FreeLists / haiku-appserver / [haiku-appserver] Re: partly paiting

[Adi Oanca <adioanca@xxxxxxxxxxxxxx>]

Stephan Aßmus wrote:
> >    Boy oh boy, I must be out of my mind!!
>
> [8<]
>
> >    Sorry about that, apologies.
>
>
> Hey, no worries Adi! It's not like we wanted to release the beast tomorrow.
>
>
>
> > > The problem that my code solved, was that B_VIEW_RESIZED was not sent for 
> > > all
> > > views that needed it. The way I solved this was to send a single message
> > > which contained the tokens from all effected views. The comment you 
> > > removed
> > > explained where this was done in Layer.cpp and what the potential problems
> > > were.
> >
> >    About this...
> >    Should we proceed like this, or should we send an individual message for
> > each view? To be compatible with R5, ::CurrentMessage() should return 
> > correct
> > values for any view, don't you think?
>
>
> If it works indeed by sending an individual message for each view in R5, then 
> I must say, this is quite an overhead. I would assume that sending a single 
> message is much cleaner and faster.

    You know I agree here. I haven't tested in R5, but I assume it's one
message per view.

> Resizing a single view (for example the 
> top window view) will result in all children with B_FOLLOW_* flags to be 
> moved and/or resized. On the app_server side, this is an "atomic" operation, 
> but when messages are sent for each view individually, I have a bad feeling 
> in my tommy, that this is a potential source of out-of-sync problems on the 
> client side. We could still emulate the single current message on the client 
> side and stick to the way I did it. What do you say?

    I agree sending one message only. Let's stay with this solution and see
if all works OK. I just am afraid of ::DetachCurrentMessage() because, if this
method is called from ::FrameMoved/Resized() the user will be dumped into 
debugger
with a seg fault; more checking is required for "case B_VIEW_RESIZED/MOVED:"


bye,
Adi.