It seems to me that pool::allocate can return without unlocking an internally locked mutex. There are a number of return statements on the way before m_mutex.unlock();. It also seems there is quite a number of if (m_free_stack.pop(b)) return b; calls. Can this be simplified?
auto allocate()
{
block_type b;
if (m_free_stack.pop(b)) return b;
if (!m_mutex.try_lock())
if (m_free_stack.pop(b)) return b;
if (m_free_stack.pop(b)) return b;
for (auto& kvp : m_segments) kvp.first->collect(m_free_stack);
if (m_free_stack.pop(b)) return b;
unsigned int counter = 0;
while (!m_free_stack.pop(b))
{
// add segments every 2nd iteration
if (++counter % 2 == 0) add_segment();
}
m_mutex.unlock();
return b;
}
It seems to me that
pool::allocatecan return without unlocking an internally locked mutex. There are a number ofreturnstatements on the way beforem_mutex.unlock();. It also seems there is quite a number ofif (m_free_stack.pop(b)) return b;calls. Can this be simplified?