Re: [v2] comctl32/taskdialog: Implement nDefaultButton and add tests

[Fabian Maurer]

I reworked my patch according to your suggestions.
> > > We already iterate over custom buttons and check every common button
> > > flag before this point, so it shouldn't be necessary to have another loop.
> >
> > I'm open to suggestion, but how would you do that?
> >
> > We could add three checks each time before taskdialog_init_button is 
> > called, 
> > or alternatively one check in taskdialog_init_button. Like giving it a 
> > pointer 
> > to default_button, and checking if the id equal the taskdialog default 
> > button.
> > But I find my current solution to be cleaner to be honest, what's your 
> > suggestion?
> > Yes, something like that. taskdialog_init_button() has all the info to
> > do that.

In the beginning I wanted to have default_button be -1 if no button was found 
yet, but as the tests show this is impossible since windows accepts a negative 
ID. Now we have to pass 2 pointers around..
Your thoughts on that? I find that the second loop makes things a lot cleaner.

Regards,
Fabian Maurer

[v2] comctl32/taskdialog: Implement nDefaultButton and add tests

[Nikolay Sivov]

On 05.11.2017 18:46, Fabian Maurer wrote:
> I reworked my patch according to your suggestions.
> > > > We already iterate over custom buttons and check every common button
> > > > flag before this point, so it shouldn't be necessary to have another loop.
> > >
> > > I'm open to suggestion, but how would you do that?
> > >
> > > We could add three checks each time before taskdialog_init_button is 
> > > called, 
> > > or alternatively one check in taskdialog_init_button. Like giving it a 
> > > pointer 
> > > to default_button, and checking if the id equal the taskdialog default 
> > > button.
> > > But I find my current solution to be cleaner to be honest, what's your 
> > > suggestion?
> > > Yes, something like that. taskdialog_init_button() has all the info to
> > > do that.
>
> In the beginning I wanted to have default_button be -1 if no button was found 
> yet, but as the tests show this is impossible since windows accepts a negative 
> ID. Now we have to pass 2 pointers around..
> Your thoughts on that? I find that the second loop makes things a lot cleaner.

I sent another version that doesn't touch helper interface and uses
single variable to mark default button, with otherwise unchanged logic
and tests.
> Regards,
> Fabian Maurer