diagnostics: make paths to external items more visible
This PR changes the reported path for an external item so that it is visible from at least one local module (i.e. it does not use any inaccessible external modules) if possible. If the external item's crate was declared with an `extern crate`, the path is guarenteed to use the `extern crate`.
Fixes#23224, fixes#23355, fixes#26635, fixes#27165.
r? @nrc
Restrict constants in patterns
This implements [RFC 1445](https://github.com/rust-lang/rfcs/blob/master/text/1445-restrict-constants-in-patterns.md). The primary change is to limit the types of constants used in patterns to those that *derive* `Eq` (note that implementing `Eq` is not sufficient). This has two main effects:
1. Floating point constants are linted, and will eventually be disallowed. This is because floating point constants do not implement `Eq` but only `PartialEq`. This check replaces the existing special case code that aimed to detect the use of `NaN`.
2. Structs and enums must derive `Eq` to be usable within a match.
This is a [breaking-change]: if you encounter a problem, you are most likely using a constant in an expression where the type of the constant is some struct that does not currently implement
`Eq`. Something like the following:
```rust
struct SomeType { ... }
const SOME_CONST: SomeType = ...;
match foo {
SOME_CONST => ...
}
```
The easiest and most future compatible fix is to annotate the type in question with `#[derive(Eq)]` (note that merely *implementing* `Eq` is not enough, it must be *derived*):
```rust
struct SomeType { ... }
const SOME_CONST: SomeType = ...;
match foo {
SOME_CONST => ...
}
```
Another good option is to rewrite the match arm to use an `if` condition (this is also particularly good for floating point types, which implement `PartialEq` but not `Eq`):
```rust
match foo {
c if c == SOME_CONST => ...
}
```
Finally, a third alternative is to tag the type with `#[structural_match]`; but this is not recommended, as the attribute is never expected to be stabilized. Please see RFC #1445 for more details.
cc https://github.com/rust-lang/rust/issues/31434
r? @pnkfelix
This change has a few parts. We introduce a new `item_path` module for
constructing item paths. The job of this module is basically to make
nice, user-readable paths -- but these paths are not necessarily 100%
unique. They meant to help a *human* find code, but not necessarily a
compute. These paths are used to drive `item_path_str` but also symbol
names.
Because the paths are not unique, we also modify the symbol name hash to
include the full `DefPath`, whereas before it included only those
aspects of the def-path that were not included in the "informative"
symbol name.
Eventually, I'd like to make the item-path infrastructure a bit more
declarative. Right now it's based purely on strings. In particular, for
impls, we should supply the raw types to the `ItemPathBuffer`, so that
symbol names can be encoded using the C++ encoding scheme for better
integration with tooling.
We used to track, for each crate, a path that led to the extern-crate
that imported it. Instead of that, track the def-id of the extern crate,
along with a bit more information, and derive the path on the fly.
We want to prevent compiling something against one version
of a dynamic library and then, at runtime accidentally
using a different version of the dynamic library. With the
old symbol-naming scheme this could not happen because every
symbol had the SVH in it and you'd get an error by the
dynamic linker when using the wrong version of a dylib. With
the new naming scheme this isn't the case any more, so this
patch adds the "link-guard" to prevent this error case.
This is implemented as follows:
- In every crate that we compile, we emit a function called
"__rustc_link_guard_<crate-name>_<crate-svh>"
- The body of this function contains calls to the
"__rustc_link_guard" functions of all dependencies.
- An executable contains a call to it's own
"__rustc_link_guard" function.
As a consequence the "__rustc_link_guard" function call graph
mirrors the crate graph and the dynamic linker will fail if a
wrong dylib is loaded somewhere because its
"__rustc_link_guard" function will contain a different SVH in
its name.
This is a [breaking-change]: according to RFC #1445, constants used as
patterns must be of a type that *derives* `Eq`. If you encounter a
problem, you are most likely using a constant in an expression where the
type of the constant is some struct that does not currently implement
`Eq`. Something like the following:
```rust
struct SomeType { ... }
const SOME_CONST: SomeType = ...;
match foo {
SOME_CONST => ...
}
```
The easiest and most future compatible fix is to annotate the type in
question with `#[derive(Eq)]` (note that merely *implementing* `Eq` is
not enough, it must be *derived*):
```rust
struct SomeType { ... }
const SOME_CONST: SomeType = ...;
match foo {
SOME_CONST => ...
}
```
Another good option is to rewrite the match arm to use an `if`
condition (this is also particularly good for floating point types,
which implement `PartialEq` but not `Eq`):
```rust
match foo {
c if c == SOME_CONST => ...
}
```
Finally, a third alternative is to tag the type with
`#[structural_match]`; but this is not recommended, as the attribute is
never expected to be stabilized. Please see RFC #1445 for more details.
This hack has long since outlived its usefulness; the transition to
trans passing around full substitutions is basically done. Instead of
`ErasedRegions`, just supply substitutions with a suitable number of
`'static` entries, and invoke `erase_regions` when needed (the latter of
which we already do).
Automated conversion using the untry tool [1] and the following command:
```
$ find -name '*.rs' -type f | xargs untry
```
at the root of the Rust repo.
[1]: https://github.com/japaric/untry
Move analysis for MIR borrowck
This PR adds code for doing MIR-based gathering of the moves in a `fn` and the dataflow to determine where uninitialized locations flow to, analogous to how the same thing is done in `borrowck`.
It also adds a couple attributes to print out graphviz visualizations of the analyzed MIR that includes the dataflow analysis results.
cc @nikomatsakis
Improve time complexity of equality relations
This PR adds a `UnificationTable` to the `TypeVariableTable` type which is used store information about variable equality instead of just storing them in a vector for later processing. By using a `UnificationTable` equality relations can be resolved in O(n) (for all realistic values of n) rather than O(n!) which can give massive speedups in certain cases (see combine as an example).
Link to combine: https://github.com/Marwes/combine
This PR adds a `UnificationTable` to the `TypeVariableTable` type which
is used store information about variable equality instead of just
storing them in a vector for later processing. By using a
`UnificationTable` equality relations can be resolved in O(n) (for all
realistic values of n) rather than O(n!) which can give massive
speedups in certain cases (see combine as an example).
Link to combine: https://github.com/Marwes/combine
emit (via debug!) scary message from `fn borrowck_mir` until basic
prototype is in place.
Gather children of move paths and set their kill bits in
dataflow. (Each node has a link to the child that is first among its
siblings.)
Hooked in libgraphviz based rendering, including of borrowck dataflow
state.
doing this well required some refactoring of the code, so I cleaned it
up more generally (adding comments to explain what its trying to do
and how it is doing it).
Update: this newer version addresses most review comments (at least
the ones that were largely mechanical changes), but I left the more
interesting revisions to separate followup commits (in this same PR).
Fix mis-uses of projection mode
A couple of places where we construct a fresh inference context were
incorrectly assuming that we were past coherence checking. This commit
corrects them to use `Topmost` rather than `AnyFinal` as the projection mode.
Fixes#32324
r? @nikomatsakis
A couple of places where we construct a fresh inference context were
incorrectly assuming that we were past coherence checking. This commit
corrects them to use `Topmost` rather than `AnyFinal` as the projection mode.
Fixes#32324
The older code would sometimes swallow errors or fail to produce a
suggestion. The newer code does not. However, just printing everything
would produce a bunch of new and kind of annoying errors, so continue
to swallow `T: 'a` errors so long as there are other things to show.
danger of inference variables floating around without their inference
context.
The main insight here is that, when we are translating substitutions
between two impls, *we already know that the more specific impl holds*,
so we do not need to add its obligations to the parameter
environment. Instead, we can just thread through the inference context
we used to show select the more specific impl in the first place.
projection sensitive to "mode" (most importantly, trans vs middle).
This commit introduces several pieces of iteration infrastructure in the
specialization graph data structure, as well as various helpers for
finding the definition of a given item, given its kind and name.
In addition, associated type projection is now *mode-sensitive*, with
three possible modes:
- **Topmost**. This means that projection is only possible if there is a
non-`default` definition of the associated type directly on the
selected impl. This mode is a bit of a hack: it's used during early
coherence checking before we have built the specialization
graph (and therefore before we can walk up the specialization
parents to find other definitions). Eventually, this should be
replaced with a less "staged" construction of the specialization
graph.
- **AnyFinal**. Projection succeeds for any non-`default` associated
type definition, even if it is defined by a parent impl. Used
throughout typechecking.
- **Any**. Projection always succeeds. Used by trans.
The lasting distinction here is between `AnyFinal` and `Any` -- we wish
to treat `default` associated types opaquely for typechecking purposes.
In addition to the above, the commit includes a few other minor review fixes.
This commit leverages the specialization graph infrastructure to allow
specializing trait implementations to leave off associated types for
which their parents have provided defaults.
It also modifies the type projection code to avoid projecting associated
types unless either (1) all input types are fully known or (2) the
available associated type is "final", i.e. not marked `default`.
This restriction is required for soundness, due to examples like:
```rust
trait Foo {
type Assoc;
}
impl<T> Foo for T {
default type Assoc = ();
}
impl Foo for u8 {
type Assoc = String;
}
fn generic<T>() -> <T as Foo>::Assoc {
() //~ ERROR
}
fn main() {
let s: String = generic::<u8>();
println!("{}", s); // bad news
}
```
This commit leverages the specialization graph infrastructure to allow
specializing trait implementations to leave off methods for which their
parents have provided defaults.
It does not yet check that the `default` keyword is appropriately used
in such cases.
- Rewrites the overlap checker to instead build up a specialization
graph, checking for overlap errors in the process.
- Use the specialization order during impl selection.
This commit does not yet handle associated types correctly, and assumes
that all items are `default` and are overridden.
The module contains a few important components:
- The `specialize` function, which determines whether one impl is a
specialization of another.
- The `SpecializationGraph`, a per-trait graph recording the
specialization tree. The main purpose of the graph is to allow
traversals upwards (to less specialized impls) for discovering
un-overridden defaults, and for ensuring that overridden items are
allowed to be overridden.
typestrong const integers
~~It would be great if someone could run crater on this PR, as this has a high danger of breaking valid code~~ Crater ran. Good to go.
----
So this PR does a few things:
1. ~~const eval array values when const evaluating an array expression~~
2. ~~const eval repeat value when const evaluating a repeat expression~~
3. ~~const eval all struct and tuple fields when evaluating a struct/tuple expression~~
4. remove the `ConstVal::Int` and `ConstVal::Uint` variants and replace them with a single enum (`ConstInt`) which has variants for all integral types
* `usize`/`isize` are also enums with variants for 32 and 64 bit. At creation and various usage steps there are assertions in place checking if the target bitwidth matches with the chosen enum variant
5. enum discriminants (`ty::Disr`) are now `ConstInt`
6. trans has its own `Disr` type now (newtype around `u64`)
This obviously can't be done without breaking changes (the ones that are noticable in stable)
We could probably write lints that find those situations and error on it for a cycle or two. But then again, those situations are rare and really bugs imo anyway:
```rust
let v10 = 10 as i8;
let v4 = 4 as isize;
assert_eq!(v10 << v4 as usize, 160 as i8);
```
stops compiling because 160 is not a valid i8
```rust
struct S<T, S> {
a: T,
b: u8,
c: S
}
let s = S { a: 0xff_ff_ff_ffu32, b: 1, c: 0xaa_aa_aa_aa as i32 };
```
stops compiling because `0xaa_aa_aa_aa` is not a valid i32
----
cc @eddyb @pnkfelix
related: https://github.com/rust-lang/rfcs/issues/1071
Do not report errors from regionck if other errors were already reported
Do not report errors from regionck if other errors were already reported during the lifetime of this inferencer. Fixes#30580.
r? @arielb1
Forbid items with the same name from appearing in overlapping inherent impl blocks
For example, the following is now correctly illegal:
```rust
struct Foo;
impl Foo {
fn id() {}
}
impl Foo {
fn id() {}
}
```
"Overlapping" here is determined the same way it is for traits (and in fact shares the same code path): roughly, there must be some way of substituting any generic types to unify the impls, such that none of the `where` clauses are provably unsatisfiable under such a unification.
Along the way, this PR also introduces an `ImplHeader` abstraction (the first commit) that makes it easier to work with impls abstractly (without caring whether they are trait or inherent impl blocks); see the first commit.
Closes#22889
r? @nikomatsakis
lint: mark associated types as live for the dead_code pass
Associated types of trait impls were being excluded from the live list. So types that only appeared in trait impls were being marked as dead code.
impl blocks.
For example, the following is now correctly illegal:
```rust
struct Foo;
impl Foo {
fn id() {}
}
impl Foo {
fn id() {}
}
```
"Overlapping" here is determined the same way it is for traits (and in
fact shares the same code path): roughly, there must be some way of
substituting any generic types to unify the impls, such that none of the
`where` clauses are provably unsatisfiable under such a unification.
Closes#22889
This commit introduces the idea of an "impl header", which consists of
everything outside the impl body: the Self type, the trait
reference (when applicable), and predicates from `where` clauses. This
type is usable with the type folding machinery, making it possible to
work with impl headers at a higher and more generic level.
Distinguish fn item types to allow reification from nothing to fn pointers.
The first commit is a rebase of #26284, except for files that have moved since.
This is a [breaking-change], due to:
* each FFI function has a distinct type, like all other functions currently do
* all generic parameters on functions are recorded in their item types, e.g.:
`size_of::<u8>` & `size_of::<i8>`'s types differ despite their identical signature.
* function items are zero-sized, which will stop transmutes from working on them
The first two cases are handled in most cases with the new coerce-unify logic,
which will combine incompatible function item types into function pointers,
at the outer-most level of if-else chains, match arms and array literals.
The last case is specially handled during type-checking such that transmutes
from a function item type to a pointer or integer type will continue to work for
another release cycle, but are being linted against. To get rid of warnings and
ensure your code will continue to compile, cast to a pointer before transmuting.
Fix incorrect trait privacy error
This PR fixes#21670 by using the crate metadata instead of `ExternalExports` to determine if an external item is public.
r? @nikomatsakis
There's a lot of stuff wrong with the representation of these types:
TyFnDef doesn't actually uniquely identify a function, TyFnPtr is used to
represent method calls, TyFnDef in the sub-expression of a cast isn't
correctly reified, and probably some other stuff I haven't discovered yet.
Splitting them seems like the right first step, though.
Add error file for E0152
It completes #31818.
However it is not complete yet:
* test will need to be updated
* the file name displayed is a bit too unclear.
I'm not sure yet what's the "correct" file name to display. If anyone has an idea on this, it'd be very appreciated.
r? @brson
implement the `?` operator
The `?` postfix operator is sugar equivalent to the try! macro, but is more amenable to chaining:
`File::open("foo")?.metadata()?.is_dir()`.
`?` is accepted on any *expression* that can return a `Result`, e.g. `x()?`, `y!()?`, `{z}?`,
`(w)?`, etc. And binds more tightly than unary operators, e.g. `!x?` is parsed as `!(x?)`.
cc #31436
---
cc @aturon @eddyb
The `?` postfix operator is sugar equivalent to the try! macro, but is more amenable to chaining:
`File::open("foo")?.metadata()?.is_dir()`.
`?` is accepted on any *expression* that can return a `Result`, e.g. `x()?`, `y!()?`, `{z}?`,
`(w)?`, etc. And binds more tightly than unary operators, e.g. `!x?` is parsed as `!(x?)`.
cc #31436
Do not trigger unused_assignments for overloaded AssignOps
If `v` were a type with some kind of indirection, so that `v += 1` would
have an effect even if `v` were not used anymore, the unused_assignments lint
would mark a false positive.
This exempts overloaded (non-primitive) assign ops from being treated as
assignments (they are method calls).
The previous compile-fail tests that ensure x += 1 can trigger for
primitive types continue to pass. Added a representative test for the
"view" indirection.
Fixes#31895
This PR implements [RFC 1192](https://github.com/rust-lang/rfcs/blob/master/text/1192-inclusive-ranges.md), which is triple-dot syntax for inclusive range expressions. The new stuff is behind two feature gates (one for the syntax and one for the std::ops types). This replaces the deprecated functionality in std::iter. Along the way I simplified the desugaring for all ranges.
This is my first contribution to rust which changes more than one character outside of a test or comment, so please review carefully! Some of the individual commit messages have more of my notes. Also thanks for putting up with my dumb questions in #rust-internals.
- For implementing `std::ops::RangeInclusive`, I took @Stebalien's suggestion from https://github.com/rust-lang/rfcs/pull/1192#issuecomment-137864421. It seemed to me to make the implementation easier and increase type safety. If that stands, the RFC should be amended to avoid confusion.
- I also kind of like @glaebhoerl's [idea](https://github.com/rust-lang/rfcs/pull/1254#issuecomment-147815299), which is unified inclusive/exclusive range syntax something like `x>..=y`. We can experiment with this while everything is behind a feature gate.
- There are a couple of FIXMEs left (see the last commit). I didn't know what to do about `RangeArgument` and I haven't added `Index` impls yet. Those should be discussed/finished before merging.
cc @Gankro since you [complained](https://www.reddit.com/r/rust/comments/3xkfro/what_happened_to_inclusive_ranges/cy5j0yq)
cc #27777#30877rust-lang/rust#1192rust-lang/rfcs#1254
relevant to #28237 (tracking issue)
If `v` were a type with some kind of indirection, so that `v += 1` would
have an effect even if `v` were not used anymore, the unused_assignments lint
would mark a false positive.
This exempts overloaded (non-primitive) assign ops from being treated as
assignments (they are method calls).
The previous compile-fail tests that ensure x += 1 can trigger for
primitive types continue to pass. Added a representative test for the
"view" indirection.
This PR privacy checks paths as they are resolved instead of in `librustc_privacy` (fixes#12334 and fixes#31779). This removes the need for the `LastPrivate` system introduced in PR #9735, the limitations of which cause #31779.
This PR also reports privacy violations in paths to intra- and inter-crate items the same way -- it always reports the first inaccessible segment of the path.
Since it fixes#31779, this is a [breaking-change]. For example, the following code would break:
```rust
mod foo {
pub use foo::bar::S;
mod bar { // `bar` should be private to `foo`
pub struct S;
}
}
impl foo::S {
fn f() {}
}
fn main() {
foo::bar::S::f(); // This is now a privacy error
}
```
r? @alexcrichton
This commit is the result of the FCPs ending for the 1.8 release cycle for both
the libs and the lang suteams. The full list of changes are:
Stabilized
* `braced_empty_structs`
* `augmented_assignments`
* `str::encode_utf16` - renamed from `utf16_units`
* `str::EncodeUtf16` - renamed from `Utf16Units`
* `Ref::map`
* `RefMut::map`
* `ptr::drop_in_place`
* `time::Instant`
* `time::SystemTime`
* `{Instant,SystemTime}::now`
* `{Instant,SystemTime}::duration_since` - renamed from `duration_from_earlier`
* `{Instant,SystemTime}::elapsed`
* Various `Add`/`Sub` impls for `Time` and `SystemTime`
* `SystemTimeError`
* `SystemTimeError::duration`
* Various impls for `SystemTimeError`
* `UNIX_EPOCH`
* `ops::{Add,Sub,Mul,Div,Rem,BitAnd,BitOr,BitXor,Shl,Shr}Assign`
Deprecated
* Scoped TLS (the `scoped_thread_local!` macro)
* `Ref::filter_map`
* `RefMut::filter_map`
* `RwLockReadGuard::map`
* `RwLockWriteGuard::map`
* `Condvar::wait_timeout_with`
Closes#27714Closes#27715Closes#27746Closes#27748Closes#27908Closes#29866
The range desugaring does not use the lang items. Hence I did not add
lang items for inclusive ranges. This cleanup commit removes the old
unused ones as well.
Whether the desugaring _should_ use lang items is another question:
see #30809. But if we decide on a strategy there we can add back these
lang items, and new ones for inclusive ranges.
For stage0 we need to keep the attributes as the lang items still exist
even if they are never used.
This is surprisingly not a breaking change. Unused #[lang] attributes do
not even trigger a lint (see #30881).
A whole bunch of stuff gets folded into struct handling! Plus, removes
an ugly hack from trans and accidentally fixes a bug with constructing
ranges from references (see later commits with tests).
When foldings Substs, we map over VecPerParamSpace instances using
EnumeratedItems which does not provide an accurate size_hint()
in its Iterator implementation. This leads to quite a large number or
reallocations. Providing a suitable size_hint() implementation reduces
the time spent in item-bodies checking quite a bit.
```
crate | before | after | ~change
-------|-------------------------
core | 7.28s | 5.44s | -25%
std | 2.07s | 1.88s | -9.2%
syntax | 8.86s | 8.30s | -6.3%
```
When foldings Substs, we map over VecPerParamSpace instances using
EnumeratedItems which does not provide an accurate size_hint()
in its Iterator implementation. This leads to quite a large number or
reallocations. Providing a suitable size_hint() implementation reduces
the time spent in item-bodies checking quite a bit.
```
crate | before | after | ~change
-------|-------------------------
core | 7.28s | 5.44s | -25%
std | 2.07s | 1.88s | -9.2%
syntax | 8.86s | 8.30s | -6.3%
```
.copy_from_slice() does the same job of .clone_from_slice(), but the
former is explicitly for Copy elements and calls `memcpy` directly, and
thus is it efficient without optimization too.
<sup>**context:** moving back to a layered approach to type checking.</sup>
It looks like they'd not ended up tightly coupled in the time one was owned by the other. Every instance outside of `FnCtxt.inh` was from an `InferCtxt` created and dropped in the same function body.
This conflicts slightly with #30652, but there too it looks like the `FulfillmentContext` is from an `InferCtxt` that is created and dropped within the same function body (across one call to a module-private function).
That said, I heard that the PR that originally moved `FulfillmentContext` into `InferCtxt` was big, which leaves me concerned that I'm missing something.
r? @nikomatsakis
Why do this: The RegionGraph representation previously conflated all
of the non-variable regions (i.e. the concrete regions such as
lifetime parameters to the current function) into a single dummy node.
A single dummy node leads DFS on a graph `'a -> '_#1 -> '_#0 -> 'b` to
claim that `'_#1` is reachable from `'_#0` (due to `'a` and `'b` being
conflated in the graph representation), which is incorrect (and can
lead to soundness bugs later on in compilation, see #30438).
Splitting the dummy node ensures that DFS will never introduce new
ancestor relationships between nodes for variable regions in the
graph.
Issue #31109 uncovered two semi-related problems:
* A panic in `str::parse::<f64>`
* A panic in `rustc::middle::const_eval::lit_to_const` where the result of float parsing was unwrapped.
This series of commits fixes both issues and also drive-by-fixes some things I noticed while tracking down the parsing panic.
The scope of these refactorings is a little bit bigger than the title implies. See each commit for details.
I’m submitting this for nitpicking now (the first 4 commits), because I feel the basic idea/implementation is sound and should work. I will eventually expand this PR to cover the translator changes necessary for all this to work (+ tests), ~~and perhaps implement a dynamic dropping scheme while I’m at it as well.~~
r? @nikomatsakis
This change also modifies the dep graph infrastructure to keep track of the number of active tasks, so that even if we are not building the full dep-graph, we still get assertions when there is no active task and one does something that would add a read/write edge. This is particularly helpful since, if the assertions are *not* active, you wind up with the error happening in the message processing thread, which is too late to know the correct backtrace.
~~Before landing, I need to do some performance measurements. Those are underway.~~
See measurements below. No real effect on time.
r? @michaelwoerister
Have the `ObligationForest` keep some per-tree state (or type `T`) and have it give a mutable reference for use when processing obligations. In this case, it will be a hashmap. This obviously affects the work that @soltanmm has been doing on snapshotting. I partly want to toss this out there for discussion.
Fixes#31157. (The test in question goes to approx. 30s instead of 5 minutes for me.)
cc #30977.
cc @aturon @arielb1 @soltanmm
r? @aturon who reviewed original `ObligationForest`
The purpose of the translation item collector is to find all monomorphic instances of functions, methods and statics that need to be translated into LLVM IR in order to compile the current crate.
So far these instances have been discovered lazily during the trans path. For incremental compilation we want to know the set of these instances in advance, and that is what the trans::collect module provides.
In the future, incremental and regular translation will be driven by the collector implemented here.
r? @nikomatsakis
cc @rust-lang/compiler
Translation Item Collection
===========================
This module is responsible for discovering all items that will contribute to
to code generation of the crate. The important part here is that it not only
needs to find syntax-level items (functions, structs, etc) but also all
their monomorphized instantiations. Every non-generic, non-const function
maps to one LLVM artifact. Every generic function can produce
from zero to N artifacts, depending on the sets of type arguments it
is instantiated with.
This also applies to generic items from other crates: A generic definition
in crate X might produce monomorphizations that are compiled into crate Y.
We also have to collect these here.
The following kinds of "translation items" are handled here:
- Functions
- Methods
- Closures
- Statics
- Drop glue
The following things also result in LLVM artifacts, but are not collected
here, since we instantiate them locally on demand when needed in a given
codegen unit:
- Constants
- Vtables
- Object Shims
General Algorithm
-----------------
Let's define some terms first:
- A "translation item" is something that results in a function or global in
the LLVM IR of a codegen unit. Translation items do not stand on their
own, they can reference other translation items. For example, if function
`foo()` calls function `bar()` then the translation item for `foo()`
references the translation item for function `bar()`. In general, the
definition for translation item A referencing a translation item B is that
the LLVM artifact produced for A references the LLVM artifact produced
for B.
- Translation items and the references between them for a directed graph,
where the translation items are the nodes and references form the edges.
Let's call this graph the "translation item graph".
- The translation item graph for a program contains all translation items
that are needed in order to produce the complete LLVM IR of the program.
The purpose of the algorithm implemented in this module is to build the
translation item graph for the current crate. It runs in two phases:
1. Discover the roots of the graph by traversing the HIR of the crate.
2. Starting from the roots, find neighboring nodes by inspecting the MIR
representation of the item corresponding to a given node, until no more
new nodes are found.
The roots of the translation item graph correspond to the non-generic
syntactic items in the source code. We find them by walking the HIR of the
crate, and whenever we hit upon a function, method, or static item, we
create a translation item consisting of the items DefId and, since we only
consider non-generic items, an empty type-substitution set.
Given a translation item node, we can discover neighbors by inspecting its
MIR. We walk the MIR and any time we hit upon something that signifies a
reference to another translation item, we have found a neighbor. Since the
translation item we are currently at is always monomorphic, we also know the
concrete type arguments of its neighbors, and so all neighbors again will be
monomorphic. The specific forms a reference to a neighboring node can take
in MIR are quite diverse. Here is an overview:
The most obvious form of one translation item referencing another is a
function or method call (represented by a CALL terminator in MIR). But
calls are not the only thing that might introduce a reference between two
function translation items, and as we will see below, they are just a
specialized of the form described next, and consequently will don't get any
special treatment in the algorithm.
A function does not need to actually be called in order to be a neighbor of
another function. It suffices to just take a reference in order to introduce
an edge. Consider the following example:
```rust
fn print_val<T: Display>(x: T) {
println!("{}", x);
}
fn call_fn(f: &Fn(i32), x: i32) {
f(x);
}
fn main() {
let print_i32 = print_val::<i32>;
call_fn(&print_i32, 0);
}
```
The MIR of none of these functions will contain an explicit call to
`print_val::<i32>`. Nonetheless, in order to translate this program, we need
an instance of this function. Thus, whenever we encounter a function or
method in operand position, we treat it as a neighbor of the current
translation item. Calls are just a special case of that.
In a way, closures are a simple case. Since every closure object needs to be
constructed somewhere, we can reliably discover them by observing
`RValue::Aggregate` expressions with `AggregateKind::Closure`. This is also
true for closures inlined from other crates.
Drop glue translation items are introduced by MIR drop-statements. The
generated translation item will again have drop-glue item neighbors if the
type to be dropped contains nested values that also need to be dropped. It
might also have a function item neighbor for the explicit `Drop::drop`
implementation of its type.
A subtle way of introducing neighbor edges is by casting to a trait object.
Since the resulting fat-pointer contains a reference to a vtable, we need to
instantiate all object-save methods of the trait, as we need to store
pointers to these functions even if they never get called anywhere. This can
be seen as a special case of taking a function reference.
Since `Box` expression have special compiler support, no explicit calls to
`exchange_malloc()` and `exchange_free()` may show up in MIR, even if the
compiler will generate them. We have to observe `Rvalue::Box` expressions
and Box-typed drop-statements for that purpose.
Interaction with Cross-Crate Inlining
-------------------------------------
The binary of a crate will not only contain machine code for the items
defined in the source code of that crate. It will also contain monomorphic
instantiations of any extern generic functions and of functions marked with
The collection algorithm handles this more or less transparently. When
constructing a neighbor node for an item, the algorithm will always call
`inline::get_local_instance()` before proceeding. If no local instance can
be acquired (e.g. for a function that is just linked to) no node is created;
which is exactly what we want, since no machine code should be generated in
the current crate for such an item. On the other hand, if we can
successfully inline the function, we subsequently can just treat it like a
local item, walking it's MIR et cetera.
Eager and Lazy Collection Mode
------------------------------
Translation item collection can be performed in one of two modes:
- Lazy mode means that items will only be instantiated when actually
referenced. The goal is to produce the least amount of machine code
possible.
- Eager mode is meant to be used in conjunction with incremental compilation
where a stable set of translation items is more important than a minimal
one. Thus, eager mode will instantiate drop-glue for every drop-able type
in the crate, even of no drop call for that type exists (yet). It will
also instantiate default implementations of trait methods, something that
otherwise is only done on demand.
Open Issues
-----------
Some things are not yet fully implemented in the current version of this
module.
Since no MIR is constructed yet for initializer expressions of constants and
statics we cannot inspect these properly.
Ideally, no translation item should be generated for const fns unless there
is a call to them that cannot be evaluated at compile time. At the moment
this is not implemented however: a translation item will be produced
regardless of whether it is actually needed or not.
<!-- Reviewable:start -->
[<img src="https://reviewable.io/review_button.png" height=40 alt="Review on Reviewable"/>](https://reviewable.io/reviews/rust-lang/rust/30900)
<!-- Reviewable:end -->
The purpose of the translation item collector is to find all monomorphic instances of functions, methods and statics that need to be translated into LLVM IR in order to compile the current crate.
So far these instances have been discovered lazily during the trans path. For incremental compilation we want to know the set of these instances in advance, and that is what the trans::collect module provides.
In the future, incremental and regular translation will be driven by the collector implemented here.
This PR adds some minor error correction to the parser - if there is a missing ident, we recover and carry on. It also makes compilation more robust so that non-fatal errors (which is still most of them, unfortunately) in parsing do not cause us to abort compilation. The effect is that a program with a missing or incorrect ident can get all the way to type checking.
In 95d904625b output was accidentally moved
from STDERR to STDOUT.
This commit also changes the order of debug output. Previously, it was:
```
/* id 22: … */ {
…
}
DEBUG:rustc::middle::dataflow:
```
Now, it is:
```
DEBUG:rustc::middle::dataflow: /* id 22: … */ {
…
}
```
This is a fix for #30741. It simplifies dep-graph tracking for trait matching. I was experimenting with having a greater resolution here, but decided to pare back to just have one dep node for "trait resolutions on trait `Foo`", which means that adding an impl to the trait `Foo` will invalidate all fns that had to do any trait matching at all on `Foo`. This seems like a reasonable starting place.
Independently, I realized I had neglected to record a dependency from trans on typeck -- this is obviously needed, since trans consumes a bunch of data structures that typeck produces (but which are not currently individually tracked) -- and because trans assumes that typeck has been done. Eventually those are going to go away and be replaced with MIR, which will be tracked, so this edge would presumably be derived automatically then, but it's an obvious enough thing to want for now.
r? @arielb1
cc @michaelwoerister -- this might indirectly fix the problem you observed with the trans cache, though it'd be nice to try and craft an independent test case for that.
was the major use-case, and to update the dep-graph. Other kinds of
predicates are now excluded from the cache because there is no easy way
to make a good dep-graph node for them, and because they are not
believed to be that useful. :)
Fixes#30741. (However, the test still gives wrong result for trans,
for an independent reason which is fixed in the next commit.)
