On Sun, Jan 23, 2011 at 6:06 PM, Justin Erenkrantz <jus...@erenkrantz.com>wrote:
> On Sun, Jan 23, 2011 at 8:42 AM, Stefan Küng <tortoise...@gmail.com> > wrote: > > Ok, tested with serf from the 0.7.x branch: memory rise is still higher > than > > with neon, indicating that there's still some (small) memory leak > somewhere. > > But checkouts and updates of even larger projects succeed without using > up > > all available memory as before. > > To compare: checkout of the TSVN source including all externals with serf > > uses up about 40MB more RAM than when using neon. I'd say that's ok. > > Does the memory keep going up? Or, does it reach a steady point? I'd > expect that ra_serf would use a bit more memory than ra_neon as it has > to manage a lot more than what neon has to do. > > As a point of reference, on Mac OS X 10.6, svn 1.6.x with ra_serf > checking out svn trunk peaks at 81MB, while ra_neon peaks at 12MB. > I have been looking at ra_serf's memory usage with svn trunk at the last svn days in Berlin and found that most memory was allocated and held in the sqlite code. Things have changed since then though. Can anyone test with svn 1.6.x to see how it uses memory? Lieven
