On 2003/03/12 16:31, Tuomo Valkonen <[EMAIL PROTECTED]> wrote:
> The programs that are "incorrectly" placed probably specify a position
> in wm size hints. Ion places the window in the frame under those
> coordinates. There used to be a CF_PLACEMENT_GEOM option that had

I didn't check that, but wouldn't this also mean that Emacs and rxvt
would be placed in the wrong frame in single-headed mode? Worked
perfectly in single-headed mode for me. 

Or did you mean viewports, not frames?

> This isn't a bug either, but a feature :). Ion doesn't check for
> EnterNotify events while the pointer is grabbed for other reasons.
> I might add some kludge to do so in this case, though.

I understand.. this bug doesn't annoy me, because I don't use the
mouse to drag windows in most cases. In fact, when I read about the
new feature that allowed me to switch workspaces, my first thought was
"There might be this bug", I tried it out, and I happened to be right
;)

> I am not such a big fan of C++. About the only usable things in the
> language (or should I say "awful kludge on top of C") are the very
[...]

I don't like the standard C++ library either. But in my opinion, your
ion code is much more an awful kludge on top of C ;) I'm only talking
about the syntax, which is OK in C++, but looks strange in your code.

> It shouldn't be too hard to convert all the class stuff to C++, but
> there would then be the memory overhead of a unused c++ library (I

You are not required to link the C++ library if you don't use it. No
overhead here.

> stick to nice c functions instaed of the c++ crap) and in some aspects
> the the dynamic function code in Ion is more convenient: no need to
> specify all the functions in one place. Also, IIRC (it's been a long

That's a good point against C++.. reminds me somehow of reflection in
Java. But I'm not sure yet if this is really required, I havn't read
the ion code completely yet.

> of an inheriting class and then the parent. But I really need to process
> "watches" first with a almost fully working object (watches are used to
> keep track of objects that could be destroyed without a pointer holder
> otherwise knowing of it). Would it be possible to overload delete to do
> this?

Overloading "delete" would be another questionable hack.. What you
need doesn't seem to be possible with standard C++
destructors. Processing watches is IMHO not the job of a destructor -
maybe a method "on_destroy" (or some other name) processes watches,
and before you delete an object, call this method. Just one idea, and
yes it's also a hack which obfuscates the code a little.

max

Reply via email to