Draft: [th/thread-local-storage-destroy]
The way how pthread_setspecific()
was used to release thread-local resources was wrong. Fix it.
Also, nm_utils_random_bytes()
used a thread local GRand
instance but leaked it when the thread exists. Fix that too.
Also, try to get a better seed for the GRand
instance in nm_utils_random_bytes()
for the fallback case where not good randomness is available.