suggested fix for the sigchld race

[Markus Friedl]

On Fri, Jan 25, 2002 at 12:07:56PM -0500, Nicolas Williams wrote:
> On Wed, Oct 31, 2001 at 05:42:01PM +0100, Markus Friedl wrote:
> > comments?
> 
> Works. The SIGCHLD/select() race is still present in 3.0.2p1. This
> patch should be applied.

wow! finally feedback on this. please check current snapshots,
it's already beed added.

> sigsetjmp() is not ugly (this is the sort of problem it's meant to be
> there for). But it's not as elegant as the pipe trick and probably much
> less portable. I prefer the the pipe trick.

sigsetjmp() is ugly as hell. think about correctly releasing resources.

> So I second the use of the pipe trick to prevent the SIGCHLD/select()
> race condition.
> 
> I'll file a bug report.

what bug report?