> On Mon, 1 Apr 2002, Bill Stoddard wrote: > > > I suspect apr_bucket_alloc_create() is called in the wrong place. > > Working on it. SHould have a patch in 30 min or so. > > A semi-related question:: > > Should apr_bucket_alloc_create() be registering a cleanup on the pool > that's passed into it so that apr_bucket_alloc_destroy() is called > automatically when that pool dies? I didn't do this before because I was > hoping to get rid of the pool parameter to apr_bucket_alloc_create() > entirely. Right now we're not using it for anything. > > --Cliff >
Cliff, Let me know what you decide. If we are removing the pool param from apr_bucket_alloc_create(), then I will add a cleanup in the Windows MPM. BTW, my windows test bed (caching enabled) can serve a 500 byte file at well over 1000 rps again (after many, many months of being in the hole). 460 rps w/o caching (compared to something around 350 rps w/o caching in 1.3). It is good to see my Windows performance improvements coming back out of the noise :-) Bill
