Fix splat ICEs and ban it in closures#158645
Open
teor2345 wants to merge 10 commits into
Open
Conversation
rustbot
added
S-waiting-on-review
Collaborator
|
r? @folkertdev rustbot has assigned @folkertdev. Use Why was this reviewer chosen?The reviewer was selected based on:
|
This was referenced Jul 1, 2026
| (*fn_ptr)(1u32, 2i8); | ||
| let fn_pp: *const fn(#[splat] (u32, i8)) = tuple_args as *const fn(#[splat] (u32, i8)); | ||
| (*fn_pp)(1, 2); | ||
| (*fn_pp)(1u32, 2i8); |
Member
There was a problem hiding this comment.
I'd expect this to need unsafe, you are reading from a raw pointer.
Contributor
Author
There was a problem hiding this comment.
I think it ICEs before the unsafe check, due to the lines above panicking. I can split the test to make that clearer.
folkertdev
requested changes
Jul 1, 2026
| } | ||
|
|
||
| if !splatted_arg_indexes.is_empty() { | ||
| if let SplatSemantic::NoClosures(closure_span) = splat_semantic { |
Contributor
There was a problem hiding this comment.
why is this not a match (together with the other if let below)?
Comment on lines
+1788
to
+1793
| Extern::Explicit(abi_str, span) | ||
| if abi_str.symbol_unescaped.as_str() == "rust-call" => | ||
| { | ||
| SplatSemantic::NoRustCall(span) | ||
| } | ||
| Extern::Explicit(_abi_str, _span) => SplatSemantic::Yes, |
Contributor
There was a problem hiding this comment.
Suggested change
| Extern::Explicit(abi_str, span) | |
| if abi_str.symbol_unescaped.as_str() == "rust-call" => | |
| { | |
| SplatSemantic::NoRustCall(span) | |
| } | |
| Extern::Explicit(_abi_str, _span) => SplatSemantic::Yes, | |
| Extern::Explicit(abi_str, span) => match abi_str.symbol_unescaped { | |
| sym::rust_dash_call => SplatSemantic::NoRustCall(span), | |
| _ => SplatSemantic::Yes, | |
| }, |
Contributor
There was a problem hiding this comment.
and add a comment on why here too perhaps?
Comment on lines
+1786
to
+1787
| // FIXME(splat): should splatting extern "C" fns be allowed? | ||
| Extern::Implicit(_) => SplatSemantic::Yes, |
Contributor
There was a problem hiding this comment.
well extern "C" probably yes? but for most other extern "CC"s probably not actually? Anyhow, not for this PR.
Contributor
Author
There was a problem hiding this comment.
I'll add this to the tracking issue as an unresolved questiom
rustbot
added
S-waiting-on-author
Collaborator
|
Reminder, once the PR becomes ready for a review, use |
This comment has been minimized.
This comment has been minimized.
Ajay-singh1
reviewed
Jul 1, 2026
| LL | dyn_asref_splat(&s); | ||
| | ^^^^^^^^^^^^^^^^^^^ | ||
|
|
||
| error[E0277]: cannot use splat attribute; the splatted argument type must be a tuple or unit, not a Trait(std::convert::AsRef<(u8 |
Contributor
There was a problem hiding this comment.
The tuple case (AsRef<(u8, f32)>) renders truncated Trait(std::convert::AsRef<(u8 with no closing , f32)>). Is this expected, or a bug in how the diagnostic formats tuple types?
Contributor
Author
There was a problem hiding this comment.
That's a bug in my normalisation regex, which stops at the first comma, rather than looking for the next field name or balanced brackets
https://github.com/rust-lang/rust/pull/158645/changes#diff-1ba06a0d11ba6d51fa60caa4eb9589af26fe6e4abc458b03ff1d96d40d444c06R8
Collaborator
|
The job Click to see the possible cause of the failure (guessed by this bot) |
teor2345
commented
Jul 1, 2026
Contributor
Author
There was a problem hiding this comment.
A first pass. Overall this PR seems kind of large, it would be easier to review if split up into smaller chunks.
Yeah I wondered about that, I'll split it up today, now I know which parts actually fail in CI.
@rustbot label +F-splat +I-ICE +T-compiler +C-bug
Comment on lines
+1786
to
+1787
| // FIXME(splat): should splatting extern "C" fns be allowed? | ||
| Extern::Implicit(_) => SplatSemantic::Yes, |
Contributor
Author
There was a problem hiding this comment.
I'll add this to the tracking issue as an unresolved questiom
| (*fn_ptr)(1u32, 2i8); | ||
| let fn_pp: *const fn(#[splat] (u32, i8)) = tuple_args as *const fn(#[splat] (u32, i8)); | ||
| (*fn_pp)(1, 2); | ||
| (*fn_pp)(1u32, 2i8); |
Contributor
Author
There was a problem hiding this comment.
I think it ICEs before the unsafe check, due to the lines above panicking. I can split the test to make that clearer.
| LL | dyn_asref_splat(&s); | ||
| | ^^^^^^^^^^^^^^^^^^^ | ||
|
|
||
| error[E0277]: cannot use splat attribute; the splatted argument type must be a tuple or unit, not a Trait(std::convert::AsRef<(u8 |
Contributor
Author
There was a problem hiding this comment.
That's a bug in my normalisation regex, which stops at the first comma, rather than looking for the next field name or balanced brackets
https://github.com/rust-lang/rust/pull/158645/changes#diff-1ba06a0d11ba6d51fa60caa4eb9589af26fe6e4abc458b03ff1d96d40d444c06R8
rustbot
added
C-bug
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
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.
This PR is part of the splat lang experiment #153629.
It fixes one ICE, and cleans up the surrounding error handling to avoid future ICEs:
Close #158482
Improves the message and error handling in another ICE (but doesn't fix it):
cc #158603
And bans splat on closures and rust-call, because they already de-tuple arguments:
Close #158605
In the process I discovered another symbol mangling bug with splatting, which I added a test for (but didn't fix yet):
cc #158644
This test might be flaky, I'll drop it from this PR if it fails in CI.