Before, there was more or less duplicated suggestions to add type hints.
Fix by clearing more generic suggestions when a more specific suggestion
is possible.
This fixes#93506 .
Track implicit `Sized` obligations in type params
When we evaluate `ty::GenericPredicates` we introduce the implicit
`Sized` predicate of type params, but we do so with only the `Predicate`
its `Span` as context, we don't have an `Obligation` or
`ObligationCauseCode` we could influence. To try and carry this
information through, we add a new field to `ty::GenericPredicates` that
tracks both which predicates come from a type param and whether that
param has any bounds already (to use in suggestions).
We also suggest adding a `?Sized` bound if appropriate on E0599.
Address part of #98539.
`cs_fold` has four distinct cases, covered by three different function
arguments:
- first field
- combine current field with previous results
- no fields
- non-matching enum variants
This commit clarifies things by replacing the three function arguments
with one that takes a new `CsFold` type with four slightly different)
cases
- single field
- combine result for current field with results for previous fields
- no fields
- non-matching enum variants
This makes the code shorter and clearer.
When deriving functions for zero-variant enums, we just generated a
function body that calls `std::instrincs::unreachable`. There is a large
comment with some not-very-useful historical discussion about
alternatives, including some discussion of feature-gating zero-variant
enums, which is clearly irrelevant today.
This commit cuts the comment down greatly.
The deriving code has some complex parts involving iterations over
selflike args and also fields within structs and enum variants.
The return types for a few functions demonstrate this:
- `TraitDef::create_{struct_pattern,enum_variant_pattern}` returns a
`(P<ast::Pat>, Vec<(Span, Option<Ident>, P<Expr>)>)`
- `TraitDef::create_struct_field_accesses` returns a `Vec<(Span,
Option<Ident>, P<Expr>)>`.
This results in per-field data stored within per-selflike-arg data, with
lots of repetition within the per-field data elements. This then has to
be "transposed" in two places (`expand_struct_method_body` and
`expand_enum_method_body`) into per-self-like-arg data stored within
per-field data. It's all quite clumsy and confusing.
This commit rearranges things greatly. Data is obtained in the needed
form up-front, avoiding the need for transposition. Also, various
functions are split, removed, and added, to make things clearer and
avoid tuple return values.
The diff is hard to read, which reflects the messiness of the original
code -- there wasn't an easy way to break these changes into small
pieces. (Sorry!) It's a net reduction of 35 lines and a readability
improvement. The generated code is unchanged.
The deriving code has inconsistent terminology to describe args.
In some places it distinguishes between:
- the `&self` arg (if present), versus
- all other args.
In other places it distinguishes between:
- the `&self` arg (if present) and any other arguments with the same
type (in practice there is at most one, e.g. in `PartialEq::eq`),
versus
- all other args.
The terms "self_args" and "nonself_args" are sometimes used for the
former distinction, and sometimes for the latter. "args" is also
sometimes used for "all other args".
This commit makes the code consistently uses "self_args"/"nonself_args"
for the former and "selflike_args"/"nonselflike_args" for the latter.
This change makes the code easier to read.
The commit also adds a panic on an impossible path (the `Self_` case) in
`extract_arg_details`.
DWARF version 5 brings a number of improvements over version 4. Quoting from
the announcement [1]:
> Version 5 incorporates improvements in many areas: better data compression,
> separation of debugging data from executable files, improved description of
> macros and source files, faster searching for symbols, improved debugging
> optimized code, as well as numerous improvements in functionality and
> performance.
On platforms where DWARF version 5 is supported (Linux, primarily), this commit
adds support for it behind a new `-Z dwarf-version=5` flag.
[1]: https://dwarfstd.org/Public_Review.php
don't succeed `evaluate_obligation` query if new opaque types were registered
fixes#98608fixes#98604
The root cause of all this is that in type flag computation we entirely ignore nongeneric things like struct fields and the signature of function items. So if a flag had to be set for a struct if it is set for a field, that will only happen if the field is generic, as only the generic parameters are checked.
I now believe we cannot use type flags to handle opaque types. They seem like the wrong tool for this.
Instead, this PR replaces the previous logic by adding a new variant of `EvaluatedToOk`: `EvaluatedToOkModuloOpaqueTypes`, which says that there were some opaque types that got hidden types bound, but that binding may not have been legal (because we don't know if the opaque type was in its defining scope or not).
Highlight conflicting param-env candidates
This could probably be further improved by noting _why_ equivalent param-env candidates (modulo regions) leads to ambiguity.
Fixes#98786
Migrate some diagnostics from `rustc_const_eval` to `SessionDiagnostic`
I'm still trying to get the hang of this, so it doesn't migrate _all_ of `rustc_const_eval`. Working on that later.
r? `@davidtwco`
Add a `-Zdump-drop-tracking-cfg` debugging flag
This is useful for debugging drop-tracking; previously, you had to recompile
rustc from source and manually add a call to `write_graph_to_file`. This
makes the option more discoverable and configurable at runtime.
I also took the liberty of making the labels for the CFG nodes much easier to read:
previously, they looked like `id(2), local_id: 48`, now they look like
```
expr from_config (hir_id=HirId { owner: DefId(0:10 ~ default_struct_update[79f9]::foo), local_id: 2})
```
r? ``@eholk``
More derive output improvements
This PR includes:
- Some test improvements.
- Some cosmetic changes to derive output that make the code look more like what a human would write.
- Some more fundamental improvements to `cmp` and `partial_cmp` generation.
r? `@Mark-Simulacrum`
Use less string interning
This removes string interning in a couple of places where doing so won't result in perf improvements. I also switched one place to use pre-interned symbols.
sess: stabilize `--terminal-width` as `--diagnostic-width`
Formerly `-Zterminal-width`, `--terminal-width` allows the user or build
tool to inform rustc of the width of the terminal so that diagnostics
can be truncated.
Pending agreement to stabilize, see tracking issue at #84673.
r? ```@oli-obk```
On partial uninit error point at where we need init
When a binding is declared without a value, borrowck verifies that all
codepaths have *one* assignment to them to initialize them fully. If
there are any cases where a condition can be met that leaves the binding
uninitialized or we attempt to initialize a field of an uninitialized
binding, we emit E0381.
We now look at all the statements that initialize the binding, and use
them to explore branching code paths that *don't* and point at them. If
we find *no* potential places where an assignment to the binding might
be missing, we display the spans of all the existing initializers to
provide some context.
Fix https://github.com/rust-lang/rust/issues/97956.
Rollup of 9 pull requests
Successful merges:
- #97917 (Implement ExitCodeExt for Windows)
- #98844 (Reword comments and rename HIR visiting methods.)
- #98979 (interpret: use AllocRange in UninitByteAccess)
- #98986 (Fix missing word in comment)
- #98994 (replace process exit with more detailed exit in src/bootstrap/*.rs)
- #98995 (Add a test for #80471)
- #99002 (suggest adding a derive for #[default] applied to variants)
- #99004 (Add a test for #70408)
- #99017 (Replace boolean argument for print_where_clause with an enum to make code more clear)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
When a binding is declared without a value, borrowck verifies that all
codepaths have *one* assignment to them to initialize them fully. If
there are any cases where a condition can be met that leaves the binding
uninitialized or we attempt to initialize a field of an unitialized
binding, we emit E0381.
We now look at all the statements that initialize the binding, and use
them to explore branching code paths that *don't* and point at them. If
we find *no* potential places where an assignment to the binding might
be missing, we display the spans of all the existing initializers to
provide some context.
interpret: use AllocRange in UninitByteAccess
also use nice new format string syntax in `interpret/error.rs`, and use the `#` flag to add `0x` prefixes where applicable.
r? ``@oli-obk``
Make lowering a query
Split from https://github.com/rust-lang/rust/pull/88186.
This PR refactors the relationship between lowering and the resolver outputs in order to make lowering itself a query.
In a first part, lowering is changed to avoid modifying resolver outputs, by maintaining its own data structures for creating new `NodeId`s and so.
Then, the `TyCtxt` is modified to allow creating new `LocalDefId`s from inside it. This is done by:
- enclosing `Definitions` in a lock, so as to allow modification;
- creating a query `register_def` whose purpose is to declare a `LocalDefId` to the query system.
See `TyCtxt::create_def` and `TyCtxt::iter_local_def_id` for more detailed explanations of the design.
Make MIR basic blocks field public
This makes it possible to mutably borrow different fields of the MIR
body without resorting to methods like `basic_blocks_local_decls_mut_and_var_debug_info`.
To preserve validity of control flow graph caches in the presence of
modifications, a new struct `BasicBlocks` wraps together basic blocks
and control flow graph caches.
The `BasicBlocks` dereferences to `IndexVec<BasicBlock, BasicBlockData>`.
On the other hand a mutable access requires explicit `as_mut()` call.
incr: cache dwarf objects in work products
Cache DWARF objects alongside object files in work products when those exist so that DWARF object files are available for thorin in packed mode in incremental scenarios.
r? `@michaelwoerister`
Finishing touches for `#[expect]` (RFC 2383)
This PR adds documentation and some functionality to rustc's lint passes, to manually fulfill expectations. This is needed for some lints in Clippy. Hopefully, it should be one of the last things before we can move forward with stabilizing this feature.
As part of this PR, I've also updated `clippy::duplicate_mod` to showcase how this new functionality can be used and to ensure that it works correctly.
---
changelog: [`duplicate_mod`]: Fixed lint attribute interaction
r? `@wesleywiser`
cc: https://github.com/rust-lang/rust/issues/97660, https://github.com/rust-lang/rust/issues/85549
And I guess that's it. Here have a magical unicorn 🦄
Improve soundness of rustc_arena
Make it runnable in miri by changing the loop iteration count for some tests in miri. Also fix a stacked borrows issue with box.
rustc_codegen_ssa: use `project_index`, not `project_field`, for array literals.
See https://github.com/rust-lang/rust/pull/98615#issuecomment-1170082774 for some context.
In short, we were using `project_field` even for array `mir::Rvalue::Aggregate`s, which results in benchmarks like `deep-vector.rs` (and presumably also some real-world usecases?) being impacted by how we handle non-array aggregate fields.
(This is a separate PR so that we can measure the perf effects in isolation)
r? `@nikic`
Use a bitset instead of a hash map in HIR ID validator
The hashset insertion was slightly hot in incr patched runs, but it seems unnecessary to use a hashset here, as it just checks that a dense set of N integers was seen.
I'm not sure if it's possible to know the amount of items beforehand to preallocate the bitset? I suppose not.
This makes it possible to mutably borrow different fields of the MIR
body without resorting to methods like `basic_blocks_local_decls_mut_and_var_debug_info`.
To preserve validity of control flow graph caches in the presence of
modifications, a new struct `BasicBlocks` wraps together basic blocks
and control flow graph caches.
The `BasicBlocks` dereferences to `IndexVec<BasicBlock, BasicBlockData>`.
On the other hand a mutable access requires explicit `as_mut()` call.
Do not fetch HIR to compute variances.
Everything can be done using higher-level queries. This simplifies the code, and should allow better incremental caching.
interpret: remove support for unsized_locals
I added support for unsized_locals in https://github.com/rust-lang/rust/pull/59780 but the current implementation is a crude hack and IMO definitely not the right way to have unsized locals in MIR. It also [causes problems](https://rust-lang.zulipchat.com/#narrow/stream/146212-t-compiler.2Fconst-eval/topic/Missing.20Layout.20Check.20in.20.60interpret.2Foperand.2Ers.60.3F). and what codegen does is unsound and has been for years since clearly nobody cares (so I hope nobody actually relies on that implementation and I'll be happy if Miri ensures they do not). I think if we want to have unsized locals in Miri/MIR we should add them properly, either by having a `StorageLive` that takes metadata or by having an `alloca` that returns a pointer (making the ptr indirection explicit) or something like that.
So, this PR removes the `LocalValue::Unallocated` hack. It adds `Immediate::Uninit`, for several reasons:
- This lets us still do fairly little work in `push_stack_frame`, in particular we do not actually have to create any allocations.
- If/when I remove `ScalarMaybeUninit`, we will need something like this to have an "optimized" representation of uninitialized locals. Without this we'd have to put uninitialized integers into the heap!
- const-prop needs some way to indicate "I don't know the value of this local'; it used to use `LocalValue::Unallocated` for that, now it can use `Immediate::Uninit`.
There is still a fundamental difference between `LocalValue::Unallocated` and `Immediate::Uninit`: the latter is considered a regular local that you can read from and write to, it just has a more optimized representation when compared with an actual `Allocation` that is fully uninit. In contrast, `LocalValue::Unallocated` had this really odd behavior where you would write to it but not read from it. (This is in fact what caused the problems mentioned above.)
While at it I also did two drive-by cleanups/improvements:
- In `pop_stack_frame`, do the return value copying and local deallocation while the frame is still on the stack. This leads to better error locations being reported. The old errors were [sometimes rather confusing](https://rust-lang.zulipchat.com/#narrow/stream/269128-miri/topic/Cron.20Job.20Failure.202022-06-24/near/287445522).
- Deduplicate `copy_op` and `copy_op_transmute`.
r? `@oli-obk`
Allow arithmetic and certain bitwise ops on AtomicPtr
This is mainly to support migrating from `AtomicUsize`, for the strict provenance experiment.
This is a pretty dubious set of APIs, but it should be sufficient to allow code that's using `AtomicUsize` to manipulate a tagged pointer atomically. It's under a new feature gate, `#![feature(strict_provenance_atomic_ptr)]`, but I'm not sure if it needs its own tracking issue. I'm happy to make one, but it's not clear that it's needed.
I'm unsure if it needs changes in the various non-LLVM backends. Because we just cast things to integers anyway (and were already doing so), I doubt it.
API change proposal: https://github.com/rust-lang/libs-team/issues/60Fixes#95492
Operand::Uninit is an *allocated* operand that is fully uninitialized.
This lets us lazily allocate the actual backing store of *all* locals (no matter their ABI).
I also reordered things in pop_stack_frame at the same time.
I should probably have made that a separate commit...
There was a problem with storing a `Box<T>` in a struct, where
the current rules would invalidate the value. this makes it store
a raw pointer instead, circumventing the aliasing problems.
Rename the `--output-width` flag to `--diagnostic-width` as this appears
to be the preferred name within the compiler team.
Signed-off-by: David Wood <david.wood@huawei.com>
Rename the `--terminal-width` flag to `--output-width` as the behaviour
doesn't just apply to terminals (and so is slightly less accurate).
Signed-off-by: David Wood <david.wood@huawei.com>
Formerly `-Zterminal-width`, `--terminal-width` allows the user or build
tool to inform rustc of the width of the terminal so that diagnostics
can be truncated.
Signed-off-by: David Wood <david.wood@huawei.com>
Compiling with `-Csplit-debuginfo=packed` was leaving behind `.dwo`
files because either the metadata or allocator module contained a DWARF
object which was not being removed by the
`maybe_remove_temps_from_module` closure.
Cache DWARF objects alongside object files in work products when those
exist so that DWARF object files are available for thorin in packed mode
in incremental scenarios.
Signed-off-by: David Wood <david.wood@huawei.com>
fix typo in note about multiple inaccessible type aliases
Mainly intended as a small typo fix ("aliass" -> "aliases") for the case where a type cannot be found in scope but there are multiple inaccessible type aliases that match the missing type.
In general this change would use the correct plural form in this scenario for base words that end with 's'.
macros: `LintDiagnostic` derive
- Move `LintDiagnosticBuilder` into `rustc_errors` so that a diagnostic derive can refer to it.
- Introduce a `DecorateLint` trait, which is equivalent to `SessionDiagnostic` or `AddToDiagnostic` but for lints. Necessary without making more changes to the lint infrastructure as `DecorateLint` takes a `LintDiagnosticBuilder` and re-uses all of the existing logic for determining what type of diagnostic a lint should be emitted as (e.g. error/warning).
- Various refactorings of the diagnostic derive machinery (extracting `build_field_mapping` helper and moving `sess` field out of the `DiagnosticDeriveBuilder`).
- Introduce a `LintDiagnostic` derive macro that works almost exactly like the `SessionDiagnostic` derive macro except that it derives a `DecorateLint` implementation instead. A new derive is necessary for this because `SessionDiagnostic` is intended for when the generated code creates the diagnostic. `AddToDiagnostic` could have been used but it would have required more changes to the lint machinery.
~~At time of opening this pull request, ignore all of the commits from #98624, it's just the last few commits that are new.~~
r? `@oli-obk`
Mainly intended as a small typo fix ("aliass" -> "aliases") for
the case where a type cannot be found in scope, and there are
multiple inaccessible type aliases that match the missing type.
In general this change would use the correct plural form in
this scenario for words that end with 's'.
Edit `rustc_mir_dataflow::framework::lattice::FlatSet` docs
Cosmetic improvements. Adds a paragraph break, and
ellipses to signify arbitrary size of a flat set.
Fix repr(align) enum handling
`enum`, for better or worse, supports `repr(align)`. That has already caused a bug in https://github.com/rust-lang/rust/issues/92464, which was "fixed" in https://github.com/rust-lang/rust/pull/92932, but it turns out that that fix is wrong and caused https://github.com/rust-lang/rust/issues/96185.
So this reverts #92932 (which fixes#96185), and attempts another strategy for fixing #92464: special-case enums when doing a cast, re-using the code to load the discriminant rather than assuming that the enum has scalar layout. This works fine for the interpreter.
However, #92464 contained another testcase that was previously not in the test suite -- and after adding it, it ICEs again. This is not surprising; codegen needs the same patch that I did in the interpreter. Probably this has to happen [around here](d32ce37a17/compiler/rustc_codegen_ssa/src/mir/rvalue.rs (L276)). Unfortunately I don't know how to do that -- the interpreter can load a discriminant from an operand, but codegen can only do that from a place. `@oli-obk` `@eddyb` `@bjorn3` any idea?
`SessionDiagnostic` isn't suitable for use on lints as whether or not it
creates an error or a warning is decided at compile-time by the macro,
whereas lints decide this at runtime based on the location of the lint
being reported (as it will depend on the user's `allow`/`deny`
attributes, etc). Re-using most of the machinery for
`SessionDiagnostic`, this macro introduces a `LintDiagnostic` derive
which implements a `DecorateLint` trait, taking a
`LintDiagnosticBuilder` and adding to the lint according to the
diagnostic struct.
`sess` field of `SessionDiagnosticDeriveBuilder` is never actually used
in the builder's member functions, so it doesn't need to be a field.
Signed-off-by: David Wood <david.wood@huawei.com>
Move the logic for building a field mapping (which is used by the
building of format strings in `suggestion` annotations) into a helper
function.
Signed-off-by: David Wood <david.wood@huawei.com>
lints: mostly translatable diagnostics
As lints are created slightly differently than other diagnostics, intended to try make them translatable first and then look into the applicability of diagnostic structs but ended up just making most of the diagnostics in the crate translatable (which will still be useful if I do make a lot of them structs later anyway).
r? ``@compiler-errors``
Change enum->int casts to not go through MIR casts.
follow-up to https://github.com/rust-lang/rust/pull/96814
this simplifies all backends and even gives LLVM more information about the return value of `Rvalue::Discriminant`, enabling optimizations in more cases.
clean up the borrowing in rustc_hir_pretty
A whole lot of the `&`s and `ref`s were redundant. I hope doing this in one big commit is fine, because all of the changes are pretty self-contained.
`@rustbot` label: +C-cleanup
fix interpreter validity check on Box
Follow-up to https://github.com/rust-lang/rust/pull/98554: avoid walking over parts of the value twice.
And then move all that logic into the general visitor so not each visitor implementation has to deal with it...
Interpret: AllocRange Debug impl, and use it more consistently
The two commits are pretty independent but it did not seem worth having two PRs for them.
r? ``@oli-obk``
more `need_type_info` improvements
this now deals with macros in suggestions and the source cost computation does what I want for `channel` 🎉
r? ``@estebank``
Add method to mutate MIR body without invalidating CFG caches.
In addition to adding this method, a handful of passes are updated to use it. There's still quite a few passes that could in principle make use of this as well, but do not at the moment because they use `VisitorMut` or `MirPatch`, which needs additional support for this.
The method name is slightly unwieldy, but I don't expect anyone to be writing it a lot, and at least it says what it does. If anyone has a suggestion for a better name though, would be happy to rename.
r? rust-lang/mir-opt
We currently do a match on the comparison of every field in a struct or
enum variant. But the last field has a degenerate match like this:
```
match ::core::cmp::Ord::cmp(&self.y, &other.y) {
::core::cmp::Ordering::Equal =>
::core::cmp::Ordering::Equal,
cmp => cmp,
},
```
This commit changes it to this:
```
::core::cmp::Ord::cmp(&self.y, &other.y),
```
This is fairly straightforward thanks to the existing `cs_fold1`
function.
The commit also removes the `cs_fold` function which is no longer used.
(Note: there is some repetition now in `cs_cmp` and `cs_partial_cmp`. I
will remove that in a follow-up PR.)
incr.comp.: Make split-dwarf commandline options [TRACKED].
This commandline options have an influence on the contents of object files (and .dwo files), so they need to be `[TRACKED]`.
r? `@davidtwco`
interpret: don't rely on ScalarPair for overflowed arithmetic
This is for https://github.com/rust-lang/rust/pull/97861.
Cc `@eddyb`
I would like to avoid making this depend on `dest.layout.abi` to avoid a branch that we are not usually covering both sides of. Though OTOH this seems like fairly straight-forward code. But let's benchmark this option first to see how bad that extra `force_allocation` really is.
It's common to see repeated assertions like this in derived `clone` and
`eq` methods:
```
let _: ::core::clone::AssertParamIsClone<u32>;
let _: ::core::clone::AssertParamIsClone<u32>;
```
This commit avoids them.
All derive ops currently use match-destructuring to access fields. This
is reasonable for enums, but sub-optimal for structs. E.g.:
```
fn eq(&self, other: &Point) -> bool {
match *other {
Self { x: ref __self_1_0, y: ref __self_1_1 } =>
match *self {
Self { x: ref __self_0_0, y: ref __self_0_1 } =>
(*__self_0_0) == (*__self_1_0) &&
(*__self_0_1) == (*__self_1_1),
},
}
}
```
This commit changes derive ops on structs to use field access instead, e.g.:
```
fn eq(&self, other: &Point) -> bool {
self.x == other.x && self.y == other.y
}
```
This is faster to compile, results in smaller binaries, and is simpler to
generate. Unfortunately, we have to keep the old pattern generating code around
for `repr(packed)` structs because something like `&self.x` (which doesn't show
up in `PartialEq` ops, but does show up in `Debug` and `Hash` ops) isn't
allowed. But this commit at least changes those cases to use let-destructuring
instead of match-destructuring, e.g.:
```
fn hash<__H: ::core:#️⃣:Hasher>(&self, state: &mut __H) -> () {
{
let Self(ref __self_0_0) = *self;
{ ::core:#️⃣:Hash::hash(&(*__self_0_0), state) }
}
}
```
There are some unnecessary blocks remaining in the generated code, but I
will fix them in a follow-up PR.
Fix bug in `rustdoc -Whelp`
Previously, this printed the debugging options, not the lint options,
and only handled `-Whelp`, not `-A/-D/-F`.
This also fixes a few other misc issues:
- Fix `// check-stdout` for UI tests; previously it only worked for run-fail and compile-fail tests
- Add lint headers for tool lints, not just builtin lints
https://github.com/rust-lang/rust/pull/98533#issuecomment-1172004197
r? ```@GuillaumeGomez```
get rid of `tcx` in deadlock handler when parallel compilation
This is a very obscure and hard-to-trace problem that affects thread scheduling. If we copy `tcx` to the deadlock handler thread, it will perform unpredictable behavior and cause very weird problems when executing `try_collect_active_jobs`(For example, the deadlock handler thread suddenly preempts the content of the blocked worker thread and executes the unknown judgment branch, like #94654).
Fortunately we can avoid this behavior by precomputing `query_map`. This change fixes the following ui tests failure on my environment when set `parallel-compiler = true`:
```
[ui] src/test\ui\async-await\no-const-async.rs
[ui] src/test\ui\infinite\infinite-struct.rs
[ui] src/test\ui\infinite\infinite-tag-type-recursion.rs
[ui] src/test\ui\issues\issue-3008-1.rs
[ui] src/test\ui\issues\issue-3008-2.rs
[ui] src/test\ui\issues\issue-32326.rs
[ui] src/test\ui\issues\issue-57271.rs
[ui] src/test\ui\issues\issue-72554.rs
[ui] src/test\ui\parser\fn-header-semantic-fail.rs
[ui] src/test\ui\union\union-nonrepresentable.rs
```
Updates #75760Fixes#94654
CTFE interning: don't walk allocations that don't need it
The interning of const allocations visits the mplace looking for references to intern. Walking big aggregates like big static arrays can be costly, so we only do it if the allocation we're interning contains references or interior mutability.
Walking ZSTs was avoided before, and this optimization is now applied to cases where there are no references/relocations either.
---
While initially looking at this in the context of #93215, I've been testing with smaller allocations than the 16GB one in that issue, and with different init/uninit patterns (esp. via padding).
In that example, by default, `eval_to_allocation_raw` is the heaviest query followed by `incr_comp_serialize_result_cache`. So I'll show numbers when incremental compilation is disabled, to focus on the const allocations themselves at 95% of the compilation time, at bigger array sizes on these minimal examples like `static ARRAY: [u64; LEN] = [0; LEN];`.
That is a close construction to parts of the `ctfe-stress-test-5` benchmark, which has const allocations in the megabytes, while most crates usually have way smaller ones. This PR will have the most impact in these situations, as the walk during the interning starts to dominate the runtime.
Unicode crates (some of which are present in our benchmarks) like `ucd`, `encoding_rs`, etc come to mind as having bigger than usual allocations as well, because of big tables of code points (in the hundreds of KB, so still an order of magnitude or 2 less than the stress test).
In a check build, for a single static array shown above, from 100 to 10^9 u64s (for lengths in powers of ten), the constant factors are lowered:
(log scales for easier comparisons)
(linear scale for absolute diff at higher Ns)
For one of the alternatives of that issue
```rust
const ROWS: usize = 100_000;
const COLS: usize = 10_000;
static TWODARRAY: [[u128; COLS]; ROWS] = [[0; COLS]; ROWS];
```
we can see a similar reduction of around 3x (from 38s to 12s or so).
For the same size, the slowest case IIRC is when there are uninitialized bytes e.g. via padding
```rust
const ROWS: usize = 100_000;
const COLS: usize = 10_000;
static TWODARRAY: [[(u64, u8); COLS]; ROWS] = [[(0, 0); COLS]; ROWS];
```
then interning/walking does not dominate anymore (but means there is likely still some interesting work left to do here).
Compile times in this case rise up quite a bit, and avoiding interning walks has less impact: around 23%, from 730s on master to 568s with this PR.
Fix FFI-unwind unsoundness with mixed panic mode
UB maybe introduced when an FFI exception happens in a `C-unwind` foreign function and it propagates through a crate compiled with `-C panic=unwind` into a crate compiled with `-C panic=abort` (#96926).
To prevent this unsoundness from happening, we will disallow a crate compiled with `-C panic=unwind` to be linked into `panic-abort` *if* it contains a call to `C-unwind` foreign function or function pointer. If no such call exists, then we continue to allow such mixed panic mode linking because it's sound (and stable). In fact we still need the ability to do mixed panic mode linking for std, because we only compile std once with `-C panic=unwind` and link it regardless panic strategy.
For libraries that wish to remain compile-once-and-linkable-to-both-panic-runtimes, a `ffi_unwind_calls` lint is added (gated under `c_unwind` feature gate) to flag any FFI unwind calls that will cause the linkable panic runtime be restricted.
In summary:
```rust
#![warn(ffi_unwind_calls)]
mod foo {
#[no_mangle]
pub extern "C-unwind" fn foo() {}
}
extern "C-unwind" {
fn foo();
}
fn main() {
// Call to Rust function is fine regardless ABI.
foo::foo();
// Call to foreign function, will cause the crate to be unlinkable to panic-abort if compiled with `-Cpanic=unwind`.
unsafe { foo(); }
//~^ WARNING call to foreign function with FFI-unwind ABI
let ptr: extern "C-unwind" fn() = foo::foo;
// Call to function pointer, will cause the crate to be unlinkable to panic-abort if compiled with `-Cpanic=unwind`.
ptr();
//~^ WARNING call to function pointer with FFI-unwind ABI
}
```
Fix#96926
`@rustbot` label: T-compiler F-c_unwind
Enable MIR inlining
Continuation of https://github.com/rust-lang/rust/pull/82280 by `@wesleywiser.`
#82280 has shown nice compile time wins could be obtained by enabling MIR inlining.
Most of the issues in https://github.com/rust-lang/rust/issues/81567 are now fixed,
except the interaction with polymorphization which is worked around specifically.
I believe we can proceed with enabling MIR inlining in the near future
(preferably just after beta branching, in case we discover new issues).
Steps before merging:
- [x] figure out the interaction with polymorphization;
- [x] figure out how miri should deal with extern types;
- [x] silence the extra arithmetic overflow warnings;
- [x] remove the codegen fulfilment ICE;
- [x] remove the type normalization ICEs while compiling nalgebra;
- [ ] tweak the inlining threshold.
Previously, this printed the debugging options, not the lint options,
and only handled `-Whelp`, not `-A/-D/-F`.
This also fixes a few other misc issues:
- Fix `// check-stdout` for UI tests; previously it only worked for run-fail and compile-fail tests
- Add lint headers for tool lints, not just builtin lints
- Remove duplicate run-make test
interpret: make a comment less scary
This slipped past my review: "has no meaning" could be read as "is undefined behavior". That is certainly not what we mean so be more clear.
cleanup mir visitor for `rustc::pass_by_value`
by changing `& $($mutability)?` to `$(& $mutability)?`
I also did some formatting changes because I started doing them for the visit methods I changed and then couldn't get myself to stop xx, I hope that's still fairly easy to review.
Improve some inference diagnostics
- Properly point out point location where "type must be known at this point", or else omit the note if it's not associated with a useful span.
- Fix up some type ambiguity diagnostics, errors shouldn't say "cannot infer type for reference `&'a ()`" when the given type has no inference variables.
Shorten def_span for more items.
The `def_span` query only returns the signature span for functions.
Struct/enum/union definitions can also have a very long body.
This PR shortens the associated span.
