Use `unwinding` crate for unwinding on Xous platform
This patch adds support for using [unwinding](https://github.com/nbdd0121/unwinding) on platforms where libunwinding isn't viable. An example of such a platform is `riscv32imac-unknown-xous-elf`.
### Background
The Rust project maintains a fork of llvm at [llvm-project](https://github.com/rust-lang/llvm-project/) where it applies patches on top of the llvm project. This mostly seems to be to get unwinding support for the SGX project, and there may be other patches that I'm unaware of.
There is a lot of machinery in the build system to support compiling `libunwind` on other platforms, and I needed to add additional patches to llvm in order to add support for Xous.
Rather than continuing down this path, it seemed much easier to use a Rust-based library. The `unwinding` crate by `@nbdd0121` fits this description perfectly.
### Future work
This could potentially replace the custom patches for `libunwind` on other platforms such as SGX, and could enable unwinding support on many more exotic platforms.
### Anti-goals
This is not designed to replace `libunwind` on tier-one platforms or those where unwinding support already exists. There is already a well-established approach for unwinding there. Instead, this aims to enable unwinding on new platforms where C++ code may be difficult to compile.
The main() function takes an argument that contains the eh_frame
address. Implement `unwinding` support by looking for unwinding data at
this address.
Signed-off-by: Sean Cross <sean@xobs.io>
std: Invert logic for inclusion of `sys_common::net`
The `library/std/src/sys_common/net.rs` module is intended to define common implementations of networking-related APIs across a variety of platforms that share similar APIs (e.g. Berkeley-style sockets and all). This module is not included for more fringe targets however such as UEFI or "unknown" targets to libstd (those classified as `restricted-std`). Previously the `sys_common/net.rs` file was set up such that an allow-list indicated it shouldn't be used. This commit inverts the logic to have an allow-list of when it should be used instead.
The goal of this commit is to make it a bit easier to experiment with a new Rust target. Currently more esoteric targets are required to get an exception in this `cfg_if` block to use `crate::sys::net` such as for unsupported targets. With this inversion of logic only targets which actually support networking will be listed, where most of those are lumped under `cfg(unix)`.
Given that this change is likely to cause some breakage for some target by accident I've attempted to be somewhat robust with this by following these steps to defining the new predicate for inverted logic.
1. Take all supported targets and filter out all `cfg(unix)` ones as these should all support `sys_common/net.rs`.
2. Take remaining targets and filter out `cfg(windows)` ones.
3. The remaining dozen-or-so targets were all audited by hand. Mostly this included `target_os = "hermit"` and `target_os = "solid_asp3"` which required an allow-list entry, but remaining targets were all already excluded (didn't use `sys_common/net.rs` so they were left out.
If this causes breakage it should be relatively easy to fix and I'd be happy to follow-up with any PRs necessary.
Split `Vec::dedup_by` into 2 cycles
First cycle runs until we found 2 same elements, second runs after if there any found in the first one. This allows to avoid any memory writes until we found an item which we want to remove.
This leads to significant performance gains if all `Vec` items are kept: -40% on my benchmark with unique integers.
Results of benchmarks before implementation (including new benchmark where nothing needs to be removed):
* vec::bench_dedup_all_100 74.00ns/iter +/- 13.00ns
* vec::bench_dedup_all_1000 572.00ns/iter +/- 272.00ns
* vec::bench_dedup_all_100000 64.42µs/iter +/- 19.47µs
* __vec::bench_dedup_none_100 67.00ns/iter +/- 17.00ns__
* __vec::bench_dedup_none_1000 662.00ns/iter +/- 86.00ns__
* __vec::bench_dedup_none_10000 9.16µs/iter +/- 2.71µs__
* __vec::bench_dedup_none_100000 91.25µs/iter +/- 1.82µs__
* vec::bench_dedup_random_100 105.00ns/iter +/- 11.00ns
* vec::bench_dedup_random_1000 781.00ns/iter +/- 10.00ns
* vec::bench_dedup_random_10000 9.00µs/iter +/- 5.62µs
* vec::bench_dedup_random_100000 449.81µs/iter +/- 74.99µs
* vec::bench_dedup_slice_truncate_100 105.00ns/iter +/- 16.00ns
* vec::bench_dedup_slice_truncate_1000 2.65µs/iter +/- 481.00ns
* vec::bench_dedup_slice_truncate_10000 18.33µs/iter +/- 5.23µs
* vec::bench_dedup_slice_truncate_100000 501.12µs/iter +/- 46.97µs
Results after implementation:
* vec::bench_dedup_all_100 75.00ns/iter +/- 9.00ns
* vec::bench_dedup_all_1000 494.00ns/iter +/- 117.00ns
* vec::bench_dedup_all_100000 58.13µs/iter +/- 8.78µs
* __vec::bench_dedup_none_100 52.00ns/iter +/- 22.00ns__
* __vec::bench_dedup_none_1000 417.00ns/iter +/- 116.00ns__
* __vec::bench_dedup_none_10000 4.11µs/iter +/- 546.00ns__
* __vec::bench_dedup_none_100000 40.47µs/iter +/- 5.36µs__
* vec::bench_dedup_random_100 77.00ns/iter +/- 15.00ns
* vec::bench_dedup_random_1000 681.00ns/iter +/- 86.00ns
* vec::bench_dedup_random_10000 11.66µs/iter +/- 2.22µs
* vec::bench_dedup_random_100000 469.35µs/iter +/- 20.53µs
* vec::bench_dedup_slice_truncate_100 100.00ns/iter +/- 5.00ns
* vec::bench_dedup_slice_truncate_1000 2.55µs/iter +/- 224.00ns
* vec::bench_dedup_slice_truncate_10000 18.95µs/iter +/- 2.59µs
* vec::bench_dedup_slice_truncate_100000 492.85µs/iter +/- 72.84µs
Resolves#77772
P.S. Note that this is same PR as #92104 I just missed review then forgot about it.
Also, I cannot reopen that pull request so I am creating a new one.
I responded to remaining questions directly by adding commentaries to my code.
Use OnceCell in cell module documentation
The spanning tree example in the std cell module implementation was created before `OnceCell` was added to Rust so it uses `RefCell`. However, in this case using `OnceCell` seems more appropriate and produces simpler code. As a bonus, this also means that all three cell types are presented in the examples of std cell module.
Add support for making lib features internal
We have the notion of an "internal" lang feature: a feature that is never intended to be stabilized, and using which can cause ICEs and other issues without that being considered a bug.
This extends that idea to lib features as well. It is an alternative to https://github.com/rust-lang/rust/pull/115623: instead of using an attribute to declare lib features internal, we simply do this based on the name. Everything ending in `_internals` or `_internal` is considered internal.
Then we rename `core_intrinsics` to `core_intrinsics_internal`, which fixes https://github.com/rust-lang/rust/issues/115597.
portable-simd: fix test suite build
``@workingjubilee`` ``@calebzulawski`` don't we run these portable-simd tests on rustc CI? Currently they don't even build here.
move exposed-provenance APIs into separate feature gate
We have already stated explicitly for all the 'exposed' functions that
> Using this method means that code is *not* following strict provenance rules.
However, they were part of the same feature gate and still described as part of the strict provenance experiment. Unfortunately, their semantics are much less clear and certainly nowhere near stabilization, so in preparation for an attempt to stabilize the strict provenance APIs, I suggest we split the things related to "exposed" into their own feature gate. I also used this opportunity to better explain how Exposed Provenance fits into the larger plan here: this is *one possible candidate* for `as` semantics, but we don't know if it is actually viable, so we can't really promise that it is equivalent to `as`. If it works out we probably want to make `as` equivalent to the 'exposed' APIs; if it doesn't, we will remove them again and try to find some other semantics for `as`.
Add substring API for `OsStr`
This adds a method for taking a substring of an `OsStr`, which in combination with [`OsStr::as_encoded_bytes()`](https://doc.rust-lang.org/std/ffi/struct.OsStr.html#method.as_encoded_bytes) makes it possible to implement most string operations in safe code.
API:
```rust
impl OsStr {
pub fn slice_encoded_bytes<R: ops::RangeBounds<usize>>(&self, range: R) -> &Self;
}
```
Motivation, examples and research at https://github.com/rust-lang/libs-team/issues/306.
Tracking issue: #118485
cc `@epage`
r? libs-api
The `library/std/src/sys_common/net.rs` module is intended to define
common implementations of networking-related APIs across a variety of
platforms that share similar APIs (e.g. Berkeley-style sockets and all).
This module is not included for more fringe targets however such as UEFI
or "unknown" targets to libstd (those classified as `restricted-std`).
Previously the `sys_common/net.rs` file was set up such that an
allow-list indicated it shouldn't be used. This commit inverts the logic
to have an allow-list of when it should be used instead.
The goal of this commit is to make it a bit easier to experiment with a
new Rust target. Currently more esoteric targets are required to get an
exception in this `cfg_if` block to use `crate::sys::net` such as for
unsupported targets. With this inversion of logic only targets which
actually support networking will be listed, where most of those are
lumped under `cfg(unix)`.
Given that this change is likely to cause some breakage for some target
by accident I've attempted to be somewhat robust with this by following
these steps to defining the new predicate for inverted logic.
1. Take all supported targets and filter out all `cfg(unix)` ones as
these should all support `sys_common/net.rs`.
2. Take remaining targets and filter out `cfg(windows)` ones.
3. The remaining dozen-or-so targets were all audited by hand. Mostly
this included `target_os = "hermit"` and `target_os = "solid_asp3"`
which required an allow-list entry, but remaining targets were all
already excluded (didn't use `sys_common/net.rs` so they were left
out.
If this causes breakage it should be relatively easy to fix and I'd be
happy to follow-up with any PRs necessary.
Use `usize::repeat_u8` instead of implementing `repeat_byte` in `memchr.rs`
It's simpler that way and the tricks don't actually make a difference: https://godbolt.org/z/zrvYY1dGx
remove the memcpy-on-equal-ptrs assumption
One of the libc we support, musl, [defines `memcpy` with `restrict` pointers](https://git.musl-libc.org/cgit/musl/tree/src/string/memcpy.c#n5). This in fact matches the definition in the C standard. Calling that `memcpy` with overlapping pointers is clearly UB, who knows what the compiler did when optimizing this `memcpy` -- it certainly assumed source and destination to be disjoint.
Lucky enough, it does not seem like we actually need this assumption that `memcpy(p, p, n)` is always allowed. clang and GCC need it since they use `memcpy` to compile C assignments, but [we use memmove for similar code](https://godbolt.org/z/bcW85WYcM). There are no known cases where LLVM introduces calls to memcpy on equal pointers itself. (And if there were, that would be a soundness bug in rustc due to the musl issue mentioned above.)
This does mean we must make sure to never call the LLVM `memcpy` builtin on equal ranges even though the LangRef says that is allowed. Currently that is the case so we just need to make sure it remains the case. :) Cc `@rust-lang/opsem` `@rust-lang/wg-llvm`
Implement thread parking for xous
This follows the pattern set by [the Windows parker](ddef56d5df/library/std/src/sys/windows/thread_parking.rs) when it uses keyed events. An atomic variable is used to track the state and optimize the fast path, while notifications are send via the ticktime server to block and unblock the thread.
ping `@xobs`
`@rustbot` label +T-libs +A-atomic
r? libs
unify read_to_end and io::copy impls for reading into a Vec
This ports over the initial probe (to avoid allocation) and the dynamic read sizing from the io::copy specialization to the `default_read_to_end` implementation which already had its own optimizations for different cases.
I think it should be a best-of-both now.
suggested by `@a1phyr` in https://github.com/rust-lang/rust/pull/117576#issuecomment-1803408492
Expand in-place iteration specialization to Flatten, FlatMap and ArrayChunks
This enables the following cases to collect in-place:
```rust
let v = vec![[0u8; 4]; 1024]
let v: Vec<_> = v.into_iter().flatten().collect();
let v: Vec<Option<NonZeroUsize>> = vec![NonZeroUsize::new(0); 1024];
let v: Vec<_> = v.into_iter().flatten().collect();
let v = vec![u8; 4096];
let v: Vec<_> = v.into_iter().array_chunks::<4>().collect();
```
Especially the nicheful-option-flattening should be useful in real code.
