Handle bindings in substructure of patterns with type ascriptions
This attempts to follow the outline described by @nikomatsakis [here](https://github.com/rust-lang/rust/issues/47184#issuecomment-420041056). Its a bit more complicated than expected for two reasons:
1. In general it handles sets of type ascriptions, because such ascriptions can be nested within patterns
2. It has a separate types in the HAIR, `PatternTypeProjections` and `PatternTypeProjection`, which are analogues to the corresponding types in the MIR.
The main reason I added the new HAIR types was because I am worried that the current implementation is inefficent, and asymptotically so: It makes copies of vectors as it descends the patterns, even when those accumulated vectors are never used.
Longer term, I would like to used a linked tree structure for the `PatternTypeProjections` and `PatternTypeProjection`, and save the construction of standalone vectors for the MIR types. I didn't want to block landing this on that hypoethetical revision; but I figured I could at least make the future change easier by differentiating between the two types now.
Oh, one more thing: This doesn't attempt to handle `ref x` (in terms of ensuring that any necessary types are ascribed to `x` in that scenario as well). We should open an issue to investigate supporting that as well. But I didn't want to block this PR on that future work.
Fix#54570
Update the existing NLL `patterns.rs` test accordingly.
includes changes addressing review feedback:
* Added example to docs for `UserTypeProjections` illustrating how we
build up multiple projections when descending into a pattern with
type ascriptions.
* Adapted niko's suggested docs for `UserTypeProjection`.
* Factored out `projection_ty` from more general `projection_ty_core`
(as a drive-by, made its callback an `FnMut`, as I discovered later
that I need that).
* Add note to docs that `PlaceTy.field_ty(..)` does not normalize its result.
* Normalize as we project out `field_ty`.
I did not think I would need this in the MIR, but in general local
decls are going to need to support this. (That, or we need to be able
define a least-upper-bound for a collection of types encountered via
the pattern compilation.)
(That is, it will pull out some component type held or referenced by
the type annotation.)
Note: this still needs to actually do projection itself. That comes in
a later commit
Rollup of 21 pull requests
Successful merges:
- #54816 (Don't try to promote already promoted out temporaries)
- #54824 (Cleanup rustdoc tests with `@!has` and `@!matches`)
- #54921 (Add line numbers option to rustdoc)
- #55167 (Add a "cheap" mode for `compute_missing_ctors`.)
- #55258 (Fix Rustdoc ICE when checking blanket impls)
- #55264 (Compile the libstd we distribute with -Ccodegen-unit=1)
- #55271 (Unimplement ExactSizeIterator for MIR traversing iterators)
- #55292 (Macro diagnostics tweaks)
- #55298 (Point at macro definition when no rules expect token)
- #55301 (List allowed tokens after macro fragments)
- #55302 (Extend the impl_stable_hash_for! macro for miri.)
- #55325 (Fix link to macros chapter)
- #55343 (rustbuild: fix remap-debuginfo when building a release)
- #55346 (Shrink `Statement`.)
- #55358 (Remove redundant clone (2))
- #55370 (Update mailmap for estebank)
- #55375 (Typo fixes in configure_cmake comments)
- #55378 (rustbuild: use configured linker to build boostrap)
- #55379 (validity: assert that unions are non-empty)
- #55383 (Use `SmallVec` for the queue in `coerce_unsized`.)
- #55391 (bootstrap: clean up a few clippy findings)
bootstrap: clean up a few clippy findings
remove useless format!()s
remove redundant field names in a few struct initializations
pass slice instead of a vector to a function
use is_empty() instead of comparisons to .len()
No functional change intended.
Compile the libstd we distribute with -Ccodegen-unit=1
This PR
- adds the `single-codegen-unit-std` option to `config.toml` which allows for setting the CGU count for `libstd` and `libtest` independently of the one for the rest of the compiler, and
- sets the new option to `true` for all dist jobs in CI.
Fixes#54872.
rustbuild: use configured linker to build boostrap
Currently, bootstrap uses it to compile everything, but x.py wasn't using it yet to build bootstrap itself.
This fixes the build for systems where the linker isn't called `cc`.
Shrink `Statement`.
This commit reduces the size of `Statement` from 80 bytes to 56 bytes on
64-bit platforms, by boxing the `AscribeUserType` variant of
`StatementKind`.
This change reduces instruction counts on most benchmarks by 1--3%.
rustbuild: fix remap-debuginfo when building a release
Fallback to the release number as we can't get the git commit sha as we're not in a git repository.
Fixes#55341
Unimplement ExactSizeIterator for MIR traversing iterators
If `root` is not `START_BLOCK`, `basic_blocks().len() - visited` does not represent their exact size.
rustc: Tweak filenames encoded into metadata
This commit is a fix for #54408 where on nightly right now whenever
generics are inlined the path name listed for the inlined function's
debuginfo is a relative path to the cwd, which surely doesn't exist!
Previously on beta/stable the debuginfo mentioned an absolute path which
still didn't exist, but more predictably didn't exist.
The change between stable/nightly is that nightly is now compiled with
`--remap-path-prefix` to give a deterministic prefix to all
rustc-generated paths in debuginfo. By using `--remap-path-prefix` the
previous logic would recognize that the cwd was remapped, causing the
original relative path name of the standard library to get emitted. If
`--remap-path-prefix` *wasn't* passed in then the logic would create an
absolute path name and then create a new source file entry.
The fix in this commit is to apply the "recreate the source file entry
with an absolute path" logic a bit more aggresively. If the source
file's name was remapped then we don't touch it, but otherwise we always
take the working dir (which may have been remapped) and then join it to
the file to ensure that we process all relative file names as well.
The end result is that the standard library should have an absolute path
for all file names in debuginfo (using our `--remap-path-prefix`
argument) as it does on stable after this patch.
Closes#54408
remove useless format!()s
remove redundant field names in a few struct initializations
pass slice instead of a vector to a function
use is_empty() instead of comparisons to .len()
No functional change intended.
This commit avoids an allocation when parsing any float and integer
literals that don't involved underscores.
This reduces the number of allocations done for the `tuple-stress`
benchmark by 10%, reducing its instruction count by just under 1%.
Report const eval error inside the query
Functional changes: We no longer warn about bad constants embedded in unused types. This relied on being able to report just a warning, not a hard error on that case, which we cannot do any more now that error reporting is consistently centralized.
r? @RalfJung
fixes#53561
Fix Rustdoc ICE when checking blanket impls
Fixes#55001, #54744
Previously, SelectionContext would unconditionally cache the selection
result for an obligation. This worked fine for most users of
SelectionContext, but it caused an issue when used by Rustdoc's blanket
impl finder.
The issue occured when SelectionContext chose a ParamCandidate which
contained inference variables. Since inference variables can change
between calls to select(), it's not safe to cache the selection result -
the chosen candidate might not be applicable for future results, leading
to an ICE when we try to run confirmation.
This commit prevents SelectionContext from caching any ParamCandidate
that contains inference variables. This should always be completely
safe, as trait selection should never depend on a particular result
being cached.
I've also added some extra debug!() statements, which I found helpful in
tracking down this bug.
Add a "cheap" mode for `compute_missing_ctors`.
`compute_missing_ctors` produces a vector. It is called a lot, but the
vector is almost always only checked for emptiness.
This commit introduces a specialized variant of `compute_missing_ctors`
(called `is_missing_ctors_empty`) that determines if the resulting set
would be empty, and changes the callsite so that `compute_missing_ctors`
is only called in the rare cases where it is needed. The code
duplication is unfortunate but I can't see a better way to do it.
This change reduces instruction counts for several benchmarks up to 2%.
r? @varkor
