rustc: Update LLVM with a ThinLTO fix
This commit updates LLVM with a patch that's landed upstream to fix an assertion
that was tripping when ThinLTO was activated. Unfortunately I wasn't able to get
a reduced test case, but I've tested manually on the affected crates and the
assertion is indeed fixed.
Closes#45131
rustc: Handle `#[no_mangle]` anywhere in a crate
This commit updates the reachability pass of the compiler to seed the local
worklist with `#[no_mangle]`-like items anywhere in a crate, not just those
reachable from public items.
Closes#45165
Better error message for comma after base struct
#41834
This adds a better error for commas after the base struct:
```
let foo = Foo {
one: 111,
..Foo::default(), // This comma is a syntax error
};
```
The current error is a generic `expected one of ...` which isn't beginner-friendly. My error looks like this:
```
error: cannot use a comma after the base struct
--> tmp/example.rs:26:9
|
26 | ..Foo::default(),
| ^^^^^^^^^^^^^^^^- help: remove this comma
|
= note: the base struct expansion must always be the last field
```
I even added a note for people who don't know why this isn't allowed.
rustc: Allow target-specific default cgus
Some targets, like msp430 and nvptx, don't work with multiple codegen units
right now for bugs or fundamental reasons. To expose this allow targets to
express a default.
Closes#45000
rustc: Add LLVM `nounwind` with `-C panic=abort`
This informs LLVM that functions can't unwind, which while it should typically
have already been inferred when necessary or otherwise not impact codegen is
apparently needed on targets like ARM to avoid references to unnecessary
symbols.
Closes#44992
MIR-borrowck: moves of prefixes invalidate uses too
I overlooked the fact that when we check if a path is moved, we need to check for interference between the (shallow) prefixes and the use in question.
~~Long term, we may want to revise how this computation is done. For example, it might be better to represent the set of invalidated prefixes in the dataflow computation (the `maybe_uninitialized` dataflow), and thus avoid one of the loops in the code here.~~
* Update: I was wrong in my original recollection of the dataflow code, which actually does the right thing, in terms of precisely tracking substructure initialization and movement.
Fix#44833
----
Update: The initial version of this PR's description (and the code as well) erroneously focused on supporting prefixes. ~~But the two main cases of interest are: 1. the *shallow* prefixes, and 2. the deref-free prefix built off a local (if the lvalue is indeed built off a local)~~
Update 2: The main cases of interest are in fact: 1. the nearest prefix with a MovePath, and 2. the suffixes.
Extend mir dump to dump each region
Building on #44878, implement the feature discussed in #44872.
Through discussions on the WG-nll-gitter, @nikomatsakis and I decided to implement this by extending `dump_mir` and all functions that it calls to take a callback of signature `FnMut(PassWhere, &mut Write) -> io::Result<()>` where `PassWhere` is an enum that represents possible locations that we may want to print out extra data in the process of dumping the MIR.
I'm not particularly wedded to the name `PassWhere`, but I felt that simply calling the enum `Where` wasn't the right thing to name it.
This work depends strongly on #44878, and should be rebased on the final version of that tree, whatever that may be.
Fix PEP8 style issues in bootstrap code
This fixes PEP8 style issues (other than line-length) in the bootstrap Python code.
The most important fix is in the `set` function where the code was indented with 6 spaces instead of 4.
let rustdoc print the crate version into docs
This PR adds a new unstable flag to rustdoc, `--crate-version`, which when present will add a new entry to the sidebar of the root module, printing the given version number:
Closes#24336
(The WIP status is because i don't want to merge this until i can get the std docs to use it, which i need help from rustbuild people to make sure i get right.)
Optimize comparison functions of Iterator
Replaced matching on tuples which led to less performant code generation. Testing on microbenchmarks consistently showed ~1.35x improvement in performance on my machine.
Fixes#44729.
This commit updates the reachability pass of the compiler to seed the local
worklist with `#[linkage]`-like items anywhere in a crate, not just those
reachable from public items.
Closes#45165
MIR borrowck: print lvalues in error messages in the same way that the AST borrowck
Fix#44974
- Print fields with `.name` rather than `.<num>`
- Autoderef values if followed by a field or an index
- Output `[..]` when borrowing inside a slice
(There are other tests that this PR also improves, but were not
completely synchronized. I chose to wait until later to pull those
into the `revisions: ast mir` testing pattern; later being either when
they *are* synchronized, or in some PR where we migrate all borrowck
tests, regardless of whether MIR-borrowck is "finished" for them or
not.)
Fix#44833 (a very specific instance of a very broad bug).
In `check_if_path_is_moved(L)`, check nearest prefix of L with
MovePath, and suffixes of L with MovePaths.
Over the course of review, ariel pointed out a number of issues that
led to this version of the commit:
1. Looking solely at supporting prefixes does not suffice: it
overlooks checking if the path was ever actually initialized in the
first place. So you need to be willing to consider non-supporting
prefixes. Once you are looking at all prefixes, you *could* just
look at the local that forms the base of the projection, but to
handle partial initialization (which still needs to be formally
specified), this code instead looks at the nearest prefix of L that
has an associated MovePath (which, in the limit, will end up being
a local).
2. You also need to consider the suffixes of the given Lvalue, due to
how dataflow is representing partial moves of individual fields out
of struct values.
3. (There was originally a third search, but ariel pointed out that
the first and third could be folded into one.)
Also includes some drive-by refactorings to simplify some method
signatures and prefer `for _ in _` over `loop { }` (at least when it
comes semi-naturally).
Improve performance of spsc_queue and stream.
This PR makes two main changes:
1. It switches the `spsc_queue` node caching strategy from keeping a shared
counter of the number of nodes in the cache to keeping a consumer only counter
of the number of node eligible to be cached.
2. It separates the consumer and producers fields of `spsc_queue` and `stream` into
a producer cache line and consumer cache line.
Overall, it speeds up `mpsc` in `spsc` mode by 2-10x.
Variance is higher than I'd like (that 2-10x speedup is on one benchmark), I believe this is due to the drop check in `send` (`fn stream::Queue::send:107`). I think this check can be combined with the sleep detection code into a version which only uses 1 shared variable, and only one atomic access per `send`, but I haven't looked through the select implementation enough to be sure.
The code currently assumes a cache line size of 64 bytes. I added a CacheAligned newtype in `mpsc` which I expect to reuse for `shared`. It doesn't really belong there, it would probably be best put in `core::sync::atomic`, but putting it in `core` would involve making it public, which I thought would require an RFC.
Benchmark runner is [here](3eca46279c/shootout), benchmarks [here](3eca46279c/queue_bench/src/lib.rs (L170-L293)).
Fixes#44512.
1. Add -f flag to curl, so when the server returns 403 or 500 it will fail
immediately.
2. Moved the checksum part into the retry loop, assuming checksum failure
is due to broken download that can be fixed by downloading again.
This commit updates LLVM with a patch that's landed upstream to fix an assertion
that was tripping when ThinLTO was activated. Unfortunately I wasn't able to get
a reduced test case, but I've tested manually on the affected crates and the
assertion is indeed fixed.
Closes#45131
This informs LLVM that functions can't unwind, which while it should typically
have already been inferred when necessary or otherwise not impact codegen is
apparently needed on targets like ARM to avoid references to unnecessary
symbols.
Closes#44992
Refactor fmt::Display and fmt::Debug impls in ppaux
Also fixes#44887.
There was a problem that unnamed late-bound regions are *always* named `'r` while they are displayed using `std::fmt::Display`.
---
```rust
fn main() {
f(|_: (), _: ()| {});
}
fn f<F>(_: F) where F: Fn(&(), for<'r> fn(&'r ())) {}
```
Before (incorrectly shadows lifetime, `for<'r>` omitted for the second argument):
```
error[E0631]: type mismatch in closure arguments
--> test.rs:2:5
|
2 | f(|_: (), _: ()| {});
| ^ ----------------- found signature of `fn((), ()) -> _`
| |
| expected signature of `for<'r> fn(&'r (), fn(&'r ())) -> _`
|
= note: required by `f`
```
After:
```
error[E0631]: type mismatch in closure arguments
--> test.rs:2:5
|
2 | f(|_: (), _: ()| {});
| ^ ----------------- found signature of `fn((), ()) -> _`
| |
| expected signature of `for<'s> fn(&'s (), for<'r> fn(&'r ())) -> _`
|
= note: required by `f`
```
r? @nikomatsakis
