On Fri, 17 Jul 2020, Zhiyi Zhang wrote: [...]
The upgrade makes the problem go away, even for debiant0615. See https://testbot.winehq.org/JobDetails.pl?Key=75654 Sadly, the real issue isn't found.
The issue did not seem to happen consistently. So it may still appear from time to time (though I hope not). It is strange that it just disappeared.
I ran user32:monitor on my box a few times too (Intel GPU). One thing I noticed is that it ChangeDisplaySettingsEx() often takes about 800 ms. Then flush_events() often adds another 200 ms. Since there are a lot of resolution changes the test takes about 85 seconds. Fortunately that's still under the timeout limit.
43 flush_events calls -> ~8.6 s not sure about the number of ChangeDisplaySettingsEx() but I expect it's more than 43 so that would account for another > 34 s
I also get a bunch of Vulkan warnings that I may be able to fix by explicitly enabling DRI3 (or upgrading to Debian Testing?). Something to test for the next time I reboot... But I don't really expect the delays to be related to these warnings.
monitor.c:651:4.581 ChangeDisplaySettingsExA(\.\DISPLAY1,pos=0,fields=0) INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete INTEL-MESA: warning: Haswell Vulkan support is incomplete monitor.c:665:5.458 Test succeeded monitor.c:674:5.458 flush_events monitor.c:676:5.659 Test succeeded