fix: Use lock_guard instead of lock/unlock calls

I just noticed it was technically possible for the "work thread" (in py
plugin manager) to be started twice, due to improper lock/unlock calls.
This commit is contained in:
John Colvin committed 2025-06-16 11:08:59 -04:00
1 parent e8bb64a80f
commit 69e338c1ca
1 file changed
+1 -3
@@ -88,13 +88,11 @@ struct __attribute__((visibility("hidden"))) PluginManager_Impl {
PluginManager_SessionData *session_data;
void start_work_thread() {
mx.lock();
std::lock_guard<std::mutex> lock{mx};
if (work_thread) {
mx.unlock();
return;
}
keep_going = true;
mx.unlock();
work_thread.emplace([&]() {
PluginManager_SessionData sd;
mx.lock();