Paul Gofman (@gofman) commented about dlls/ntdll/tests/sync.c:
+ + if (tests[i].queue_apc) + ok( apc_count > 0, "no APCs executed.\n" ); + + /* test Sleep (non-alertable only) */ + if (!tests[i].alertable) + { + SetLastError( 0xdeadbeef ); + Sleep( timeout.QuadPart ? (-timeout.QuadPart / 10000) : 0 ); + ok( GetLastError() == 0xdeadbeef, "got error %#lx.\n", GetLastError() ); + } + + /* break out early once we see the expected result */ + if (!tests[i].alertable && !tests[i].timeout && noyields[0]) + break; +#if 0 /* TODO */ We place things which are to be skipped by stay for documentation in 'if (0)', not #ifdef. But I still think it is not needed and it is unclear what is TODO there, I think you might just remove those iterations, count arrays and allow both results for random part (_SUCCESS vs _NO_YIELD_PERFORMED).
-- https://gitlab.winehq.org/wine/wine/-/merge_requests/7169#note_92950