Repository navigation
LLVM ERROR Attribute 'align' exceed the max size 2^14 #121444
Description
auto-reduced (treereduce-rust):
// #45662
#[repr(align(536870912))]
pub struct A(i64);
pub extern "C" fn foo(x: A) {}
fn main() {
foo(A(0));
}
original:
//@ run-pass
#![allow(stable_features)]
#![allow(unused_variables)]
// #45662
#![feature(repr_align)]
#[repr(align(536870912))]
pub struct A(#[allow(dead_code)] i64);
#[allow(improper_ctypes_definitions)]
pub extern "C" fn foo(x: A) {}
fn main() {
foo(A(0));
}Version information
rustc 1.78.0-nightly (f8131a48a 2024-02-21)
binary: rustc
commit-hash: f8131a48a46ac3bc8a3d0fe0477055b132cffdc3
commit-date: 2024-02-21
host: x86_64-unknown-linux-gnu
release: 1.78.0-nightly
LLVM version: 18.1.0
Command:
/home/matthias/.rustup/toolchains/master/bin/rustc -Zincremental-verify-ich=yes -Cincremental=<dir> -Cdebuginfo=2
Program output
warning: unused variable: `x`
--> /tmp/icemaker_global_tempdir.YLDguKS9zOCA/rustc_testrunner_tmpdir_reporting.HqYrVsxsA6S8/mvce.rs:6:23
|
6 | pub extern "C" fn foo(x: A) {}
| ^ help: if this is intentional, prefix it with an underscore: `_x`
|
= note: `#[warn(unused_variables)]` on by default
warning: field `0` is never read
--> /tmp/icemaker_global_tempdir.YLDguKS9zOCA/rustc_testrunner_tmpdir_reporting.HqYrVsxsA6S8/mvce.rs:4:14
|
4 | pub struct A(i64);
| - ^^^
| |
| field in this struct
|
= note: `#[warn(dead_code)]` on by default
help: consider changing the field to be of unit type to suppress this warning while preserving the field numbering, or remove the field
|
4 | pub struct A(());
| ~~
warning: `extern` fn uses type `A`, which is not FFI-safe
--> /tmp/icemaker_global_tempdir.YLDguKS9zOCA/rustc_testrunner_tmpdir_reporting.HqYrVsxsA6S8/mvce.rs:6:26
|
6 | pub extern "C" fn foo(x: A) {}
| ^ not FFI-safe
|
= help: consider adding a `#[repr(C)]` or `#[repr(transparent)]` attribute to this struct
= note: this struct has unspecified layout
note: the type is defined here
--> /tmp/icemaker_global_tempdir.YLDguKS9zOCA/rustc_testrunner_tmpdir_reporting.HqYrVsxsA6S8/mvce.rs:4:1
|
4 | pub struct A(i64);
| ^^^^^^^^^^^^
= note: `#[warn(improper_ctypes_definitions)]` on by default
Attribute 'align' exceed the max size 2^14
ptr @_ZN4mvce3foo17h02ccc75a6bd1c329E
in function _ZN4mvce3foo17h02ccc75a6bd1c329E
LLVM ERROR: Broken function found, compilation aborted!
Metadata
Metadata
Assignees
Labels
Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues.Category: This is a bug.Issue: The compiler panicked, giving an Internal Compilation Error (ICE) ❄️Status: This bug is tracked inside the repo by a `known-bug` test.Status: A Minimal Complete and Verifiable Example has been found for this issueRelevant to the compiler team, which will review and decide on the PR/issue.Issue expected to be fixed by the next major LLVM upgrade, or backported fixesPerformance or correctness regression from one stable version to another.
Activity
matthiaskrgr commentedon Feb 22, 2024
Regression in nightly-2023-07-16
commit[0] 2023-07-14: Auto merge of #113471 - compiler-errors:new-solver-norm-escaping, r=lcnr
commit[1] 2023-07-15: Auto merge of #113514 - jyn514:more-gha-groups, r=oli-obk
commit[2] 2023-07-15: Auto merge of #112157 - erikdesjardins:align, r=nikic
commit[3] 2023-07-15: Auto merge of #113732 - matthiaskrgr:rollup-nm5qy4i, r=matthiaskrgr
commit[4] 2023-07-15: Auto merge of #113697 - GuillaumeGomez:rm-unneeded-externallocation-handling, r=lqd
ERROR: no CI builds available between ad96323 and 4c8bb79 within last 167 days
I suspect #112157 cc @erikdesjardins , @nikic
matthiaskrgr commentedon Feb 22, 2024
fwiw there is nothing about this that requires nightly
matthiaskrgr commentedon Feb 22, 2024
This is a 1.72 -> 1.73 regression (as in, build pass -> llvm error)
erikdesjardins commentedon Feb 23, 2024
Yeah, this is almost certainly because of #112157.
The build passed in 1.72 because we ignored alignment for
byvalarguments. That caused us to miscompile such code, which was the motivation for #112157.I don't think we can drop or clamp the alignment attribute when it's too high for LLVM, since that would reintroduce the miscompile. (So I don't really see any option besides providing a nicer error diagnostic.)
Although...Clang just ignores the requested alignment entirely if it's absurdly high, which also affects struct layout (https://godbolt.org/z/fjzfcsWda), which means we already differ from its ABI...
erikdesjardins commentedon Feb 23, 2024
(When that PR was merged I thought the nightmare was over, but it turns out the ride never ends...)
Auto merge of rust-lang#127513 - nikic:llvm-19, r=cuviper
Auto merge of rust-lang#127513 - nikic:llvm-19, r=cuviper
Auto merge of rust-lang#127513 - nikic:llvm-19, r=cuviper
Auto merge of #127513 - nikic:llvm-19, r=cuviper
Auto merge of #127513 - nikic:llvm-19, r=cuviper