Re: [AD] 4.3 graphic drivers |
[ Thread Index |
Date Index
| More lists.liballeg.org/allegro-developers Archives
]
On 2005-11-10, Elias Pschernig <elias@xxxxxxxxxx> wrote:
>
> Bob's idea clearly is the most robust - even if the driver was compiled
> against another version of Allegro, it would still work (since we build
> the vtable at runtime out of name strings and versions.. so Allegro
> always has complete knowledge).
Strings.. ugh. And maintaining all the compatibility.. double ugh.
> Evert's way allows binary compatibility, as long as prototypes don't
> change, and no crucial new entries are added.
>
> With Peter's way, the driver needs to be compiled against the new
> Allegro version if the vtable entries change. It's not very different
> from the current system in how it works, but cleaner than the completely
> static vtables we have in 4.2.
Plus my one is the only one that preserves type safety.
I think the static vtables are cleanest. The only reason any of our
three suggestion make any sense at all is the assumption that NULL is a
decent default for vtable methods.
Peter