On Tue Oct 8 12:33:36 2024 +0000, Zhiyi Zhang wrote:
Are you suggesting checking all the pixels or just some?
One pixel is probably enough. test_themed_background() is such an example of a visual test.
The problem is that I don't know exact placement of sibling windows that get corrupted. As I mentioned I'm using winecfg (where this bug was noticed originally), and replicating what winecfg doesn't seem reasonable. Why not use winecfg as an interactive test to check the correctness of listview repainting?
Besides, the patch fixes an obvious bug, user32 painting mechanisms are unversal and comctl32 controls are just normal user32 windows that must follow the user32 painting rules. So, I don't get where the speculation about comctl32 strange behaviour comes from.