[ENet-discuss] 1.0 imminent - any requests?

[Lee Salzman]

So, I am basically almost going to go with this interface:

typedef enum
{
  ENET_VERSION = 1 /* or subsequent other point releases */
} ENetVersion;

typedef struct
{
  void * (* malloc) (size_t);
  void (* free) (void *);
  int (* rand) (void);
} ENetCallbacks;

void enet_initialize_with_callbacks(ENetVersion, const ENetCallbacks *);

Possibly with a macro ENET_INITIALIZE_WITH_CALLBACKS(callbacks) to
automatically handle passing in the version. I will figure out some way
to deal with the calling convention stuff.

Any objections to this?

Lee

On Tue, Dec 23, 2003 at 03:01:21PM -0500, Brian Hook wrote:
> My belief is that an API should have a "simple default" initializer 
> that does the standard stuff you'd expect, and then an enhanced one 
> that allows the user to specify specific parameters.
> 
> For example:
> 
> int api_initialize( void );
> int api_initialize_ex( const api_callbacks *cb );
> 
> typedef void *(*API_ALLOC)( size_t size );
> typedef void  (*API_FREE)( void *p );
> 
> typedef struct
> {
>    int cb_size;
>    API_ALLOC cb_alloc;
>    API_FREE  cb_free;
>   /* whatever else is necessary */
> } api_callbacks;
> 
> The "cb_size" allows you to rev this later without breaking binary 
> compatibility.
> 
> The default api_initialize would be something like:
> 
> void api_initialize( void )
> {
>    api_callbacks cb;
> 
>    cb.cb_size  = sizeof( cb );
>    cb.cb_alloc = __api_default_alloc;
>    cb.cb_free  = __api_default_free;
>  
>    return api_initialize_ex( &cb );
> }
> 
> You'll probably want to explicitly define the calling convention if 
> possible to avoid weird problems if the library is compiled for cdecl 
> but someone has their compiler default to fastcall or stdcall or some 
> other convention.
> 
> Brian
> 
> 
> 

> _______________________________________________
> ENet-discuss mailing list
> ENet-discuss at cubik.org
> http://lists.cubik.org/mailman/listinfo/enet-discuss