-
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
Stabilize const_type_id feature #72488
Conversation
(rust_highfive has picked a reviewer for you, use r? to override) |
I don't see any hazards from this. The result of cc @rust-lang/wg-const-eval |
Cc @rust-lang/lang because this is about const-stabilizing an intrinsic. |
What about collisions #10389? Currently this API is really dangerous to use and there is no way to detect collisions even without using dylibs. Not really related to const-stabilization but in current shape it's pretty useless both in const and in runtime. |
In my opinion, the fact that |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Unfortunately it's possible to hit collisions unintentionally.
Because there will be more ways to hit it. That's what I wanted to say. I'd rather see that feature deprecated and switch |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Ah looks like I’ve got a test to fix up:
|
☔ The latest upstream changes (presumably #73265) made this pull request unmergeable. Please resolve the merge conflicts. |
@bors r+ |
📌 Commit 9d4818c has been approved by |
☀️ Test successful - checks-actions, checks-azure |
Stability attribute is 1.46, has to be updated to 1.47. |
Just came across this, some of the people involved in this might be interested in #75923 (comment). |
Interesting, eddyb. FWIW I'd be fine with reverting this on beta under a "new information came in that made us not confident in the promise any more" justification. Wouldn't be the first rolled-back stabilization... |
See this discussion for why stabilization was likely premature, and we should take it back before the release. @KodrAus can you create the PR that reverts this, or should I? |
revert const_type_id stabilization This reverts rust-lang#72488, which is currently on beta and scheduled to stabilize in `1.47.0`, based on rust-lang#75923 (comment) It turns out we might not be quite ready to stabilize `TypeId` in const contexts before having a chance to rework its internals. Since `TypeId` is a bit of an oddity we want to be careful about how those internals are currently being relied on while making changes. That will be easier to do without having to also consider compile-time contexts. r? `@eddyb`
Pkgsrc changes: * Remove patches now integrated upstream, many related to SunOS / Illumos. * The LLVM fix for powerpc is also now integrated upstream. * Adapt those patches where the source has moved or parts are integrated. * The randomness patches no longer applies, and I could not find where those files went... * Provide a separate bootstrap for NetBSD/powerpc 9.0, since apparently the C++ ABI is different from 8.0. Yes, this appears to be specific to the NetBSD powerpc ports. Upstream changes: Version 1.47.0 (2020-10-08) ========================== Language -------- - [Closures will now warn when not used.][74869] Compiler -------- - [Stabilized the `-C control-flow-guard` codegen option][73893], which enables [Control Flow Guard][1.47.0-cfg] for Windows platforms, and is ignored on other platforms. - [Upgraded to LLVM 11.][73526] - [Added tier 3\* support for the `thumbv4t-none-eabi` target.][74419] - [Upgrade the FreeBSD toolchain to version 11.4][75204] - [`RUST_BACKTRACE`'s output is now more compact.][75048] \* Refer to Rust's [platform support page][forge-platform-support] for more information on Rust's tiered platform support. Libraries --------- - [`CStr` now implements `Index<RangeFrom<usize>>`.][74021] - [Traits in `std`/`core` are now implemented for arrays of any length, not just those of length less than 33.][74060] - [`ops::RangeFull` and `ops::Range` now implement Default.][73197] - [`panic::Location` now implements `Copy`, `Clone`, `Eq`, `Hash`, `Ord`, `PartialEq`, and `PartialOrd`.][73583] Stabilized APIs --------------- - [`Ident::new_raw`] - [`Range::is_empty`] - [`RangeInclusive::is_empty`] - [`Result::as_deref`] - [`Result::as_deref_mut`] - [`Vec::leak`] - [`pointer::offset_from`] - [`f32::TAU`] - [`f64::TAU`] The following previously stable APIs have now been made const. - [The `new` method for all `NonZero` integers.][73858] - [The `checked_add`,`checked_sub`,`checked_mul`,`checked_neg`, `checked_shl`, `checked_shr`, `saturating_add`, `saturating_sub`, and `saturating_mul` methods for all integers.][73858] - [The `checked_abs`, `saturating_abs`, `saturating_neg`, and `signum` for all signed integers.][73858] - [The `is_ascii_alphabetic`, `is_ascii_uppercase`, `is_ascii_lowercase`, `is_ascii_alphanumeric`, `is_ascii_digit`, `is_ascii_hexdigit`, `is_ascii_punctuation`, `is_ascii_graphic`, `is_ascii_whitespace`, and `is_ascii_control` methods for `char` and `u8`.][73858] Cargo ----- - [`build-dependencies` are now built with opt-level 0 by default.][cargo/8500] You can override this by setting the following in your `Cargo.toml`. ```toml [profile.release.build-override] opt-level = 3 ``` - [`cargo-help` will now display man pages for commands rather just the `--help` text.][cargo/8456] - [`cargo-metadata` now emits a `test` field indicating if a target has tests enabled.][cargo/8478] - [`workspace.default-members` now respects `workspace.exclude`.][cargo/8485] - [`cargo-publish` will now use an alternative registry by default if it's the only registry specified in `package.publish`.][cargo/8571] Misc ---- - [Added a help button beside Rustdoc's searchbar that explains rustdoc's type based search.][75366] - [Added the Ayu theme to rustdoc.][71237] Compatibility Notes ------------------- - [Bumped the minimum supported Emscripten version to 1.39.20.][75716] - [Fixed a regression parsing `{} && false` in tail expressions.][74650] - [Added changes to how proc-macros are expanded in `macro_rules!` that should help to preserve more span information.][73084] These changes may cause compiliation errors if your macro was unhygenic or didn't correctly handle `Delimiter::None`. - [Moved support for the CloudABI target to tier 3.][75568] - [`linux-gnu` targets now require minimum kernel 2.6.32 and glibc 2.11.][74163] Internal Only -------- - [Improved default settings for bootstrapping in `x.py`.][73964] You can read details about this change in the ["Changes to `x.py` defaults"](https://blog.rust-lang.org/inside-rust/2020/08/30/changes-to-x-py-defaults.html) post on the Inside Rust blog. - [Added the `rustc-docs` component.][75560] This allows you to install and read the documentation for the compiler internal APIs. (Currently only available for `x86_64-unknown-linux-gnu`.) [1.47.0-cfg]: https://docs.microsoft.com/en-us/windows/win32/secbp/control-flow-guard [76980]: rust-lang/rust#76980 [75048]: rust-lang/rust#75048 [74163]: rust-lang/rust#74163 [71237]: rust-lang/rust#71237 [74869]: rust-lang/rust#74869 [73858]: rust-lang/rust#73858 [75716]: rust-lang/rust#75716 [75908]: rust-lang/rust#75908 [75516]: rust-lang/rust#75516 [75560]: rust-lang/rust#75560 [75568]: rust-lang/rust#75568 [75366]: rust-lang/rust#75366 [75204]: rust-lang/rust#75204 [74650]: rust-lang/rust#74650 [74419]: rust-lang/rust#74419 [73964]: rust-lang/rust#73964 [74021]: rust-lang/rust#74021 [74060]: rust-lang/rust#74060 [73893]: rust-lang/rust#73893 [73526]: rust-lang/rust#73526 [73583]: rust-lang/rust#73583 [73084]: rust-lang/rust#73084 [73197]: rust-lang/rust#73197 [72488]: rust-lang/rust#72488 [cargo/8456]: rust-lang/cargo#8456 [cargo/8478]: rust-lang/cargo#8478 [cargo/8485]: rust-lang/cargo#8485 [cargo/8500]: rust-lang/cargo#8500 [cargo/8571]: rust-lang/cargo#8571 [`Ident::new_raw`]: https://doc.rust-lang.org/nightly/proc_macro/struct.Ident.html#method.new_raw [`Range::is_empty`]: https://doc.rust-lang.org/nightly/std/ops/struct.Range.html#method.is_empty [`RangeInclusive::is_empty`]: https://doc.rust-lang.org/nightly/std/ops/struct.RangeInclusive.html#method.is_empty [`Result::as_deref_mut`]: https://doc.rust-lang.org/nightly/std/result/enum.Result.html#method.as_deref_mut [`Result::as_deref`]: https://doc.rust-lang.org/nightly/std/result/enum.Result.html#method.as_deref [`TypeId::of`]: https://doc.rust-lang.org/nightly/std/any/struct.TypeId.html#method.of [`Vec::leak`]: https://doc.rust-lang.org/nightly/std/vec/struct.Vec.html#method.leak [`f32::TAU`]: https://doc.rust-lang.org/nightly/std/f32/consts/constant.TAU.html [`f64::TAU`]: https://doc.rust-lang.org/nightly/std/f64/consts/constant.TAU.html [`pointer::offset_from`]: https://doc.rust-lang.org/nightly/std/primitive.pointer.html#method.offset_from
The tracking issue for
const_type_id
points to the ill-fated #41875. So I'm re-energizingTypeId
shenanigans by opening this one up to see if there's anything blocking us from stabilizing the constification of type ids.Will wait for CI before pinging teams/groups.
This PR stabilizes the
const_type_id
feature, which allowsTypeId::of
(and the underlying unstable intrinsic) to be called in constant contexts.There are some sanity tests that demonstrate its usage, but I’ve included some more below.
As a simple example, you could create a constant item that contains some type ids:
Type ids can also now appear in associated constants. You could create a trait that associates each type with its constant type id:
TypeId::of
is generic, which we saw above in the way the genericSelf
argument was used. This has some implications for const evaluation. It means we can make trait impls evaluate differently depending on information that wasn't directly passed through the trait system. This violates the parametricity property, which requires all instances of a generic function to behave the same way with respect to its generic parameters. That's not unique toTypeId::of
, other generic const functions based on compiler intrinsics likemem::align_of
can also violate parametricity. In practice Rust doesn't really have type parametricity anyway since it monomorphizes generics into concrete functions, so violating it using type ids isn’t new.As an example of how impls can behave differently, you could combine constant type ids with the
const_if_match
feature to dispatch calls based on the type id of the genericSelf
, rather than based on information aboutSelf
that was threaded through trait bounds. It's like a rough-and-ready form of specialization:Type ids have some edges that this stabilization exposes to more contexts. It's possible for type ids to collide (but this is a bug). Since they can change between compiler versions, it's never valid to cast a type id to its underlying value.