I can't say for certain but I suspect if you're burying the machine that badly then 'unexpected' results will arise. Is there any chance of trying a different perl distro? Is there any chance of a fault in your code? <just>asking</just>. Win98, for all it's other ugliness never did anything like this to me, nor NT4.0. Are you testing -e locally or to network mapped files? What process is burying your box so drasticlly?

It totally could be my code, but I've seen the same behavior in 2 very different programs I'm running, so that's why I'm leaning towards blaming the system. Unfortunately I can't upgrade my perl distro b/c our software is only compatible with this one (oops). The -e test is over a network mapped drive, if that helps at all. In general, the reason the box is so hosed right now is b/c im running something that takes hours and hours, but i don't have hours and hours, so I'm just threading the hell out of it (oops again). I guess I should just be more patient and see if the problem goes away in a more sane environment. As a test case though, I think the problem shows up if you write a program that continuously puts stuff in a hash, and then every 100k or 1million or so, do a -e test. I think the -e will fail before the perl proc crashes. -k