Rollup of 4 pull requests
Successful merges:
- #118908 (Add all known `target_feature` configs to check-cfg)
- #118933 (Cleanup errors handlers even more)
- #118943 (update `measureme` to 10.1.2 to deduplicate `parking_lot`)
- #118948 (Use the `Waker::noop` API in tests)
r? `@ghost`
`@rustbot` modify labels: rollup
update `measureme` to 10.1.2 to deduplicate `parking_lot`
This PR updates `measureme` to the latest release to remove the last duplicates of `parking_lot` 0.11 we had in our dependency tree.
```console
Updating measureme v10.1.1 -> v10.1.2
Removing parking_lot v0.11.2
Removing parking_lot_core v0.8.6
```
Also removes `instant` from the allowed list of dependencies, as it's no longer used.
r? `@michaelwoerister` (Thanks for the release in the first place 🙏)
Add all known `target_feature` configs to check-cfg
This PR adds all the known `target_feature` from ~~`rustc_codegen_ssa`~~ `rustc_target` to the well known list of check-cfg.
It does so by moving the list from `rustc_codegen_ssa` to `rustc_target` ~~`rustc_session` (I not sure about this, but some of the moved function take a `Session`)~~, then using it the `fill_well_known` function.
This already proved to be useful since portable-simd had a bad cfg.
cc `@nnethercote` (since we discussed it in https://github.com/rust-lang/rust/pull/118494)
Currently, `emit_diagnostic` takes `&mut self`.
This commit changes it so `emit_diagnostic` takes `self` and the new
`emit_diagnostic_without_consuming` function takes `&mut self`.
I find the distinction useful. The former case is much more common, and
avoids a bunch of `mut` and `&mut` occurrences. We can also restrict the
latter with `pub(crate)` which is nice.
Compare `Handler::warn` and `Handler::span_warn`. Conceptually they are
almost identical. But their implementations are weirdly different.
`warn`:
- calls `DiagnosticBuilder::<()>::new(self, Warning(None), msg)`, then `emit()`
- which calls `G::diagnostic_builder_emit_producing_guarantee(self)`
- which calls `handler.emit_diagnostic(&mut db.inner.diagnostic)`
`span_warn`:
- calls `self.emit_diag_at_span(Diagnostic::new(Warning(None), msg), span)`
- which calls `self.emit_diagnostic(diag.set_span(sp))`
I.e. they both end up at `emit_diagnostic`, but take very different
routes to get there.
This commit changes `span_*` and similar ones to not use
`emit_diag_at_span`. Instead they just call `struct_span_*` + `emit`.
Some nice side-effects of this:
- `span_fatal` and `span_fatal_with_code` don't need
`FatalError.raise()`, because `emit` does that.
- `span_err` and `span_err_with_code` doesn't need `unwrap`.
- `struct_span_note`'s `span` arg type is changed from `Span` to
`impl Into<MultiSpan>` like all the other functions.
The `Handler` functions that directly emit diagnostics can be more
easily implemented using `struct_foo(msg).emit()`. This mirrors
`Handler::emit_err` which just does `create_err(err).emit()`.
`Handler::bug` is not converted because of weirdness involving
conflation bugs and fatal errors with `EmissionGuarantee`. I'll fix that
later.
rustc_codegen_ssa: Remove trailing spaces in Display impl for CguReuse
Otherwise errors will look like this:
error: CGU-reuse for `cgu_invalidated_via_import-bar` is `PreLto ` but should be `PostLto `
### Background
I noticed that error messages looked wonky while investigating if
529047cfc3/compiler/rustc_codegen_ssa/src/assert_module_sources.rs (L281-L287)
should not be wrapped by `sess.emit_err(...)`. Right now it looks like the error is accidentally ignored. It looks like 706452eba7 might have accidentally started ignoring it (by removing the `diag.span_err()` call). I am still investigating, but regardless of the outcome we should fix the trailing whitespace.
Unconditionally register alias-relate in projection goal
Follow-up to #118725, which subtly broke closure signature inference on combinators like `Result::map` which I noticed in syn.
Essentially, instead of using `eq` which will eagerly infer `?1 := <?2 as Trait>::Assoc`, we can directly emit an alias-relate goal, which will stay ambiguous for as long as `?2` is ambiguous.
This also more closely models the conceptual framing that projects-to acts like an alias-relate when solving, and like a normalizes-to when in a param env.
r? lcnr
[rustdoc] Use Map instead of Object for source files and search index
It's cleaner and is also easier to manipulate `Map` rather than `Object` types.
r? `@notriddle`
Add -Zunpretty=stable-mir output test
As strongly suggested here https://github.com/rust-lang/rust/pull/118364#issuecomment-1827974148 this adds output test for `-Zunpretty=stable-mir`, added test shows almost all the functionality of the current printer.
r? `@compiler-errors`
fix --dry-run when the change-id warning is printed
previously:
```
Building bootstrap
Compiling bootstrap v0.0.0 (/home/jyn/src/rust2/src/bootstrap)
Finished dev [unoptimized] target(s) in 4.23s
thread 'main' panicked at src/bin/main.rs:147:17:
fs::write(warned_id_path, latest_change_id.to_string()) failed with No such file or directory (os error 2)
```
fix dynamic size/align computation logic for packed types with dyn trait tail
This logic was never updated to support `packed(N)` where `N > 1`, and it turns out to be wrong for that case.
Fixes https://github.com/rust-lang/rust/issues/80925
`@bjorn3` I have not looked at cranelift; I assume it basically copied the size-of-val logic and hence could use much the same patch.
It's unclear why this is used here. All entries in the third column of
`UNICODE_ARRAY` are covered by `ASCII_ARRAY`, so if the lookup fails
it's a genuine compiler bug. It was added way back in #29837, for no
clear reason.
This commit changes it to `span_bug`, which is more typical.
It's necessary for `derive(Diagnostic)`, but is best avoided elsewhere
because there are clearer alternatives.
This required adding `Handler::struct_almost_fatal`.
cache param env canonicalization
Canonicalize ParamEnv only once and store it. Then whenever we try to canonicalize `ParamEnvAnd<'tcx, T>` we only have to canonicalize `T` and then merge the results.
Prelimiary results show ~3-4% savings in diesel and serde benchmarks.
Best to review commits individually. Some commits have a short description.
Initial implementation had a soundness bug (https://github.com/rust-lang/rust/pull/117749#issuecomment-1840453387) due to cache invalidation:
- When canonicalizing `Ty<'?0>` we first try to resolve region variables in the current InferCtxt which may have a constraint `?0 == 'static`. This means that we register `Ty<'?0> => Canonical<Ty<'static>>` in the cache, which is obviously incorrect in another inference context.
- This is fixed by not doing region resolution when canonicalizing the query *input* (vs. response), which is the only place where ParamEnv is used, and then in a later commit we *statically* guard against any form of inference variable resolution of the cached canonical ParamEnv's.
r? `@ghost`
This doesn't change behavior.
It should prevent unintentional resolution of inference variables
during canonicalization, which previously caused a soundness bug.
See PR description for more.
Enable stack probes on aarch64 for LLVM 18
I tested this on `aarch64-unknown-linux-gnu` with LLVM main (~18).
cc #77071, to be closed once we upgrade our LLVM submodule.
Tweak `short_ty_string` to reduce number of files
When shortening types and writing them to disk, make `short_ty_string` capable of reusing the same file, instead of writing a file per shortened type.
rustdoc-search: use set ops for ranking and filtering
This commit adds ranking and quick filtering to type-based search, improving performance and having it order results based on their type signatures.
Preview
-------
Profiler output: https://notriddle.com/rustdoc-html-demo-6/profile-8/index.html
Preview: https://notriddle.com/rustdoc-html-demo-6/ranking-and-filtering-v2/std/index.html
Motivation
----------
If I write a query like `str -> String`, a lot of functions come up. That's to be expected, but `String::from` should come up on top, and it doesn't right now. This is because the sorting algorithm is based on the functions name, and doesn't consider the type signature at all. `slice::join` even comes up above it!
To fix this, the sorting should take into account the function's signature, and the closer match should come up on top.
Guide-level description
-----------------------
When searching by type signature, types with a "closer" match will show up above types that match less precisely.
Reference-level explanation
---------------------------
Functions signature search works in three major phases:
* A compact "fingerprint," based on the [bloom filter] technique, is used to check for matches and to estimate the distance. It sometimes has false positive matches, but it also operates on 128 bit contiguous memory and requires no backtracking, so it performs a lot better than real unification.
The fingerprint represents the set of items in the type signature, but it does not represent nesting, and it ignores when the same item appears more than once.
The result is rejected if any query bits are absent in the function, or if the distance is higher than the current maximum and 200 results have already been found.
* The second step performs unification. This is where nesting and true bag semantics are taken into account, and it has no false positives. It uses a recursive, backtracking algorithm.
The result is rejected if any query elements are absent in the function.
[bloom filter]: https://en.wikipedia.org/wiki/Bloom_filter
Drawbacks
---------
This makes the code bigger.
More than that, this design is a subtle trade-off. It makes the cases I've tested against measurably faster, but it's not clear how well this extends to other crates with potentially more functions and fewer types.
The more complex things get, the more important it is to gather a good set of data to test with (this is arguably more important than the actual benchmarking ifrastructure right now).
Rationale and alternatives
--------------------------
Throwing a bloom filter in front makes it faster.
More than that, it tries to take a tactic where the system can not only check for potential matches, but also gets an accurate distance function without needing to do unification. That way it can skip unification even on items that have the needed elems, as long as they have more items than the currently found maximum.
If I didn't want to be able to cheaply do set operations on the fingerprint, a [cuckoo filter] is supposed to have better performance. But the nice bit-banging set intersection doesn't work AFAIK.
I also looked into [minhashing], but since it's actually an unbiased estimate of the similarity coefficient, I'm not sure how it could be used to skip unification (I wouldn't know if the estimate was too low or too high).
This function actually uses the number of distinct items as its "distance function." This should give the same results that it would have gotten from a Jaccard Distance $1-\frac{|F\cap{}Q|}{|F\cup{}Q|}$, while being cheaper to compute. This is because:
* The function $F$ must be a superset of the query $Q$, so their union is just $F$ and the intersection is $Q$ and it can be reduced to $1-\frac{|Q|}{|F|}.
* There are no magic thresholds. These values are only being used to compare against each other while sorting (and, if 200 results are found, to compare with the maximum match). This means we only care if one value is bigger than the other, not what it's actual value is, and since $Q$ is the same for everything, it can be safely left out, reducing the formula to $1-\frac{1}{|F|} = \frac{|F|}{|F|}-\frac{1}{|F|} = |F|-1$. And, since the values are only being compared with each other, $|F|$ is fine.
Prior art
---------
This is significantly different from how Hoogle does it.
It doesn't account for order, and it has no special account for nesting, though `Box<t>` is still two items, while `t` is only one.
This should give the same results that it would have gotten from a Jaccard Distance $1-\frac{|A\cap{}B|}{|A\cup{}B|}$, while being cheaper to compute.
Unresolved questions
--------------------
`[]` and `()`, the slice/array and tuple/union operators, are ignored while building the signature for the query. This is because they match more than one thing, making them ambiguous. Unfortunately, this also makes them a performance cliff. Is this likely to be a problem?
Right now, the system just stashes the type distance into the same field that levenshtein distance normally goes in. This means exact query matches show up on top (for example, if you have a function like `fn nothing(a: Nothing, b: i32)`, then searching for `nothing` will show it on top even if there's another function with `fn bar(x: Nothing)` that's technically a closer match in type signature.
Future possibilities
--------------------
It should be possible to adopt more sorting criteria to act as a tie breaker, which could be determined during unification.
[cuckoo filter]: https://en.wikipedia.org/wiki/Cuckoo_filter
[minhashing]: https://en.wikipedia.org/wiki/MinHash
