On Nov 27, 2007 1:04 PM, Jed Wing <[EMAIL PROTECTED]> wrote: > I'm investigating it now. I'll let you know what I find. I've given it > a quick test, and the bug seems to be replicatable.
What I'm seeing is that after some number of downloads, accept stops returning. The process doesn't seem to be leaking memory or file descriptors, and there isn't, as far as I can tell, a good reason for accept to start misbehaving. One solution is to further simplify chm_http and not spawn off threads to handle the requests, instead handling them directly. Since chm_http is mainly intended as a simple example, rather than a production-quality HTTP server, I'm tempted to take that approach, even though it completely serializes access to pages. (This could cause some trouble if you were to, say, try to download a very large file from a chm archive over a slow link, during which time no other requests could be serviced.) At any rate, on my machine, chm_http serves pages up at least as rapidly without spawning threads as with, and the problem seems to go away. This is probably what I'll do. -- - jed -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
