Don't acquire the mutex during the destructor of PassRegistry.

This destructor is run as part of static program termination, and
so all ManagedStatics (including this lock) will have been
destroyed by llvm_shutdown.  Furthermore, if there is actually
a race condition during static program termination, then we are
just hiding a bug somewhere else, because other threads should
not be running at this point.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@210717 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/IR/PassRegistry.cpp b/lib/IR/PassRegistry.cpp
index 6a5bee2f57b16da0a7d0ff1da047210261c2db63..e85fabffb8ce4377d739a742aec8d6b9cb0628b2 100644 (file)
--- a/lib/IR/PassRegistry.cpp
+++ b/lib/IR/PassRegistry.cpp
@@ -73,7 +73,10 @@ void *PassRegistry::getImpl() const {
 //
 
 PassRegistry::~PassRegistry() {
-  sys::SmartScopedWriter<true> Guard(*Lock);
+  // Don't acquire the mutex here.  This is destroyed during static execution of
+  // static destructors, after llvm_shutdown() has been called, so all instances
+  // of all ManagedStatics (including the Mutex), will have been destroyed as
+  // well.
   PassRegistryImpl *Impl = static_cast<PassRegistryImpl*>(pImpl);
   delete Impl;
   pImpl = nullptr;