-
Notifications
You must be signed in to change notification settings - Fork 12.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
make Instant::{duration_since, elapsed, sub}
saturating and remove workarounds
#89926
Conversation
(rust-highfive has picked a reviewer for you, use r? to override) |
This comment has been minimized.
This comment has been minimized.
03e1ed4
to
3145a12
Compare
If I have As mentioned,
I don't think anyone would expect that the two sides would be unequal in any situation. While I understand the motivation, I don't think breaking logical invariants (even if not explicitly stated) is a reasonable solution. |
I love the idea of making this change, and removing all the associated complexity and panics. I think in the common case people won't get two instants backwards, because they can call I don't think this would affect any invariants people rely on in practice. And people can always call |
Is there any reason why we might want to keep |
Currently there is no wrapping subtraction available on the |
☔ The latest upstream changes (presumably #88652) made this pull request unmergeable. Please resolve the merge conflicts. |
Initially, my main concern was that this silences potentially important failure modes. Namely, I can imagine someone relying on the fact that Changing algebraic laws also doesn't feel that great to me either. Technically they aren't documented, although a reasonable person might infer them from the panicking conditions of For those that understand this issue better than me: what is the cost of not doing this? |
|
I think this is the one that interests me to the most. Although I am perhaps wrong to find this one interesting. In particular---while again acknowledging that I don't work in such a domain---I would imagine that such a panic would be highly interesting to someone that wants a strong guarantee about monotonicity. That is, they might have written their code in a way where they want to see a panic if the underlying clock source didn't produce monotonic values. Maybe they should have used With that said, I am pretty much reasoning in the abstract here. I don't have enough experience to put my feet on the ground here. The silent failure mode does concern me, but maybe my abstract reasoning is too much of a stretch. If we have folks working with the guarantees of monotonic clocks in production that we could ping, getting their input on a change like this would probably be wise. |
0f1b815
to
777e339
Compare
As discussed in the libs-api meeting I have changed the methods to panic in debug mode (via overflow-checks) and saturate in release mode. Discussion on exactly which of |
I think we hit "time goes backwards in a VM" bug in NEAR: https://near.zulipchat.com/#narrow/stream/295302-dev-general/topic/general/near/248247028. So, for us, such a change would be very much welcome. However, there's a wrinkle -- we compile our code in
For Ideally, I think we shouldn't be piggybacking on To sum up:
|
If you're already hitting those panics that should mean that Also, that link would require signup to read it. Depending on how much control you have over where the Instant's are compared maybe you could selectively change the overflow-checks flag for those crates via profile overrides? |
Yeah, practically for near specifically I think we currently only run on
Oups, good catch! That's the usual Zulip's "login with GitHub" thing, we should set-up zulip archive. The short discussion here basically just links to tokio-rs/tokio#3619 |
☔ The latest upstream changes (presumably #90437) made this pull request unmergeable. Please resolve the merge conflicts. |
777e339
to
56a59c0
Compare
This comment has been minimized.
This comment has been minimized.
…askrgr Rollup of 9 pull requests Successful merges: - rust-lang#89926 (make `Instant::{duration_since, elapsed, sub}` saturating and remove workarounds) - rust-lang#90532 (More informative error message for E0015) - rust-lang#93810 (Improve chalk integration) - rust-lang#93851 (More practical examples for `Option::and_then` & `Result::and_then`) - rust-lang#93885 (bootstrap.py: Suggest disabling download-ci-llvm option if url fails to download) - rust-lang#93886 (Stabilise inherent_ascii_escape (FCP in rust-lang#77174)) - rust-lang#93930 (add link to format_args! when mention it in docs) - rust-lang#93936 (Couple of driver cleanups) - rust-lang#93944 (Don't relabel to a team if there is already a team label) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup
Seeing this on Rust Weekly News gave me a shock. So I just checked my code, and I don't use I'm trying to think of the cases where this could break. Maybe when two ordered events come in from an external source, e.g. from hardware or over a network stream, and two timers are set from those events, and then the timers execute in the opposite order, breaking the coder's assumptions. For this to have worked in the case of the previous Edit: Really it comes down to whether there are scenarios that would have worked with the previous implementation that won't work now. Thinking about it, in practice the time getting stuck on the same value for Z ms can also give unexpected problems. |
From the updated docs
From an API perspective Rust already relied on the OS guarantees to provide monotonic time. It additionally applies some workarounds that are only relevant in situations when operating system fails to uphold its contract. What changed are the workarounds. What doesn't change is that any violation of monotonicity already is a platform bug, i.e. running rust on a broken system. This is akin to, albeit less severe, a filesystem not upholding its durability guarantees for example. |
I understand that. However if a coder requires a strictly non-decreasing |
This is labeled |
IMO what is missing from this thread is any mention of pressuring vendors that have known broken VM/hypervisors in production (AWS). Hiding this is enabling their continued bad behavior. |
If it is a strict requirement then using a system that isn't broken should be preferable. As it is preferable to put some critical transactional database on storage which honors FLUSH CACHE commands rather than putting a on a drive that YOLOs things (and thus isn't spec-compliant).
Well, users did encounter panics and filed bug reports with Rust but not with AWS. Or if they did they went unheard I guess. If there's an interest in stress-testing clocks on new hardware this could be done with a standalone program, similar to memory or GPU stress testing programs. It doesn't have to happen in production software. |
3308: chore: fix instant panic on some case r=zhangsoledad a=driftluo ### What problem does this PR solve? ref: rust-lang/rust#89926 upgrade tokio/tentacle/parking_lot tokio: ``` ### Fixed - time: prevent panicking in `sleep` with large durations ([#4495]) - time: eliminate potential panics in `Instant` arithmetic on platforms where `Instant::now` is not monotonic ([#4461]) - io: fix `DuplexStream` not participating in cooperative yielding ([#4478]) - rt: fix potential double panic when dropping a `JoinHandle` ([#4430]) ### Changed - update minimum supported Rust version to 1.49 ([#4457]) - update `parking_lot` dependency to v0.12.0 ([#4459]) - update `mio` dependency to v0.8 ([#4449]) - rt: remove an unnecessary lock in the blocking pool ([#4436]) - rt: remove an unnecessary enum in the basic scheduler ([#4462]) - time: use bit manipulation instead of modulo to improve performance ([#4480]) - net: use `std::future::Ready` instead of our own `Ready` future ([#4271]) - replace deprecated `atomic::spin_loop_hint` with `hint::spin_loop` ([#4491]) - fix miri failures in intrusive linked lists ([#4397]) ``` ### Check List Tests - Unit test - Integration test ### Release note ```release-note Title Only: Include only the PR title in the release note. ``` Co-authored-by: driftluo <[email protected]>
As of rust-lang/rust#89926, this test behaves differently. We support rust versions before and after this change, so both behaviors are possible, so just disable this test for now.
As of rust-lang/rust#89926, this test behaves differently. We support rust versions before and after this change, so both behaviors are possible, so just disable this test for now.
Pkgsrc changes: * Bump available bootstraps to 1.59.0. * Delete one patch which no longer applies, adjust another. Upstream changes: Version 1.60.0 (2022-04-07) ========================== Language -------- - [Stabilize `#[cfg(panic = "...")]` for either `"unwind"` or `"abort"`.] [93658] - [Stabilize `#[cfg(target_has_atomic = "...")]` for each integer size and `"ptr"`.][93824] Compiler -------- - [Enable combining `+crt-static` and `relocation-model=pic` on `x86_64-unknown-linux-gnu`][86374] - [Fixes wrong `unreachable_pub` lints on nested and glob public reexport][87487] - [Stabilize `-Z instrument-coverage` as `-C instrument-coverage`][90132] - [Stabilize `-Z print-link-args` as `--print link-args`][91606] - [Add new Tier 3 target `mips64-openwrt-linux-musl`\*][92300] - [Add new Tier 3 target `armv7-unknown-linux-uclibceabi` (softfloat)\*][92383] - [Fix invalid removal of newlines from doc comments][92357] - [Add kernel target for RustyHermit][92670] - [Deny mixing bin crate type with lib crate types][92933] - [Make rustc use `RUST_BACKTRACE=full` by default][93566] - [Upgrade to LLVM 14][93577] \* Refer to Rust's [platform support page][platform-support-doc] for more information on Rust's tiered platform support. Libraries --------- - [Guarantee call order for `sort_by_cached_key`][89621] - [Improve `Duration::try_from_secs_f32`/`f64` accuracy by directly processing exponent and mantissa][90247] - [Make `Instant::{duration_since, elapsed, sub}` saturating][89926] - [Remove non-monotonic clocks workarounds in `Instant::now`][89926] - [Make `BuildHasherDefault`, `iter::Empty` and `future::Pending` covariant][92630] Stabilized APIs --------------- - [`Arc::new_cyclic`][arc_new_cyclic] - [`Rc::new_cyclic`][rc_new_cyclic] - [`slice::EscapeAscii`][slice_escape_ascii] - [`<[u8]>::escape_ascii`][slice_u8_escape_ascii] - [`u8::escape_ascii`][u8_escape_ascii] - [`Vec::spare_capacity_mut`][vec_spare_capacity_mut] - [`MaybeUninit::assume_init_drop`][assume_init_drop] - [`MaybeUninit::assume_init_read`][assume_init_read] - [`i8::abs_diff`][i8_abs_diff] - [`i16::abs_diff`][i16_abs_diff] - [`i32::abs_diff`][i32_abs_diff] - [`i64::abs_diff`][i64_abs_diff] - [`i128::abs_diff`][i128_abs_diff] - [`isize::abs_diff`][isize_abs_diff] - [`u8::abs_diff`][u8_abs_diff] - [`u16::abs_diff`][u16_abs_diff] - [`u32::abs_diff`][u32_abs_diff] - [`u64::abs_diff`][u64_abs_diff] - [`u128::abs_diff`][u128_abs_diff] - [`usize::abs_diff`][usize_abs_diff] - [`Display for io::ErrorKind`][display_error_kind] - [`From<u8> for ExitCode`][from_u8_exit_code] - [`Not for !` (the "never" type)][not_never] - [_Op_`Assign<$t> for Wrapping<$t>`][wrapping_assign_ops] - [`arch::is_aarch64_feature_detected!`][is_aarch64_feature_detected] Cargo ----- - [Port cargo from `toml-rs` to `toml_edit`][cargo/10086] - [Stabilize `-Ztimings` as `--timings`][cargo/10245] - [Stabilize namespaced and weak dependency features.][cargo/10269] - [Accept more `cargo:rustc-link-arg-*` types from build script output.][cargo/10274] - [cargo-new should not add ignore rule on Cargo.lock inside subdirs][cargo/10379] Misc ---- - [Ship docs on Tier 2 platforms by reusing the closest Tier 1 platform docs][92800] - [Drop rustc-docs from complete profile][93742] - [bootstrap: tidy up flag handling for llvm build][93918] Compatibility Notes ------------------- - [Remove compiler-rt linking hack on Android][83822] - [Mitigations for platforms with non-monotonic clocks have been removed from `Instant::now`][89926]. On platforms that don't provide monotonic clocks, an instant is not guaranteed to be greater than an earlier instant anymore. - [`Instant::{duration_since, elapsed, sub}` do not panic anymore on underflow, saturating to `0` instead][89926]. In the real world the panic happened mostly on platforms with buggy monotonic clock implementations rather than catching programming errors like reversing the start and end times. Such programming errors will now results in `0` rather than a panic. - In a future release we're planning to increase the baseline requirements for the Linux kernel to version 3.2, and for glibc to version 2.17. We'd love your feedback in [PR #95026][95026]. Internal Changes ---------------- These changes provide no direct user facing benefits, but represent significant improvements to the internals and overall performance of rustc and related tools. - [Switch all libraries to the 2021 edition][92068] [83822]: rust-lang/rust#83822 [86374]: rust-lang/rust#86374 [87487]: rust-lang/rust#87487 [89621]: rust-lang/rust#89621 [89926]: rust-lang/rust#89926 [90132]: rust-lang/rust#90132 [90247]: rust-lang/rust#90247 [91606]: rust-lang/rust#91606 [92068]: rust-lang/rust#92068 [92300]: rust-lang/rust#92300 [92357]: rust-lang/rust#92357 [92383]: rust-lang/rust#92383 [92630]: rust-lang/rust#92630 [92670]: rust-lang/rust#92670 [92800]: rust-lang/rust#92800 [92933]: rust-lang/rust#92933 [93566]: rust-lang/rust#93566 [93577]: rust-lang/rust#93577 [93658]: rust-lang/rust#93658 [93742]: rust-lang/rust#93742 [93824]: rust-lang/rust#93824 [93918]: rust-lang/rust#93918 [95026]: rust-lang/rust#95026 [cargo/10086]: rust-lang/cargo#10086 [cargo/10245]: rust-lang/cargo#10245 [cargo/10269]: rust-lang/cargo#10269 [cargo/10274]: rust-lang/cargo#10274 [cargo/10379]: rust-lang/cargo#10379 [arc_new_cyclic]: https://doc.rust-lang.org/stable/std/sync/struct.Arc.html#method.new_cyclic [rc_new_cyclic]: https://doc.rust-lang.org/stable/std/rc/struct.Rc.html#method.new_cyclic [slice_escape_ascii]: https://doc.rust-lang.org/stable/std/slice/struct.EscapeAscii.html [slice_u8_escape_ascii]: https://doc.rust-lang.org/stable/std/primitive.slice.html#method.escape_ascii [u8_escape_ascii]: https://doc.rust-lang.org/stable/std/primitive.u8.html#method.escape_ascii [vec_spare_capacity_mut]: https://doc.rust-lang.org/stable/std/vec/struct.Vec.html#method.spare_capacity_mut [assume_init_drop]: https://doc.rust-lang.org/stable/std/mem/union.MaybeUninit.html#method.assume_init_drop [assume_init_read]: https://doc.rust-lang.org/stable/std/mem/union.MaybeUninit.html#method.assume_init_read [i8_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i8.html#method.abs_diff [i16_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i16.html#method.abs_diff [i32_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i32.html#method.abs_diff [i64_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i64.html#method.abs_diff [i128_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i128.html#method.abs_diff [isize_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.isize.html#method.abs_diff [u8_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u8.html#method.abs_diff [u16_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u16.html#method.abs_diff [u32_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u32.html#method.abs_diff [u64_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u64.html#method.abs_diff [u128_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u128.html#method.abs_diff [usize_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.usize.html#method.abs_diff [display_error_kind]: https://doc.rust-lang.org/stable/std/io/enum.ErrorKind.html#impl-Display [from_u8_exit_code]: https://doc.rust-lang.org/stable/std/process/struct.ExitCode.html#impl-From%3Cu8%3E [not_never]: https://doc.rust-lang.org/stable/std/primitive.never.html#impl-Not [wrapping_assign_ops]: https://doc.rust-lang.org/stable/std/num/struct.Wrapping.html#trait-implementations [is_aarch64_feature_detected]: https://doc.rust-lang.org/stable/std/arch/macro.is_aarch64_feature_detected.html
Pkgsrc changes: * Bump available bootstraps to 1.59.0. * Adjust line number in patches which had non-zero offsets. Upstream changes: Version 1.60.0 (2022-04-07) =========================== Language -------- - [Stabilize `#[cfg(panic = "...")]` for either `"unwind"` or `"abort"`.] [93658] - [Stabilize `#[cfg(target_has_atomic = "...")]` for each integer size and `"ptr"`.][93824] Compiler -------- - [Enable combining `+crt-static` and `relocation-model=pic` on `x86_64-unknown-linux-gnu`][86374] - [Fixes wrong `unreachable_pub` lints on nested and glob public reexport][87487] - [Stabilize `-Z instrument-coverage` as `-C instrument-coverage`][90132] - [Stabilize `-Z print-link-args` as `--print link-args`][91606] - [Add new Tier 3 target `mips64-openwrt-linux-musl`\*][92300] - [Add new Tier 3 target `armv7-unknown-linux-uclibceabi` (softfloat)\*][92383] - [Fix invalid removal of newlines from doc comments][92357] - [Add kernel target for RustyHermit][92670] - [Deny mixing bin crate type with lib crate types][92933] - [Make rustc use `RUST_BACKTRACE=full` by default][93566] - [Upgrade to LLVM 14][93577] \* Refer to Rust's [platform support page][platform-support-doc] for more information on Rust's tiered platform support. Libraries --------- - [Guarantee call order for `sort_by_cached_key`][89621] - [Improve `Duration::try_from_secs_f32`/`f64` accuracy by directly processing exponent and mantissa][90247] - [Make `Instant::{duration_since, elapsed, sub}` saturating][89926] - [Remove non-monotonic clocks workarounds in `Instant::now`][89926] - [Make `BuildHasherDefault`, `iter::Empty` and `future::Pending` covariant][92630] Stabilized APIs --------------- - [`Arc::new_cyclic`][arc_new_cyclic] - [`Rc::new_cyclic`][rc_new_cyclic] - [`slice::EscapeAscii`][slice_escape_ascii] - [`<[u8]>::escape_ascii`][slice_u8_escape_ascii] - [`u8::escape_ascii`][u8_escape_ascii] - [`Vec::spare_capacity_mut`][vec_spare_capacity_mut] - [`MaybeUninit::assume_init_drop`][assume_init_drop] - [`MaybeUninit::assume_init_read`][assume_init_read] - [`i8::abs_diff`][i8_abs_diff] - [`i16::abs_diff`][i16_abs_diff] - [`i32::abs_diff`][i32_abs_diff] - [`i64::abs_diff`][i64_abs_diff] - [`i128::abs_diff`][i128_abs_diff] - [`isize::abs_diff`][isize_abs_diff] - [`u8::abs_diff`][u8_abs_diff] - [`u16::abs_diff`][u16_abs_diff] - [`u32::abs_diff`][u32_abs_diff] - [`u64::abs_diff`][u64_abs_diff] - [`u128::abs_diff`][u128_abs_diff] - [`usize::abs_diff`][usize_abs_diff] - [`Display for io::ErrorKind`][display_error_kind] - [`From<u8> for ExitCode`][from_u8_exit_code] - [`Not for !` (the "never" type)][not_never] - [_Op_`Assign<$t> for Wrapping<$t>`][wrapping_assign_ops] - [`arch::is_aarch64_feature_detected!`][is_aarch64_feature_detected] Cargo ----- - [Port cargo from `toml-rs` to `toml_edit`][cargo/10086] - [Stabilize `-Ztimings` as `--timings`][cargo/10245] - [Stabilize namespaced and weak dependency features.][cargo/10269] - [Accept more `cargo:rustc-link-arg-*` types from build script output.][cargo/10274] - [cargo-new should not add ignore rule on Cargo.lock inside subdirs][cargo/10379] Misc ---- - [Ship docs on Tier 2 platforms by reusing the closest Tier 1 platform docs][92800] - [Drop rustc-docs from complete profile][93742] - [bootstrap: tidy up flag handling for llvm build][93918] Compatibility Notes ------------------- - [Remove compiler-rt linking hack on Android][83822] - [Mitigations for platforms with non-monotonic clocks have been removed from `Instant::now`][89926]. On platforms that don't provide monotonic clocks, an instant is not guaranteed to be greater than an earlier instant anymore. - [`Instant::{duration_since, elapsed, sub}` do not panic anymore on underflow, saturating to `0` instead][89926]. In the real world the panic happened mostly on platforms with buggy monotonic clock implementations rather than catching programming errors like reversing the start and end times. Such programming errors will now results in `0` rather than a panic. - In a future release we're planning to increase the baseline requirements for the Linux kernel to version 3.2, and for glibc to version 2.17. We'd love your feedback in [PR #95026][95026]. Internal Changes ---------------- These changes provide no direct user facing benefits, but represent significant improvements to the internals and overall performance of rustc and related tools. - [Switch all libraries to the 2021 edition][92068] [83822]: rust-lang/rust#83822 [86374]: rust-lang/rust#86374 [87487]: rust-lang/rust#87487 [89621]: rust-lang/rust#89621 [89926]: rust-lang/rust#89926 [90132]: rust-lang/rust#90132 [90247]: rust-lang/rust#90247 [91606]: rust-lang/rust#91606 [92068]: rust-lang/rust#92068 [92300]: rust-lang/rust#92300 [92357]: rust-lang/rust#92357 [92383]: rust-lang/rust#92383 [92630]: rust-lang/rust#92630 [92670]: rust-lang/rust#92670 [92800]: rust-lang/rust#92800 [92933]: rust-lang/rust#92933 [93566]: rust-lang/rust#93566 [93577]: rust-lang/rust#93577 [93658]: rust-lang/rust#93658 [93742]: rust-lang/rust#93742 [93824]: rust-lang/rust#93824 [93918]: rust-lang/rust#93918 [95026]: rust-lang/rust#95026 [cargo/10086]: rust-lang/cargo#10086 [cargo/10245]: rust-lang/cargo#10245 [cargo/10269]: rust-lang/cargo#10269 [cargo/10274]: rust-lang/cargo#10274 [cargo/10379]: rust-lang/cargo#10379 [arc_new_cyclic]: https://doc.rust-lang.org/stable/std/sync/struct.Arc.html#method.new_cyclic [rc_new_cyclic]: https://doc.rust-lang.org/stable/std/rc/struct.Rc.html#method.new_cyclic [slice_escape_ascii]: https://doc.rust-lang.org/stable/std/slice/struct.EscapeAscii.html [slice_u8_escape_ascii]: https://doc.rust-lang.org/stable/std/primitive.slice.html#method.escape_ascii [u8_escape_ascii]: https://doc.rust-lang.org/stable/std/primitive.u8.html#method.escape_ascii [vec_spare_capacity_mut]: https://doc.rust-lang.org/stable/std/vec/struct.Vec.html#method.spare_capacity_mut [assume_init_drop]: https://doc.rust-lang.org/stable/std/mem/union.MaybeUninit.html#method.assume_init_drop [assume_init_read]: https://doc.rust-lang.org/stable/std/mem/union.MaybeUninit.html#method.assume_init_read [i8_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i8.html#method.abs_diff [i16_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i16.html#method.abs_diff [i32_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i32.html#method.abs_diff [i64_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i64.html#method.abs_diff [i128_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.i128.html#method.abs_diff [isize_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.isize.html#method.abs_diff [u8_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u8.html#method.abs_diff [u16_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u16.html#method.abs_diff [u32_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u32.html#method.abs_diff [u64_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u64.html#method.abs_diff [u128_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.u128.html#method.abs_diff [usize_abs_diff]: https://doc.rust-lang.org/stable/std/primitive.usize.html#method.abs_diff [display_error_kind]: https://doc.rust-lang.org/stable/std/io/enum.ErrorKind.html#impl-Display [from_u8_exit_code]: https://doc.rust-lang.org/stable/std/process/struct.ExitCode.html#impl-From%3Cu8%3E [not_never]: https://doc.rust-lang.org/stable/std/primitive.never.html#impl-Not [wrapping_assign_ops]: https://doc.rust-lang.org/stable/std/num/struct.Wrapping.html#trait-implementations [is_aarch64_feature_detected]: https://doc.rust-lang.org/stable/std/arch/macro.is_aarch64_feature_detected.html
As of rust-lang/rust#89926, this test behaves differently. We support rust versions before and after this change, so both behaviors are possible, so just disable this test for now.
This removes all mutex/atomic-based workarounds for non-monotonic clocks and makes the previously panicking methods saturating instead. Additionally
saturating_duration_since
becomes deprecated sinceduration_since
now fills that role.Effectively this moves the fixup from
Instant
construction to the comparisons.This has some observable effects, especially on platforms without monotonic clocks:
checked_duration_since
will now returnNone
in more cases. Previously it only happened when one compared instants obtained in the wrong order or manually created ones. Now it also does on backslides.b.duration_since(a) + c.duration_since(b) != c.duration_since(a)
The upsides are reduced complexity and lower overhead of
Instant::now
.Motivation
Currently we must choose between two poisons. One is high worst-case latency and jitter of
Instant::now()
due to explicit synchronization; see #83093 for benchmarks, the worst-case overhead is > 100x. The other is sporadic panics on specific, rare combinations of CPU/hypervisor/operating system due to platform bugs.Use-cases where low-overhead, fine-grained timestamps are needed - such as syscall tracing, performance profiles or sensor data acquisition (drone flight controllers were mentioned in a libs meeting) in multi-threaded programs - are negatively impacted by the synchronization.
The panics are user-visible (program crashes), hard to reproduce and can be triggered by any dependency that might be using Instants for any reason.
A solution that is fast and doesn't panic is desirable.
closes #84448
closes #86470