Rename `~const Drop` to `~const Destruct` #94901

fee1-dead · 2022-03-13T04:11:34Z

Migration notes: We have renamed ~const Drop to ~const std::marker::Destruct. When you bump the pinned nightly version, please make sure that all usages of ~const Drop are changed to ~const Destruct!

r? @oli-obk

Completely switching to ~const Destructible would be rather complicated, so it seems best to add it for now and wait for it to be backported to beta in the next release.

The rationale is to prevent complications such as #92149 and #94803 by introducing an entirely new trait. And ~const Destructible reads a bit better than ~const Drop. Name Bikesheddable.

fee1-dead · 2022-03-13T04:14:01Z

wait, even if I just add this as a lang item now it wouldn't work because the logic for handling them would still need to be backported... I guess we could go with a compatible approach by ignoring ~const Drop first and do something like ~const Drop + ~const Destructible

@rustbot author

oli-obk · 2022-03-13T11:17:24Z

You could implement Destructible for all T on bootstrap

fee1-dead · 2022-03-13T11:18:59Z

yeah that works, which means we could land this now and I will implement the logic later

shamatar · 2022-03-13T12:42:49Z

If the purpose of this trait to mark the types that can be trivially dropped for the purposes of CTFE, shouldn't it be "auto" trait or some other special "const auto" trait, or may be simultaneously marker trait (never manually implementable, but not sure about it)?

fee1-dead · 2022-03-13T12:50:12Z

Yeah if we can implement it for all T in bootstrap it wouldn't make sense to just land this now. I will prepare an implementation when I have time.

fmease · 2022-03-14T14:05:08Z

Minor bikeshedding: In a lot of cases, trait names use the imperative form of a verb (Copy, Send, Unsize, Borrow, Display, CoerceUnsized) instead of an adjective ending in -able or -ible. The latter is quite discouraged in Rust AFAIK. I suggest renaming Destructible to Destruct while still possible.

clarfonthey · 2022-03-15T02:36:08Z

Is there ever going to be a case where types don't implement Destructible, and only that they might not be const-destructible?

fee1-dead · 2022-03-15T05:16:04Z

Don't think so. The rationale behind the use of the word destructible is as another word for drop, where any type can be dropped/destructed in runtime, but not in compile time.

clarfonthey · 2022-03-17T03:01:08Z

My main concern is that it seems like this is just a renamed version of ~const Drop. Like, I get that it will internally work differently, but I fail to see the main differences for an end user, other than confusion.

fee1-dead · 2022-03-18T10:15:39Z

~const Drop is as confusing as any other names. If we use another trait we can write the documentation to be specifically about dropping a type in compile time and say that it is implemented non-const-ly for all T. The main rationale is to prevent any more special casing to make it not conflict with : Drop bounds.

oli-obk · 2022-03-21T11:12:16Z

compiler/rustc_trait_selection/src/traits/select/confirmation.rs

-        if obligation.param_env.constness() == Constness::NotConst {
-            return Ok(ImplSourceConstDropData { nested: vec![] });
+    ) -> Result<ImplSourceConstDestructData<PredicateObligation<'tcx>>, SelectionError<'tcx>> {
+        // `~const Destruct` in a non-const environment is always trivially true, since our type is `Drop`


Well... for now it's true because all types implement Destruct. Unless we get those weird kind of types that may never get destructed, outside of constness all types implement Destruct

oli-obk · 2022-03-21T12:01:59Z

compiler/rustc_trait_selection/src/traits/select/confirmation.rs

@@ -1054,9 +1062,29 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
        let cause = obligation.derived_cause(BuiltinDerivedObligation);

        // If we have a custom `impl const Drop`, then
-        // first check it like a regular impl candidate
+        // first check it like a regular impl candidate.
+        // This is copied from confirm_impl_candidate but remaps the predicate to `~const Drop` beforehand.


What does this code make work that wouldn't otherwise?

So.. basically the nested obligation causes must be correctly ~const Destruct, but we want the obligation that is matched to actually be ~const Drop such that match_impl wouldn't fail.

oli-obk · 2022-03-23T12:13:57Z

@bors r+

bors · 2022-03-23T12:13:59Z

📌 Commit fe5b813 has been approved by oli-obk

bors · 2022-03-23T14:04:46Z

⌛ Testing commit fe5b813 with merge 9280445...

bors · 2022-03-23T16:53:41Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing 9280445 to master...

bors · 2022-03-23T16:55:38Z

☀️ Test successful - checks-actions
Approved by: oli-obk
Pushing 9280445 to master...

rust-timer · 2022-03-23T18:28:56Z

Finished benchmarking commit (9280445): comparison url.

Summary: This benchmark run did not return any relevant results.

If you disagree with this performance assessment, please file an issue in rust-lang/rustc-perf.

@rustbot label: -perf-regression

This change was driven by [rust-lang/rust#94901][1]. It's actually more than just renaming - it fixes the irregular meaning of `~const Drop` (“the type can be destructed in a const context”, instead of the regular meaning of “`<T as Drop>::drop is `const fn`”), which the compiler was unable to reconcile in `Drop` bounds and thus necessitated the work- around explained in `[ref:drop_const_bounds]`. This is no longer an issue because `~const Destruct` has a regular meaning. [1]: rust-lang/rust#94901

This change was driven by [rust-lang/rust#94901][1]. It's actually more than just renaming - it fixes the irregular meaning of `~const Drop` (“the type can be destructed in a const context”, instead of the regular meaning of “`<T as Drop>::drop is `const fn`”), which the compiler was unable to reconcile in `Drop` bounds and thus necessitated for us to implement the work-around explained in `[ref:drop_const_bounds]`. This is no longer an issue because `~const Destruct` has a regular meaning. [1]: rust-lang/rust#94901

rust-highfive assigned oli-obk Mar 13, 2022

rustbot added the T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. label Mar 13, 2022

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Mar 13, 2022

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Mar 13, 2022

fee1-dead force-pushed the destructable branch from 9fc7cb3 to dff3984 Compare March 20, 2022 11:12

fee1-dead changed the title ~~Add Destructible trait for replacing ~const Drop bounds in the future~~ Rename ~const Drop to ~const Destruct Mar 21, 2022

fee1-dead added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Mar 21, 2022

This comment has been minimized.

Sign in to view

fee1-dead added 2 commits March 21, 2022 17:04

Add Destructible for replacing ~const Drop

4df2a28

Rename ~const Drop to ~const Destruct

1f3ee7f

fee1-dead force-pushed the destructable branch from a8b0960 to 1f3ee7f Compare March 21, 2022 06:04

Tidy up

fe5b813

oli-obk requested changes Mar 21, 2022

View reviewed changes

fee1-dead requested a review from oli-obk March 23, 2022 09:47

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Mar 23, 2022

bors added the merged-by-bors This PR was explicitly merged by bors. label Mar 23, 2022

bors merged commit 9280445 into rust-lang:master Mar 23, 2022

This was referenced Mar 23, 2022

Fix ice when error reporting recursion errors #94391

Merged

Relax Allocator bounds into pin-safe trait #94114

Closed

rustbot added this to the 1.61.0 milestone Mar 23, 2022

This was referenced Mar 23, 2022

Properly track ImplObligations #91030

Merged

Relax A: 'static bound for boxed Pin APIs #90822

Closed

Implement #[deprecated_safe] #95025

Closed

Add PeekableIterator trait as an experiment #95195

Closed

fee1-dead deleted the destructable branch March 24, 2022 07:24

fee1-dead mentioned this pull request Mar 24, 2022

New nightly r3-os/r3#5

Closed

fee1-dead mentioned this pull request Apr 13, 2022

~const Drop bounds are broken / wrong / unsound #94803

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Rename `~const Drop` to `~const Destruct` #94901

Rename `~const Drop` to `~const Destruct` #94901

fee1-dead commented Mar 13, 2022 •

edited

Loading

fee1-dead commented Mar 13, 2022 •

edited

Loading

oli-obk commented Mar 13, 2022

fee1-dead commented Mar 13, 2022

shamatar commented Mar 13, 2022

fee1-dead commented Mar 13, 2022

fmease commented Mar 14, 2022 •

edited

Loading

clarfonthey commented Mar 15, 2022

fee1-dead commented Mar 15, 2022

clarfonthey commented Mar 17, 2022

fee1-dead commented Mar 18, 2022

This comment has been minimized.

oli-obk Mar 21, 2022

oli-obk Mar 21, 2022

fee1-dead Mar 21, 2022

oli-obk commented Mar 23, 2022

bors commented Mar 23, 2022

bors commented Mar 23, 2022

bors commented Mar 23, 2022

bors commented Mar 23, 2022

rust-timer commented Mar 23, 2022

Rename ~const Drop to ~const Destruct #94901

Rename ~const Drop to ~const Destruct #94901

Conversation

fee1-dead commented Mar 13, 2022 • edited Loading

fee1-dead commented Mar 13, 2022 • edited Loading

oli-obk commented Mar 13, 2022

fee1-dead commented Mar 13, 2022

shamatar commented Mar 13, 2022

fee1-dead commented Mar 13, 2022

fmease commented Mar 14, 2022 • edited Loading

clarfonthey commented Mar 15, 2022

fee1-dead commented Mar 15, 2022

clarfonthey commented Mar 17, 2022

fee1-dead commented Mar 18, 2022

This comment has been minimized.

oli-obk Mar 21, 2022

Choose a reason for hiding this comment

oli-obk Mar 21, 2022

Choose a reason for hiding this comment

fee1-dead Mar 21, 2022

Choose a reason for hiding this comment

oli-obk commented Mar 23, 2022

bors commented Mar 23, 2022

bors commented Mar 23, 2022

bors commented Mar 23, 2022

bors commented Mar 23, 2022

rust-timer commented Mar 23, 2022

Rename `~const Drop` to `~const Destruct` #94901

Rename `~const Drop` to `~const Destruct` #94901

fee1-dead commented Mar 13, 2022 •

edited

Loading

fee1-dead commented Mar 13, 2022 •

edited

Loading

fmease commented Mar 14, 2022 •

edited

Loading