c-variadic: document `Clone` and `Drop` instances and require `VaArgSafe: Copy` by folkertdev · Pull Request #155821 · rust-lang/rust

folkertdev · 2026-04-26T12:07:27Z

tracking issue: #44930

Fixing some things that came up in the stabilization PR

r? tgross35
cc @kpreid

rustbot · 2026-04-26T12:07:32Z

tgross35 is currently at their maximum review capacity.
They may take a while to respond.

joshtriplett · 2026-04-29T22:13:33Z

This is still unstable, so it doesn't need an FCP.

LGTM.

@bors r+ rollup

rust-bors · 2026-04-29T22:13:36Z

📌 Commit d143239 has been approved by joshtriplett

It is now in the queue for this repository.

programmerjake · 2026-04-30T00:44:55Z

+    /// Corresponds to `va_end` in C.
    #[inline] // Avoid codegen when not used to help backends that don't support VaList.
    fn drop(&mut self) {
+        // Call the rust `va_end` intrinsic, which is a no-op and does not map to LLVM `va_end`.


this is incorrect on latest nightly, it does call llvm.va_end. see #155697 (comment)

View changes since the review

programmerjake · 2026-04-30T00:47:23Z

de-approving so this can get looked at:
@bors r-

rust-bors · 2026-04-30T00:47:27Z

This pull request was unapproved.

View changes since this unapproval

folkertdev · 2026-04-30T09:02:43Z

huh, that never got updated apparently. It doesn't really matter because it's a no-op, but we should clean that up.

programmerjake · 2026-04-30T09:38:32Z

huh, that never got updated apparently. It doesn't really matter because it's a no-op, but we should clean that up.

when you or someone else cleans it up, please put the llvm.va_end in VaList::drop and not in the fn f(args: ...) { ... } implementation, that way it properly calls llvm.va_end on cloned VaLists too. that's needed if we gain any targets that need to free memory in va_end.

folkertdev · 2026-04-30T09:52:59Z

It's more complicated than that: va_end in C must be called in the same frame that called va_start, hence its ties to .... The situation in LLVM is a bit murky, but so far we've followed that, though the knowledge that it's a no-op gives us some flexibility.

The drop instance is for tracking that a VaList is not used after going out of scope, but that does not need to correspond to va_end being called.

programmerjake · 2026-04-30T15:07:11Z

It's more complicated than that: va_end in C must be called in the same frame that called va_start, hence its ties to .... The situation in LLVM is a bit murky, but so far we've followed that, though the knowledge that it's a no-op gives us some flexibility.

IMO it's unnecessary to follow the must-be-in-same-function rule, we did talk about that before along with why Drop should call va_end: https://rust-lang.zulipchat.com/#narrow/channel/146212-t-compiler.2Fconst-eval/topic/c-variadics.20in.20const-eval/near/563785065

folkertdev · 2026-04-30T18:08:38Z

That's in line with that I'm proposing: the rust va_end is already a no-op, we just didn't remove the llvm backend providing an implementation. If we remove that, the fallback body, a no-op, will be used instead. The Drop implementation calls the rust va_end intrinsic.

The use of the LLVM va_end is tied to ..., and we can remove it too, although given that it's a no-op that isn't crucial.

because a VaList can be cloned, the same c-variadic argument can be read twice and that is only safe if the argument type is copy

…ent `VaArgSafe`

it should use the fallback body instead

rustbot · 2026-04-30T18:32:55Z

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 · 2026-04-30T18:38:34Z

@programmerjake let us know if you think the latest commit unblocks this PR at least

programmerjake · 2026-04-30T18:58:39Z

@programmerjake let us know if you think the latest commit unblocks this PR at least

yeah, now that the comment is accurate I think it's good enough for this PR.

That's in line with that I'm proposing: the rust va_end is already a no-op, we just didn't remove the llvm backend providing an implementation. If we remove that, the fallback body, a no-op, will be used instead. The Drop implementation calls the rust va_end intrinsic.

The use of the LLVM va_end is tied to ..., and we can remove it too, although given that it's a no-op that isn't crucial.

I'm proposing what I thought we had agreed on in the Zulip thread, which is kinda the opposite of what this PR does, though it doesn't matter until we add a target where llvm.va_end is not a no-op.

folkertdev · 2026-04-30T19:03:56Z

I'm proposing what I thought we had agreed on in the Zulip thread, which is kinda the opposite of what this PR does, though it doesn't matter until we add a target where llvm.va_end is not a no-op.

I don't really see an LLVM target ever doing that, but perhaps the gcc backend will need it one day for some obscure target. Even so I suspect at that point that we'd pull va_start and va_end into the rust compiler for that target.

@bors r=joshtriplett

rust-bors · 2026-04-30T19:03:59Z

📌 Commit ac12c69 has been approved by joshtriplett

It is now in the queue for this repository.

Rollup of 19 pull requests Successful merges: - #155237 (Disentangle AST crates and error crates) - #155249 (Fix: On wasm targets, call `panic_in_cleanup` if panic occurs in cleanup) - #155853 (Use `_mcount` as the mcount symbol name on RISC-V Linux GNU targets) - #155919 (simplify `ast_fragments!`) - #155939 (Add feature gate for view_types experiment) - #155954 (rustdoc: preserve parent doc cfg for `macro_export` macros) - #155974 (add `c_variadic_experimental_arch` feature) - #155991 (Catch unwinds from the global ctxt callback to complete queries profiling data in more cases) - #156003 (Pass Session to optimize_and_codegen_fat_lto) - #153566 (Add suggestion for E0401 on inner const items) - #154610 (Suggest public re-exports when a private module makes an import path inaccessible) - #155523 (Reorganize `tests/ui/issues/` - 02) - #155821 (c-variadic: document `Clone` and `Drop` instances and require `VaArgSafe: Copy`) - #155980 (Move `feature*` methods from `parse` mod to `errors` mod.) - #155987 (Make lifting infallible) - #155988 (tests/run-make/print-cfg: add Android target_env case) - #156000 (Fix ICE when using -Zinstrument-mcount and -Clinker-flavor=lld) - #156002 (Allow to use `Diagnostic` directly in `SharedContext::emit_lint`) - #156015 (rustc-dev-guide subtree update)

@kpreid

Rollup merge of #155821 - folkertdev:doc-va-list-clone, r=joshtriplett c-variadic: document `Clone` and `Drop` instances and require `VaArgSafe: Copy` tracking issue: #44930 Fixing some things that came up in the stabilization PR r? tgross35 cc @kpreid

folkertdev added the F-c_variadic `#![feature(c_variadic)]` label Apr 26, 2026

rustbot assigned tgross35 Apr 26, 2026

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-libs Relevant to the library team, which will review and decide on the PR/issue. labels Apr 26, 2026

kpreid reviewed Apr 26, 2026

View reviewed changes

Comment thread library/core/src/ffi/va_list.rs Outdated

Comment thread library/core/src/ffi/va_list.rs Outdated

folkertdev force-pushed the doc-va-list-clone branch from 2538ff2 to d143239 Compare April 26, 2026 15:51

folkertdev mentioned this pull request Apr 26, 2026

Stabilize c-variadic function definitions #155697

Open

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 Apr 29, 2026

programmerjake reviewed Apr 30, 2026

View reviewed changes

rust-bors Bot added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Apr 30, 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 Apr 30, 2026

folkertdev added 4 commits April 30, 2026 20:09

c-variadic: document Clone and Drop instances

eb8f85e

c-variadic: add a Copy bound to VaArgSafe

ab19cd3

because a VaList can be cloned, the same c-variadic argument can be read twice and that is only safe if the argument type is copy

c-variadic: test that const and mutable void and char pointers implem…

ba9444c

…ent `VaArgSafe`

remove custom va_end implementation in the LLVM backend

ac12c69

it should use the fallback body instead

folkertdev force-pushed the doc-va-list-clone branch from d143239 to ac12c69 Compare April 30, 2026 18:32

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 Apr 30, 2026

jhpratt mentioned this pull request May 1, 2026

Rollup of 19 pull requests #156023

Merged

rust-bors Bot merged commit 0ab1a47 into rust-lang:main May 1, 2026
11 checks passed

rustbot added this to the 1.97.0 milestone May 1, 2026

Uh oh!

Conversation

folkertdev commented Apr 26, 2026

Uh oh!

rustbot commented Apr 26, 2026

Uh oh!

Uh oh!

Uh oh!

joshtriplett commented Apr 29, 2026

Uh oh!

rust-bors Bot commented Apr 29, 2026

Uh oh!

programmerjake Apr 30, 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!

programmerjake commented Apr 30, 2026

Uh oh!

rust-bors Bot commented Apr 30, 2026

Uh oh!

folkertdev commented Apr 30, 2026

Uh oh!

programmerjake commented Apr 30, 2026

Uh oh!

folkertdev commented Apr 30, 2026

Uh oh!

programmerjake commented Apr 30, 2026

Uh oh!

folkertdev commented Apr 30, 2026

Uh oh!

rustbot commented Apr 30, 2026

Uh oh!

folkertdev commented Apr 30, 2026

Uh oh!

programmerjake commented Apr 30, 2026

Uh oh!

folkertdev commented Apr 30, 2026

Uh oh!

rust-bors Bot commented Apr 30, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

programmerjake Apr 30, 2026 •

edited by rustbot

Loading