feat: Index workspace symbols at startup rather than on the first symbol search.
This will eliminate potential many-second delays when performing the first search, at the price of making cache priming (“Indexing N/M” in the VS Code status bar) take a little longer in total. Hopefully this additional time is insignificant because a typical session will involve at least one symbol search.
Further improvement would be to do this as a separate parallel task (which will be beneficial if the workspace contains a small number of large crates), but that would require significant additional refactoring of the progress-reporting mechanism to understand multiple tasks per crate. Happy to tackle that in this PR if desired, but I thought I'd propose the minimal change first.
internal: add tracing to project discovery and VFS loading
With `"env RA_PROFILE=vfs_load|parallel_prime_caches|discover_command>500`, this results in the following output:
```
21888ms discover_command
11627ms vfs_load @ total = 701
1503ms vfs_load @ total = 701
30211ms parallel_prime_caches
```
As a followup, I'd like to make hprof emit the information above as JSON.
rustdoc: inherit parent's stability where applicable
It is currently not possible for a re-export to have a different stability (https://github.com/rust-lang/rust/issues/30827). Therefore the standard library uses a hack when moving items like `std::error::Error` or `std::net::IpAddr` into `core` by marking the containing module (`core::error` / `core::net`) as unstable or stable in a later version than the items the module contains.
Previously, rustdoc would always show the *stability as declared* for an item rather than the *stability as publicly reachable* (i.e. the features required to actually access the item), which could be confusing when viewing the docs. This PR changes it so that we show the stability of the first unstable parent or the most recently stabilized parent instead, to hopefully make things less confusing.
fixes https://github.com/rust-lang/rust/issues/130765
screenshots:
 
Separate collection of crate-local inherent impls from error tracking
#119895 changed the return type of the `crate_inherent_impls` query from `CrateInherentImpls` to `Result<CrateInherentImpls, ErrorGuaranteed>` to avoid needing to use the non-parallel-friendly `track_errors()` to track if an error was reporting from within the query... This was mostly fine until #121113, which stopped halting compilation when we hit an `Err(ErrorGuaranteed)` in the `crate_inherent_impls` query.
Thus we proceed onwards to typeck, and since a return type of `Result<CrateInherentImpls, ErrorGuaranteed>` means that the query can *either* return one of "the list inherent impls" or "error has been reported", later on when we want to assemble method or associated item candidates for inherent impls, we were just treating any `Err(ErrorGuaranteed)` return value as if Rust had no inherent impls defined anywhere at all! This leads to basically every inherent method call failing with an error, lol, which was reported in #127798.
This PR changes the `crate_inherent_impls` query to return `(CrateInherentImpls, Result<(), ErrorGuaranteed>)`, i.e. returning the inherent impls collected *and* whether an error was reported in the query itself. It firewalls the latter part of that query into a new `crate_inherent_impls_validity_check` just for the `ensure()` call.
This fixes#127798.
Fix cargo staging for run-make tests
Follow-up to https://github.com/rust-lang/rust/pull/130642#issuecomment-2366891866 to make sure that when
```
$ COMPILETEST_FORCE_STAGE0=1 ./x test run-make --stage 0
```
is used, bootstrap cargo is used in order to avoid building stage 1 rustc. Note that run-make tests are usually not written with `--stage 0` in mind and some tests may rely on stage1 rustc (nightly) behavior, and it is expected that some tests will fail under this invocation.
This PR also fixes `tool::Cargo` staging in compiletest when preparing for `run-make` test mode, by chopping off a stage from the `compiler` passed to `tool::Cargo` such that when the user invokes with stage `N`
```
./x test run-make --stage N
```
the `run-make` test suite will be tested against the cargo built by stage `N` compiler. Let's take `N=1`, i.e. `--stage 1`, without chopping off a stage, previously `./x test run-make --stage 1` will cause stage 1 rustc + std to be built, then stage 2 rustc, and cargo will be produced by the stage 2 rustc, which is clearly not what we want. By chopping off a stage, it means that cargo will be produced by the stage 1 rustc.
cc #119946, #59864.
See discussions regarding the tool staging at https://rust-lang.zulipchat.com/#narrow/stream/326414-t-infra.2Fbootstrap/topic/.E2.9C.94.20stage1.20run-make.20tests.20now.20need.20stage2.20rustc.20built.20for.20c.2E.2E.2E.
Previously if you pass compiler stage 1 to `tool::Cargo`, it will build
stage2 rustc and give you back a cargo built with stage2 rustc, which is
not what we want.
This commit adds a hack that chops off a stage from the compiler passed
to `tool::Cargo`, meaning that we will get a cargo built with stage 1
compiler, avoiding unnecessary and incorrect build of stage2 rustc and
the cargo built by that.
fix: Fix a bug in span map merge, and add explanations of how span maps are stored
Because it took me hours to figure out that contrary to common sense, the offset stored is the *end* of the node, and we search by the *start*. Which is why we need a convoluted `partition_point()` instead of a simple `binary_search()`. And this was not documented at all. Which made me make mistakes with my implementation of `SpanMap::merge()`.
The other bug fixed about span map merging is correctly keeping track of the current offset in presence of multiple sibling macro invocations. Unrelated, but because of the previous issue it took me hours to debug, so I figured out I'll put them together for posterity.
Fixes#18163.
fix: Fix name resolution when an import is resolved to some namespace and then later in the algorithm another namespace is added
The import is flagged as "indeterminate", and previously it was re-resolved, but only at the end of name resolution, when it's already too late for anything that depends on it.
This issue was tried to fix in https://github.com/rust-lang/rust-analyzer/pull/2466, but it was not fixed fully.
That PR is also why IDE features did work: the import at the end was resolved correctly, so IDE features that re-resolved the macro path resolved it correctly.
I was concerned about the performance of this, but this doesn't seem to regress `analysis-stats .`, so I guess it's fine to land this. I have no idea about the incremental perf however and I don't know how to measure that, although when typing in `zbus` (including creating a new function, which should recompute the def map) completion was fast enough.
I didn't check what rustc does, so maybe it does something more performant, like keeping track of only possibly problematic imports.
Fixes#18138.
Probably fixes#17630.
analysis-stats: respect `--disable-proc-macros` flag
I noticed that this flag wasn't being respected by `analysis-stats` when profiling proc macro expansion, so here's a small fix.
internal: Make COMPLETION_MARKER more explicitly r-a
If a user ever sees the completion marker, it's confusing to see text about IntelliJ. Use a string that's more explicitly about completion for rust-analyzer.
Check vtable projections for validity in miri
Currently, miri does not catch when we transmute `dyn Trait<Assoc = A>` to `dyn Trait<Assoc = B>`. This PR implements such a check, and fixes https://github.com/rust-lang/miri/issues/3905.
To do this, we modify `GlobalAlloc::VTable` to contain the *whole* list of `PolyExistentialPredicate`, and then modify `check_vtable_for_type` to validate the `PolyExistentialProjection`s of the vtable, along with the principal trait that was already being validated.
cc ``@RalfJung``
r? ``@lcnr`` or types
I also tweaked the diagnostics a bit.
---
**Open question:** We don't validate the auto traits. You can transmute `dyn Foo` into `dyn Foo + Send`. Should we check that? We currently have a test that *exercises* this as not being UB:
6c6d210089/src/tools/miri/tests/pass/dyn-upcast.rs (L14-L20)
I'm not actually sure if we ever decided that's actually UB or not 🤔
We could perhaps still check that the underlying type of the object (i.e. the concrete type that was unsized) implements the auto traits, to catch UB like:
```rust
fn main() {
let x: &dyn Trait = &std::ptr::null_mut::<()>();
let _: &(dyn Trait + Send) = std::mem::transmute(x);
//~^ this vtable is not allocated for a type that is `Send`!
}
```
If a user ever sees the completion marker, it's confusing to see text
about IntelliJ. Use a string that's more explicitly about completion
for rust-analyzer.
add unqualified_local_imports lint
This lint helps deal with https://github.com/rust-lang/rustfmt/issues/4709 by having the compiler detect imports of local items that are not syntactically distinguishable from imports from other cates. Making them syntactically distinguishable ensures rustfmt can consistently apply the desired import grouping.
bootstrap: Set the dylib path when building books with rustdoc
The library path is needed when the toolchain has been configured with
`[rust] rpath = false`. Otherwise, building the reference book will get
an error when it tries to run rustdoc, like:
rustdoc: error while loading shared libraries: librustc_driver-2ec457c3b8826b72.so
std: implement the `random` feature (alternative version)
Implements the ACP rust-lang/libs-team#393.
This PR is an alternative version of #129120 that replaces `getentropy` with `CCRandomGenerateBytes` (on macOS) and `arc4random_buf` (other BSDs), since that function is not suited for generating large amounts of data and should only be used to seed other CPRNGs. `CCRandomGenerateBytes`/`arc4random_buf` on the other hand is (on modern platforms) just as secure and uses its own, very strong CPRNG (ChaCha20 on the BSDs, AES on macOS) periodically seeded with `getentropy`.
