When an OS reboots or shuts down, it sends all processes SIGTERM before
sending SIGKILL. This allows btcd to do a proper shutdown which most
importantly closes the database.
This commit prevents a race that could happen on shutdown if a sigint was
received between adding the first the interrupt handler and and future
handlers. This code is loosely based on initial pull request #154 which
was not accepted due to introducing a race of its own.
Thanks to @tuxcanfly for the intial pull request and finding the issue.
Also, make every subsystem within btcd use its own logger instance so each
subsystem can have its own level specified independent of the others.
This is work towards #48.
Since the main SIGINT handler is running as a goroutine, the main
goroutine must be kept active long enough for it to finish or it will be
nuked when the main goroutine exits. This commit makes that happen by
slightly modifying how it waits for shutdown.
Rather than having the main goroutine only wait for the server to
shutdown, there is now a shutdown channel that is used to signal the main
goroutine to shutdown for all cases such as a graceful shutdown, a
scheduled shutdown, an RPC stop, or a SIGINT.
While here, also add a few prints to indicate a SIGINT was received and
the shutdown progress.
