"Daniel P. Berrange" <berrange@redhat.com> wrote:
There statstest test case is expected to raise a number of libvirt errors. Since these are now stored in thread locals, it is expected that this won't be free'd automatically at system shutdown. This patch adds a call to virResetError() at the end to release the memory associated with the stored error object. There is still one block of memory alocated though, which is the thread local error error object itself. We can't free this easily, so I just add a valgrind suppression rule instead. This lets 'make valgrind' pass again
Also since the test case doesn't call virInitialize()as a normal app would do, we need to explicitly initialize the thread local storage to ensure things work correctly.
valgrind++ ACK.