Re: [AD] Proposal for a set_icon function

[Angelo Mottola <a.mottola@xxxxxxxxxx>]

Sorry to all for letting the discussion to fall down the last week, I've been 
somewhat busy in the meantime...

Anyway, I'll restart from where we stopped, adding a summary of our 
possibilities on the set_icon() function proposal:

1) set_icon(BITMAP *large_icon, BITMAP *small_icon) where large_icon and 
small_icon have to be 32x32 and 16x16 versions of the icon to set. If one of 
them is not present, it is generated by stretching the other one. If both are 
unavailable, the default platform app icon is chosen.

2) set_icon(BITMAP *large_icon, BITMAP *small_icon) where the two icons can 
be any size/depth, and are then stretched/color converted to fit into the 
platform dependent icon sizes the OS provides. Obviously these icons would 
look best if directly passed with the proper sizes the running OS expects.

3) same as 2) but just with one icon, simply stretching to generate the other 
missing one.

4) set_icon(BITMAP *icon, int w, int h) is too much elaborated IMHO (when one 
is setting an icon he probably doesn't care about the size, and just wants to 
pass it a bitmap) and doesn't provide two versions of the same icon.

I'd go for a version with the 2 icons possibility, as stretching generally 
doesn't generate good looking pictures when you're handling small images as 
icons.

In any case, the function would set the icon for the window/taskbar only; 
under BeOS I'm not sure this is possible, as when you change a program icon 
this is probably automatically stored as an attribute for the file, thus 
remaining attached to it even when the program has exited.
The BeOS behaviour could be handled as an exception though, specifying it in 
the docs...

Now I'll wait your comments, then if you agree I could start coding it.

-- 
Angelo Mottola
a.mottola@xxxxxxxxxx
http://www.ecplusplus.com