1. Unzip attached file and load HTML page in WebKit or Safari 3.x 2. Press Start Test button 3. Monitor application RSIZE in Activity Monitor as test runs RSIZE steadily increases as multiple background images are swapped repeatedly with the same image urls. The more elements that swap background images, the faster memory usage increases. This only seems to occur when setting the background url(), not when setting background colors or other background properties. It looks like the same images are being allocated repeatedly (and never deallocated) for every JS call to "element.background = url(image)" I ran leaks from Terminal but didn't see anything unusual. FWIW, I see the same memory increase in Safari 3 and 3.1 but NOT in Safari 2.x or Firefox 2.x.
Created attachment 21069 [details] Test case
I cannot reproduce any memory growth on the test case with a recent build of WebKit.
Created attachment 21099 That's strange... I'm able to duplicate it consistently on an Intel MBP and a PowerPC iMac both running Leopard. Tested on r33029 build of WebKit on both machines and saw the same memory growth immediately (the longer the test runs the higher it goes). I've attached a screen capture so you can see what I'm seeing.
I am not able to see Safari process increasing memory (I might be reading the Activity Monitor wrong here) but I do see that this test uses a lot of CPU time and while everything closed in the background except few apps, it pushed the website testpage to consume 50% plus CPU in activity monitor while opening GitHub reduced it to 30% range. Is it something worth investigating further? Thanks!
Yeah, I can't reproduce this.