Re: [AD] END_OF_MAIN removal patch for msvc

["Stefanos A." <stapostol@xxxxxxxxxx>]

On Tue, 2008-11-18 at 18:25 -0500, Evert Glebbeek wrote:
> On 18 Nov 2008, at 18:07, Thomas Fjellstrom wrote:
> > Yeah, I'm not sure going whole hog on the event system is the right  
> > direction,
> > but neither are hacks like END_OF_MAIN.
> 
> END_OF_MAIN() actually works fairly well though, and it has the  
> distinct advantage of being very easy to change into a no-op on  
> platforms that dont (or no longer) need it. So unless there's a way  
> to truely make it a no-op on all platforms, maybe the best course of  
> action is the conservative course of action: not touch it.
> As hacks go it also isn't horrible, I've seen far worse. The ucontext  
> stuff sounds like a bad hack, but if it's actually part of the OS...  
> I don't know.

There are worse hacks that END_OF_MAIN, but not a whole lot :)

The biggest problem is that it is very user-visible. The user
(probably) doesn't understand why he has to type this. Even worse, the
program may still work without the macro, but then fail on other
platforms / configurations. This is the worst kind of bug, right in the
hands of every single Allegro user: obscure, hard to reproduce and very
easy to make.

In that light, SDL is better: the implementation may be even more hacky,
but at least it's not (much) visible to the user. GLFW is the best here:
no magic sauce, the user simply has to poll for events (an easy concept
to grasp).

We are all long-time users of A4 and there are things we take for
granted, but that shouldn't prevent positive changes from happening.


> 
> Anyway, while A5 is still in WIP stage at the moment, it should be  
> getting close to feature freeze or we're never going to make a spring  
> release date. Do we really want to discuss large changes to the  
> design (eg, changing the way Allegro works, mandating a message system)?

Which is better: delaying a release for a month or a bad API that sticks
for the next 5 years?