[PATCH] comctl32: Send EVENT_OBJECT_STATECHANGE when checking a checkbox

[Nikolay Sivov]

On 2/8/22 11:23, Alex Henrie wrote:
> On Tue, Feb 8, 2022 at 12:08 AM Alex Henrie <alexhenrie24 at gmail.com> wrote:
>> On Mon, Feb 7, 2022 at 11:48 PM Alex Henrie <alexhenrie24 at gmail.com> wrote:
>>> On Mon, Feb 7, 2022 at 11:27 PM Nikolay Sivov <nsivov at codeweavers.com> wrote:
>>>>> --- a/dlls/user32/tests/msg.c
>>>>> +++ b/dlls/user32/tests/msg.c
>>>>> @@ -1774,7 +1774,7 @@ static const struct message WmModalDialogSeq[] = {
>>>>>        { EVENT_OBJECT_FOCUS, winevent_hook|wparam|lparam, OBJID_CLIENT, 0 },
>>>>>        { WM_KILLFOCUS, sent|parent },
>>>>>        { WM_IME_SETCONTEXT, sent|parent|wparam|optional, 0 },
>>>>> -    { EVENT_OBJECT_STATECHANGE, winevent_hook|wparam|lparam|winevent_hook_todo, 0, 0 },
>>>>> +    { EVENT_OBJECT_STATECHANGE, winevent_hook|wparam|lparam, 0, 0 },
>>>>>        { WM_ENABLE, sent|parent|wparam, 0 },
>>>>>        { HCBT_CREATEWND, hook },
>>>>>        { EVENT_OBJECT_REORDER, winevent_hook|wparam|lparam|optional, 0, 0 }, /* Not sent on Win10. */
>>>>>
>>>> Doesn't this still pass in wine, without comctl32 change, with or
>>>> without winevent_hook_todo? I see we have winetest_platform checks
>>>> around winevent_hook_todo.
>>> I don't think the platform check has anything to do with it, but I
>>> just noticed that all of these tests pass todo=TRUE to ok_sequence,
>>> which means that they do indeed pass whether or not
>>> EVENT_OBJECT_STATECHANGE is specifically marked winevent_hook_todo. So
>>> you're right, these are not good tests for the feature I'm trying to
>>> implement.
>> This must be what you were looking at:
>>
>>          /*
>>           * silently drop hook messages if there is no support for them, mark
>>           * winevent todo's.
>>           */
>>          else if ((expected->flags & optional) ||
>>                   ((expected->flags & hook) && !hCBT_hook) ||
>>                   ((expected->flags & winevent_hook) && !hEvent_hook) ||
>>                   ((expected->flags & kbd_hook) && !hKBD_hook) ||
>>                   ((expected->flags & winevent_hook_todo) &&
>> !strcmp(winetest_platform, "wine")))
>>
>> Apparently ok_sequence treats all hooks as optional? That doesn't seem
>> helpful...
> It took me a minute but I think I get it: If we're expecting a hook
> and don't get one, it's OK. But if we do get a hook, it had better be
> the right one.
>
> More importantly, I missed the fact that there are two (entirely?)
> separate checkbox implementations in Windows: One in user32 and one in
> comctl32. The one that NVDA uses is definitely in comctl32, but
> comctl32's checkbox tests will have to be expanded to test hooks.

Yes, that's what caught my attention, why comctl32 button would affect 
user32 tests.

>
> -Alex