Lift the same-signature restriction for `extern "tail"` by folkertdev · Pull Request #157983 · rust-lang/rust

folkertdev · 2026-06-16T18:42:45Z

tracking issue: #157427

The extern "tail" calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply.

We need the ABI to deviate from extern "Rust" to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for extern "tail" that doesn't work reliably because the signatures can be different.

I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in extern "tail" calls.

Also we can't really promote using this yet due to tailcc being a bit broken on LLVM 22 on x86_64.

r? WaffleLapkin
cc @bjorn3

rustbot · 2026-06-16T18:42:52Z

WaffleLapkin is not on the review rotation at the moment.
They may take a while to respond.

bjorn3 · 2026-06-16T18:51:37Z

I don't think it is possible supporting unsized arguments with tail calls except for the narrow case where every tail call passes the unsized argument it received from its own caller, in which case Indirect { on_stack: false } would work even with extern "tail". I don't think it is worth supporting this narrow of a rule, so I think disallowing unsized arguments on extern "tail" entirely is the right call.

WaffleLapkin · 2026-06-22T10:47:36Z

                        arg.cast_to_with_attrs(Reg { kind: RegKind::Integer, size }, attr.into());
                    }
+
+                    if self.conv == CanonAbi::RustTail && arg.layout.is_sized() {


extern "tail" requires that args are sized, so I think arg.layout.is_sized() should be an assert instead

View changes since the review

right, fixed

WaffleLapkin

r=me with green CI

View changes since this review

rustbot · 2026-06-22T15:01:53Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

folkertdev · 2026-06-22T16:31:34Z

@bors r=WaffleLapkin

rust-bors · 2026-06-22T16:31:37Z

📌 Commit d8d4b89 has been approved by WaffleLapkin

It is now in the queue for this repository.

🌲 The tree is currently closed for pull requests below priority 5. This pull request will be tested once the tree is reopened.

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

rust-bors · 2026-06-23T02:38:15Z

This pull request was unapproved.

This PR was contained in a rollup (#158277), which was unapproved.

View changes since this unapproval

folkertdev · 2026-06-23T08:42:29Z

Turns out x86 fails on LLVM 22 even in the simple case. It should be fixed by LLVM 23, I'll validate that locally (but the outcome should not be blocking here).

@bors try jobs=i686-gnu-nopt-1,x86_64-gnu-nopt

…tion, r=<try> Lift the same-signature restriction for `extern "tail"` try-job: i686-gnu-nopt-1 try-job: x86_64-gnu-nopt

rust-bors · 2026-06-23T11:15:24Z

☀️ Try build successful (CI)
Build commit: 79c979d (79c979da0163c369152caeca08bb9256b5f79533)
Base parent: 4429659 (4429659e4745016bd3f26a4a421843edc7fbc422)

folkertdev · 2026-06-23T12:42:45Z

@bors r=WaffleLapkin

rust-bors · 2026-06-23T12:42:48Z

📌 Commit e638ebe has been approved by WaffleLapkin

It is now in the queue for this repository.

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

…uwer Rollup of 23 pull requests Successful merges: - #158315 (`rust-analyzer` subtree update) - #155739 (Add temporary scope to assert_eq and assert_ne) - #156885 (Fix misattributed type inference error span for index expressions) - #157271 (simplify some `proc_macro` things) - #157883 (Remove strict invariant node_type on hir_type during ty privacy visit) - #157921 (trait solver: Resolve region vars in max universe) - #157960 (delegation: add support for infers in generics) - #157983 (Lift the same-signature restriction for `extern "tail"`) - #158105 (Extract all instance shim variants into new `ShimKind` enum) - #158207 (Resolver: local/external split of `resolve_ident_in_module_non_globs_unadjusted` ) - #158279 (Follow goto and drop when linting unreachable code) - #157527 (Move derive tests into their dedicated folder) - #157807 (don't ice on non-lifetime binders under `-Zassumptions-on-binders`) - #158020 (Update mingw-w64 C toolchain) - #158222 (format: ignore println newline in foreign format hints) - #158223 (Move target checking for #[lang] to the attribute parser) - #158252 (Use `cfg_select` in `std::os`) - #158257 ( fix escaping placeholder check in next solver normalization folder) - #158263 (Only load the feature list once in the entire resolver) - #158274 (triagebot: Stop pinging myself) - #158282 (slice_split_once: bounds check optimization note) - #158300 (Improve unknown crate_type diagnostic suggestions) - #158304 (mailmap: update mu001999) Failed merges: - #158256 (Avoid parser panics bubbling out to proc macros)

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

Rollup of 27 pull requests Successful merges: - #158315 (`rust-analyzer` subtree update) - #155739 (Add temporary scope to assert_eq and assert_ne) - #156885 (Fix misattributed type inference error span for index expressions) - #157271 (simplify some `proc_macro` things) - #157883 (Remove strict invariant node_type on hir_type during ty privacy visit) - #157921 (trait solver: Resolve region vars in max universe) - #157960 (delegation: add support for infers in generics) - #157983 (Lift the same-signature restriction for `extern "tail"`) - #158053 (Optimize network address parser) - #158105 (Extract all instance shim variants into new `ShimKind` enum) - #158207 (Resolver: local/external split of `resolve_ident_in_module_non_globs_unadjusted` ) - #158279 (Follow goto and drop when linting unreachable code) - #157527 (Move derive tests into their dedicated folder) - #157807 (don't ice on non-lifetime binders under `-Zassumptions-on-binders`) - #158020 (Update mingw-w64 C toolchain) - #158039 (c-variadic: test that we use equality up to free lifetimes) - #158222 (format: ignore println newline in foreign format hints) - #158223 (Move target checking for #[lang] to the attribute parser) - #158252 (Use `cfg_select` in `std::os`) - #158257 ( fix escaping placeholder check in next solver normalization folder) - #158263 (Only load the feature list once in the entire resolver) - #158267 (FromUtf8Error::into_utf8_lossy better example and suggest use) - #158274 (triagebot: Stop pinging myself) - #158282 (slice_split_once: bounds check optimization note) - #158300 (Improve unknown crate_type diagnostic suggestions) - #158304 (mailmap: update mu001999) - #158309 (Update `rustc-literal-escaper` version to `0.0.8`) Failed merges: - #158256 (Avoid parser panics bubbling out to proc macros)

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

Rollup of 35 pull requests Successful merges: - #158315 (`rust-analyzer` subtree update) - #158336 (Stop excluding `stdarch` test crates from `rust-src`) - #155739 (Add temporary scope to assert_eq and assert_ne) - #156885 (Fix misattributed type inference error span for index expressions) - #157271 (simplify some `proc_macro` things) - #157419 (move rustc_type_ir Term things to term_kind.rs) - #157883 (Remove strict invariant node_type on hir_type during ty privacy visit) - #157921 (trait solver: Resolve region vars in max universe) - #157960 (delegation: add support for infers in generics) - #157983 (Lift the same-signature restriction for `extern "tail"`) - #158053 (Optimize network address parser) - #158105 (Extract all instance shim variants into new `ShimKind` enum) - #158207 (Resolver: local/external split of `resolve_ident_in_module_non_globs_unadjusted` ) - #158279 (Follow goto and drop when linting unreachable code) - #157527 (Move derive tests into their dedicated folder) - #157807 (don't ice on non-lifetime binders under `-Zassumptions-on-binders`) - #157939 (Reorganize `tests/ui/issues` [8/N]) - #157946 (Make `char::is_private_use` and `char::is_assigned` unstably public) - #158003 (Reorganize `tests/ui/issues` [9/N]) - #158020 (Update mingw-w64 C toolchain) - #158039 (c-variadic: test that we use equality up to free lifetimes) - #158060 (Reorganize `tests/ui/issues` [10/N]) - #158222 (format: ignore println newline in foreign format hints) - #158223 (Move target checking for #[lang] to the attribute parser) - #158252 (Use `cfg_select` in `std::os`) - #158263 (Only load the feature list once in the entire resolver) - #158267 (FromUtf8Error::into_utf8_lossy better example and suggest use) - #158272 (Reorganize `tests/ui/issues` [13/N]) - #158274 (triagebot: Stop pinging myself) - #158282 (slice_split_once: bounds check optimization note) - #158300 (Improve unknown crate_type diagnostic suggestions) - #158304 (mailmap: update mu001999) - #158309 (Update `rustc-literal-escaper` version to `0.0.8`) - #158314 (Fix incorrect unsafe debug assertion in unchecked_div_exact) - #158326 (Add `io::ErrorKind::TooManyOpenFiles`)

@bjorn3

…ure-restriction, r=WaffleLapkin Lift the same-signature restriction for `extern "tail"` tracking issue: rust-lang#157427 The `extern "tail"` calling convention uses callee cleanup (i.e. the callee restores the stack, not the caller). Hence, the same-signature restriction that is normally required to codegen tail calls does not apply. We need the ABI to deviate from `extern "Rust"` to make sure indirect arguments are passed by stack offset, not via a pointer into the caller's stack frame (the value would potentially be overwritten). For standard tail calls we work around this problem by storing the value in the caller's caller, but for `extern "tail"` that doesn't work reliably because the signatures can be different. I'm not sure about unsized arguments. That feature seems really broken, so I'm not sure how much work we should put into trying to do something reasonable there. Fundamentally I don't think we can support unsized arguments in `extern "tail"` calls. Also we can't really promote using this yet due to `tailcc` being a bit broken on LLVM 22 on x86_64. r? WaffleLapkin cc @bjorn3

folkertdev added the F-rust_tail_cc `#![feature(rust_tail_cc)]` label Jun 16, 2026

rustbot assigned WaffleLapkin Jun 16, 2026

rustbot added F-explicit_tail_calls `#![feature(explicit_tail_calls)]` S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jun 16, 2026

This comment has been minimized.

Sign in to view

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch from 9ff5f86 to 0a30594 Compare June 16, 2026 19:41

This comment has been minimized.

Sign in to view

bjorn3 reviewed Jun 16, 2026

View reviewed changes

Comment thread compiler/rustc_mir_build/src/check_tail_calls.rs

folkertdev commented Jun 16, 2026

View reviewed changes

Comment thread compiler/rustc_hir_typeck/src/check.rs

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch 2 times, most recently from ce0a4ed to 73d1dea Compare June 16, 2026 20:59

This comment has been minimized.

Sign in to view

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch from 73d1dea to 2a61ca9 Compare June 17, 2026 09:04

This comment has been minimized.

Sign in to view

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch from 2a61ca9 to 0aa786a Compare June 17, 2026 10:39

WaffleLapkin reviewed Jun 22, 2026

View reviewed changes

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch from 0aa786a to f6c3707 Compare June 22, 2026 13:24

This comment has been minimized.

Sign in to view

WaffleLapkin approved these changes Jun 22, 2026

View reviewed changes

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch from f6c3707 to d8d4b89 Compare June 22, 2026 15:01

rust-bors Bot 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 Jun 22, 2026

rust-bors Bot removed the S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. label Jun 23, 2026

Lift the same-signature restriction for extern "tail"

e638ebe

folkertdev force-pushed the tailcc-lift-same-signature-restriction branch from 637142b to e638ebe Compare June 23, 2026 08:39

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Jun 23, 2026

Auto merge of #157983 - folkertdev:tailcc-lift-same-signature-restric…

79c979d

…tion, r=<try> Lift the same-signature restriction for `extern "tail"` try-job: i686-gnu-nopt-1 try-job: x86_64-gnu-nopt

rust-bors Bot 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-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Jun 23, 2026

JonathanBrouwer mentioned this pull request Jun 23, 2026

Rollup of 23 pull requests #158316

Closed

JonathanBrouwer mentioned this pull request Jun 23, 2026

Rollup of 23 pull requests #158324

Closed

jhpratt mentioned this pull request Jun 24, 2026

Rollup of 27 pull requests #158335

Closed

jhpratt mentioned this pull request Jun 24, 2026

Rollup of 34 pull requests #158337

Closed

jhpratt mentioned this pull request Jun 24, 2026

Rollup of 35 pull requests #158342

Open

JonathanBrouwer mentioned this pull request Jun 24, 2026

[CI] Rollup of 35 pull requests #158354

Draft

Uh oh!

Conversation

folkertdev commented Jun 16, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jun 16, 2026

Uh oh!

bjorn3 commented Jun 16, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

WaffleLapkin Jun 22, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

folkertdev Jun 22, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

WaffleLapkin left a comment • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rustbot commented Jun 22, 2026

Uh oh!

folkertdev commented Jun 22, 2026

Uh oh!

rust-bors Bot commented Jun 22, 2026

Uh oh!

rust-bors Bot commented Jun 23, 2026

Uh oh!

folkertdev commented Jun 23, 2026

Uh oh!

This comment has been minimized.

rust-bors Bot commented Jun 23, 2026

Uh oh!

folkertdev commented Jun 23, 2026

Uh oh!

rust-bors Bot commented Jun 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

folkertdev commented Jun 16, 2026 •

edited by rustbot

Loading

WaffleLapkin Jun 22, 2026 •

edited by rustbot

Loading

WaffleLapkin left a comment •

edited by rustbot

Loading