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! 🎉), 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").
Fixes: #54130
Signed-off-by: Levente Kurusa \<lkurusa@acm.org\>
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`.
Fixes#53097.
r? @nrc
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.
This concerns #52042.
Migrate `src/test/ui/run-pass/*` back to `src/test/run-pass/`.
Moves all the tests from `src/test/ui/run-pass/**` back to `src/test/run-pass/`.
This should have no impact on our overall testing completeness due to PR #54223Fix#54047
[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
r? @nikomatsakis
NLL: regression test for "dropck: track order of destruction for r-value temporaries"
Once this lands, we can remove the E-needstest from #22323.
(We shouldn't close the bug itself, however, because we are leaving the NLL-fixed-by-NLL bugs open until NLL is turned on by default.)
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.
* 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.
Before, if we had a projection like `<T as Foo<'0>>::Bar: 'x` and a
where clause like `<T as Foo<'a>>::Bar: 'a`, we considered those to
have nothing to do with one another. Therefore, we would use the
"overconstrained" path of adding `T: 'x` and `'0: 'x` requirements. We
now do a "fuzzy" match where we erase regions first and hence we see
the env bound `'a`.
We used to apply it repeatedly as we went, relying on the current
value of the `region_bound_pairs_accum` vector. But now we save those
values into a map, so we can just process all the registered region
obligations at the end.
Stabilize pattern_parentheses feature
Addresses #51087 .
Stabilizes the previously unstable feature `pattern_parentheses` which enables the use of `()` in match patterns.
`impl trait` in bindings (feature: impl-trait-existential-types)
This PR enables `impl Trait` syntax (opaque types) to be used in bindings, e.g.
* `let foo: impl Clone = 1;`
* `static foo: impl Clone = 2;`
* `const foo: impl Clone = 3;`
This is part of [RFC 2071](https://github.com/rust-lang/rfcs/blob/master/text/2071-impl-trait-existential-types.md) ([tracking issue](https://github.com/rust-lang/rust/issues/34511)), but exists behind the separate feature gate `impl_trait_in_bindings`.
CC @cramertj @oli-obk @eddyb @Centril @varkor
Make "await" a pseudo-edition keyword
This change makes "await" ident an error in 2018 edition without async_await
feature and adds "await" to the 2018 edition keyword lint group that
suggest migration on the 2015 edition.
cc https://github.com/rust-lang/rust/issues/53834
r? @nikomatsakis
Support an explicit annotation for marker traits
From the tracking issue for rust-lang/rfcs#1268:
> It seems obvious that we should make a `#[marker]` annotation. ~ https://github.com/rust-lang/rust/issues/29864#issuecomment-368959441
This PR allows you to put `#[marker]` on a trait, at which point:
- [x] The trait must not have any items ~~All of the trait's items must have defaults~~
- [x] Any impl of the trait must be empty (not override any items)
- [x] But impls of the trait are allowed to overlap
r? @nikomatsakis
This change makes "await" ident an error in 2018 edition without async_await
feature and adds "await" to the 2018 edition keyword lint group that
suggest migration on the 2015 edition.
[NLL] Rework checking for borrows conflicting with drops
Previously, we would split the drop access into multiple checks for each
field of a struct/tuple/closure and through `Box` dereferences. This
changes this to check if the borrow is accessed by the drop in
`places_conflict`.
We also now handle enums containing `Drop` types.
Closes#53569
r? @nikomatsakis
cc @pnkfelix
Previously, we would split the drop access into multiple checks for each
field of a struct/tuple/closure and through `Box` dereferences. This
changes this to check if the borrow is accessed by the drop in
places_conflict.
This also allows us to handle enums in a simpler way, since we don't
have to construct any new places.
When dropping a self-borrowing struct we shouldn't add a "values in a
scope are dropped in the opposite order they are defined" message,
since there is only one value being dropped.
Error now correctly checks whether the borrow that does not live
long enough is being returned before annotating the error with the
arguments and return type from the signature - as this would not be
relevant if the borrow was not being returned.
Enhances annotation logic to properly consider named lifetimes where
lifetime elision rules that were previously implemented would not apply.
Further, adds new help and note messages to diagnostics and highlights
only lifetime when dealing with named lifetimes.
This error can only occur within a function when a borrow of data owned
within the function is returned; and when there are arguments that could
have been returned instead. Therefore, it is always applicable to add a
specific note that links to the relevant rust documentation about
dangling references.
Report when borrow could cause `&mut` aliasing during Drop
We were already issuing an error for the cases where this cropped up, so this is not fixing any soundness holes. The previous diagnostic just wasn't accurately describing the problem in the user's code.
Fix#52059
[NLL] Record more infomation on free region constraints in typeck
Changes:
* Makes the span of the MIR return place point to the return type
* Don't try to use a path to a type alias as a path to the adt it aliases (fixes an ICE)
* Don't claim that `self` is declared outside of the function. [see this test](f2995d5b1a (diff-0c9e6b1b204f42129b481df9ce459d44))
* Remove boring/interesting distinction and instead add a `ConstraintCategory` to the constraint.
* Add categories for implicit `Sized` and `Copy` requirements, for closure bounds, for user type annotations and `impl Trait`.
* Don't use the span of the first statement for Locations::All bounds (even if it happens to work on the tests we have)
Future work:
* Fine tuning the heuristic used to choose the place the report the error.
* Reporting multiple places (behind a flag)
* Better closure bounds reporting. This probably requires some discussion.
r? @nikomatsakis
NLL: disallow creation of immediately unusable variables
Fix#53695
Original description follows
----
This WIP PR is for discussing the impact of fixing #53695 by injecting a fake read in let patterns.
(Travis will fail, at least the `mir-opt` suite is failing in its current state)
Rollup of 16 pull requests
Successful merges:
- #53652 (define copy_within on slices)
- #54261 (Make `dyn` a keyword in the 2018 edition)
- #54280 (remove (more) CAS API from Atomic* types where not natively supported)
- #54323 (rustbuild: drop color handling)
- #54350 (Support specifying edition in doc test)
- #54370 (Improve handling of type bounds in `bit_set.rs`.)
- #54371 (add -Zui-testing to rustdoc)
- #54374 (Make 'proc_macro::MultiSpan' public.)
- #54402 (Use no_default_libraries for all NetBSD flavors)
- #54409 (Detect `for _ in in bar {}` typo)
- #54412 (add applicability to span_suggestion call)
- #54413 (Add UI test for deref recursion limit printing twice)
- #54415 (parser: Tweak function parameter parsing to avoid rollback on succesfull path)
- #54420 (Compress `Liveness` data some more.)
- #54422 (Simplify slice's first(_mut) and last(_mut) with get)
- #54446 (Unify christianpoveda's emails)
Failed merges:
- #54058 (Introduce the partition_dedup/by/by_key methods for slices)
r? @ghost
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.
yet there's no concrete suggestion emitted.
r? @estebank
`src/test/ui/extern/extern-const.rs` seems to have an inconsistent behaviour across build configurations, possibly non-deterministically. This is tracked in issue 54388.
For this PR, disable running rustfix on it because it failed on CI, until that issue is fixed.
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.
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`.
Fixes#53097.
By introducing a new map that tracks the errors reported and the
`Place`s that spawned those errors against the move out that the error
was referring to, we are able to silence duplicate errors by emitting
only the error which corresponds to the most specific `Place` (that which
other `Place`s which reported errors are prefixes of).
This generally is an improvement, however there is a case -
`liveness-move-in-while` - where the output regresses.
Issue 54246
I added the option of providing a help message for deprecated features, that takes precedence over the default `help: remove this attribute` message, along with messages for the features that mention replacements in the reason for deprecation.
Fixes#54246.
Regression test for rust-lang/rust#53675.
(Includes a couple variations on the theme. I confirmed that the ones
in `in_expression_position` and `what_if_we_use_panic_directly_in_expr`
both failed back on "rustc 1.30.0-nightly (0f063aef6 2018-09-03)".)
Fix#53675
It is worth pointing out that the reason that so few diagnostics are
effected is because of the filter I put in where it only goes down the
new path if the borrowed place is *not* a prefix of the dropped place.
(Without that filter, a *lot* of the tests would need this change, and
it would probably be a net loss for the UX, since you'd see it even in
cases like borrows of generic types where there is no explicit mention
of `Drop`.)
use structured suggestion for "missing mut" label
Fixes#54133 for both NLL and non-NLL.
r? @estebank
I'm not super happy with the existing wording here, since it's now a suggestion. I wonder if the message would work better as something like "help: make binding mutable: `mut foo`"?
Also, are the `HELP` and `SUGGESTION` comments necessary?
(Includes a couple variations on the theme. I confirmed that the ones
in `in_expression_position` and `what_if_we_use_panic_directly_in_expr`
both failed back on "rustc 1.30.0-nightly (0f063aef6 2018-09-03)".)
resolve: Introduce two sub-namespaces in macro namespace
Two sub-namespaces are introduced in the macro namespace - one for bang macros and one for attribute-like macros (attributes, derives).
"Sub-namespace" means this is not a newly introduced full namespace, the single macro namespace is still in place.
I.e. you still can't define/import two macros with the same name in a single module, `use` imports still import only one name in macro namespace (from any sub-namespace) and not possibly two.
However, when we are searching for a name used in a `!` macro call context (`my_macro!()`) we skip attribute names in scope, and when we are searching for a name used in attribute context (`#[my_macro]`/`#[derive(my_macro)]`) we are skipping bang macro names in scope.
In other words, bang macros cannot shadow attribute macros and vice versa.
For a non-macro analogy, we could e.g. skip non-traits when searching for `MyTrait` in `impl MyTrait for Type { ... }`.
However we do not do it in non-macro namespaces because we don't have practical issues with e.g. non-traits shadowing traits with the same name, but with macros we do, especially after macro modularization.
For `#[test]` and `#[bench]` we have a hack in the compiler right now preventing their shadowing by `macro_rules! test` and similar things. This hack was introduced after making `#[test]`/`#[bench]` built-in macros instead of built-in attributes (https://github.com/rust-lang/rust/pull/53410), something that needed to be done from the start since they are "active" attributes transforming their inputs.
Now they are passed through normal name resolution and can be shadowed, but that's a breaking change, so we have a special hack basically applying this PR for `#[test]` and `#[bench]` only.
Soon all potentially built-in attributes will be passed through normal name resolution (https://github.com/rust-lang/rust/pull/53913) and that uncovers even more cases where the strict "macro namespace is a single namespace" rule needs to be broken.
For example, with strict rules, built-in macro `cfg!(...)` would shadow built-in attribute `#[cfg]` (they are different things), standard library macro `thread_local!(...)` would shadow built-in attribute `#[thread_local]` - both of these cases are covered by special hacks in https://github.com/rust-lang/rust/pull/53913 as well.
Crater run uncovered more cases of attributes being shadowed by user-defined macros (`warn`, `doc`, `main`, even `deprecated`), we cannot add exceptions in the compiler for all of them.
Regressions with user-defined attributes like https://github.com/rust-lang/rust/issues/53583 and https://github.com/rust-lang/rust/issues/53898 also appeared after enabling macro modularization.
People are also usually confused (https://github.com/rust-lang/rust/issues/53205#issuecomment-411552763, https://github.com/rust-lang/rust/issues/53583#issuecomment-415447800) when they see conflicts between attributes and non-attribute macros for the first time.
So my proposed solution is to solve this issue by introducing two sub-namespaces and thus skipping resolutions of the wrong kind and preventing more error-causing cases of shadowing.
Fixes https://github.com/rust-lang/rust/issues/53583
rustc_resolve: don't treat uniform_paths canaries as ambiguities unless they resolve to distinct Def's.
In particular, this allows this pattern that @cramertj mentioned in https://github.com/rust-lang/rust/issues/53130#issuecomment-420848814:
```rust
use log::{debug, log};
fn main() {
use log::{debug, log};
debug!(...);
}
```
The canaries for the inner `use log::...;`, *in the macro namespace*, see the `log` macro imported at the module scope, and the (same) `log` macro, imported in the block scope inside `main`.
Previously, these two possible (macro namspace) `log` resolutions would be considered ambiguous (from a forwards-compat standpoint, where we might make imports aware of block scopes).
With this PR, such a case is allowed *if and only if* all the possible resolutions refer to the same definition (more specifically, because the *same* `log` macro is being imported twice).
This condition subsumes previous (weaker) checks like #54005 and the second commit of #54011.
Only the last commit is the main change, the other two are cleanups.
r? @petrochenkov cc @Centril @joshtriplett
[NLL] Suggest let binding
Closes#49821
Also adds an alternative to `explain_why_borrow_contains_point` that allows changing error messages based on the reason that will be given. This will also be useful for #51026, #51169 and maybe further changes to does not live long enough messages.
Suggest valid crate type if invalid crate type is found
This adds a suggestion to the `invalid_crate_types` lint.
The suggestion is based on the Levenshtein distance to existing crate
types. If no suggestion is found it will show the lint without any
suggestions.
Closes#53958
This adds a suggestion to the `invalid_crate_types` lint.
The suggestion is based on the Levenshtein distance to existing crate
types. If no suggestion is found it will show the lint without any
suggestions.
Add rustc SHA to released DWARF debuginfo
This commit updates the debuginfo that is encoded in all of our released
artifacts by default. Currently it has paths like `/checkout/src/...` but these
are a little inconsistent and have changed over time. This commit instead
attempts to actually define the file paths in our debuginfo to be consistent
between releases.
All debuginfo paths are now intended to be `/rustc/$sha` where `$sha` is the git
sha of the released compiler. Sub-paths are all paths into the git repo at that
`$sha`.
