http://bugs.winehq.org/show_bug.cgi?id=4964
------- Additional Comments From juris.smotrovs@sets.lv 2006-26-05 06:27 ------- Created an attachment (id=2528) --> (http://bugs.winehq.org/attachment.cgi?id=2528&action=view) Patch correcting the problem (enabled by solution of bug 4960)
This is a smaller patch dealing only with this bug's problem. It uses the undocumented SWP flag for window state change introduced with the solution for bug 4960. It defines it as a constant and uses it in default window procedure's WM_WINDOWPOSCHANGED processing, to detect that a window changed state, and send WM_SIZE in that case. Additionally, it enhances checking for todo flag within dlls/user/tests/msg.c test, since otherwise a message sequence marked with "todo" starts producing errors (because currently there is no checking for todo at defwinproc flag comparing), which it should not. In fact, the situation in the test improves: the sequence of expected messages actually sent increases, however that leads at coincidence at the second part of the test: expected ...0x46 by defwinproc, 0x83 by defwinproc, 0x47 by defwinproc, 0x46, 0x83, 0x47; actual ...0x46, 0x83, 0x47. The problem is that some (defwinproc) messages are not sent, but the message comparer in the test thinks that messages are correct, but have incorrect sender. (A solution for this is included in the bigger patch earlier sent.)