linuxnewbie: clean handling of SIGTERM

[Randall Kern]

This is hopefully a simple "everybody has done that" kind of question, but
my Linux (unix) experience is sort of lacking.

I have a daemon that runs a loop, waiting for work.  When a request arrives,
it processes the request, and goes back into the loop.

Now I want to check for shutdown on each iteration of the loop.  I thought I
could register a handler for SIGTERM, and simply set a variable to tell my
loop to terminate:

_shutdown = 0
def sigterm(sig, frame):
    global _shutdown
    _shutdown = 1

signal.signal(signal.SIGTERM, sigterm)
while not _shutdown:
    request = get_request()    # returns None if _shutdown gets set (select
based loop)
    if request is not None:
        process_request(request)

The problem is that I get various exceptions when the signal is raised
(EINTR, but in various guises depending on the code that's running).  It
seems to me I could fix this if I could use reliable (BSD) signals from
python, but I don't see a way to do that.

Otherwise, I'll have to trap each and every EINTR derived exception, which
will mean copying a lot of library code (generally, a big hairy mess).

Is there a better way?

Thanks,
-Randy

P.S. My system does recover from a forced shutdown, but I would like to
complete any running requests before shutting down.