https://bugs.winehq.org/show_bug.cgi?id=55396
--- Comment #6 from François Gouget fgouget@codeweavers.com --- Created attachment 74973 --> https://bugs.winehq.org/attachment.cgi?id=74973 Minimal advapi32:registry to break the other tests
* Looking at the w8adm test configuration I don't see anything wrong: Windows update is disabled, there is no process that should not be running, etc. * I also restored the VM from backup but the tests fail in the same way even with the old amount of memory (but note the live snapshot is always rebuilt by the TestBot). * I also verified that before the update the tests were running without elevated privileges. The old logs on test.winehq.org have the right set of skips indicating that this was the case. The current logs have the same set of skips too (particularly in advapi32:registry). * So now I'm thinking these tests should have failed all along and we somehow got lucky.
Then I hacked advapi32:registry to find which part causes the other tests to fail. This shows that the line below in test_reg_open_key() is sufficient to cause the other tests to crash (see the attached patch).
ret = RegCreateKeyExA(HKEY_LOCAL_MACHINE, "Software\Wine", 0, NULL, 0, KEY_WOW64_32KEY | KEY_ALL_ACCESS, NULL, &hkRoot32, NULL);
Also this call is preceded by a check for the presence of IsWow64Process which does not skip: w8adm is 32-bit only but the WoW APIs are present regardless. Maybe the fix is to tweak that check to skip in the non-elevated privileges case? Or maybe testing WoW-stuff on a non-WoW configuration is just wrong even though it does not seem to break things in the w8 (elevated privileges) configuration.