows that this does not seem to affect performance in a measurable way:
I ran `cargo clean && cargo build` 20 times on actix-web (84b27db) and these are the results:
```text
rustc master:
Mean Std.Dev. Min Median Max
real 66.637 2.996 57.220 67.714 69.314
user 307.293 14.741 258.093 312.209 320.702
sys 12.524 0.653 10.499 12.726 13.193
rustc fix-bug-overflow-send:
Mean Std.Dev. Min Median Max
real 66.297 4.310 53.532 67.516 70.348
user 306.812 22.371 236.917 314.748 326.229
sys 12.757 0.952 9.671 13.125 13.544
```
I will do a more comprehensive benchmark (compiling rustc stage1) and post the results.
Auto merge of #53816 - zackmdavis:elided_lifetimes_in_paths_field_day, r=nikomatsakis
don't elide lifetimes in paths in librustc/
In light of the "Apply to rustc" checkbox on #44524 and @nikomatsakis's [recent comment about regularly wanting visual indication of elided lifetimes in types](https://github.com/rust-lang/rust/issues/44524#issuecomment-414663773), I was curious to see what it would look like if we turned the `elided_lifetimes_in_path` lint on in at least one crate in the codebase (I chose librustc). Given that I couldn't figure out how to get `cargo fix` work with the build system, this arguably wasn't a very efficient use of my time, but once I started, the conjunction of moral law and the sunk cost fallacy forced me to continue.
This is mostly applying the `<'_>` suggestions issued by the lint, but there were a few places where I named the lifetimes (_e.g._, `<'a, 'gcx, 'tcx>` on `TyCtxt`) in order to match style with surrounding code.
Zack M. Davis [Thu, 30 Aug 2018 05:02:42 +0000 (22:02 -0700)]
don't elide lifetimes in paths in librustc/
This seemed like a good way to kick the tires on the
elided-lifetimes-in-paths lint (#52069)—seems to work! This was also
pretty tedious—it sure would be nice if `cargo fix` worked on this
codebase (#53896)!
Auto merge of #54591 - ljedrz:cleanup_typeck_rest, r=zackmdavis
A few cleanups and minor improvements to typeck
This PR complements https://github.com/rust-lang/rust/pull/54533, which was limited to `check`.
- change a few `push` loops to `extend`s
- prefer `to_owned` to `to_string` for string literals
- prefer `if let` to `match` where only one branch matters
- a few other minor improvements
- whitespace fixes
Alex Crichton [Sat, 29 Sep 2018 18:03:59 +0000 (11:03 -0700)]
rust: Add a `-C default-linker-libraries` option
This commit adds a new codegen option for the compiler which disables
rustc's passing of `-nodefaultlibs` by default on relevant platforms.
Sometimes Rust is linked with C code which fails to link with
`-nodefaultlibs` and is unnecessarily onerous to get linking correctly
with `-nodefaultlibs`.
An example of this is that when you compile C code with sanitizers and
then pass `-fsanitize=address` to the linker, it's incompatible with
`-nodefaultlibs` also being passed to the linker.
In these situations it's easiest to turn off Rust's default passing of
`-nodefaultlibs`, which was more ideological to start with than
anything! Preserving the default is somewhat important but having this
be opt-in shouldn't cause any breakage.
Auto merge of #54599 - nikomatsakis:issue-54593-impl-Trait, r=eddyb
use closure def-id in returns, but base def-id in locals
The refactorings to handle `let x: impl Trait` wound up breaking `impl Trait` in closure return types. I think there are some deeper problems with the code in question, but this a least should make @eddyb's example work.
- #54564 (Add 1.29.1 release notes)
- #54567 (Include path in stamp hash for debuginfo tests)
- #54577 (rustdoc: give proc-macros their own pages)
- #54590 (std: Don't let `rust_panic` get inlined)
- #54598 (Remove useless lifetimes from `Pin` `impl`s.)
- #54604 (Added help message for `self_in_typedefs` feature gate)
- #54635 (Improve docs for std::io::Seek)
- #54645 (Compute Android gdb version in compiletest)
Auto merge of #54554 - RalfJung:maybe-uninit, r=nagisa
Revert most of MaybeUninit, except for the new API itself
This reverts most of https://github.com/rust-lang/rust/pull/53508/ for perf reasons (first commit reverts that entire PR), except for the new API itself (added back in 2nd commit).
Rollup merge of #54577 - QuietMisdreavus:docs-for-procs, r=GuillaumeGomez
rustdoc: give proc-macros their own pages
related to https://github.com/rust-lang/rust/issues/49553 but i don't think it'll fix it
Currently, rustdoc doesn't expose proc-macros all that well. In the source crate, only their definition function is exposed, but when re-exported, they're treated as a macro! This is an awkward situation in all accounts. This PR checks functions to see whether they have any of `#[proc_macro]`, `#[proc_macro_attribute]`, or `#[proc_macro_derive]`, and exposes them as macros instead. In addition, attributes and derives are exposed differently than other macros, getting their own item-type, CSS class, and module heading.
Function-like proc-macros are lumped in with `macro_rules!` macros, but they get a different declaration block (i'm open to tweaking this, it's just what i thought of given how function-proc-macros operate):
There's one wrinkle which this PR doesn't address, which is why i didn't mark this as fixing the linked issue. Currently, proc-macros don't expose their attributes or source span across crates, so while rustdoc knows they exist, that's about all the information it gets. This leads to an "inlined" macro that has absolutely no docs on it, and no `[src]` link to show you where it was declared.
The way i got around it was to keep proc-macro re-export disabled, since we do get enough information across crates to properly link to the source page:
Rollup merge of #54645 - tromey:android-gdb-version, r=alexcrichton
Compute Android gdb version in compiletest
compiletest has special code for running gdb for Android targets. In
particular it computes a different path to gdb. However, this gdb is
not used for the version test, which results in some tests being run
when they should not be. You can see this in #54004.
This patch moves the special case to analyze_gdb and a new helper
function to decide whether the case applies. This causes the version
check to work properly.
Note that the bulk of the runtest.rs change is just reindentation
caused by moving from a "match" to an "if" -- but there is a (small)
change buried in there.
Because `call_once` is generic, but `is_completed` is not, we need
`#[inline]` annotation to allow LLVM to inline `is_completed` into
`call_once` in downstream crates.
Rollup merge of #54567 - tromey:paths-in-stamp-hashes, r=Mark-Simulacrum
Include path in stamp hash for debuginfo tests
The debuginfo tests are exposed to the environment in a couple of
ways: the path to the gdb executable matters, as does the Python path
used when loading lldb.
This patch incorporates these paths into the hash that is written to
the stamp file, so that changing the path will cause the tests to be
re-run.
Auto merge of #54278 - eddyb:spanned-generic-predicates, r=nikomatsakis
rustc: keep a Span for each predicate in ty::GenericPredicates.
This should allow finer-grained diagnostics, including migration suggestions for #54090.
(Note that I haven't changed most of the users of `predicates_of` to use the new spans)
Auto merge of #53013 - zackmdavis:infer_outlints, r=nikomatsakis
in which inferable outlives-requirements are linted
RFC 2093 (tracking issue #44493) lets us leave off these
commonsensically inferable `T: 'a` outlives requirements. (A separate
feature-gate was split off for the case of 'static lifetimes, for
which questions still remain.) Detecting these was requested as an
idioms-2018 lint.
Resolves #52042, an item under the fabulous metaïssue #52047.
It's plausible that this shouldn't land until after `infer_outlives_requirements` has been stabilized ([final comment period started](https://github.com/rust-lang/rust/issues/44493#issuecomment-408852946) 4 days ago), but I think there's also a strong case to not-wait in order to maximize the time that [Edition Preview 2](https://internals.rust-lang.org/t/rust-2018-release-schedule-and-extended-beta/8076) users have to kick at it. (It's allow by default, so there's no impact unless you explicitly turn it or the rust-2018-idioms group up to `warn` or higher.)
Questions—
* Is `explicit-outlives-requirements` a good name? (I chose it as an [RFC 344](https://github.com/rust-lang/rfcs/blob/master/text/0344-conventions-galore.md#lints)-compliant "inversion" of the feature-gate name, `infer_outlives_requirements`, but I could imagine someone arguing that the word `struct` should be part of the name somewhere, for specificity.)
* Are there any false-positives or false-negatives? @nikomatsakis [said that](https://github.com/rust-lang/rust/issues/52042#issuecomment-406409795) getting this right would be "fairly hard", which makes me nervous that I'm missing something. The UI test in the initial submission of this pull request just exercises the examples [given in the Edition Guide](https://rust-lang-nursery.github.io/edition-guide/2018/transitioning/ownership-and-lifetimes/struct-inference.html).
Fixes toolstate regression caused by https://github.com/rust-lang/rust/pull/54356.
The `save::process_crate` now needs to be passed an additional `&Input`, this change contains the RLS equivalent of [this](https://github.com/rust-lang/rust/pull/54356/files#diff-707a0eda6b2f1a0537abc3d23133748cR983).
Tom Tromey [Fri, 28 Sep 2018 16:32:59 +0000 (10:32 -0600)]
Compute Android gdb version in compiletest
compiletest has special code for running gdb for Android targets. In
particular it computes a different path to gdb. However, this gdb is
not used for the version test, which results in some tests being run
when they should not be. You can see this in #54004.
This patch moves the special case to analyze_gdb and a new helper
function to decide whether the case applies. This causes the version
check to work properly.
Note that the bulk of the runtest.rs change is just reindentation
caused by moving from a "match" to an "if" -- but there is a (small)
change buried in there.
Auto merge of #54356 - Xanewok:save-analysis-invocation, r=nrc
Emit used rustc invocation in the save-analysis file
Blocked on https://github.com/nrc/rls-data/pull/19. (I'm guessing it won't pass CI due to an out-of-tree git dependency)
This should allow RLS to recreate a Rust compilation build plan from the save-analysis files alone, which should be useful when fetching those from external build systems, most notably Buck now.
Also this includes some more potentially useful compilation-specific options (e.g. sysroot or the actual path to extern crates) but that's not required for the build plan bits.
Auto merge of #54568 - levex:issue-54130, r=nagisa
codegen_llvm: check inline assembly constraints with LLVM
---%<---
Hey all,
As issue #54130 highlights, constraints are not checked and passing bad constraints to LLVM can crash it since a `Verify()` call is placed inside an assertion (see: `src/llvm/lib/IR/InlineAsm.cpp:39`).
As this is my first PR to the Rust compiler (woot! :tada:), there might be better ways of achieving this result. In particular, I am not too happy about generating an error in codegen; it would be much nicer if we did it earlier. However, @rkruppe [noted on IRC](https://botbot.me/mozilla/rustc/2018-09-25/?msg=104791581&page=1) that this should be fine for an unstable feature and a much better solution than the _status quo_, which is an ICE.
Thanks!
--->%---
LLVM provides a way of checking whether the constraints and the actual
inline assembly make sense. This commit introduces a check before
emitting code for the inline assembly. If LLVM rejects the inline
assembly (or its constraints), then the compiler emits an error E0668
("malformed inline assembly").
Auto merge of #54338 - orium:fix-macro-inc-comp, r=nrc
Use full name to identify a macro in a `FileName`.
Before this two macros with same name would be indistinguishable inside a `FileName`. This caused a bug in incremental compilation (see #53097) since two different macros would map out to the same `StableFilemapId`.
Zack M. Davis [Sun, 26 Aug 2018 19:22:04 +0000 (12:22 -0700)]
in which inferable outlives-requirements are linted
RFC 2093 (tracking issue #44493) lets us leave off
commonsensically inferable `T: 'a` outlives requirements. (A separate
feature-gate was split off for the case of 'static lifetimes, for
which questions still remain.) Detecting these was requested as an
idioms-2018 lint.
It turns out that issuing a correct, autofixable suggestion here is
somewhat subtle in the presence of other bounds and generic
parameters. Basically, we want to handle these three cases:
• One outlives-bound. We want to drop the bound altogether, including
the colon—
MyStruct<'a, T: 'a>
^^^^ help: remove this bound
• An outlives bound first, followed by a trait bound. We want to
delete the outlives bound and the following plus sign (and
hopefully get the whitespace right, too)—
MyStruct<'a, T: 'a + MyTrait>
^^^^^ help: remove this bound
• An outlives bound after a trait bound. We want to delete the
outlives lifetime and the preceding plus sign—
MyStruct<'a, T: MyTrait + 'a>
^^^^^ help: remove this bound
This gets (slightly) even more complicated in the case of where
clauses, where we want to drop the where clause altogether if there's
just the one bound. Hopefully the comments are enough to explain
what's going on!
A script (in Python, sorry) was used to generate the
hopefully-sufficiently-exhaustive UI test input. Some of these are
split off into a different file because rust-lang-nursery/rustfix#141
(and, causally upstream of that, #53934) prevents them from being
`run-rustfix`-tested.
We also make sure to include a UI test of a case (copied from RFC
2093) where the outlives-bound can't be inferred. Special thanks to
Niko Matsakis for pointing out the `inferred_outlives_of` query,
rather than blindly stripping outlives requirements as if we weren't a
production compiler and didn't care.
Igor Matuszewski [Sat, 22 Sep 2018 20:24:32 +0000 (22:24 +0200)]
Move `filename_for_metadata` to codegen_utils
This function isn't strictly tied to LLVM (it's more of a utility) and
it's now near an analogous, almost identical `filename_for_input` (for
rlibs and so forth).
Also this means not depending on the backend when one wants to know the
accurate .rmeta output filename.
Auto merge of #54468 - matthewjasper:fix-polonius, r=nikomatsakis
[NLL] Get Polonius borrow check to work in simple cases
* Restores the generation of outlives facts from subtyping.
* Restore liveness facts.
* Generate invalidates facts at the start point of each location,
where we check for errors.
* Add a small test for simple cases (previously these cases have compiled, and more recently ICEd).
Closes #54212
cc #53142 (will need test)
### Known limitations
* Two phase borrows aren't implemented for Polonius yet
* Invalidation facts haven't been updated for some of the recent changes to make `Drop` terminators access fewer things.
* Fact generation is not as optimized as it could be.
* Around 30 tests fail in compare mode, often tests that are ignored in nll compare mode
Auto merge of #54533 - ljedrz:cleanup_librustc_typeck_check, r=davidtwco
A few cleanups and minor improvements to typeck/check
- turn a `loop` into a `while let`
- turn a `push_back` loop into an `extend`
- turn a few `push` loops into collected iterators
- prefer `vec![x; n]` to `(0..n).map(|_| x).collect()`
- combine two loops doing the same thing on 2 data sets using `chain`
- use `unwrap_or` where applicable and readable
- add a `potentially_plural_count` helper function to simplify several `format!()` calls
- prefer `to_owned` to `to_string` for string literals
- change `match` to `if let` where only one branch matters
- a few other minor improvements
- whitespace fixes
Auto merge of #52319 - tinco:issue_12590, r=pnkfelix
Track whether module declarations are inline (fixes #12590)
To track whether module declarations are inline I added a field `inline: bool` to `ast::Mod`. The main use case is for pretty to know whether it should render the items associated with the module, but perhaps there are use cases for this information to not be forgotten in the AST.
Niko Matsakis [Wed, 26 Sep 2018 20:37:31 +0000 (16:37 -0400)]
use the closure def-id in returns, but closure-base def-id in locals
Using the `closure_base_def_id` indiscriminantely, as we were doing
before, winds up "going wrong" if the closure type includes the `impl
Trait` from the parent. The problem arises because the return value
for closures is inferred and meant to treat the return
type *opaquely*, so we don't want to be "desugaring" it into the
underlying type.
Matthew Jasper [Wed, 26 Sep 2018 19:41:14 +0000 (20:41 +0100)]
Get Polonius borrow check to work in simple cases
* Restores the generation of outlives facts from subtyping.
* Restore liveness facts.
* Generate invalidates facts at the start point of each location,
where we check for errors.
* Add a small test for simple cases.
Auto merge of #54453 - nikomatsakis:nll-issue-53121-shred-outlives, r=pnkfelix
rework how we handle outlives relationships
When we encounter an outlives relationship involving a projection, we use to over-constrain in some cases with region constraints. We also used to evaluate whether the where-clauses in the environment might apply **before** running inference.
We now avoid doing both of those things:
- If there are where-clauses in the environment that might be useful, we add no constraints.
- After inference is done, we check if we wound up inferring values compatible with the where-clause, and make use of them if so.
I realize now that this PR includes some meandering commits and refactorings from when I expected to go in a different direction. If desired, I could try to remove some of those.
Auto merge of #51946 - japaric:emit-stack-sizes, r=nikomatsakis
[eRFC] add -Z emit-stack-sizes
# What
This PR exposes LLVM's ability to report the stack usage of each function through the unstable /
experimental `-Z emit-stack-sizes` flag.
# Motivation
The end goal is to enable whole program analysis of stack usage to prove absence of stack overflows
at compile time. Such property is important in systems that lack a MMU / MPU and where stack
overflows can corrupt memory. And in systems that have protection against stack overflows such proof
can be used to opt out of runtime checks (e.g. stack probes or the MPU).
Such analysis requires the call graph of the program, which can be obtained from MIR, and the stack
usage of each function in the program. Precise information about the later later can only be
obtained from LLVM as it depends on the optimization level and optimization options like LTO.
This PR does **not** attempt to add the ability to perform such whole program analysis to rustc;
it simply does the minimal amount of work to enable such analysis to be implemented out of tree.
# Implementation
This PR exposes a way to set LLVM's `EmitStackSizeSection` option from the command line. The option
is documented [here]; the documentation is copied below for convenience and posteriority:
> A section containing metadata on function stack sizes will be emitted when
> TargetLoweringObjectFile::StackSizesSection is not null, and TargetOptions::EmitStackSizeSection
> is set (-stack-size-section). The section will contain an array of pairs of function symbol values
> (pointer size) and stack sizes (unsigned LEB128). The stack size values only include the space
> allocated in the function prologue. Functions with dynamic stack allocations are not included.
Where the LLVM feature is not available (e.g. LLVM version < 6.0) or can't be applied (e.g. the
output format doesn't support sections e.g. .wasm files) the flag does nothing -- i.e. no error or
warning is emitted.
#[inline(never)]
fn stack() {
unsafe {
// array allocated on the stack
let array: [i32; 4] = mem::uninitialized();
for elem in &array {
ptr::read_volatile(&elem);
}
}
}
EOF
$ # we need a custom linking step to preserve the .stack_sizes section
$ # (see unresolved questions for a solution that doesn't require custom linking)
$ cat > keep-stack-sizes.x <<'EOF'
SECTIONS
{
.stack_sizes :
{
KEEP(*(.stack_sizes));
}
}
EOF
Like `-Z sanitize` this is a re-export of an LLVM feature. To me knowledge, we don't have a policy
about stabilization of such features as they are incompatible with, or demand extra implementation
effort from, alternative backends (e.g. cranelift). As such this feature will remain experimental /
unstable for the foreseeable future.
# Unresolved questions
## Section name
Should we rename the `.stack_sizes` section to `.debug_stacksizes`?
With the former name linkers will strip the section unless told otherwise using a linker script,
which means getting this information requires both knowledge about linker scripts and a custom
linker invocation (see example above).
If we use the `.debug_stacksizes` name (I believe) linkers will always keep the section, which means
`-Z emit-stack-sizes` is the only thing required to get the stack usage information.
# ~TODOs~
~Investigate why this doesn't work with the `thumb` targets. I get the LLVM error shown below:~
``` console
$ cargo new --lib foo && cd $_
$ echo '#![no_std] pub fn foo() {}' > src/lib.rs
$ cargo rustc --target thumbv7m-none-eabi -- -Z emit-stack-sizes
LLVM ERROR: unsupported relocation on symbol
```
~which sounds like it might be related to the `relocation-model` option. Maybe `relocation-model =
static` is not supported for some reason?~
This fixed itself after the LLVM upgrade.
---
r? @nikomatsakis
cc @rust-lang/compiler @perlindgren @whitequark