On Fri, May 08, 2009 at 05:08:35PM -0400, Jorge Arellano Cid wrote:
On Thu, Apr 23, 2009 at 06:01:56PM +0200, Hofmann Johannes wrote:
Hi,
I did some more tests regarding that image related memory leak, Jeremy reported some time ago. It turns out that memory related to aborted images is not released. With CSS and the related repushs this seems to happen frequently. But I can also trigger it without CSS by hitting Ctrl-R while loading a large image. Attached diagnostic patch shows the issue. Load a page with a single large JPEG image and hit Ctrl-R while it is loading. In that case only the second instance of DilloJpeg is deleted.
I tried to figure out, why the first cache client is not getting aborted properly but the cache related code is quite complex... Does anyone know how this should work?
The repository has a patch for this bug. It has the debug messages for easy testing. I've given it some stress but would like to see how it works in the outside world.
Please give it some testing.
Cool! Also partially loaded images I saw sometimes seem to be fixed now. Cheers, Johannes