On Mon Jun 26 10:29:19 2023 +0000, Vladislav Timonin wrote:
Added missing period in message of first commit, and rebase while I'm at it. Also tried looking into the test crash. Out of 16 test bot runs, [only 1 failed](https://testbot.winehq.org/JobDetails.pl?Key=134116&f301=exe64.report#k3...). Thankfully, with a slightly useful log:
comdlg32:itemdlg start - itemdlg.c:346: test_instantiation itemdlg.c:531: test_instantiation end itemdlg.c:561: test_basics itemdlg.c:1050: test_basics end itemdlg.c:1153: Testing FileOpenDialog (advise) itemdlg.c:1161: Testing FileSaveDialog (advise) itemdlg.c:1169: test_advise end itemdlg.c:1213: test_events itemdlg.c:1334: test_events end itemdlg.c:1531: test_filename 0894:itemdlg: unhandled exception c0000005 at 000007FEFD1E90A9 comdlg32:itemdlg:0894 done (-1073741819) in 4sUnfortunately, still doesn't tell why, things of note here are the crash happened in `test_filename` (at least this specific crash) and exception handler failed to print last seen `winetest_set_location`.
It is probably unrelated, but shouldn't the WM_TIMER handler return 0 if the timer message was handled by the window proc, instead of TRUE like in tests/itemdlg.c, in both the existing cases and the new case? (Also one case with {} and the other two without {} looks a bit odd in my opinion)
I find MS docs difficult to interpret sometimes, but [this](https://learn.microsoft.com/en-us/windows/win32/winmsg/wm-timer) seems like it is the corresponding documentation, and I've found some example code snippets that return 0:
An application should return zero if it processes this message.
I've used the "WINETEST_REPORT_SUCCESS" checkbox on the testbot to get an idea where crashes occur. I believe it doesn't work when there are lots of successes to report though, I believe the log was cut or something.