On Fri Nov 10 15:33:38 2023 +0000, Jacek Caban wrote:
The change keeping `outer_window` pointer alive for longer affects tens of NULL checks that we have in the code. Did you actually review the impact? It seems to me that a number of those checks are less correct after the change. I really think it needs closer look and probably some refactoring first so that the change has less impact. I'd suggest dropping all but first two patches for now.
What kind of impact are you expecting? Of course, places where we didn't check for it, are still probably going to crash (there probably are some of them), but that can't be *worse* than before, since it will actually happen less than before.
Is there some code that actually needs to know when it has been detached? I'll try to review all of them but if you know one off the top of your head, it would be interesting to see why.