The wording of RFC #495 enables moves out of slices. Unfortuantely, non-zeroing
moves out of slices introduce a very annoying complication: as slices can
vary in their length, indexes from the start and end may or may not overlap
depending on the slice's exact length, which prevents assigning a particular
drop flag for each individual element.
For example, in the code
```Rust
fn foo<T>(a: Box<[Box<[T]>]>, c: bool) -> T {
match (a, c) {
(box [box [t, ..], ..], true) => t,
(box [.., box [.., t]], false) => t,
_ => panic!()
}
}
```
If the condition is false, we have to drop the first element
of `a`, unless `a` has size 1 in which case we drop all the elements
of it but the last.
If someone comes with a nice way of handling it, we can always re-allow
moves out of slices.
This is a [breaking-change], but it is behind the `slice_patterns` feature
gate and was not allowed until recently.
Update lifetime errors to specifically note temporaries
This PR updates the error message we give in the case of a temporary value not living long enough.
Before:
<img width="497" alt="screen shot 2016-08-31 at 10 02 47 am" src="https://cloud.githubusercontent.com/assets/547158/18138551/27a06794-6f62-11e6-9ee2-bdf8bed75ca7.png">
Now:
<img width="488" alt="screen shot 2016-08-31 at 10 03 01 am" src="https://cloud.githubusercontent.com/assets/547158/18138557/2e5cf322-6f62-11e6-9047-4a78abf3d78c.png">
Specifically, it makes the following changes:
* Detects if a temporary is being used. If so, it changes the labels to mention that a temporary value specifically is in question
* Simplifies wording of the existing labels to focus on lifetimes rather than values being valid
* Changes the help to a note, since the help+span wasn't as helpful (and sometimes more confusing) than just a note.
r? @nikomatsakis
rustc_borrowck: Don't hash types in loan paths
1) Types for equal loan paths are not always equal, they can sometimes differ in lifetimes, making equal loan paths hash differently.
Example:
71bdeea561/src/libcollections/linked_list.rs (L835-L856)
One of `self.list`s has type
```
&ReFree(CodeExtent(15013/CallSiteScope { fn_id: 18907, body_id: 18912 }), BrNamed(0:DefIndex(3066), 'a(397), WontChange)) mut linked_list::LinkedList<T>
```
and other has type
```
&ReScope(CodeExtent(15018/Remainder(BlockRemainder { block: 18912, first_statement_index: 0 }))) mut linked_list::LinkedList<T>
```
(... but I'm not sure it's not a bug actually.)
2) Not hashing types is faster than hashing types.
r? @arielb1
[MIR] track Location in MirVisitor, combine Location
All the users of MirVisitor::visit_statement implement their own statement index tracking. This PR move the tracking into MirVisitor itself.
Also, there were 2 separate implementations of Location that were identical. This PR eliminates one of them.
Move 'doesn't live long enough' errors to labels
This patch moves the "doesn't live long enough" region-style errors to instead use labels.
An example follows.
Before:
```
error: `x` does not live long enough
--> src/test/compile-fail/send-is-not-static-ensures-scoping.rs:26:18
|
26 | let y = &x;
| ^
|
note: reference must be valid for the block at 23:10...
--> src/test/compile-fail/send-is-not-static-ensures-scoping.rs:23:11
|
23 | fn main() {
| ^
note: ...but borrowed value is only valid for the block suffix following statement 0 at 25:18
--> src/test/compile-fail/send-is-not-static-ensures-scoping.rs:25:19
|
25 | let x = 1;
| ^
```
After:
```
error: `x` does not live long enough
--> src/test/compile-fail/send-is-not-static-ensures-scoping.rs:26:18
|
26 | let y = &x;
| ^ does not live long enough
...
32 | };
| - borrowed value only valid until here
...
35 | }
| - borrowed value must be valid until here
```
r? @nikomatsakis
Implement the `!` type
This implements the never type (`!`) and hides it behind the feature gate `#[feature(never_type)]`. With the feature gate off, things should build as normal (although some error messages may be different). With the gate on, `!` is usable as a type and diverging type variables (ie. types that are unconstrained by anything in the code) will default to `!` instead of `()`.
[MIR] Add explicit SetDiscriminant StatementKind for deaggregating enums
cc #35186
To deaggregate enums, we need to be able to explicitly set the discriminant. This PR implements a new StatementKind that does that.
I think some of the places that have `panics!` now could maybe do something smarter.
refactor lvalue_ty to be method of lvalue
Currently `Mir` (and `MirContext`) implement a method `lvalue_ty` (and actually many more `foo_ty`). But this should be a method of `Lvalue`.
If you have an `lvalue` and you want to get its type, the natural thing to write is:
```
lvalue.ty()
```
Of course it needs context, but still:
```
lvalue.ty(mir, tcx)
```
Makes more sense than
```
mir.lvalue_ty(lvalue, tcx)
```
I actually think we should go a step farther and have traits so we could get the type of some value generically, but that's up for debate. The thing I'm running into a lot in the compiler is I have a value of type `Foo` and I know that there is some related type `Bar` which I can get through some combination of method calls, but it's often not as direct as I would imagine. Unless you already know the code, its not clear why you would look in `Mir` for a method to get the type of an `Lvalue`.
Address ICEs running w/ incremental compilation and building glium
Fixes for various ICEs I encountered trying to build glium with incremental compilation enabled. Building glium now works. Of the 4 ICEs, I have test cases for 3 of them -- I didn't isolate a test for the last commit and kind of want to go do other things -- most notably, figuring out why incremental isn't saving much *effort*.
But if it seems worthwhile and I can come back and try to narrow down the problem.
r? @michaelwoerister
Fixes#34991Fixes#32015
Per the discussion on #34765, we make one `DepNode::Mir` variant and use
it to represent both the MIR tracking map as well as passes that operate
on MIR. We also track loads of cached MIR (which naturally comes from
metadata).
Note that the "HAIR" pass adds a read of TypeckItemBody because it uses
a myriad of tables that are not individually tracked.
This is a spiritual succesor to #34268/8531d581, in which we replaced a
number of matches of None to the unit value with `if let` conditionals
where it was judged that this made for clearer/simpler code (as would be
recommended by Manishearth/rust-clippy's `single_match` lint). The same
rationale applies to matches of None to the empty block.
Casual grepping revealed some places in the codebase (some of which
antedated `if let`'s December 2014 stabilization in c200ae5a) where we
were using a match with a `None => ()` arm where (in the present
author's opinion) an `if let` conditional would be more readable. (Other
places where matching to the unit value did seem to better express the
intent were left alone.)
It's likely that we don't care about making such trivial,
non-functional, sheerly æsthetic changes.
But if we do, this is a patch.
Add explanations for E0503 and E0508.
(cannot use `..` because it was mutably borrowed, cannot move out of type `..`, a non-copy fixed-size array)
Part of #32777.
MIR cleanups and predecessor cache
This PR cleans up a few things in MIR and adds a predecessor cache to allow graph algorithms to be run easily.
r? @nikomatsakis
Use it instead of a `panic` for inexhaustive matches and correct the
comment. I think we trust our match-generation algorithm enough to
generate these blocks, and not generating an `unreachable` means that
LLVM won't optimize `match void() {}` to an `unreachable`.
Fix issue #34101
Fix issue #34101: do not track subcontent of type with dtor nor gather flags for untracked content.
(Includes a regression test, which needed to go into `compile-fail/`
due to weaknesses when combining `#[deny(warnings)]` with
`tcx.sess.span_warn(..)`)
(Includes a regression test, which needed to go into `compile-fail/`
due to weaknesses when combining `#[deny(warnings)]` with
`tcx.sess.span_warn(..)`)
(updated with review feedback from arielb1.)
rustc: add ReErased to be used by trait selection, MIR and trans.
`ReErased` replaces `ReStatic` (i.e. `'static`) for erasing regions.
Using a distinct lifetime helps prevent accidental mix-ups between the two.
It also allows cleaner type printing (see test changes), including in symbol names:
```rust
str..pattern..CharSearcher$LT$$u27$static$GT$::drop.30560::h840c2f2afc03bbea // before
str..pattern..CharSearcher::drop.30561::h6bd31d2af614377a // after
```
Not that we should be producing symbols this way, but it's still better.
MSVC requires unwinding code to be split to a tree of *funclets*, where each funclet
can only branch to itself or to to its parent.
Luckily, the code we generates matches this pattern. Recover that structure in
an analyze pass and translate according to that.
this introduces a DropAndReplace terminator as a fix to #30380. That terminator
is suppsoed to be translated by desugaring during drop elaboration, which is
not implemented in this commit, so this breaks `-Z orbit` temporarily.
Separate bindings from other patterns in HIR
Now when name resolution is done on AST, we can avoid dumping everything that looks like an identifier into `PatKind::Ident` in HIR.
`hir::PatKind::Ident` is removed, fresh bindings are now called `hir::PatKind::Binding`, everything else goes to `hir::PatKind::Path`.
I intend to do something with `PatKind::Path`/`PatKind::QPath` as well using resolution results, but it requires some audit and maybe some deeper refactoring of relevant resolution/type checking code to do it properly.
I'm submitting this part of the patch earlier to notify interested parties that I'm working on this.
cc @jseyfried
r? @eddyb
Increase spacing in error format for readability.
Two small tweaks that seem to help readability quite a bit:
* Add spacing header<->snippet, but use the |> on the side for visual consistency
* Fix#33819
* Fix#33763
* Move format-sensitive test (issue-26480 in cfail) to ui test
r? @nikomatsakis
Fixes to mir dataflow
Fixes to mir dataflow
This collects a bunch of changes to `rustc_borrowck::borrowck::dataflow` (which others have pointed out should probably migrate to some crate that isn't tied to the borrow-checker -- but I have not attempted that here, especially since there are competing approaches to dataflow that we should also evaluate).
These changes:
1. Provide a family of related analyses: MovingOutStatements (which is what the old AST-based dataflo computed), as well as MaybeInitialized, MaybeUninitalized, and DefinitelyInitialized.
* (The last two are actually inverses of each other; we should pick one and drop the other.)
2. Fix bugs in the pre-existing analysis implementation, which was untested and thus some obvious bugs went unnoticed, which brings us to the third point:
3. Add a unit test infrastructure for the MIR dataflow analysis.
* The tests work by adding a new intrinsic that is able to query the analysis state for a particular expression (technically, a particular L-value).
* See the examples in compile-fail/mir-dataflow/inits-1.rs and compile-fail/mir-dataflow/uninits-1.rs
* These tests are only checking the results for MaybeInitialized, MaybeUninitalized, and DefinitelyInitialized; I am not sure if it will be feasible to generalize this testing strategy to the MovingOutStatements dataflow operator.
Also got rid of the `trait HasMoveData`, since I am now just imposing
the constraint that `BitDenotation<Ctxt=MoveData<'tcx>>` where
necessary instead.
As some drive-by's:
* moved bitwise operators into `mod bitslice`
* factored out `fn gen` and `fn kill` methods on `BlockSets` type
* removed outdated comment about `fn propagate_call_return`
* removed `on_all_children_bits`, rewriting calls to use `super::on_all_children_bits`
* moved `fn path` helper routine out of `impl MirBorrowckCtxtPreDataflow`
Add error explanation for E0502
I am questioning the order of presentation on the suggested code fixes, but I'm not sure what would be best. Thoughts?
r? @GuillaumeGomez
Fix for old school error issues, improvements to new school
This PR:
* Fixes some old school error issues, specifically #33559, #33543, #33366
* Improves wording borrowck errors with match patterns
* De-emphasize multi-line spans, so we don't color the single source character when we're trying to say "span starts here"
* Rollup of #33392 (which should help fix#33390)
r? @nikomatsakis
(The static semantics of `rustc_peek` is derived from attributes
attached to the function being compiled; in this case,
`rustc_peek(&expr)` observes the dataflow state for the l-value
`expr`.)
Incorporates many fixes contributed by arielb1.
----
revise borrowck::mir::dataflow code to allow varying domain for bitvectors.
This particular code implements the `BitDenotation` trait for three
analyses:
* `MovingOutStatements`, which, like `borrowck::move_data`, maps each
bit-index to a move instruction, and a 1 means "the effect of this
move reaches this point" (and the assigned l-value, if a scoped
declaration, is still in scope).
* `MaybeInitializedLvals`, which maps each bit-index to an l-value.
A 1 means "there exists a control flow path to this point that
initializes the associated l-value."
* `MaybeUninitializedLvals`, which maps each bit-index to an l-value
A 1 means "there exists a control flow path to this point that
de-initializes the associated l-value."
----
Revised `graphviz` dataflow-rendering support in `borrowck::mir`.
One big difference is that this code is now parameterized over the
`BitDenotation`, so that it can be used to render dataflow results
independent of how the dataflow bitvectors are interpreted; see where
reference to `MoveOut` is replaced by the type parameter `D`.
----
Factor out routine to query subattributes in `#[rustc_mir(..)]`.
(Later commits build upon this for some unit testing and instrumentation.)
----
thread through a tcx so that I can query types of lvalues as part of analysis.
----
Revised `BitDenotation::Ctxt`, allowing variation beyond `MoveData`.
The main motivation is to ease threading through a `TyCtxt`.
(In hindsight it might have been better to instead attach the `TyCtxt`
to each of the different dataflow implementations, but that would
require e.g. switching away from having a `Default` impl, so I am
leaving that experiment for another time.)
Batch of improvements to errors for new error format
This is a batch of improvements to existing errors to help get the most out of the new error format.
* Added labels to primary spans (^^^) for a set of errors that didn't currently have them
* Highlight the source blue under the secondary notes for better readability
* Move some of the "Note:" into secondary spans+labels
* Fix span_label to take &mut instead, which makes it work the same as other methods in that set
rustbuild: Add support for crate tests + doctests
This commit adds support to rustbuild to run crate unit tests (those defined by
`#[test]`) as well as documentation tests. All tests are powered by `cargo test`
under the hood.
Each step requires the `libtest` library is built for that corresponding stage.
Ideally the `test` crate would be a dev-dependency, but for now it's just easier
to ensure that we sequence everything in the right order.
Currently no filtering is implemented, so there's not actually a method of
testing *only* libstd or *only* libcore, but rather entire swaths of crates are
tested all at once.
A few points of note here are:
* The `coretest` and `collectionstest` crates are just listed as `[[test]]`
entires for `cargo test` to naturally pick up. This mean that `cargo test -p
core` actually runs all the tests for libcore.
* Libraries that aren't tested all mention `test = false` in their `Cargo.toml`
* Crates aren't currently allowed to have dev-dependencies due to
rust-lang/cargo#860, but we can likely alleviate this restriction once
workspaces are implemented.
cc #31590
This commit adds support to rustbuild to run crate unit tests (those defined by
`#[test]`) as well as documentation tests. All tests are powered by `cargo test`
under the hood.
Each step requires the `libtest` library is built for that corresponding stage.
Ideally the `test` crate would be a dev-dependency, but for now it's just easier
to ensure that we sequence everything in the right order.
Currently no filtering is implemented, so there's not actually a method of
testing *only* libstd or *only* libcore, but rather entire swaths of crates are
tested all at once.
A few points of note here are:
* The `coretest` and `collectionstest` crates are just listed as `[[test]]`
entires for `cargo test` to naturally pick up. This mean that `cargo test -p
core` actually runs all the tests for libcore.
* Libraries that aren't tested all mention `test = false` in their `Cargo.toml`
* Crates aren't currently allowed to have dev-dependencies due to
rust-lang/cargo#860, but we can likely alleviate this restriction once
workspaces are implemented.
cc #31590
