Fixing devPrivates

[tiago.vignatti at nokia.com]

On Thu, Apr 29, 2010 at 04:43:47PM +0200, ext Alan Coopersmith wrote:
> Tiago Vignatti wrote:
> > 
> > Reviewing your proposal made me think if we really need devPrivates mechanism
> > at all.
> > 
> > It only exists to not change ABI all the time on data structures. But hey, is
> > this a _real_ problem? I mean, for who cares about ABI, we have a way to track
> > the control changes just going to xf86Module.h and bumping ABI_*. 
> 
> You'd prefer having all the structures allocate space for extensions that may
> not be enabled (either at runtime or at compile time)?   And having to recompile
> everything every time an extension needs to add a new member?

yes. yes.

And when extensions are not used then would be just nil pointers in those
structures, which doesn't cost much.

 
> Because having the ABI depend on #ifdefs of whether the server was compiled with
> DRI2 or XF86DGA is effectively having no ABI at all, and killing the ability to
> provide driver binaries, whether closed source like nvidia, or open source like
> the vmware & virtual box guest drivers they include in their guest additions kits.

it wouldn't kill the ability of provide binaries. Would difficult though.


And you totally ignored my readability argument, which you know counts a lot
in terms of development.

                            Tiago