sys._exit()

[Donn Cave]

Quoth Ignacio Vazquez-Abrams <ignacio at openservices.net>:
| On Wed, 22 Aug 2001, Chuck Esterbrook wrote:
...
|> What would be the motivation for a child process of a fork to _exit()
|> instead of exit()? e.g., why wouldn't the child wish to cleanup handlers,
|> flush stdio buffers, etc.? Does anyone use _exit()?

| Because then it would step all over the data of the parent, if it is still
| running.

Well, note that the fork runs in a separate address space and can't
modify the parent's state.  But there are still issues.  For example,
after a failed execve(2) you wouldn't want to flush C stdio buffers
on the way out, because that buffered data was copied from the parent -
imagine a bunch of disk files with extra copies of random chunks of
output.  If the fork is actually writing its own files, then those
should be flushed explicitly prior to _exit.

And sys.exit() raises an exception that you probably don't want to
catch in the fork.

By the way, I think we're really talking about posix._exit() here,
there is no sys._exit().

	Donn Cave, donn at drizzle.com