Re: XP [was: Commit -- Insert symbol dialog Modeless for WIN32]


Subject: Re: XP [was: Commit -- Insert symbol dialog Modeless for WIN32]
From: Martin Sevior (msevior@mccubbin.ph.unimelb.edu.au)
Date: Sun May 21 2000 - 06:53:55 CDT


On Sun, 21 May 2000, Mike Nordell wrote:

> Martin Sevior wrote:
>
> > The widget pointer is cast as a (void *) and stored that way in
> > xap_App. If there is no running id the value of the pointer
> > is (void *) NULL, which is valid C++ on all platforms.
>
> So is storing a Win32 HWND or a BeOS BWindow* as a void* in the
> baseclass, but it's still platform dependent data (albeit in "hidden"
> form) in XP code.

Yes, that was EXACTLY the idea. Every platform needed to be able to get to
a centrally stored widget pointer. The best place to put them all was in
XAP_App since then I only need to write the code once, not N times for
XAP_UnixApp, XAP_WinApp etc. I needed
this at first when I was doing my bottom-up coding to make the thing work.
Now I can see I don't need to store platform-specfic widget pointers in
XAP_App. However I will rewrite isRunning so that it is only in
XAP_Dialog_Modeless and not a virtual function without refering to widget
pointers. It doesn't need to be virtual since if it's "ID" is in the
modelessID table in xap_App it running.

I've always been a bottom-up programmer. I think the best way to
write C++ is do bottom up coding to work out how to code it and to find
out just what you need to make a task work. After that the bottom up
design can be rewritten into a proper class heiracy. Of course this won't
always work but it is a good way for me to procede right now.

> can't really *get* to that data, much less *set* it, without violating
> some very basic design principles, and in that case that void* is
> probably the lesser of the problems.
>
 While I agree with the asthetics of this sometimes pragmatism has its
place. That code got the modeless dialogs up and running solid. I can see
this point now and in fact there is a fair bit of extraneous code left
over from the original implementation. I'll clean all that up and
hopefully fix Jaoquin's segfault by tomorrow night.

Cheers

Martin



This archive was generated by hypermail 2b25 : Sun May 21 2000 - 06:54:06 CDT