Warn about unreachable code following an expression with an uninhabited type
This pull request fixes#85071. The issue is that liveness analysis currently is "smarter" than reachability analysis when it comes to detecting uninhabited types: Unreachable code is detected during type checking, where full type information is not yet available. Therefore, the check for type inhabitedness is quite crude:
fc81ad22c4/compiler/rustc_typeck/src/check/expr.rs (L202-L205)
i.e. it only checks for `!`, but not other, non-trivially uninhabited types, such as empty enums, structs containing an uninhabited type, etc. By contrast, liveness analysis, which runs after type checking, can benefit from the more sophisticated `tcx.is_ty_uninhabited_from()`:
fc81ad22c4/compiler/rustc_passes/src/liveness.rs (L981)fc81ad22c4/compiler/rustc_passes/src/liveness.rs (L996)
This can lead to confusing warnings when a variable is reported as unused, but the use of the variable is not reported as unreachable. For instance:
```rust
enum Foo {}
fn f() -> Foo {todo!()}
fn main() {
let x = f();
let _ = x;
}
```
currently leads to
```
warning: unused variable: `x`
--> t1.rs:5:9
|
5 | let x = f();
| ^ help: if this is intentional, prefix it with an underscore: `_x`
|
= note: `#[warn(unused_variables)]` on by default
warning: 1 warning emitted
```
which is confusing, because `x` _appears_ to be used in line 6. With my changes, I get:
```
warning: unreachable expression
--> t1.rs:6:13
|
5 | let x = f();
| --- any code following this expression is unreachable
6 | let _ = x;
| ^ unreachable expression
|
= note: `#[warn(unreachable_code)]` on by default
note: this expression has type `Foo`, which is uninhabited
--> t1.rs:5:13
|
5 | let x = f();
| ^^^
warning: unused variable: `x`
--> t1.rs:5:9
|
5 | let x = f();
| ^ help: if this is intentional, prefix it with an underscore: `_x`
|
= note: `#[warn(unused_variables)]` on by default
warning: 2 warnings emitted
```
My implementation is slightly inelegant because unreachable code warnings can now be issued in two different places (during type checking and during liveness analysis), but I think it is the solution with the least amount of unnecessary code duplication, given that the new warning integrates nicely with liveness analysis, where unreachable code is already implicitly detected for the purpose of finding unused variables.
Changes from 81473 extended the dead code lint with an ability to detect
fields that are written to but never read from. The implementation skips
over fields on the left hand side of an assignment, without marking them
as live.
A field access might involve an automatic dereference and de-facto read
the field. Conservatively mark expressions with deref adjustments as
live to avoid generating false positive warnings.
Address inconsistency in using "is" with "declared here"
"is" was generally used for NLL diagnostics, but not other diagnostics. Using "is" makes the diagnostics sound more natural and readable, so it seems sensible to commit to them throughout.
r? @Centril
According to @estebank, def_span scans forward on the line until it finds a {,
and if it can't find one, fallse back to the span for the whole item. This
was apparently written before the identifier span was explicitly tracked on
each node.
This means that if an unused function signature spans multiple lines, the
entire function (potentially hundreds of lines) gets flagged as dead code.
This could, for example, cause IDEs to add error squiggly's to the whole
function.
By using the span from the ident instead, we narrow the scope of this in
most cases. In a wider sense, it's probably safe to use ident.span
instead of def_span in most locations throughout the whole code base,
but since this is my first contribution, I kept it small.
Some interesting points that came up while I was working on this:
- I reorganized the tests a bit to bring some of the dead code ones all
into the same location
- A few tests were for things unrelated to dead code (like the
path-lookahead for parens), so I added #![allow(dead_code)] and
cleaned up the stderr file to reduce noise in the future
- The same fix doesn't apply to const and static declarations. I tried
adding these cases to the match expression, but that created a much
wider change to tests and error messages, so I left it off until I
could get some code review to validate the approach.
This helps organize the tests better. I also renamed several of the tests to remove redundant dead-code in the path, and better match what they're testing
