Auto merge of #53108 - RalfJung:mutex, r=alexcrichton

clarify partially initialized Mutex issues

Using a `sys_common::mutex::Mutex` without calling `init` is dangerous, and yet there are some places that do this. I tried to find all of them and add an appropriate comment about reentrancy.

I found two places where (I think) reentrancy can actually occur, and was not able to come up with an argument for why this is okay. Someone who knows `io::lazy` and/or `sys_common::at_exit_imp` should have a careful look at this.

1 files changed, 4 insertions, 2 deletions

diff --git a/src/libstd/sys/unix/mutex.rs b/src/libstd/sys/unix/mutex.rs
index 60b03cdbeb0..1d447de1134 100644
--- a/src/libstd/sys/unix/mutex.rs
+++ b/src/libstd/sys/unix/mutex.rs
@@ -25,8 +25,10 @@ unsafe impl Sync for Mutex {}
 #[allow(dead_code)] // sys isn't exported yet
 impl Mutex {
     pub const fn new() -> Mutex {
-        // Might be moved and address is changing it is better to avoid
-        // initialization of potentially opaque OS data before it landed
+        // Might be moved to a different address, so it is better to avoid
+        // initialization of potentially opaque OS data before it landed.
+        // Be very careful using this newly constructed `Mutex`, reentrant
+        // locking is undefined behavior until `init` is called!
         Mutex { inner: UnsafeCell::new(libc::PTHREAD_MUTEX_INITIALIZER) }
     }
     #[inline]