On Thu, Jun 26, 2014 at 05:46:56PM +0100, Eric Anholt wrote: > Emil Velikov <[email protected]> writes: > > Not aiming at it atm, but just curious - which one would you consider as the > > better option: > > - Convert waffle to use libepoxy, or > > - clone the current implementation into waffle and work from there ? > > Make waffle link against libepoxy.
I don't want to import the libepoxy code into waffle. But I'm also not fundamentally opposed to adding a written-from-scratch dispatch mechanism to waffle. But, we should avoid adding any written-from-scratch dispatch until (1) waffle's users (piglit, apitrace, maybe a few more?) really need dispatch and (2) we've made a real effort to make libepoxy work. (There does exist a waffle user, Piglit, who *does* need dispatch, and libepoxy today doesn't work for it, and I just recently rewrote large hunks of Piglit's homebrew dispatch, but I digress... I'm still waiting for libepoxy to become Piglit-friendly). As for why Eric didn't add dispatch to waffle, it's because most users just want dispatch and not the GLX/CGL/Android/EGL abstraction that waffle provides. Any realworld app, like a game, needs much more from a platform abstraction layer than what waffle provides and will ever provide, such as input and audio. Since realworld games already have chosen their own platform abstraction layer, it doesn't make sense to require them to add another platform abstraction (waffle) to be able to get sane GL dispatch. _______________________________________________ waffle mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/waffle
