Valgrind reports memory leak in runtime with current beta compiler

[jfrimmel]

I maintain the cargo-valgrind-crate, which is a wrapper for running binaries & tests automatically within valgrind for memory analysis. As a sanity check, a periodic job runs some basic smoke tests against the current beta compiler. This smoke test was recently triggered, indicating a regression. In order to outrule the aforementioned tool, I present a reproducer with plain valgrind invocations below.

Code

I tried the default test code generated by cargo new --lib:

pub fn add(left: u64, right: u64) -> u64 {
    left + right
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn it_works() {
        let result = add(2, 2);
        assert_eq!(result, 4);
    }
}

I expected to see this happen: when running the tests inside valgrind, no memory errors/leaks should be reported. This can be tested using the following commands:

$ cargo new --lib some-tests
$ cd some-tests
$ cargo +stable test --lib  # run the tests normally
$ valgrind --leak-check=full target/debug/deps/some_tests-fa75b02a65f3d9a8  # now run the test executable within valgrind, adjust hash as necessary
<snip>
==42061== 
==42061== HEAP SUMMARY:
==42061==     in use at exit: 0 bytes in 0 blocks
==42061==   total heap usage: 637 allocs, 637 frees, 77,942 bytes allocated
==42061== 
==42061== All heap blocks were freed -- no leaks are possible
==42061== 
==42061== For lists of detected and suppressed errors, rerun with: -s
==42061== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

There are no leaks reported with the current stable compiler (see version info below).

Instead, this happened: when performing the same commands with the current beta compiler (see version info below), there is a leak reported:

$ cargo +beta test --lib
$ valgrind --leak-check=full target/debug/deps/some_tests-865f7506563a5a65  # adjust the hash as necessary
<snip>
==42409== 
==42409== HEAP SUMMARY:
==42409==     in use at exit: 48 bytes in 1 blocks
==42409==   total heap usage: 638 allocs, 637 frees, 77,968 bytes allocated
==42409== 
==42409== 48 bytes in 1 blocks are possibly lost in loss record 1 of 1
==42409==    at 0x48457A8: malloc (vg_replace_malloc.c:446)
==42409==    by 0x1A5E77: alloc (alloc.rs:96)
==42409==    by 0x1A5E77: alloc_impl (alloc.rs:192)
==42409==    by 0x1A5E77: allocate (alloc.rs:254)
==42409==    by 0x1A5E77: {closure#0}<std::thread::Inner> (sync.rs:484)
==42409==    by 0x1A5E77: allocate_for_layout<core::mem::maybe_uninit::MaybeUninit<std::thread::Inner>, alloc::sync::{impl#14}::new_uninit::{closure_env#0}<std::thread::Inner>, fn(*mut u8) -> *mut alloc::sync::ArcInner<core::mem::maybe_uninit::MaybeUninit<std::thread::Inner>>> (sync.rs:1952)
==42409==    by 0x1A5E77: new_uninit<std::thread::Inner> (sync.rs:482)
==42409==    by 0x1A5E77: std::thread::Thread::new (mod.rs:1429)
==42409==    by 0x1A5499: std::thread::current::init_current (current.rs:227)
==42409==    by 0x1AEC23: current_or_unnamed (current.rs:184)
==42409==    by 0x1AEC23: std::sync::mpmc::context::Context::new (context.rs:72)
==42409==    by 0x140CD1: __init (context.rs:43)
==42409==    by 0x140CD1: call_once<fn() -> core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>, ()> (function.rs:250)
==42409==    by 0x140CD1: unwrap_or_else<core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>, fn() -> core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>> (option.rs:1023)
==42409==    by 0x140CD1: std::sys::thread_local::native::lazy::Storage<T,D>::initialize (lazy.rs:64)
==42409==    by 0x142485: get_or_init<core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>, (), fn() -> core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>> (lazy.rs:56)
==42409==    by 0x142485: {closure#0} (mod.rs:94)
==42409==    by 0x142485: call_once<std::sync::mpmc::context::{impl#0}::with::CONTEXT::{constant#0}::{closure_env#0}, (core::option::Option<&mut core::option::Option<core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>>>)> (function.rs:250)
==42409==    by 0x142485: try_with<core::cell::Cell<core::option::Option<std::sync::mpmc::context::Context>>, std::sync::mpmc::context::{impl#0}::with::{closure_env#1}<std::sync::mpmc::list::{impl#3}::recv::{closure_env#1}<test::event::CompletedTest>, ()>, ()> (local.rs:309)
==42409==    by 0x142485: with<std::sync::mpmc::list::{impl#3}::recv::{closure_env#1}<test::event::CompletedTest>, ()> (context.rs:52)
==42409==    by 0x142485: std::sync::mpmc::list::Channel<T>::recv (list.rs:437)
==42409==    by 0x15BBBF: recv_deadline<test::event::CompletedTest> (mod.rs:1119)
==42409==    by 0x15BBBF: recv_timeout<test::event::CompletedTest> (mod.rs:1051)
==42409==    by 0x15BBBF: recv_timeout<test::event::CompletedTest> (mpsc.rs:905)
==42409==    by 0x15BBBF: run_tests<test::console::run_tests_console::{closure_env#2}> (lib.rs:430)
==42409==    by 0x15BBBF: test::console::run_tests_console (console.rs:323)
==42409==    by 0x1791F6: test::test_main (lib.rs:150)
==42409==    by 0x179B7A: test::test_main_static (lib.rs:172)
==42409==    by 0x13F522: some_tests::main (lib.rs:0)
==42409==    by 0x13F45A: core::ops::function::FnOnce::call_once (function.rs:250)
==42409==    by 0x13F55D: std::sys::backtrace::__rust_begin_short_backtrace (backtrace.rs:152)
==42409== 
==42409== LEAK SUMMARY:
==42409==    definitely lost: 0 bytes in 0 blocks
==42409==    indirectly lost: 0 bytes in 0 blocks
==42409==      possibly lost: 48 bytes in 1 blocks
==42409==    still reachable: 0 bytes in 0 blocks
==42409==         suppressed: 0 bytes in 0 blocks
==42409== 
==42409== For lists of detected and suppressed errors, rerun with: -s
==42409== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

Version it worked on

It most recently worked on: Rust 1.85
rustc --version --verbose:

rustc 1.85.0 (4d91de4e4 2025-02-17)
binary: rustc
commit-hash: 4d91de4e48198da2e33413efdcd9cd2cc0c46688
commit-date: 2025-02-17
host: x86_64-unknown-linux-gnu
release: 1.85.0
LLVM version: 19.1.7

Version with regression

rustc +beta --version --verbose:

rustc 1.86.0-beta.5 (b2af9a598 2025-03-08)
binary: rustc
commit-hash: b2af9a598e793de1f65afa6c14f7eb7545332ac1
commit-date: 2025-03-08
host: x86_64-unknown-linux-gnu
release: 1.86.0-beta.5
LLVM version: 19.1.7

Also happens with the current nightly:
rustc +nightly --version --verbose:

rustc 1.87.0-nightly (665025243 2025-03-11)
binary: rustc
commit-hash: 6650252439d4e03368b305c42a10006e36f1545e
commit-date: 2025-03-11
host: x86_64-unknown-linux-gnu
release: 1.87.0-nightly
LLVM version: 20.1.0

@rustbot modify labels: +regression-from-stable-to-beta -regression-untriaged

[jfrimmel]

Might be a duplicate of #135608.

[jieyouxu]

I think this is #135608 indeed.

[jieyouxu]

Closing as a duplicate of #135608, thanks for the report.