On 20 October 2011 21:05, Noel Grandin <noelgran...@gmail.com> wrote: > Just checked in another fix, this one seems to behave at least as well as the > old code on my machine. > > Sorry about the bumpy ride :-; > > Regards, Noel. >
Thanks for sticking with it, Noel. This version at least appears to work for the particular app where I first noticed the problems. The main app (mostly standard Pivot Components & skins) paints properly as far as I can see, and the transparent overlay also now updates as expected. I don't have time to test this any more thoroughly at the moment, so will back out the change locally because I can't afford to lose time tracking down issues that may or may not be due to this optimization. Unless we can get a decent amount of user testing and feedback on the change, and we can be confident that there are no more issues with it, then I propose pushing it back to a release *after* 2.0.1. A patch of the most recent version of the change can be attached to the JIRA ticket to be applied by any interested parties. As this is nominally a performance enhancement, then we should also get a small test app to use for benchmarking before & after versions of the code. (There are currently 2 unanswered requests for sample code on the ticket) As arbitrary as benchmarks can be, I'd prefer one to simply assuming that a planned change ends up being as beneficial as intended. The ticket mentions that a profiler highlighted the 'problem' area, so any instructions for reproducing the profiler & test configuration would also be useful. I can move my comments/concerns to the JIRA ticket after any further discussion here. Chris