Always create elided lifetime parameters for functions #97720

cjgillot · 2022-06-03T23:36:04Z

Anonymous and elided lifetimes in functions are sometimes (async fns) --and sometimes not (regular fns)-- desugared to implicit generic parameters.

This difference of treatment makes it some downstream analyses more complicated to handle. This step is a pre-requisite to perform lifetime elision resolution on AST.

There is currently an inconsistency in the treatment of argument-position impl-trait for functions and async fns:

trait Foo<'a> {}
fn foo(t: impl Foo<'_>) {} //~ ERROR missing lifetime specifier
async fn async_foo(t: impl Foo<'_>) {} //~ OK
fn bar(t: impl Iterator<Item = &'_ u8>) {} //~ ERROR missing lifetime specifier
async fn async_bar(t: impl Iterator<Item = &'_ u8>) {} //~ OK

The current implementation reports "missing lifetime specifier" on foo, but accepts it in async_foo.
This PR proposes to accept the anonymous lifetime in both cases as an extra generic lifetime parameter.
This change would be insta-stable, so let's ping t-lang.
Anonymous lifetimes in GAT bindings keep being forbidden:

fn foo(t: impl Foo<Assoc<'_> = Bar<'_>>) {}
                         ^^        ^^
                       forbidden   ok

I started a discussion here: https://rust-lang.zulipchat.com/#narrow/stream/213817-t-lang/topic/Anonymous.20lifetimes.20in.20universal.20impl-trait/near/284968606

r? @petrochenkov

cjgillot · 2022-06-05T08:19:29Z

@rust-lang/lang

This PR proposes to accept having anonymous lifetimes inside argument-position impl-trait, and to treat them as an extra generic lifetime parameter on the function.

There is currently an inconsistency in the treatment of argument-position impl-trait for functions and async fns:

trait Foo<'a> {}
fn foo(t: impl Foo<'_>) {} //~ ERROR missing lifetime specifier
async fn async_foo(t: impl Foo<'_>) {} //~ OK (since 1.39.0)

The semantics would be:

fn foo<'1, _T>(t: _T) where _T: Foo<'1> {}

Anonymous lifetimes in GAT bindings keep being forbidden:

fn foo(t: impl Foo<Assoc<'_> = Bar<'_>>) {}
                         ^^        ^^
                         ERROR     OK

I started a discussion here: https://rust-lang.zulipchat.com/#narrow/stream/213817-t-lang/topic/Anonymous.20lifetimes.20in.20universal.20impl-trait/near/284968606
where I expose what I understood of the anonymous lifetime resolution algorithm.

petrochenkov · 2022-06-05T14:10:43Z

I've added an example with '_ in associated item constraints to the PR description, otherwise it wasn't clear whether it was accepted previously or not.

src/test/ui/static/static-reference-to-fn-2.stderr

compiler/rustc_ast_lowering/src/lib.rs

compiler/rustc_resolve/src/late.rs

petrochenkov · 2022-06-05T16:50:24Z

(I still need to review some parts of compiler/rustc_resolve/src/late.rs, will finish a bit later.)

compiler/rustc_resolve/src/late.rs

petrochenkov · 2022-07-13T14:35:20Z

@bors r+

bors · 2022-07-13T14:35:22Z

📌 Commit 5a20834 has been approved by petrochenkov

It is now in the queue for this repository.

klensy · 2022-07-13T18:11:20Z

compiler/rustc_feature/src/active.rs

@@ -148,6 +148,8 @@ declare_features! (
    /// below (it has to be checked before expansion possibly makes
    /// macros disappear).
    (active, allow_internal_unstable, "1.0.0", None, None),
+    /// Allows using anonymous lifetimes in argument-position impl-trait.
+    (active, anonymous_lifetime_in_impl_trait, "1.63.0", None, None),


Always create elided lifetime parameters for functions Anonymous and elided lifetimes in functions are sometimes (async fns) --and sometimes not (regular fns)-- desugared to implicit generic parameters. This difference of treatment makes it some downstream analyses more complicated to handle. This step is a pre-requisite to perform lifetime elision resolution on AST. There is currently an inconsistency in the treatment of argument-position impl-trait for functions and async fns: ```rust trait Foo<'a> {} fn foo(t: impl Foo<'_>) {} //~ ERROR missing lifetime specifier async fn async_foo(t: impl Foo<'_>) {} //~ OK fn bar(t: impl Iterator<Item = &'_ u8>) {} //~ ERROR missing lifetime specifier async fn async_bar(t: impl Iterator<Item = &'_ u8>) {} //~ OK ``` The current implementation reports "missing lifetime specifier" on `foo`, but **accepts it** in `async_foo`. This PR **proposes to accept** the anonymous lifetime in both cases as an extra generic lifetime parameter. This change would be insta-stable, so let's ping t-lang. Anonymous lifetimes in GAT bindings keep being forbidden: ```rust fn foo(t: impl Foo<Assoc<'_> = Bar<'_>>) {} ^^ ^^ forbidden ok ``` I started a discussion here: https://rust-lang.zulipchat.com/#narrow/stream/213817-t-lang/topic/Anonymous.20lifetimes.20in.20universal.20impl-trait/near/284968606 r? `@petrochenkov`

Rollup of 5 pull requests Successful merges: - rust-lang#97720 (Always create elided lifetime parameters for functions) - rust-lang#98315 (Stabilize `core::ffi:c_*` and rexport in `std::ffi`) - rust-lang#98705 (Implement `for<>` lifetime binder for closures) - rust-lang#99126 (remove allow(rustc::potential_query_instability) in rustc_span) - rust-lang#99139 (Give a better error when `x dist` fails for an optional tool) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup

Always create elided lifetime parameters for functions Anonymous and elided lifetimes in functions are sometimes (async fns) --and sometimes not (regular fns)-- desugared to implicit generic parameters. This difference of treatment makes it some downstream analyses more complicated to handle. This step is a pre-requisite to perform lifetime elision resolution on AST. There is currently an inconsistency in the treatment of argument-position impl-trait for functions and async fns: ```rust trait Foo<'a> {} fn foo(t: impl Foo<'_>) {} //~ ERROR missing lifetime specifier async fn async_foo(t: impl Foo<'_>) {} //~ OK fn bar(t: impl Iterator<Item = &'_ u8>) {} //~ ERROR missing lifetime specifier async fn async_bar(t: impl Iterator<Item = &'_ u8>) {} //~ OK ``` The current implementation reports "missing lifetime specifier" on `foo`, but **accepts it** in `async_foo`. This PR **proposes to accept** the anonymous lifetime in both cases as an extra generic lifetime parameter. This change would be insta-stable, so let's ping t-lang. Anonymous lifetimes in GAT bindings keep being forbidden: ```rust fn foo(t: impl Foo<Assoc<'_> = Bar<'_>>) {} ^^ ^^ forbidden ok ``` I started a discussion here: https://rust-lang.zulipchat.com/#narrow/stream/213817-t-lang/topic/Anonymous.20lifetimes.20in.20universal.20impl-trait/near/284968606 r? ``@petrochenkov``

Rollup of 5 pull requests Successful merges: - rust-lang#97720 (Always create elided lifetime parameters for functions) - rust-lang#98315 (Stabilize `core::ffi:c_*` and rexport in `std::ffi`) - rust-lang#98705 (Implement `for<>` lifetime binder for closures) - rust-lang#99126 (remove allow(rustc::potential_query_instability) in rustc_span) - rust-lang#99139 (Give a better error when `x dist` fails for an optional tool) Failed merges: r? `@ghost` `@rustbot` modify labels: rollup

…henkov Resolve function lifetime elision on the AST ~Based on rust-lang#97720 Lifetime elision for functions is purely syntactic in nature, so can be resolved on the AST. This PR replicates the elision logic and diagnostics on the AST, and replaces HIR-based resolution by a `delay_span_bug`. This refactor allows for more consistent diagnostics, which don't have to guess the original code from HIR. r? `@petrochenkov`

Resolve function lifetime elision on the AST ~Based on rust-lang/rust#97720 Lifetime elision for functions is purely syntactic in nature, so can be resolved on the AST. This PR replicates the elision logic and diagnostics on the AST, and replaces HIR-based resolution by a `delay_span_bug`. This refactor allows for more consistent diagnostics, which don't have to guess the original code from HIR. r? `@petrochenkov`

rust-highfive assigned petrochenkov Jun 3, 2022

rustbot added the T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. label Jun 3, 2022

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Jun 3, 2022

This comment has been minimized.

Sign in to view

cjgillot force-pushed the all-fresh branch from e442727 to e149c30 Compare June 4, 2022 13:19

This comment has been minimized.

Sign in to view

cjgillot added the T-lang Relevant to the language team, which will review and decide on the PR/issue. label Jun 5, 2022

cjgillot added the I-lang-nominated Nominated for discussion during a lang team meeting. label Jun 5, 2022

petrochenkov reviewed Jun 5, 2022

View reviewed changes

petrochenkov 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 Jun 5, 2022

cjgillot force-pushed the all-fresh branch 2 times, most recently from cacdbf7 to 5983b4a Compare June 6, 2022 07:23

petrochenkov 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 Jun 6, 2022

petrochenkov approved these changes Jun 6, 2022

View reviewed changes

petrochenkov added S-waiting-on-team Status: Awaiting decision from the relevant subteam (see the T-<team> label). and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Jun 6, 2022

This comment was marked as resolved.

Sign in to view

cjgillot mentioned this pull request Jun 7, 2022

Move object lifetime default computation #97839

Closed

cjgillot force-pushed the all-fresh branch from 5983b4a to 3d17886 Compare June 9, 2022 20:33

This comment has been minimized.

Sign in to view

This comment was marked as resolved.

Sign in to view

cjgillot mentioned this pull request Jun 12, 2022

In-band lifetimes: Lint against single-use lifetime names #44752

Open

2 tasks

cjgillot force-pushed the all-fresh branch from 1de1add to b86f3f0 Compare June 19, 2022 12:01

petrochenkov added the S-blocked Status: Blocked on something else such as an RFC or other implementation work. label Jul 2, 2022

cjgillot added 4 commits July 13, 2022 14:14

Always use CreateParameter mode for function definitions.

031b2c5

Clippy fallout.

c7ac816

Bless ui-fulldeps tests.

3b1b38d

Add feature gate.

5a20834

cjgillot force-pushed the all-fresh branch from a66baca to 5a20834 Compare July 13, 2022 12:39

cjgillot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-blocked Status: Blocked on something else such as an RFC or other implementation work. labels Jul 13, 2022

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 Jul 13, 2022

klensy reviewed Jul 13, 2022

View reviewed changes

Dylan-DPC mentioned this pull request Jul 14, 2022

Rollup of 5 pull requests #99229

Closed

Dylan-DPC mentioned this pull request Jul 14, 2022

Rollup of 5 pull requests #99231

Merged

bors merged commit f5e9cb5 into rust-lang:master Jul 14, 2022

rustbot added this to the 1.64.0 milestone Jul 14, 2022

cjgillot deleted the all-fresh branch July 14, 2022 16:47

lqd mentioned this pull request Jul 15, 2022

Regression when compiling with 07-14 nightly #99276

Closed

apiraino mentioned this pull request Sep 1, 2022

Anon lifetime in impl Trait no longer suggests adding a lifetime parameter #100615

Closed

cjgillot mentioned this pull request Sep 1, 2022

Is this difference in lifetime elision for impl Trait arguments between async and non-async function intended? #85417

Open

csmoe mentioned this pull request Oct 31, 2022

Broken compilation with &(dyn Trait + '_) #103762

Open

c410-f3r mentioned this pull request Jan 27, 2023

Stabilize anonymous_lifetime_in_impl_trait #107378

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Always create elided lifetime parameters for functions #97720

Always create elided lifetime parameters for functions #97720

cjgillot commented Jun 3, 2022 •

edited by petrochenkov

Loading

This comment has been minimized.

This comment has been minimized.

cjgillot commented Jun 5, 2022

petrochenkov commented Jun 5, 2022

petrochenkov commented Jun 5, 2022

This comment was marked as resolved.

This comment has been minimized.

This comment has been minimized.

This comment was marked as resolved.

petrochenkov commented Jul 13, 2022

bors commented Jul 13, 2022

klensy Jul 13, 2022

Always create elided lifetime parameters for functions #97720

Always create elided lifetime parameters for functions #97720

Conversation

cjgillot commented Jun 3, 2022 • edited by petrochenkov Loading

This comment has been minimized.

This comment has been minimized.

cjgillot commented Jun 5, 2022

petrochenkov commented Jun 5, 2022

petrochenkov commented Jun 5, 2022

This comment was marked as resolved.

This comment has been minimized.

This comment has been minimized.

This comment was marked as resolved.

petrochenkov commented Jul 13, 2022

bors commented Jul 13, 2022

klensy Jul 13, 2022

Choose a reason for hiding this comment

cjgillot commented Jun 3, 2022 •

edited by petrochenkov

Loading