This uses a (per-trait) hash-table to separate impls from different TraitDefs, and makes coherence go so much quicker. I will post performance numbers tomorrow.
This is still WIP, as when there's an overlap error, impls can get printed in the wrong order, which causes a few issues. Should I pick the local impl with the smallest NodeId to print?
Could you take a look at this @nikomatsakis?
This commit brings the `Error` trait in line with the [Error interoperation
RFC](https://github.com/rust-lang/rfcs/pull/201) by adding downcasting,
which has long been intended. This change means that for any `Error`
trait objects that are `'static`, you can downcast to concrete error
types.
To make this work, it is necessary for `Error` to inherit from
`Reflect` (which is currently used to mark concrete types as "permitted
for reflection, aka downcasting"). This is a breaking change: it means
that impls like
```rust
impl<T> Error for MyErrorType<T> { ... }
```
must change to
```rust
impl<T: Reflect> Error for MyErrorType<T> { ... }
```
This commit furthermore marks `Reflect` as stable, since we are already
essentially committed to it via `Any`. Note that in the future, if we
determine that the parametricity aspects of `Reflect` are not needed, we
can deprecate the trait and provide a blanket implementation for it
for *all* types (rather than by using OIBIT), which would allow all
mentions of `Reflect` to be dropped over time. So there is not a strong
commitment here.
[breaking-change]
r? @alexcrichton
This commit brings the `Error` trait in line with the [Error interoperation
RFC](https://github.com/rust-lang/rfcs/pull/201) by adding downcasting,
which has long been intended. This change means that for any `Error`
trait objects that are `'static`, you can downcast to concrete error
types.
To make this work, it is necessary for `Error` to inherit from
`Reflect` (which is currently used to mark concrete types as "permitted
for reflection, aka downcasting"). This is a breaking change: it means
that impls like
```rust
impl<T> Error for MyErrorType<T> { ... }
```
must change to something like
```rust
impl<T: Reflect> Error for MyErrorType<T> { ... }
```
except that `Reflect` is currently unstable (and should remain so for
the time being). For now, code can instead bound by `Any`:
```rust
impl<T: Any> Error for MyErrorType<T> { ... }
```
which *is* stable and has `Reflect` as a super trait. The downside is
that this imposes a `'static` constraint, but that only
constrains *when* `Error` is implemented -- it does not actually
constrain the types that can implement `Error`.
[breaking-change]
Apparently implementations are allowed to return EDEADLK instead of blocking
forever, in which case this can lead to unsafety in the `RwLock` primitive
exposed by the standard library. A debug-build of the standard library would
have caught this error (due to the debug assert), but we don't ship debug
builds right now.
This commit adds explicit checks for the EDEADLK error code and triggers a panic
to ensure the call does not succeed.
Closes#25012
Explanations for E0079, E0080, E0081, E0082, E0083 and E0084 as part of #24407.
All the errors concern the use of `#[repr(X)]` with enum types.
I also updated the short description for E0079 so that it takes sign into account.
I'm interested in helping out with #16676 but more in the grammar than the reference-- here's my first chunk, more to come!! 🎉
I did pull a bit *out* of the reference, though, that was more relevant to the grammar but wasn't moved over as part of #24729.
I'm looking at, e.g. https://github.com/rust-lang/rust/blob/master/src/libsyntax/ast.rs, as the source of truth, please let me know if I should be checking against something else instead/in addition.
r? @steveklabnik
Puts implementations in bins hashed by the fast-reject key, and
only looks up the relevant impls, reducing O(n^2)-ishness
Before: 688.92user 5.08system 8:56.70elapsed 129%CPU (0avgtext+0avgdata 1208164maxresident)k, LLVM 379.142s
After: 637.78user 5.11system 8:17.48elapsed 129%CPU (0avgtext+0avgdata 1201448maxresident)k LLVM 375.552s
Performance increase is +7%-ish
The former stopped making sense when we started interning substs and made
TraitRef a 2-word copy type, and I'm moving the latter into an arena as
they live as long as the type context.
I've been working on improving the diagnostic registration system so that it can:
* Check uniqueness of error codes *across the whole compiler*. The current method using `errorck.py` is prone to failure as it relies on simple text search - I found that it breaks when referencing an error's ident within a string (e.g. `"See also E0303"`).
* Provide JSON output of error metadata, to eventually facilitate HTML output, as well as tracking of which errors need descriptions. The current schema is:
```
<error code>: {
"description": <long description>,
"use_site": {
"filename": <filename where error is used>,
"line": <line in file where error is used>
}
}
```
[Here's][metadata-dump] a pretty-printed sample dump for `librustc`.
One thing to note is that I had to move the diagnostics arrays out of the diagnostics modules. I really wanted to be able to capture error usage information, which only becomes available as a crate is compiled. Hence all invocations of `__build_diagnostics_array!` have been moved to the ends of their respective `lib.rs` files. I tried to avoid moving the array by making a plugin that expands to nothing but couldn't invoke it in item position and gave up on hackily generating a fake item. I also briefly considered using a lint, but it seemed like it would impossible to get access to the data stored in the thread-local storage.
The next step will be to generate a web page that lists each error with its rendered description and use site. Simple mapping and filtering of the metadata files also allows us to work out which error numbers are absent, which errors are unused and which need descriptions.
[metadata-dump]: https://gist.github.com/michaelsproul/3246846ff1bea71bd049
