Why is the process of going into daemon mode (on Unix) split into two functions, start_daemon and finish_daemon?
zw
On Wed, Jun 13, 2012 at 7:26 PM, Zack Weinberg zackw@panix.com wrote:
Why is the process of going into daemon mode (on Unix) split into two functions, start_daemon and finish_daemon?
It's been so long since we added that code; I hope I remember.
If I've got it right, the idea is that you call start_daemon() to begin daemonizing. Then you do a bunch of stuff in the child process that needs to happen in the child process, but which might fail. If it fails, you can just exit. If it succeeds, you inform the parent process of your success, and then exit. That way:
* The parent process can learn about and report some common kinds of initialization failure, and * The parent process doesn't exit until the child is well and truly launched.
It's been that way since cacacfe2b1 (originally svn:969). Unfortunately, the only commit message we have from back from when we added the commit (8.5 years ago!) is "Integrate new daemon code, adapted from submission by christian grothoff".
yrs,
On Wed, Jun 13, 2012 at 9:52 PM, Nick Mathewson nickm@alum.mit.edu wrote:
* The parent process can learn about and report some common kinds of initialization failure, and * The parent process doesn't exit until the child is well and truly launched.
Thanks, that makes a lot of sense.
I ask because I was using util.c as a reference for adding daemon mode support to Stegotorus; I might adopt a similar scheme (there are similar initialization-sequencing headaches) but probably not till I get around to the badly-needed Real Command Line Parser.
zw
-
Nick Mathewson -
Zack Weinberg