OK, a month late, but... On Sat, Dec 18, 2004 at 11:40:03PM -0800, Justin Erenkrantz wrote: > What is *supposed* to be the behavior of long-running CGI scripts that have > their connections closed on them before they complete? Are they supposed > to finish, or abort prematurely? Do they receive some signal when the > parent closes its handle?
You're seeing expected behaviour, I think. httpd only finds out that the TCP connection was closed as and when it tries to write to it. So if the CGI script doesn't write anything for > Timeout seconds, the server never finds out that the connection was aborted, and the normal timeout handling applies. i.e. the script is ignored until r->pool is destroyed, when the pipes are closed then the script is SIGTERMed (or SIGKILLed if it hangs around ignoring SIGTERM) is that any use? joe
