-
Notifications
You must be signed in to change notification settings - Fork 12.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Miscompilation on aarch64-apple-darwin (likely a calling convention bug) #112548
Comments
@rustbot label O-aarch64 O-macos |
WG-prioritization assigning priority (Zulip discussion). @rustbot label -I-prioritize +P-high |
I have another reproduction of what I believe is the same bug, but this one is only 34 lines, has only safe code and doesn't need a hasher: use std::fmt::Debug;
pub fn dump_var<T: Debug>(
f: usize,
var0: usize,
val0: T,
var1: usize,
val1: i32,
var2: usize,
val2: i32,
var3: usize,
val3: i32,
) {
println!(
"fn{f}:_{var0} = {val0:?}\n_{var1} = {val1:?}\n_{var2} = {val2:?}\n_{var3} = {val3:?}"
);
}
fn fn10() {
let _2 = 302053680377554952743809083898710279615_u128;
let mut _5 = _2.wrapping_shl(_2 as u32);
let mut _18: (u128, [isize; 3]) = (0, [0; 3]);
_18.1 = [9223372036854775807_isize; 3];
dump_var(11, 0, 0, 0, 0, 0, 0, 0, 0);
dump_var(11, 0, 0, 0, 0, 0, 0, /* addend */ 1, 0);
let _1 = _5 | _18.0;
let _11 = _5.wrapping_mul(_1);
_5 = _11;
dump_var(10, 5, _5, 0, 0, 0, 0, 0, 0);
}
pub fn main() {
fn10();
} This should print fn11:_0 = 0
_0 = 0
_0 = 0
_0 = 0
fn11:_0 = 0
_0 = 0
_0 = 0
_1 = 0
fn10:_5 = 85070591730234615865843651857942052864
_0 = 0
_0 = 0
_0 = 0 but with % rustc -Zmir-opt-level=0 -Copt-level=3 repro.rs && ./repro
fn11:_0 = 0
_0 = 0
_0 = 0
_0 = 0
fn11:_0 = 0
_0 = 0
_0 = 0
_1 = 0
fn10:_5 = 85070591730234615865843651857942052865
_0 = 0
_0 = 0
_0 = 0 The value of |
Any chance that the fix for #112767 also fixes this one? Looks kinda similar to me. |
No it fixes neither program in this issue. llvm/llvm-project#63475 might be related, since it also passes a vector on the stack |
Does this issue still reproduce? |
Running on an MBP M1, with rustc 1.81.0-nightly (8337ba9 2024-06-12), I get the same hash output for both of the executions listed:
|
@alex Thanks! Would you mind also trying the other test case in #112548 (comment)? |
That version does reproduce the issue:
|
Thanks! |
I can reproduce this. |
Upstream issue: llvm/llvm-project#97147. |
@rustbot label +llvm-fixed-upstream |
Update LLVM submodule Fixes rust-lang#112548 and unlock rust-lang#125642. r? `@cuviper` or `@nikic`
Rollup merge of rust-lang#127190 - DianQK:llvm-backport, r=nikic Update LLVM submodule Fixes rust-lang#112548 and unlock rust-lang#125642. r? ``@cuviper`` or ``@nikic``
Fuzzer generated code:
This should output
15917201677548574216
: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=a51ce49127180037639732b4a990ddbe.But on
-Zmir-opt-level=0 -Copt-level>=1
it prints something different.-Zmir-opt-level>=1
masks the bug.This is only reproducible on
aarch64-apple-darwin
, but not onx86_64-apple-darwin
oraarch64-unknown-linux-gnu
. The reproducer is very sensitive to small changes. For instance, changing the length of_23
infn16
to8
prevents the bug. For these reasons I suspect it's an Apple Silicon specific calling convention bug leading to some stack corruption.The text was updated successfully, but these errors were encountered: