extend NLL checker to understand `'empty` combined with universes
This PR extends the NLL region checker to understand `'empty` combined with universes. In particular, it means that the NLL region checker no longer considers `exists<R2> { forall<R1> { R1: R2 } }` to be provable. This is work towards https://github.com/rust-lang/rust/issues/59490, but we're not all the way there. One thing in particular it does not address is error messages.
The modifications to the NLL region inference code turned out to be simpler than expected. The main change is to require that if `R1: R2` then `universe(R1) <= universe(R2)`.
This constraint follows from the region lattice (shown below), because we assume then that `R2` is "at least" `empty(Universe(R2))`, and hence if `R1: R2` (i.e., `R1 >= R2` on the lattice) then `R1` must be in some universe that can name `'empty(Universe(R2))`, which requires that `Universe(R1) <= Universe(R2)`.
```
static ----------+-----...------+ (greatest)
| | |
early-bound and | |
free regions | |
| | |
scope regions | |
| | |
empty(root) placeholder(U1) |
| / |
| / placeholder(Un)
empty(U1) -- /
| /
... /
| /
empty(Un) -------- (smallest)
```
I also made what turned out to be a somewhat unrelated change to add a special region to represent `'empty(U0)`, which we use (somewhat hackily) to indicate well-formedness checks in some parts of the compiler. This fixes#68550.
I did some investigation into fixing the error message situation. That's a bit trickier: the existing "nice region error" code around placeholders relies on having better error tracing than NLL currently provides, so that it knows (e.g.) that the constraint arose from applying a trait impl and things like that. I feel like I was hoping *not* to do such fine-grained tracing in NLL, and it seems like we...largely...got away with that. I'm not sure yet if we'll have to add more tracing information or if there is some sort of alternative.
It's worth pointing out though that I've not kind of shifted my opinion on whose job it should be to enforce lifetimes: I tend to think we ought to be moving back towards *something like* the leak-check (just not the one we *had*). If we took that approach, it would actually resolve this aspect of the error message problem, because we would be resolving 'higher-ranked errors' in the trait solver itself, and hence we wouldn't have to thread as much causal information back to the region checker. I think it would also help us with removing the leak check while not breaking some of the existing crates out there.
Regardless, I think it's worth landing this change, because it was relatively simple and it aligns the set of programs that NLL accepts with those that are accepted by the main region checker, and hence should at least *help* us in migration (though I guess we still also have to resolve the existing crates that rely on leak check for coherence).
r? @matthewjasper
Address the diagnostics part of #70844.
```
error[E0597]: `counter` does not live long enough
--> $DIR/issue-54556-niconii.rs:22:20
|
LL | if let Ok(_) = counter.lock() { }
| ^^^^^^^-------
| |
| borrowed value does not live long enough
| a temporary with access to the borrow is created here ...
...
LL | }
| -
| |
| `counter` dropped here while still borrowed
| ... and the borrow might be used here, when that temporary is dropped and runs the destructor for type `std::result::Result<MutexGuard<'_>, ()>`
|
help: consider adding semicolon after the expression so its temporaries are dropped sooner, before the local variables declared by the block are dropped
|
LL | if let Ok(_) = counter.lock() { };
| ^
```
Use TypeRelating for instantiating query responses
`eq` can add constraints to `RegionConstraintData`, which isn't allowed during borrow checking outside of a `CustomTypeOp`. Use `TypeRelating` instead to always push constraints to the obligations list.
closes#69490
replace the leak check with universes, take 2
This PR is an attempt to revive the "universe-based region check", which is an important step towards lazy normalization. Unlike before, we also modify the definition of `'empty` so that it is indexed by a universe. This sidesteps some of the surprising effects we saw before -- at the core, we no longer think that `exists<'a> { forall<'b> { 'b: 'a } }` is solveable. The new region lattice looks like this:
```
static ----------+-----...------+ (greatest)
| | |
early-bound and | |
free regions | |
| | |
scope regions | |
| | |
empty(root) placeholder(U1) |
| / |
| / placeholder(Un)
empty(U1) -- /
| /
... /
| /
empty(Un) -------- (smallest)
```
This PR has three effects:
* It changes a fair number of error messages, I think for the better.
* It fixes a number of bugs. The old algorithm was too conservative and caused us to reject legal subtypings.
* It also causes two regressions (things that used to compile, but now do not).
* `coherence-subtyping.rs` gets an additional error. This is expected.
* `issue-57639.rs` regresses as before, for the reasons covered in #57639.
Both of the regressions stem from the same underlying property: without the leak check, the instantaneous "subtype" check is not able to tell whether higher-ranked subtyping will succeed or not. In both cases, we might be able to fix the problem by doing a 'leak-check like change' at some later point (e.g., as part of coherence).
This is a draft PR because:
* I didn't finish ripping out the leak-check completely.
* We might want to consider a crater run before landing this.
* We might want some kind of design meeting to cover the overall strategy.
* I just remembered I never finished 100% integrating this into the canonicalization code.
* I should also review what happens in NLL region checking -- it probably still has a notion of bottom (empty set).
r? @matthewjasper
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
The polonius output has one more error which should be displayed
in the regular case, but error reporting in the regular case stopped
at the first error.
Admittedly it would be nice to combine suggestions for the same source
lifetime so that `'a: 'b` and `'a: 'c` are not bothsuggested, but instead
a single `'a: 'b + 'c` is.
In most places, we use a span when emitting `expected...found` errors.
However, there were a couple of places where we didn't use any span,
resulting in hard-to-interpret error messages.
This commit attaches the relevant span to these notes, and additionally
switches over to using `note_expected_found` instead of manually
formatting the message
Add outlives suggestions for some lifetime errors
This PR implements suggestion diagnostics for some lifetime mismatch errors. When the borrow checker finds that some lifetime 'a doesn't outlive some other lifetime 'b that it should outlive, then in addition to the current lifetime error, we also emit a suggestion for how to fix the problem by adding a bound:
- If a and b are normal named regions, suggest to add the bound `'a: 'b`
- If b is static, suggest to replace a with static
- If b also needs to outlive a, they must be the same, so suggest unifying them
We start with a simpler implementation that avoids diagnostic regression or implementation complexity:
- We only makes suggestions for lifetimes the user can already name (eg not closure regions or elided regions)
- For now, we only emit a help note, not an actually suggestion because it is significantly easier.
Finally, there is one hack: it seems that implicit regions in async fn are given the name '_ incorrectly. To avoid suggesting '_: 'x, we simply filter out such lifetimes by name.
For more info, see this internals thread:
https://internals.rust-lang.org/t/mechanical-suggestions-for-some-borrow-checker-errors/9049/3
TL;DR Make suggestions to add a `where 'a: 'b` constraint for some lifetime errors. Details are in the paper linked from the internals thread above.
r? @estebank
TODO
- [x] Clean up code
- [x] Only make idiomatic suggestions
- [x] don't suggest naming `&'a self`
- [x] rather than `'a: 'static`, suggest replacing `'a` with `'static`
- [x] rather than `'a: 'b, 'b: 'a`, suggest replacing `'a` with `'b` or vice versa
- [x] Performance (maybe need a perf run when this is closer to the finish line?)
- perf run was clean...
- EDIT: perf run seems to only check non-error performance... How do we check that error performance didn't regress?
- [x] Needs ui tests
- [x] Integrate the `help` message into the main lifetime `error`
