[RFC] GLX dispatch rewrite

[Jon TURNEY]

On 30/11/2013 00:07, Keith Packard wrote:
> Jeremy Huddleston Sequoia writes:
> 
>> We can't guarantee what symbols (ie OpenGL extensions) will be
>> available in libGL at link time and runtime which is why I wrote
>> _glapi_create_table_from_handle.  This fills out the dispatch table
>> with entry points into our libGL (or no-op entries if unavailable).
> 
> Ok, I think I'm getting a better idea of what's going on at least,
> thanks for answering over the weekend :-)
> 
>> Specifically, we have these linking issues due to capabilities not in
>> our libGL at build time.
> 
> Talking with Jon on IRC, it sounds like he really does want the old
> dispatch tables so that he can flip between Mesa swrast and native GL on
> the fly.
> 
> Ideally, I think what we want is a header file that would #define the
> gl names to go through a dispatch table, so that only systems that
> needed GL thunks would end up using them. And, for the dispatch table to
> be generated from mesa source, of course.

This seems entirely plausible, but also a bit of a reversion.  It also seems a
bit superfluous to write this when there is already a perfectly good dispatch
table implementation inside mesa's libGL.

I've written an implementation which uses the mesa (private) glapi to create a
dispatch table, and that seems to work, but a patch to mesa to install glapi.h
seems unlikely to be accepted.

> Is this a simple matter of pulling bits out of the old tree and sticking
> them somewhere? If we can limit the effect of these changes to OS X and
> Windows, I'm reasonably comfortable closing the non-critical window down
> and having work on the fix go on in parallel with other testing.