On Mon, 2004-01-26 at 14:33, Henri Asseily wrote: > Sorry to get back to the old signal handling thing, but I was wondering > about what was said below, a couple of weeks ago: Tim answered your questions. FWIW, the alarm(0) I was missing in the initial fragment, was the one INSIDE the eval. BOTH are needed, otherwise, if hte connect were to succeed one wants to cancel it, and if it dies for some other reason, one wants to cancel it.
BTW: have a look at http://search.cpan.org/~lbaxter/Sys-SigAction-0.01/ which grew out of this discussion. If you want to have your code move forward to newer perls and still work the same way without modification (and months of hunting for whats wrong), I could consider this. I will probably update to 0.02 in the next week or so, I want to fix an annoying "undefined" warning and update the pod. Lincoln
