Major revision to the dropck_legal_cycles test.
1. Added big comment block explaining the test framework.
2. Added tests exericising Rc and Arc. This was inspired by a comment
from eefriedman on PR #28861.
3. Made the cycle-detection not issue false-positives on acyclic dags.
Doing this efficiently required revising the framework; instead of
visiting all children (i.e. doing a traversal), now each test is
responsible for supplying the path that will act as a witness to
the cycle.
Luckily for me, all of the pre-existing tests worked with a trivial
path built from "always tke your first left", but new tests I added
did require other input paths (i.e., "first turn right, then left".
(The path representation is a bit-string and its branches are
n-ary, not word phrases and binary branches as you might think
from the outline above.)
cc PR #27185
1. Added big comment block explaining the test framework.
2. Added tests exericising Rc and Arc. This was inspired by a comment
from eefriedman on PR #28861.
3. Made the cycle-detection not issue false-positives on acyclic dags.
Doing this efficiently required revising the framework; instead of
visiting all children (i.e. doing a traversal), now each test is
responsible for supplying the path that will act as a witness to
the cycle.
Luckily for me, all of the pre-existing tests worked with a trivial
path built from "always tke your first left", but new tests I added
did require other input paths (i.e., "first turn right, then left".
(The path representation is a bit-string and its branches are
n-ary, not word phrases and binary branches as you might think
from the outline above.)
This new version takes inspiration from the C implementation of the benchmark, but instead of explicitly using SIMD operations which can't be done on stable, it instead arranges everything the same way and leaves the actual vectorization up to LLVM.
In addition to the ~20% speed gains (see below), this PR also adds some general niceties which showcase the language a little bit: a `Vec3` type to cut down on `(x, y, z)` repetition, using `while let` instead of `loop-if-break`, iterator adapters instead of for loops etc.
Here are the times in seconds of 10 runs each on my workstation:
```
before: 6.254, 6.260, 6.263, 6.264, 6.265, 6.267, 6.334, 6.341, 6.441, 6.509
before-min: 6.254
before-median: 6.266
before-max: 6.509
after: 4.823, 4.824, 4.826, 4.827, 4.837, 4.839, 4.881, 4.959, 4.990, 5.377
after-min: 4.823
after-median: 4.838
after-max: 5.377
gcc: 4.674, 4.676, 4.680, 4.682, 4.695, 4.696, 4.701, 4.708, 4.794, 5.297
gcc-min: 4.674
gcc-median: 4.696
gcc-max: 5.297
```
On my i7 laptop the speed up is less impressive, from ~5.4s to ~4.7s, but still significant. On my Vultr VPS the numbers look closer to the workstation results. Surprisingly my laptop beats both office workstation and VPS...
In #28864, @aarzee submitted some whitespace fixes. I r+'d it. But
@retp998 noticed[1] that this file is explicitly a test of this kind of
whitespace, and so I shouldn't have changed it. This restores the old
line endings.
1: https://github.com/rust-lang/rust/pull/28864#discussion_r41332279
In #28864, @aarzee submitted some whitespace fixes. I r+'d it. But
@retp998 noticed[1] that this file is explicitly a test of this kind of
whitespace, and so I shouldn't have changed it. This restores the old
line endings.
1: https://github.com/rust-lang/rust/pull/28864#discussion_r41332279
This turned up as part of #3170. When constructing an `undef` value to
return in the error case, we were trying to get the element type of the
Rust-level value being indexed instead of the underlying array; when
indexing a slice, that's not an array and the LLVM assertion failure
reflects this.
The regression test is a lightly altered copy of `const-array-oob.rs`.
This lint warning was originally intended to help against misuse of the old Rust
`int` and `uint` types in FFI bindings where the Rust `int` was not equal to the
C `int`. This confusion no longer exists (as Rust's types are now `isize` and
`usize`), and as a result the need for this lint has become much less over time.
Additionally, starting with [the RFC for libc][rfc] it's likely that `isize` and
`usize` will be quite common in FFI bindings (e.g. they're the definition of
`size_t` and `ssize_t` on many platforms).
[rfc]: https://github.com/rust-lang/rfcs/pull/1291
This commit disables these lints to instead consider `isize` and `usize` valid
types to have in FFI signatures.
This turned up as part of #3170. When constructing an `undef` value to
return in the error case, we were trying to get the element type of the
Rust-level value being indexed instead of the underlying array; when
indexing a slice, that's not an array and the LLVM assertion failure
reflects this.
The regression test is a lightly altered copy of `const-array-oob.rs`.
The reason this was not failing is fascinating. The variable $(rustc)
is empty, so the make recipe was expanded as " -o foo foo.rs". make
interpreted this as an instruction to run the command "o foo foo.rs"
and ignore any failure that occurred, because it uses a leading '-' on
a command to signal that behavior.
By RFC1214:
> Before calling a fn, we check that its argument and return types are WF.
The previous code only checked the trait-ref, which was not enough
in several cases.
As this is a soundness fix, it is a [breaking-change]. Some new annotations are needed, which I think are because of #18653 and the imperfection of `projection_must_outlive` (that can probably be worked around by moving the wf obligation later).
Fixes#28609
r? @nikomatsakis
The reason this was not failing is fascinating. The variable $(rustc)
is empty, so the make recipe was expanded as " -o foo foo.rs". make
interpreted this as an instruction to run the command "o foo foo.rs"
and ignore any failure that occurred, because it uses a leading '-' on
a command to signal that behavior.
By RFC1214:
Before calling a fn, we check that its argument and return types are WF. This check takes place after all higher-ranked lifetimes have been instantiated. Checking the argument types ensures that the implied bounds due to argument types are correct. Checking the return type ensures that the resulting type of the call is WF.
The previous code only checked the trait-ref, which was not enough
in several cases.
As this is a soundness fix, it is a [breaking-change].
Fixes#28609
This PR closes out #28716 and #28735 by making two changes to the compiler:
1. The `--emit` flag to the compiler now supports the ability to specify the output file name of a partuclar emit type. For example `--emit dep-info=bar.d,asm=foo.s,link` is now accepted.
2. The dep-info emission now emits a dummy target for all input file names to protect against deleted files.
The behavior here isn't really ideal, but we can't really do much better
given the current state of constant evaluation.
The changes to ExprUseVisitor are to avoid a compile error; apparently
that bit of code is extremely sensitive to changes in other areas of the
compiler.
Fixes#28670, and probably a bunch of duplicates.
This commit swaps the order of linking local native libraries and upstream
native libraries on the linker command line. Detail of bugs this can cause can
be found in #28595, and this change also invalidates the test case that was
added for #12446 which is now considered a bug because the downstream dependency
would need to declare that it depends on the native library somehow.
Closes#28595
