fix: Some file watching related vfs fixes
Fixes https://github.com/rust-lang/rust-analyzer/issues/15554, additionally it seems that client side file watching was broken on windows this entire time, this PR switches `DidChangeWatchedFilesRegistrationOptions` to use relative glob patterns which do work on windows in VSCode.
Have Derive Attribute share a token tree with it's proc macros.
The goal of this PR is to stop creating a token tree for each derive proc macro.
This is done by giving the derive proc macros an id to its parent derive element.
From running the analysis stat on the rust analyzer project I did see a small memory decrease.
```
Inference: 42.80s, 362ginstr, 591mb
MIR lowering: 8.67s, 67ginstr, 291mb
Mir failed bodies: 18 (0%)
Data layouts: 85.81ms, 609minstr, 8mb
Failed data layouts: 135 (6%)
Const evaluation: 440.57ms, 5235minstr, 13mb
Failed const evals: 1 (0%)
Total: 64.16s, 552ginstr, 1731mb
```
After Change
```
Inference: 40.32s, 340ginstr, 593mb
MIR lowering: 7.95s, 62ginstr, 292mb
Mir failed bodies: 18 (0%)
Data layouts: 87.97ms, 591minstr, 8mb
Failed data layouts: 135 (6%)
Const evaluation: 433.38ms, 5226minstr, 14mb
Failed const evals: 1 (0%)
Total: 60.49s, 523ginstr, 1680mb
```
Currently this breaks the expansion for the actual derive attribute.
## TODO
- [x] Pick a better name for the function `smart_macro_arg`
fix: Fix projects depending on `rustc_private` hanging
If loading the root fails, we'll hang up in this loop as we never inserted the entry that asserts we already visited a package. This fixes that
Fixes https://github.com/rust-lang/rust-analyzer/issues/16902
internal: Enforce utf8 paths
Cargo already requires this, and I highly doubt r-a works with non-utf8 paths generally either. This just makes dealing with paths a lot easier.
Add fuel to match checking
Exhaustiveness checking is NP-hard hence can take extremely long to check some specific matches. This PR makes ehxaustiveness bail after a set number of steps. I chose a bound that takes ~100ms on my machine, which should be more than enough for normal matches.
I'd like someone with less recent hardware to run the test to see if that limit is low enough for them. Also curious if the r-a team thinks this is a good ballpark or if we should go lower/higher. I don't have much data on how complex real-life matches get, but we can definitely go lower than `500 000` steps.
The second commit is a drive-by soundness fix which doesn't matter much today but will matter once `min_exhaustive_patterns` is stabilized.
Fixes https://github.com/rust-lang/rust-analyzer/issues/9528 cc `@matklad`
fix: Skip problematic cyclic dev-dependencies
Implements a workaround for https://github.com/rust-lang/rust-analyzer/issues/14167, notably it does not implement the ideas surfaced in the issue, but takes a simpler to implement approach (and one that is more consistent).
Effectively, all this does is discard dev-dependency edges that go from a workspace library target to another workspace library target. This means, using a dev-dependency to another workspace member inside unit tests will always fail to resolve for r-a now, (instead of being order dependent and causing problems elsewhere) while things will work out fine in integration tests, benches, examples etc. This effectively acknowledges package cycles to be okay, but crate graph cycles to be invalid:
Quoting https://github.com/rust-lang/rust-analyzer/issues/14167#issuecomment-1864145772
> Though, if you have “package cycle” in integration tests, you’d have “crate cycle” in unit test.
We disallow the latter here, while continuing to support the former
(What's missing is to supress diagnostics for such unit tests, though not doing so might be a good deterrent, making devs avoid the pattern altogether)
feat: Implement ATPIT
Resolves#16584
Note: This implementation only works for ATPIT, not for TAIT.
The main hinderence that blocks the later is the defining sites of TAIT can be inner blocks like in;
```rust
type X = impl Default;
mod foo {
fn bar() -> super::X {
()
}
}
```
So, to figure out we are defining it or not, we should recursively probe for nested modules and bodies.
For ATPIT, we can just look into current body because `error[E0401]: can't use 'Self' from outer item` prevent such nested structures;
```rust
trait Foo {
type Item;
fn foo() -> Self::Item;
}
struct Bar;
impl Foo for Bar {
type Item = impl Default;
fn foo() -> Self::Item {
fn bar() -> Self::Item {
^^^^^^^^^^
|
use of `Self` from outer item
refer to the type directly here instead
5
}
bar()
}
}
```
But this implementation does not checks for unification of same ATPIT between different bodies, monomorphization, nor layout for similar reason. (But these can be done with lazyness if we can utilize something like "mutation of interned value" with `db`. I coundn't find such thing but I would appreciate it if such thing exists and you could let me know 😅)
