[lazarus] Alignment

[Micha Nelissen]

Mattias Gaertner wrote:

> On Tue, 12 Aug 2003 16:34:24 +0200
> Micha Nelissen <M.Nelissen at student.tue.nl> wrote:
> 
> 
>>Hi,
>>
>>Currently aligning is done recursively, if you setbounds on a form which 
>>has a nested tabcontrol with a nested form with a nested form then the 
>>recursion goes down to the last form setting width/height and back up. 
>>The problem is that the win32 interface have differences between 
>>window/client size and this cannot be compensated if this resizing is 
>>done in one step. You can see this bug in the object inspector, 
>>differences accumulate causing the scrollbar to be partly visible among 
>>other things. It could be done if the form aligned it's controls and 
>>those controls call the interface immediately with their new 
>>position/size. Then the interface will report the new size back to the 
>>lcl which then continues aligning it's controls until all is done.
>>
>>How about it?
> 
> 
> We can add an interface function, to ask if bounds should be cached or sent
> directly. But before doing this, we should make sure that the win32 intf is
> doing, what it is told do to.
> I have created a new example with a three nested groupboxes, all
> Align=alClient. (Unable to commit, because now the name server does not even
> know cvs.freepascal.org).

I just keep retrying until it works, usually it does within a couple of 
tries.

> When you do this and add writeln in TWinControl.DoSendBoundsToInterface; and
> TWinControl.InvalidateClientRectCache;, you can see at least three bugs:
> 
> - The form is not resized after the LCL asked 
>   the win32 interface to do so.
> - the groupboxes are not resized after the LCL
>   asked the second time the win32 interface to do so
> - The win32 intf does not tell the LCL, that
>   the clientrect of the groupboxes have changed 
> 
> I think, these bugs must be fixed first before optimizing the LCL.

Ok, good point. Awaiting the example to appear in CVS...

Micha.