On 24/01/06, Oleg Kalnichevski <[EMAIL PROTECTED]> wrote: > On Tue, 2006-01-24 at 13:10 +0000, sebb wrote: > > On 23/01/06, Oleg Kalnichevski <[EMAIL PROTECTED]> wrote: > > > On Mon, 2006-01-23 at 19:47 +0000, sebb wrote: > > > > Might be useful to add a finalizer which logs a warning if the Manager > > > > has not been shut down properly? > > > > > > > > Sebastian > > > > > > > > > > Sebastian, > > > There are situations where such a warning would be inappropriate. > > > Consider the case of HttpClient being used as a singleton. We might, > > > > Not sure I understand what you mean here - surely even a singleton > > needs to be closed properly? > > Probably not, as long as the connection manager is meant to be active > for the whole lifespan of the JVM process
I see - it would perhaps be annoying to see the message in this case. Can one distinguish GC occurring JVM shutdown from regular GC? If so, one could just warn if a regular GC occurred. > The real trouble with the connection managers is their reliance on the > garbage collector to close persistent connections rather the fact that > connection managers necessarily need to be shut down. The existing > connection managers work fine as long as one does not do something > exceptionally silly, such as creating a new connection manager per > request in a tight loop, which in no time results in thousands of > connection managers waiting to be GC-ed, each holding on open > connection. > > Hope that explains it. Yes, thanks. > > Oleg > > > > > > however, run a simple static counter of how many connection managers > > > have been GC-ed without having been properly shut down and start > > > printing warnings beyond a certain threshold in order to circumvent the > > > problem. > > > > > > > OK. > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
