On Tue Dec 10 08:03:08 2024 +0000, Michele Dionisio wrote:
do you have any idea? I cleanup this merge rewuest removing the test that not works or rewriting to never use the default crtl-c handler?
The best forward IMO is to mark the failing test with flaky_wine, and keep timeout as it should be (don't try to emulate STATUS_CONTROL_C_EXIT). From examination of the test results, the function behavior is correct (timeout is stopped by ctrl-c cf previous post), it's only the exit code which isn't the correct one. As of today we don't have applications expecting the STATUS_CONTROL_C_EXIT exit code (it's something that has been discovered by the timeout tests). Marking the test as flaky_wine: 1) clearly marks that Wine behavior is not exactly what native does 2) and lists the expected behavior
For the cause, it could be a race between the two threads when exiting the process, and also a race with conhost termination. Or something else. Unfortunately, I can't repro locally. Unless you want to keep investigating, I could take over on this precise point.