Why does Valgrind not detect a memory leak in a Rust program using nightly 1.29.0?

问题

I'm trying to detect a memory leak in a Rust program using Valgrind following this blog post. My source code is simply:

#![feature(alloc_system)]
extern crate alloc_system;

use std::mem;

fn allocate() {
    let bad_vec = vec![0u8; 1024*1024];
    mem::forget(bad_vec);
}

fn main() {
    allocate();
}

I expect the call to mem::forget() to generate a memory leak that Valgrind would be able to pick up on. However, when I run Valgrind, it reports that no leaks are possible:

[memtest]> cargo run
   Compiling memtest v0.1.0 (file:///home/icarruthers/memtest)
    Finished dev [unoptimized + debuginfo] target(s) in 0.28s
     Running `target/debug/memtest`
[memtest]> valgrind target/debug/memtest
==18808== Memcheck, a memory error detector
==18808== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==18808== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
==18808== Command: target/debug/memtest
==18808== 
==18808== 
==18808== HEAP SUMMARY:
==18808==     in use at exit: 0 bytes in 0 blocks
==18808==   total heap usage: 0 allocs, 0 frees, 0 bytes allocated
==18808== 
==18808== All heap blocks were freed -- no leaks are possible
==18808== 
==18808== For counts of detected and suppressed errors, rerun with: -v
==18808== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

I am upgraded to the latest nightly (1.29.0-nightly (6a1c0637c 2018-07-23)).

What am I missing?

回答1:

Rust 1.32

As of Rust 1.32, the default allocator for an executable is now the system allocator, so you don't need to set anything by default.

Previous versions

You aren't using the global allocator setting correctly. This is a nightly feature, which means that it's prone to change at any time. Your blog post is out of date.

Check the module docs for std::alloc to see the correct usage:

#![feature(alloc_system)]
extern crate alloc_system;

#[global_allocator]
static GLOBAL: alloc_system::System = alloc_system::System;

use std::mem;

fn allocate() {
    let bad_vec = vec![0u8; 1024*1024];
    mem::forget(bad_vec);
}

fn main() {
    allocate();
}

root@3fb431791293:/tmp/vg# valgrind target/debug/vg
==6326== Memcheck, a memory error detector
==6326== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
==6326== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright info
==6326== Command: target/debug/vg
==6326==
==6326==
==6326== HEAP SUMMARY:
==6326==     in use at exit: 1,048,576 bytes in 1 blocks
==6326==   total heap usage: 12 allocs, 11 frees, 1,050,753 bytes allocated
==6326==
==6326== LEAK SUMMARY:
==6326==    definitely lost: 1,048,576 bytes in 1 blocks
==6326==    indirectly lost: 0 bytes in 0 blocks
==6326==      possibly lost: 0 bytes in 0 blocks
==6326==    still reachable: 0 bytes in 0 blocks
==6326==         suppressed: 0 bytes in 0 blocks
==6326== Rerun with --leak-check=full to see details of leaked memory
==6326==
==6326== For counts of detected and suppressed errors, rerun with: -v
==6326== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

来源：https://stackoverflow.com/questions/51509314/why-does-valgrind-not-detect-a-memory-leak-in-a-rust-program-using-nightly-1-29