This commit upgrades the main LLVM submodule to LLVM's current master branch.
The LLD submodule is updated in tandem as well as compiler-builtins.
Along the way support was also added for LLVM 7's new features. This primarily
includes the support for custom section concatenation natively in LLD so we now
add wasm custom sections in LLVM IR rather than having custom support in rustc
itself for doing so.
Some other miscellaneous changes are:
* We now pass `--gc-sections` to `wasm-ld`
* The optimization level is now passed to `wasm-ld`
* A `--stack-first` option is passed to LLD to have stack overflow always cause
a trap instead of corrupting static data
* The wasm target for LLVM switched to `wasm32-unknown-unknown`.
* The syntax for aligned pointers has changed in LLVM IR and tests are updated
to reflect this.
* The `thumbv6m-none-eabi` target is disabled due to an [LLVM bug][llbug]
Nowadays we've been mostly only upgrading whenever there's a major release of
LLVM but enough changes have been happening on the wasm target that there's been
growing motivation for quite some time now to upgrade out version of LLD. To
upgrade LLD, however, we need to upgrade LLVM to avoid needing to build yet
another version of LLVM on the builders.
The revision of LLVM in use here is arbitrarily chosen. We will likely need to
continue to update it over time if and when we discover bugs. Once LLVM 7 is
fully released we can switch to that channel as well.
[llbug]: https://bugs.llvm.org/show_bug.cgi?id=37382
Move self trait predicate to items
This is a "reimagination" of @tmandry's PR #50183. The main effect is described in this comment from one of the commits:
---
Before we had the following results for `predicates_of`:
```rust
trait Foo { // predicates_of: Self: Foo
fn bar(); // predicates_of: Self: Foo (inherited from trait)
}
```
Now we have removed the `Self: Foo` from the trait. However, we still
add it to the trait ITEM. This is because when people do things like
`<T as Foo>::bar()`, they still need to prove that `T: Foo`, and
having it in the `predicates_of` seems to be the cleanest way to
ensure that happens right now (otherwise, we'd need special case code
in various places):
```rust
trait Foo { // predicates_of: []
fn bar(); // predicates_of: Self: Foo
}
```
However, we sometimes want to get the list of *just* the predicates
truly defined on a trait item (e.g., for chalk, but also for a few
other bits of code). For that, we define `predicates_defined_on`,
which does not contain the `Self: Foo` predicate yet, and we plumb
that through metadata and so forth.
---
I'm assigning @eddyb as the main reviewer, but I thought I might delegate to scalexm for this one in any case. I also want to post an alternative that I'll leave in the comments; it occurred to me as I was writing. =)
r? @eddyb
cc @scalexm @tmandry @leodasvacas
This new query returns only the predicates *directly defined* on an
item (in contrast to the more common `predicates_of`, which returns
the predicates that must be proven to reference an item). These two
sets are almost always identical except for traits, where
`predicates_of` includes an artificial `Self: Trait<...>` predicate
(basically saying that you cannot use a trait item without proving
that the trait is implemented for the type parameters).
This new query is only used in chalk lowering, where this artificial
`Self: Trait` predicate is problematic. We encode it in metadata but
only where needed since it is kind of repetitive with existing
information.
Co-authored-by: Tyler Mandry <tmandry@gmail.com>
It was pointed out in review that the glob-exported
underscore-suffixed convention for `Spanned` HIR nodes is no longer
preferred: see February 2016's #31487 for AST's migration away from
this style towards properly namespaced NodeKind enums.
This concerns #51968.
There are at least a couple (and plausibly even three) diagnostics that
could use the spans of visibility modifiers in order to be reliably
correct (rather than hacking and munging surrounding spans to try to
infer where the visibility keyword must have been).
We follow the naming convention established by the other `Spanned` HIR
nodes: the "outer" type alias gets the "prime" node-type name, the
"inner" enum gets the name suffixed with an underscore, and the variant
names are prefixed with the prime name and `pub use` exported from here
(from HIR).
Thanks to veteran reviewer Vadim Petrochenkov for suggesting this
uniform approach. (A previous draft, based on the reasoning that
`Visibility::Inherited` should not have a span, tried to hack in a named
`span` field on `Visibility::Restricted` and a positional field on
`Public` and `Crate`. This was ... not so uniform.)
Namely: labels, type parameters, bindings in patterns, parameter names in functions without body.
All of these do not need hygiene after lowering to HIR, only span locations.
This is gated on edition 2018 & the `async_await` feature gate.
The parser will accept `async fn` and `async unsafe fn` as fn
items. Along the same lines as `const fn`, only `async unsafe fn`
is permitted, not `unsafe async fn`.The parser will not accept
`async` functions as trait methods.
To do a little code clean up, four fields of the function type
struct have been merged into the new `FnHeader` struct: constness,
asyncness, unsafety, and ABI.
Also, a small bug in HIR printing is fixed: it previously printed
`const unsafe fn` as `unsafe const fn`, which is grammatically
incorrect.
Remove attribute_cache from CrateMetadata
This PR will fix#50508 by removing the `attribute_cache` from the `CrateMetadata` struct. Seeing as performance was referenced in the original issue, I also cleaned up a `self.entry(node_id);` call which might have occasionally happened redundantly.
r? @michaelwoerister
