rustc crashes with segmentation fault in 1.73.0 #116668
Assignees
Labels
A-LLVM
Area: Code generation parts specific to LLVM. Both correctness bugs and optimization-related issues.
C-bug
Category: This is a bug.
I-crash
Issue: The compiler crashes (SIGSEGV, SIGABRT, etc). Use I-ICE instead when the compiler panics.
P-high
High priority
regression-from-stable-to-stable
Performance or correctness regression from one stable version to another.
T-compiler
Relevant to the compiler team, which will review and decide on the PR/issue.
Comments
rustbot
added
the
needs-triage
nikic
added
I-crash
rustbot
added
the
I-prioritize
|
Backtrace from a rustc build, tip of master with LLVM assertions and Rust debug assertions. No assertion of any kind is hit, we still segfault. |
|
|
Here's the config I'm using: profile = "compiler"
change-id = 115898
[llvm]
assertions = true
download-ci-llvm = true
[rust]
incremental = false
debug = true
[build]
sanitizers = trueAm I missing something to enable sanitizers in LLVM? |
Sorry, I'm actually not sure about the exact behavior of sanitizers here. |
|
It's probably obvious, but I managed to bisect it to Edit: managed to finish the bisection. searched nightlies: from nightly-2023-07-07 to nightly-2023-10-13 bisected with cargo-bisect-rustc v0.6.7Host triple: x86_64-unknown-linux-gnu cargo bisect-rustc --preserve -- build --release --no-default-features -F parking_lot --bin qdrant |
|
Minimizing this is proving to be quite difficult, but I managed to pinpoint a change. In impl OperationDurationsAggregator {
pub fn new() -> Arc<Mutex<Self>> {
Arc::new(Mutex::new(Self {
ok_count: 0,
fail_count: 0,
timings: todo!(), //[0.; AVG_DATASET_LEN],
timing_index: 0,
timing_loops: 0,
min_value: None,
max_value: None,
last_response_date: Some(Utc::now().round_subsecs(2)),
}))
}
}Interestingly, while trying to go from I'm not nearly enough of an LLVM expert to understand if this is relevant, but I thought I'd include it in the event it is. |
|
WG-prioritization assigning priority (Zulip discussion). @rustbot label -I-prioritize +P-high cc @nikic |
rustbot
added
P-high
|
Preliminary reduction: ; RUN: opt -S -passes=gvn < %s
define void @ham() {
bb:
br label %bb1
bb1: ; preds = %bb5, %bb
%phi = phi i64 [ 0, %bb ], [ %add, %bb5 ]
%phi2 = phi ptr [ null, %bb ], [ %select, %bb5 ]
%icmp = icmp eq i64 %phi, 0
br i1 %icmp, label %bb8, label %bb3
bb3: ; preds = %bb1
%zext = zext i16 poison to i64
%sub = sub nsw i64 0, %zext
%getelementptr = getelementptr inbounds { { { { ptr, i64 }, i64 } }, i64 }, ptr null, i64 %sub
%add = add i64 %phi, -1
%icmp4 = icmp eq ptr %getelementptr, null
br i1 %icmp4, label %bb8, label %bb5
bb5: ; preds = %bb3
%load = load i64, ptr %phi2, align 8
%load6 = load i64, ptr null, align 8
%icmp7 = icmp ugt i64 %load, %load6
%select = select i1 %icmp7, ptr %phi2, ptr null
br label %bb1
bb8: ; preds = %bb3, %bb1
%load9 = load i64, ptr %phi2, align 8
%mul = mul i64 %load9, 0
%getelementptr10 = getelementptr inbounds ptr, ptr null, i64 29
%getelementptr11 = getelementptr inbounds ptr, ptr null, i64 27
unreachable
} |
|
Upstream issue: llvm/llvm-project#69301 |
|
This consistently fails with (signal: 11, SIGSEGV: invalid memory reference) for me on 1.73.0 (stable-aarch64-apple-darwin) #[test]
fn test_sigsegv() {
let data:[f32; 1000*1000] = vec![1.0; 1000*1000]
.try_into()
.expect("Wrong matrix size!");
} |
|
@kuzemchik Can you please file a separate issue for that? It doesn't look related to this one. |
bors
added a commit
to rust-lang-ci/rust
that referenced
this issue
Nov 1, 2023
Update to LLVM 17.0.4 Fixes rust-lang#116668. Fixes rust-lang#116941. Fixes rust-lang#116976. r? `@cuviper`
3tilley
pushed a commit
to 3tilley/rust
that referenced
this issue
Nov 1, 2023
Update to LLVM 17.0.4 Fixes rust-lang#116668. Fixes rust-lang#116941. Fixes rust-lang#116976. r? `@cuviper`
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
When I compile a project with rust 1.73.0 on linux, I get a segmentation fault error.
When I compile it with 1.72.1 all is well.
How to reproduce
on linux
If reproduced, you should get this error
If I do the same, but use 1.72.1, all is well.
The text was updated successfully, but these errors were encountered: