On Mon, Dec 14, 2009 at 10:34 AM, C. Michael Pilato <cmpil...@collab.net> wrote: > Mark Phippard tipped me off to this. > > Glad to hear that APR folk took the report seriously enough to investigate > the matter. Did anyone there happen to look at the fact that MaxMemFree > seems to not work at all, or that pools appear not to be automagically > defragmentable?
The core issue that Ben identified is still there - but, in practice, most web servers tend to have very repetitive data flows and memory allocation patterns, so it's not something that anyone but SVN users (who have big commits and big memory peaks) would ever encounter. We could make the fragmentation a bit smarter - that is, not jump the power-of-2 boundary automatically. But, this now seems more appropriate for d...@apr and not d...@subversion. -- justin
