C-variadic functions must be unsafe #141733
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
r? @nnethercote rustbot has assigned @nnethercote. Use |
rustbot
added
S-waiting-on-review
folkertdev
commented
May 29, 2025
folkertdev
force-pushed
the
c-variadic-unsafe-at-any-speed
branch
from
27bb41f to
2bab33c
Compare
tests/ui/c-variadic/feature-gate-extended_varargs_abi_support.rs
Outdated
Show resolved
Hide resolved
tests/ui/c-variadic/feature-gate-extended_varargs_abi_support.stderr
Outdated
Show resolved
Hide resolved
|
@folkertdev This seems semantically reasonable to me, with a lang hat on. I posted a few comments on the implementation. With those fixed, the implementation looks reasonable to me, but I'm not a compiler reviewer, so someone else should take a look at the implementation. |
folkertdev
force-pushed
the
c-variadic-unsafe-at-any-speed
branch
from
c02225d to
c31ba7b
Compare
nnethercote
reviewed
Jun 1, 2025
There was a problem hiding this comment.
Looks good, a couple of tiny nits to fix, then r=me.
@bors delegate=folkertdev
| @@ -83,3 +83,8 @@ trait T { | |||
| //~^ ERROR only foreign, `unsafe extern "C"`, or `unsafe extern "C-unwind"` functions may have a C-variadic arg | |||
| //~| ERROR `...` must be the last argument of a C-variadic function | |||
| } | |||
|
|
|||
| unsafe extern "C" { | |||
| safe fn s_f1(...); | |||
There was a problem hiding this comment.
Given that this is machine applicable, it would be good to have one or two examples with longer signatures containing additional arguments. A printf-style signature would be a good example.
There was a problem hiding this comment.
(And how did I get this far without knowing that safe was a keyword?)
folkertdev
force-pushed
the
c-variadic-unsafe-at-any-speed
branch
2 times, most recently
from
c067ed5 to
604a56b
Compare
This comment has been minimized.
This comment has been minimized.
folkertdev
force-pushed
the
c-variadic-unsafe-at-any-speed
branch
from
604a56b to
30ddd3c
Compare
|
The job Click to see the possible cause of the failure (guessed by this bot) |
|
@folkertdev: 🔑 Insufficient privileges: Not in reviewers |
|
hmm, apparently closing/reopening (to get CI to re-run) removes the delegation. Anyway, I think this is ready. |
|
@bors r+ |
bors
added
S-waiting-on-bors
matthiaskrgr
added a commit
to matthiaskrgr/rust
that referenced
this pull request
Jun 2, 2025
…y-speed, r=nnethercote C-variadic functions must be unsafe tracking issue: rust-lang#44930 A function that uses `...` is always unsafe to call, because it is UB to provide the wrong number of arguments, or arguments of an unexpected type. Hence, an `unsafe extern "C" { /* ... */ }` block should not be able to declare a `safe fn` that uses `...`. cc `@joshtriplett` `@workingjubilee` I'm not really sure who'd be a good reviewer for the actual parser code. `@rustbot` label: +F-c_variadic
bors
added a commit
that referenced
this pull request
Jun 2, 2025
Rollup of 6 pull requests Successful merges: - #140715 (Clarify &mut-methods' docs on sync::OnceLock) - #141309 (x86 (32/64): go back to passing SIMD vectors by-ptr) - #141677 (Async drop - type instead of async drop fn, fixes #140484) - #141733 (C-variadic functions must be unsafe) - #141858 (Fix typo in `StructuralPartialEq` docs) - #141874 (add f16_epsilon and f128_epsilon diagnostic items) r? `@ghost` `@rustbot` modify labels: rollup
workingjubilee
added a commit
to workingjubilee/rustc
that referenced
this pull request
Jun 3, 2025
…y-speed, r=nnethercote C-variadic functions must be unsafe tracking issue: rust-lang#44930 A function that uses `...` is always unsafe to call, because it is UB to provide the wrong number of arguments, or arguments of an unexpected type. Hence, an `unsafe extern "C" { /* ... */ }` block should not be able to declare a `safe fn` that uses `...`. cc ``@joshtriplett`` ``@workingjubilee`` I'm not really sure who'd be a good reviewer for the actual parser code. ``@rustbot`` label: +F-c_variadic
workingjubilee
added a commit
to workingjubilee/rustc
that referenced
this pull request
Jun 3, 2025
…y-speed, r=nnethercote C-variadic functions must be unsafe tracking issue: rust-lang#44930 A function that uses `...` is always unsafe to call, because it is UB to provide the wrong number of arguments, or arguments of an unexpected type. Hence, an `unsafe extern "C" { /* ... */ }` block should not be able to declare a `safe fn` that uses `...`. cc ```@joshtriplett``` ```@workingjubilee``` I'm not really sure who'd be a good reviewer for the actual parser code. ```@rustbot``` label: +F-c_variadic
|
Isn't this a breaking change? Can we go through something closer to our regular processes for doing that? |
|
Actually, yes. I think this is straightforward though. I propose to error on the following construction, that declares a function using C variadic arguments as unsafe extern "C" {
safe fn printf(format: *const u8, ...);
}This seems like an oversight from when I suspect, because @rustbot label +I-lang-nominated +I-lang-easy-decision |
rustbot
added
I-lang-easy-decision
|
Except for our documentation. Can you please send a PR to update https://github.com/rust-lang/reference/blob/master/src/items/external-blocks.md#variadic-functions with a more suitable example? |
|
And then send a second PR that would add a new rule for this restriction. |
This was referenced Jun 3, 2025
|
@bors r- |
bors
added
S-waiting-on-author
traviscross
added
S-waiting-on-documentation
|
@rustbot labels -I-lang-easy-decision This turned out to not be an easy decision. More details from discussion to come. |
rustbot
removed
the
I-lang-easy-decision
|
We discussed this in today's @rust-lang/lang meeting. Several people raised the point that calling a C varargs function is safe if you never look at the arguments, which technically makes it "library UB", not "language UB". On all targets I'm aware of, this property holds; I've pinged @Amanieu for additional review and confirmation for current and anticipated future targets. Given this, lang does not want to prohibit this at the language level. Several of us were amenable to a lint, though we didn't attempt to establish consensus on that in this meeting, and would need to confirm that before someone should invest any efforts in writing one. |
|
Fun edge case. Assuming there are no weird targets where never looking at the arguments is still UB, as a user I'd expect the reference to at least make a note on this (that |
traviscross
removed
I-lang-nominated
It's in scope for a non-normative admonition, e.g. |
|
With rust-lang/reference#1839 merged I think we're done here. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
tracking issue: #44930
A function that uses
...is always unsafe to call, because it is UB to provide the wrong number of arguments, or arguments of an unexpected type. Hence, anunsafe extern "C" { /* ... */ }block should not be able to declare asafe fnthat uses....cc @joshtriplett @workingjubilee
I'm not really sure who'd be a good reviewer for the actual parser code.
@rustbot label: +F-c_variadic