Re: [AD] END_OF_MAIN removal patch for msvc

[Evert Glebbeek <eglebbk@xxxxxxxxxx>]

On 16 Nov 2008, at 18:09, torhu wrote:
> I've had some trouble building allegro 5 on msvc, but now I've got  
> it working.  So I've attached a patch which makes END_OF_MAIN just  
> an empty macro on msvc.

Cool. Is there a similar way to do it for MinGW?

> 1) This means that apps built with allegro 5 on msvc will by  
> default have a console window.  But this is what happens on other  
> Windows compilers already, at least mingw does this.  So how to  
> disable the console will definitely become a FAQ for msvc users.   
> Adding the required linker option to allegro headers is an option.   
> See below.
>
> #ifdef ALLEGRO_USE_CONSOLE
>   #pragma comment(linker,"/SUBSYSTEM:windows").
> #endif
>
> Not sure I like this, though.  Seems a bit over the top, or what?

Shouldn't that be #ifndef?
Anyway, I personally generally dislike #pragma's, but in this case  
I'll go practical and say I have no opinion either way. If we want to  
be newbie-friendly it might be a good idea to have it (my suspicion  
is there'll be more newbies using MSVC than MinGW).

> 2) This opens up the possibility to remove ALLEGRO_USE_CONSOLE and  
> ALLEGRO_CONSOLE_OK.  They're only useful on Windows.  Other  
> platforms always define ALLEGRO_CONSOLE_OK, and ignore  
> ALLEGRO_USE_CONSOLE.

ALLEGRO_CONSOLE_OK is defined if printf() can be meaningfully used,  
right? (Obviously subject to whether the program was started from a  
GUI or not). I can't say I've ever cared about or used either, so the  
implication of pulling these isn't very clear to me.

> 3) OS X and QNX uses END_OF_MAIN to get hold of the address of  
> _mangled_main.  A different solution needed?

As far as I understand the situation on OS X (someone correct me if  
I'm wrong), the problem is that on OS X messages can only be received  
from the main thread. So either we need to have the user respond to  
the messages themselves (obviously not desirable for a multi-platform  
library like Allegro), or we hijack the main thread and run the user  
main() from a thread we create ourselves. This is what the magic main  
does on OS X. As far as I know, there have been a few attempts to  
work around this, but there doesn't seem to be a good solution.


>   And is ALLEGRO_WITH_MAGIC_MAIN needed?  It's off by default.

It might be on *NIX, if there's no other way to grab the commandline  
arguments (which are needed to get the executable path). Disclaimer:  
this is the case in 4.2/3, I don't know if this is still true of 4.9  
of the top of my head.

Evert