Detecting problems in a forked process

[Donn Cave]

Quoth Jim Segrave <jes at nl.demon.net>:
| In article <mailman.2690.1135894161.18701.python-list at python.org>,
| James Colannino  <james at colannino.org> wrote:
|> Hey everyone.  I'm writing a small application in Python that uses 
|> os.fork() to create a separate process in which another application is 
|> run in the background.  The problem is that I need to know whether or 
|> not that separate application managed to start and return from within 
|> the parent appropriately.
...
|> Is there a way that I can do this?  I thought, only for a *VERY* brief 
|> second, about cheating my way around this with a global variable, but 
|> then I realized that this wouldn't work due to the fact that I will have 
|> multiple forks doing the same thing at the same time.  Thanks in advance :)

If your thought there had continued for another second or two, you
probably would have wondered whether variables are shared between
forks.  (Actually they are, but only as an optimization - changes
are not shared.)

| options:
|
| Have the child set it's exit code to indicate success or failure and
| use one of the various os.wait functions in the parent to retrieve it.

This is the simplest, as long as it's easy to wait long enough to reliably
catch the failure case.

| create a pipe before forking and you can pass data back and forth
| between the parent and child

If you want the parent to wait for errors right up to the successful
execve(), and then continue, you can set the 'close on exec' flag on
the pipe write end file descriptor and make sure no process holds
this open but the child fork.  See the subprocess module for example
code (or just use the subprocess module, if it's supported in the
deployed Python version.)

	Donn Cave, donn at drizzle.com