nodejitsu / aeternum Goto Github PK

gcc -O2 -g -Wall -Ideps/libuv/include -framework CoreServices -o aeternum aeternum.c options.c deps/libuv/libuv.a
aeternum.c: In function ‘set_pidfile_path’:
aeternum.c:147: error: ‘WCHAR’ undeclared (first use in this function)
aeternum.c:147: error: (Each undeclared identifier is reported only once
aeternum.c:147: error: for each function it appears in.)
aeternum.c:147: error: expected ‘;’ before ‘homedir’
aeternum.c:148: warning: implicit declaration of function ‘FAILED’
aeternum.c:148: warning: implicit declaration of function ‘SHGetFolderPathW’
aeternum.c:148: error: ‘CSIDL_PROFILE’ undeclared (first use in this function)
aeternum.c:148: error: ‘path’ undeclared (first use in this function)
aeternum.c:148: error: expected ‘)’ before ‘{’ token
aeternum.c:172: error: expected expression before ‘}’ token
make: *** [aeternum] Error 1

Those appear to be some Windows constants. WTF.

aeternum behaves properly if it receives one SIGUSR1 signal, but on the second attempt it kills the aeternum process and leaves its child running.

cc @mmalecki

Would be nice if this worked.

A daemon should change to / before it runs, at least optionally. That way it does not block an unmount that might be needed.

Using SIGKILL gives the process no chance for cleanup. I propose sending SIGTERM initially, waiting a reasonable amount of time (possibly user defined and defaulting to x) for the process to end itself, and then finally send SIGKILL if necessary. I have some potentially longer-running operations (like sending emails) in my nodejitsu processes and my app will wait and exit cleanly when sent SIGTERM, but who knows what will be finished with a SIGKILL. Unfortunately I'm not great with C, or I'd send a PR.

In node.js...

var kill = function (child) {
  var wait = 5 * 60 * 1000;
  var timeout = setTimeout(child.kill.bind(child, 'SIGKILL'), wait);
  child.on('close', clearTimeout.bind(null, timeout);
  child.kill();
};