On Thu, 2007-03-29 at 10:32 +0200, Donn wrote:
What - was this not helpful? Perhaps someone can tell me at what figure these numbers would show a leak?
I think most of the developers are still asleep, though they may also be puzzled. It doesn't look like a particularly severe leak if there is one. ~300MB virtual size shouldn't be forcing a machine with 1GB of RAM into swap.
If the problem is indeed Inkscape, the only thing I can think of would be something like massive spikes in memory consumption, where the memory was still quickly freed again, and therefore missed at the sample intervals we took.
Since this behavior seems to be related to the general slowness-on-zoom issues, it may be that livarot's (our renderer's) tessellation code is exhibiting such behavior. That's a tenuous guess on my part right now though.
-mental