* On Tue, 8 Jan 2008, Eric Pouech wrote: > * Andrew Riedi a écrit : > > > > dlls/user32/tests/cursoricon.c | 201 > > ++++++++++++++++++++++++++++++++++++++++ ... > > +static void do_child(void) > > +{ > > + WNDCLASS class; > > + MSG msg; > > + BOOL ret; > > + > > + /* Register a new class. */ > > + class.style = CS_GLOBALCLASS; > > + class.lpfnWndProc = callback_child; > > + class.cbClsExtra = 0; > > + class.cbWndExtra = 0; > > + class.hInstance = GetModuleHandle(NULL); > > + class.hIcon = NULL; > > + class.hCursor = NULL; > > + class.hbrBackground = NULL; > > + class.lpszMenuName = NULL; > > + class.lpszClassName = "cursor_child"; > > + > > + SetLastError(0xdeadbeef); > > + ret = RegisterClass(&class); > > + ok(ret, "Failed to register window class. Error: %d\n", > > GetLastError()); ... > > IMO, the ok() tests in the child process are a bad idea (they won't be > counted, nor returned as errors, by the parent process)
Was Wine test framework architecture done such way on a purpose? Why would it be a bad idea to take into account a child output also?