On 15 Dec 2011, at 14:29, David FLEURY wrote:

> 
>> Sorry, I should have said "seevral places in the flush() method without
>> checking..." though mostly to call the draw() method, so if draw() is
>> internally always checking that the xid is non-null it may well be
>> fine...!
> 
> Good. For now, I will keep my patch until my tests are ok for a next fltk release.

OK, though I still think there should be a Better Way!


>> Ah, yes. I see - so the issue is perhaps that we should prevent Fl_Tile
>> from resizing its subwindows to zero, we need to force a minimum
>> dimension, that sort of thing?


There's been discussion very recently as to whether a Fl_Tile ought to be allowed to resize down to zero or not at all, and I think the  emerging opinion os that it should not.

However, at present, it is allowed to become zero...

>>> 
>>> On thing I see is the code under WIN32 don't seem to be reach at all
>>> because draw() is not called for a 0 size window.
>> 
>> OK - but be aware that the WIN32 code is not in play here, you are
>> running the X11 code, so that's the code path that is failing for you.
> 
> 
> Not sure to understand the meaning of your sentence, but I have
> tested and debug my test code under Windows too.
> And the Fl_Double_Window::draw() is never called (breakpoint never reach)
> May be a solution would be to find why draw() is not done (w() != 0 && h() != 0 may be for children ?) under windows and done under Linux.

Sorry, what I meant was that when you are using an X-server (Exceed or otherwise) to run the apps on a Linux host, but displaying on a Windows host (which was the configuration you described initially) then, in that case, the WIN32 code is not in play, only the X11 code, since it is the X11 system that is rendering the display, not the WIN32 system.

However, from your subsequent posts, I think you are also testing the code locally on the Windows machine, and as a result we may have been talking at crossed purposes?