First round of small crash fixes for svn

[Brad Fitzpatrick]

On Mon, 6 Nov 2006, Brad Fitzpatrick wrote:

> Alan,
>
> > I noticed if a lot of jobs were dying or errors were flying about, the
> > parent has a tendency to crash, and the children don't necessarily
> > notice :) I haven't tracked down how/why this happens yet.
>
> An strace of the parent would almost certainly give the answer.

Cancel the strace request... I found it.

It's fixed on both sides now:

The parent was missing an event_err handler (only had event_hup), so it
would die in Danga::Socket's base class when the socket to the child
closed.  That caused the parent crash.

And when the parent crashes (kill -9 or whatever), the children no longer
can go crazy/detached.  They reliably die when their parent dies.  (this
one was fixed last week)

- Brad